Proceedings of a Workshop

INTRODUCTION

Pet dogs and cats breathe the same air, sit on the same couch, play in the same yard, drink the same water, and often sleep in the same bed as their human companions. Because they share the environment of humans, these companion animals are exposed to the same environmental agents. Cats and dogs are also susceptible to many of the same age-related diseases as humans, including cancers associated with environmental risk factors, although they age more rapidly and show the effects of exposures sooner. Given the lifelong immersion of companion animals in the natural human environment, data collected on their exposures and health outcomes have the potential to provide new insights into environmentally induced disease that complement traditional laboratory, clinical, and public health studies. Furthermore, because companion animals experience a similar spectrum of disease as humans, with earlier onset, they could serve as sentinels for environmental risks to human health.

To examine the potential role of companion animals as sentinels of relevant, shared environmental exposures that may affect human aging and cancer, the National Cancer Policy Forum held a workshop¹ in collaboration

___________________

¹ The planning committee’s role was limited to planning the workshop, and the Proceedings of a Workshop has been prepared by the workshop rapporteurs as a factual summary of what occurred at the workshop. Statements, recommendations, and opinions expressed are those of individual presenters and participants, and are not necessarily endorsed or verified by the National Academies of Sciences, Engineering, and Medicine, and they should not be construed as reflecting any group consensus.

with the Forum on Aging, Disability, and Independence and the Standing Committee on the Use of Emerging Science for Environmental Health Decisions to explore this promising and underutilized pathway for research. This hybrid workshop was held at the Keck Center in Washington, DC, and concurrently online on December 1–3, 2021. The planning committee developed the agenda for the workshop sessions, selected and invited speakers and discussants, and moderated the panel discussions. In designing the workshop, the planning committee focused on identifying gaps in the research and solutions to close them.

The workshop convened an array of experts in diverse fields representing academic veterinary and medical centers, universities and schools of public health, government agencies, industry, and pet owners. The expertise of the invited speakers included environmental exposures, comparative oncology, biomonitoring, data management, veterinary bioethics, and environmental justice, as well as other areas. The workshop was open to the public and audience members also included a variety of perspectives. Presentations and panel discussions covered the current state of the science and pathways for accelerating research, along with opportunities and challenges for using this novel translational approach to exposure science to advance human health. This workshop proceedings is the rapporteurs’ summary of the speakers’ presentations and the moderated panel discussions. The moderated discussions included panelists’ responses to questions from both the in-person and virtual audience, as well as audience comments; when identified, audience comments are attributed by name and affiliation. The workshop statement of task is in Appendix A, the agenda in Appendix B, and the biosketches of workshop planning committee members and invited speakers in Appendix C. The web-cast and speakers’ presentations have been archived online.²

Planning committee chair Linda S. Birnbaum (National Institute of Environmental Health Sciences [NIEHS] and National Toxicology Program [NTP], Emeritus; Duke University) opened the workshop by outlining the topics to be explored, which included:

• Potential data sources needed to assess whether companion animals may serve as sentinels for human environmental exposures;
• The state of the science for biomarkers of exposure and use of biosensors for application to companion animal populations of interest;
• Best practices for collection, storage, and analysis of biosamples to assess exposures;

___________________

² See https://www.nationalacademies.org/event/12-01-2021/the-role-of-companion-animals-as-sentinels-for-predicting-environmental-exposure-effects-on-aging-and-cancer-susceptibility-in-humans-a-workshop (accessed December 10, 2021).

• Strategies for standardizing, sharing, and aggregating health records and relevant metadata across species; and
• Current policies and regulations related to monitoring and mitigating environmental exposures and the role for prospective interventions based on companion animal data.

Over the course of 3 days, participants offered many suggestions to advance the use of companion animals as sentinels for human environmental exposures, cancer, and aging, and these are summarized in Box 1. Additional details regarding these suggestions can be found throughout the proceedings.

BACKGROUND ON CANCER, AGING, AND ENVIRONMENTAL EXPOSURE RESEARCH

Ned Sharpless (National Cancer Institute [NCI]) placed this workshop into context with respect to the current state of cancer research and environmental exposures, as well as research needs. There have been significant advances in cancer treatment in recent years, most dramati-

cally the development of effective treatments for some melanomas and lung cancers. He noted that while he served as Acting Commissioner at the U.S. Food and Drug Administration (FDA), one-third of newly licensed therapeutics and devices were cancer related. Nonetheless, social

determinants of health,³ including exposure to environmental carcinogens, can lead to significant disparities in cancer incidence and outcomes. Although biological research tends to focus on understanding the precise mechanism by which a given agent causes a particular cancer, the FDA and the U.S. Environmental Protection Agency (EPA) face a different challenge, which is the practical need to regulate multiple exposures on a daily basis. This is of pressing importance, and the NCI and NIEHS can provide the scientific underpinning needed to help inform these policy decisions, said Sharpless.

The NCI currently funds roughly 80 extramural grants to study environmental impacts on cancer, covering a wide range of topics, said Sharpless. In addition, environmental agents are a focus of larger extramural and intramural studies, including many cohorts within the 20-year-old NCI Cohort Consortium.⁴ The NCI supports the monographs program of the World Health Organization’s (WHO’s) International Agency for Research on Cancer (IARC). The IARC aims to classify environmental carcinogens;⁵ recently initiated the Connect for Cancer Prevention study, which follows 200,000 individuals over time with collections of biospecimens⁶ and detailed surveys of environmental exposures;⁷ and participates in the landmark NIEHS/NCI Agricultural Health study, which monitors the effects of pesticides.⁸ Intramural research within the NCI’s Division of Cancer Epidemiology and Genetics⁹ includes:

• The Diesel Exhaust in Miners Study (DEMS), focused on lung cancer (Attfield et al., 2012);
• The New England Bladder Cancer Study, monitoring the effects of arsenic and other contaminants in drinking water;¹⁰

___________________

³ Social determinants of health are the conditions in the environments where people are born, live, learn, work, play, worship, and age that affect a wide range of health, functioning, and quality-of-life outcomes and risks. See https://health.gov/healthypeople/objectives-anddata/social-determinants-health (accessed March 28, 2022).

⁴ See https://epi.grants.cancer.gov/cohort-consortium/cohort_projects.html (accessed February 3, 2022).

⁵ See https://monographs.iarc.who.int (accessed February 3, 2022).

⁶ Biospecimens are also known as biological samples.

⁷ See https://www.cancer.gov/connect-prevention-study (accessed February 3, 2022).

⁸ See https://aghealth.nih.gov (accessed February 3, 2022).

⁹ See https://dceg.cancer.gov (accessed February 3, 2022).

¹⁰ See https://dceg.cancer.gov/research/cancer-types/bladder/bladder-new-england (accessed February 3, 2022).

• A study examining the impact of coal-fired stoves and heaters on the development of lung cancer in never-smoking women in China;¹¹ and
• An examination of the effects of pesticides on multiple myeloma (Hofmann et al., 2021).

The NCI also chose environmental carcinogens as one of six topics to focus on for increased cooperation with the United Kingdom. The NCI has extensive interactions with industry, including supporting the development of novel therapeutics and devices, publishing data sets through NCI portals, and funding mechanisms to support the development of new companies, said Sharpless.

A recent NCI meeting, Breast Cancer and the Environment,¹² highlighted many promising new tools for studying the cancer–environment connection, including ultrasensitive detection, data aggregation, and exposure monitoring. Referencing the emerging science of mutational signatures, Sharpless envisioned modern, molecularly precise Koch’s postulates that could ascribe a particular cancer to a particular agent based on the pattern of somatic DNA mutations in the tumor. James DeGregori (University of Colorado) noted the role played by “selectogens”—chemicals that induce cancer by altering the context around the tumor rather than by mutagenesis. There will likely be many such compounds with a variety of mechanisms, such as the induction of chronic inflammation, and no clear mutation signature, agreed Sharpless. “They are environmental agents that clearly induce cancer . . . but they are not mutagens. They do not damage DNA. They do not break DNA. They do not bind to DNA. They do not have an ascertainable signature,” said Sharpless, adding that a scientific approach is needed to study them. Birnbaum suggested standardizing in vitro bioassays and defining a set of key characteristics for evaluating potential carcinogens that do not damage DNA.

New approaches are needed to study the links between environmental agents and cancer, said Sharpless. The time-honored troika of biological plausibility, rodent research, and cohort studies is useful, but many rodent studies have failed to predict outcomes in humans, he said. Companion animals, which share their human’s environment, could potentially bridge this gap. The NCI is also very interested in studying cancer immunotherapies in clinical trials for companion dogs with naturally occurring tumors, said Sharpless. Because rodents tend to be poor models for evaluating immuno-oncology agents, these treatments often go straight to human clinical testing, which is very inefficient, and Sharpless expressed hope that dog trials could advance treatments for both

___________________

¹¹ See https://dceg.cancer.gov/research/what-we-study/indoor-air-pollution-lung (accessed February 3, 2022).

¹² See https://www.cancer.gov/research/areas/causes/breast-cancer-and-the-environmentcontroversial-and-emerging-exposures.pdf (accessed December 12, 2021).

animals and humans. The NCI funds a canine immunotherapy trials network¹³ and has developed dog-specific antibodies for use in cancer immunotherapy.

The NCI also created a large Integrated Canine Data Commons (ICDC)¹⁴ that maintains data from clinical trials for dogs with cancer. The ICDC is a node in the NCI’s Cancer Research Data Commons (CRDC).¹⁵ The ICDC contains a wide variety of genetic, clinical, immunological, and imaging sets in a cloud-based resource containing computational tools, said Amy LeBlanc (NCI), who encouraged researchers to use it for sharing and analyzing data. Researchers can bring their own data and tools to the cloud and combine them with data in the CRDC for integrative analysis, explained LeBlanc.

HISTORY AND CURRENT STATE OF THE SCIENCE OF ENVIRONMENTAL EXPOSURE EFFECTS ON AGING AND CANCER SUSCEPTIBILITY

Speakers offered background on the data linking environmental exposures to cancer and aging in humans and companion animals and considered the relevance of dog genetics to the study of human age-related disease. Speaker presentations focused on the latest science related to understanding the relationship between environmental exposures and cancer and aging in humans and in companion animals, respectively.

Environmental Exposure and Cancer

Cancer Demographics

Cancer is a complex group of over 100 diseases in which cells grow uncontrollably, said Gary Ellison (NIEHS/NCI), who summarized the science linking environmental exposures to cancer in humans. Cancer can occur and spread anywhere in the body, and one tissue can be the source of multiple types of cancer—for example, there are several genetically distinct subtypes of breast cancer, and each could have a different etiology. Approximately, 5 to 10 percent of cancers are due to inherited mutations, with the remaining due to errors that occur during cell division or as a result of DNA damage from the environment.¹⁶ The incidence of most cancers increases with age.

___________________

¹³ See https://dctd.cancer.gov/NewsEvents/20190327_canine_immunotherapy.htm (accessed February 3, 2022).

¹⁴ See https://caninecommons.cancer.gov (accessed December 6, 2021).

¹⁵ See https://datascience.cancer.gov/data-commons (accessed December 15, 2021).

¹⁶ See https://www.cancer.gov/about-cancer/understanding/what-is-cancer and https://www.cancer.gov/about-cancer/causes-prevention/genetics (both accessed March 30, 2022).

Cancer mortality rates have decreased in both males and females in recent years, likely due to advances in detection and treatment (Islami et al., 2021). Nonetheless, melanoma, myeloma, and kidney and liver cancers have been increasing, and environmental factors may play a role. Colorectal cancers have been increasing in those born after 1960, with approximately 10 percent of new diagnoses in individuals under age 50 (Stoffel and Murphy, 2020). The incidence of these cancers is 1.75- to 3-fold higher in those born after 1990 than in 1930, suggesting a possible effect of diet or early life exposure, said Ellison (Figure 1).

Characterizing Environmental Carcinogens

Ellison offered a broad definition of “environment” that comprises all nongenetic factors, including pollutants in air and water; chemicals in building materials, cosmetics, sunscreens, and medicines; pesticides and agricultural chemicals; synthetic additives in food; and lifestyle factors such as nutrition, smoking, and stress. Any of these factors may convey an increased risk of cancer, either by inducing growth-promoting changes in a cell or by altering the cell’s milieu (e.g., immunosuppression). Carcinogens are characterized based on the types of genetic, epigenetic, and functional changes they induce in cells (Smith et al., 2016), and exposures to different types of carcinogens lead to different types of cancer. For example, physical exposures such as radiation and ultraviolet light are associated with stomach cancer and melanoma, respectively; pesticides with lung cancer, leukemia/lymphoma, and bladder cancer; diet and exercise with liver, pancreatic, and colorectal cancer; infectious agents with cervical and liver cancer, among others; and reproductive hormones with breast cancer in females and prostate cancer in males. Tobacco smoking is associated with many cancer types.¹⁷

Two agencies, the NTP¹⁸ and IARC, are tasked with evaluating the carcinogenicity of potentially hazardous substances. As of 2016, the NTP listed 62 known and 186 likely human carcinogens. The IARC has identified 121 human carcinogens, 90 probable carcinogens, and 322 possible carcinogens of over 1,000 agents evaluated.¹⁹ A determination of carcinogenicity is based on evidence that exposure leads to cancer in humans and experimental animals, as well as on mechanistic studies, explained Ellison.

Documenting Environmental Exposures

The complexity of documenting exposure to environmental carcinogens is a particular challenge, said Ellison (Figure 2).²⁰ The association of asbestos with lung cancer and benzene with leukemia are well-documented examples of occupational exposures to carcinogens; in these cases, exposures were high dose, long in duration, and easy to document. However, many environmental

___________________

¹⁷ See https://monographs.iarc.who.int/human_cancer_known_causes_and_prevention_organ_site (accessed February 3, 2022).

¹⁸ See https://ntp.niehs.nih.gov/whoweare/index.html?utm_source=direct&utm_medium=prod&utm_campaign=ntpgolinks&utm_term=whoweare (accessed December 16, 2021).

¹⁹ See https://ntp.niehs.nih.gov/ntp/roc/content/introduction_508.pdf (accessed March 30, 2022).

²⁰ See https://www.epa.gov/risk/guidelines-human-exposure-assessment (accessed December 6, 2021).

exposures occur in lower doses in nonoccupational settings and have long latency. Exposures can occur in complex mixtures, with doses and interactions among agents mediating their effects. The timing of exposure is important, as there may be windows of greater susceptibility to both genetic and epigenetic changes across the life span. This is particularly true during gestation and early life, when exposures to environmental toxins may produce lifelong effects. Compounding these difficulties is the fact that retrospective assessment of exposure is not very reliable. Other challenges include variation among individuals in genetic risk factors and windows of susceptibility, tissue-specific effects, and effects that are only observed in subsequent generations, said Ellison.

Cheryl Lyn Walker (Baylor College of Medicine) and Ellison both noted that significant racial disparities exist in cancer incidence and mortality and that these are likely due to racialized social structures, such as segregation in historically redlined neighborhoods, where environmental health disparities are pervasive and affect the quality of food, medical care, and proximity to environmental toxicants. It is also important to consider the impact on cancer of trauma associated with racism, said Ellison. Race is a social construct, he noted, adding that “we need to get away from looking at race and ethnicity, and look at other factors that are associated with racial differences in this country.” These factors can lead to epigenetic changes, which are heritable, so the effects of environmental health disparities may be transmitted through the generations, added Walker.

Ellison identified the need to further evaluate probable and possible environmental carcinogens, to evaluate agents linked to cancer in animals when human evidence is insufficient, to consider variations in genetic susceptibility and windows of exposure in prospective study design, to evaluate the cancer risk associated with low levels of agents and with complex mixtures, and to investigate exposures in diverse populations. He described three National Institutes of Health (NIH) efforts to support research into human environmental exposures:

• The NCI Cohort Consortium,²¹ an extramural–intramural partnership, consists of 61 high-quality, large cohorts representing diverse populations from at least 15 countries and 4 continents (North America, Europe, China, and Australia). Extensive risk factor data are available from more than 7 million study participants and baseline biospecimens from more than 2 million. Within this consortium, the NCI funds well-characterized aging cohorts consisting of 1.1 million individuals.

___________________

²¹ See https://epi.grants.cancer.gov/cohort-consortium (accessed December 17, 2021).

• A new partnership between the NCI and the NIEHS²² recently funded five new cohorts to study the effects of emerging and important environmental exposures on cancer risk. These cohorts are racially/ethnically diverse, will generate data on a wide variety of exposures, and will enable the study of associations between exposures and biomarkers such as immune modulation, oxidation, hormone-mediated effects, and cell immortalization.
• The Human Health Exposure Analysis Resource: 2019–2024 (HHEAR),²³ led by the NIEHS, supports investigators who want to add or expand exposure analysis to their human health research. The HHEAR offers three areas of support: a network of laboratories to analyze biological and environmental specimens using both targeted and untargeted approaches, a coordinating center, and a data center. The HHEAR currently has more than 40 studies in its data repository, said Yuxia Cui (NIEHS), and all the exposure data and associated phenotypic information will eventually be made publicly accessible to researchers.

Environmental Exposure, Cancer, and Aging in Companion Animals: The Companion Dog Model

Dogs as Sentinels for Human Disease Agents: Shared DNA, Environment, and Disease

The utility of animal sentinels for environmental exposures is illustrated by the classic example of “the canary in the coal mine”, where canaries were brought into mines early in the 20th century to sense carbon monoxide (CO) exposure, said Peter Rabinowitz (University of Washington).²⁴ Canaries displayed the three qualities of an animal sentinel needed for assessing the health risks posed to humans by environmental hazards: greater susceptibility to the hazard; greater exposure (in the case of canaries, more rapid metabolism of CO); and shorter latency.

The usefulness of dogs and cats as sentinels of harmful environmental exposures has been demonstrated through several public health calamities over

___________________

²² See https://grants.nih.gov/grants/guide/rfa-files/RFA-CA-20-049.html (accessed December 17, 2021).

²³ See https://grantome.com/grant/NIH/U24-ES026539-02 (accessed December 17, 2021).

²⁴ See https://www.smithsonianmag.com/smart-news/story-real-canary-coal-mine-180961570 (accessed December 15, 2021).

the last century, said Kurunthachalam Kannan (New York University School of Medicine). A classic example arose from industrial mercury dumping in the fishing village of Minamata, Japan, which contaminated the local fish supply; local cats, who ate leftover fish, displayed “dancing cat fever” years before the neurological symptoms of mercury poisoning became evident in people (James et al., 2020). The deliberate addition of melamine to pet food in 2007 caused kidney failure in cats and dogs, and its addition to infant formula in China the following year caused kidney failure in children (Skinner et al., 2010); thousands of children were affected and several died. In the 1940s and 1950s, the high rate of bladder cancer in dogs exposed to aromatic amines linked these compounds to the high rate of bladder cancers in textile industry workers (Dietrich and Golka, 2012; Hayes et al., 1981).

The etiology of cancer involves both genetic and environmental factors, said Audrey Ruple (Virginia Tech), so the best sentinels for carcinogens would be similar to humans in terms of both genetics and exposure. Dogs share about 650 million base pairs of ancestral DNA sequence with humans, and dog orthologous genes are more similar to humans than are the same genes from mice (Kirkness et al., 2003; Paoloni and Khanna, 2008). The convergent evolution of dog and human genes is likely due to similar selective pressure from shared environmental exposures over many millennia, said Ruple. The geological record documents human exposures to dust and silica going back millions of years, joined by wood smoke roughly a million years ago, both of which left an imprint in the human genetic code, said Caleb “Tuck” Finch (University of Southern California) (Trumble and Finch, 2019). Domestication of dogs roughly 20,000 years ago may have brought them into the smoky environment of small human dwellings of the time, said Finch. Shared exposures are still the norm in the 40 percent of U.S. homes that include pet dogs, said Ruple. Not only do pets share a common living space with humans, but their behavioral patterns resemble those of toddlers, including mouthing objects and contact with the ground, noted Kannan. Dogs are a fantastic model species from a genetics standpoint, added Adam Boyko (Embark Veterinary, Inc./Cornell University). There are hundreds of dog breeds and they are much more differentiated than human populations, with larger ranges of phenotypic diversity, including a 50-fold difference in body size that correlates with a 2-fold difference in aging rate. This provides a useful model for genetic association studies and for mapping the mutations that underlie specific phenotypes, said Boyko.

Sharpless, Ruple, Rabinowitz, and Kannan listed multiple advantages to using companion dogs as a model for environmentally induced cancers, many of which are summarized in Figure 3 from Sharpless’s presentation. Ruple emphasized the difficulty in identifying environmental carcinogens in humans due to a long latency period, whereby decades may elapse between exposure and disease. In contrast, dogs are short lived and their cancers have

a much shorter latency. Matthew Breen (North Carolina State University) noted that the roughly 7-fold difference in longevity between humans and dogs translates into potential differences in latency of exposure, said Breen, allowing scientists to “harness the impacts of environmental influences on our pet dogs, living in our environment but over a much shorter period of time.” Rabinowitz and Kannan both highlighted the example of mesothelioma, which is largely caused by asbestos exposure and has a latency of 30–35 years in humans, with a mean age of about 8 years in dogs (Glickman et al., 1983). Ruple noted that the incidence of cancer is 10 times higher in dogs than humans. She added that studying the tumors that develop spontaneously in companion dogs (Schiffman and Breen, 2015) can avoid the ethical concerns associated with inducing tumors in laboratory animals. Rabinowitz added that some breeds have increased susceptibility to certain cancers, such as osteosarcomas in golden retrievers and histiocytic sarcomas in Bernese mountain dogs, and might therefore be more sensitive to certain carcinogens. Many studies have demonstrated the utility of dogs as sentinels for the reproductive effects of environmental contaminants (Sumner et al., 2020). Richard Lea (University of Nottingham) commented that dog testes and ovaries, which are disposed of in large quantities following spay/neuter-

ing procedures, are a valuable resource that can be used both to screen for environmental contaminants and to study the connection of these agents to specific pathologies. Breen emphasized that the different types of cancer shared between humans and dogs show “a huge amount of similarity” in both anatomic pathology and genetics (Figure 4).

Why Pet Dogs with Spontaneous Tumors are Good Models for Human Disease

“What is very striking is the sheer number of dogs that get cancer each year in the United States and the overall incidence that translates into,” added Breen. Out of the roughly 60 million pet dogs that visit a veterinarian for health care each year in the United States, 6 million are diagnosed with cancer. “This represents an enormous opportunity to provide health data and biological specimens from dogs over time and geographical space over the United States,” said Breen.

Using advances in data collection, researchers have recently begun to develop and analyze large veterinary data sets (Paynter et al., 2021). These include prospective longitudinal studies in dogs, such as the Golden Retriever Lifetime Study (Box 3) and the Dog Aging Project (Box 4); see section on Ongoing Canine Population Studies.

Ruple noted several limitations to the use of animals as sentinels of environmental exposures. There is no nationwide census of dogs in the United States and no dog cancer registry, which makes it difficult to assess the actual incidence of cancers in dogs. The absence of information regarding the numbers and types of pets in U.S. households was referred to throughout the meeting as “the denominator problem.” Because dogs do not receive death certificates, the burden of cancers remains unknown and is not captured by veterinary records that list the cause of death as euthanasia. The failure to record the reason why these dogs are being euthanized is a flaw in the current system, said Ruple.

Ellison noted that companion dog research lends itself well to citizen science approaches and that the NIEHS is a leader in this approach, with most funding opportunities now requiring community engagement. Citizen scientists collect biospecimens like toenails, hair, and teeth from their pets to capture data on exposures. Ruple cautioned that there are considerable health disparities in veterinary medicine, with the enrolled dogs primarily from mostly affluent White people, in research studies aimed at investigating environmental impacts on cancer that she has been involved with. These are the same individuals already represented in the existing data sources. These disparities in participation need to be understood as “a threat to the validity of the work we are doing” and actively addressed by engaging communities that are not well represented. Otherwise, this research will only exacerbate the inequities that already exist in human medicine, warned Ruple. Ellison raised the possibility of leveraging data from existing human cohort studies supported by the NCI to include dogs to increase representation for both human and dog research.

Animals as Sentinels of Environmental Health Hazards: A Brief History

Ruple briefly reviewed the findings from a 1991 NASEM consensus report entitled “Animals as Sentinels of Environmental Health Hazards,” which focused on efforts to monitor cancer-related health outcomes in dogs, primarily through retrospective, observational case-control studies assessing exposures in healthy and sick animals (NRC, 1991). One such study found an 8-fold increase in dog mesothelioma if the owner was exposed to asbestos (Glickman et al., 1983), while others found positive associations between a range of cancers in dogs and the use of herbicides or pesticides. Ruple noted that studies on the carcinogenic effects of pesticide use in humans, which

have begun to appear in the last 5 years, took 20 years to complete, whereas the dog studies covered in the 1991 report were done in considerably less time.

Studies published since that 1991 report have demonstrated further associations between cancer in companion dogs and environmental pollutants, as well as between dog cancer and smoking by the owner. Search-and-rescue dogs deployed to the World Trade Center following the 9/11 attacks did not have statistically significant increased cancer rates compared to controls 15 years later, which could offer reassurance to humans who worked on the site (Otto et al., 2020). Studies cited by Ruple and Rabinowitz in their presentations are summarized in Table 1. Additional canine and feline cancer sentinel studies are reviewed in Ruple et al. (2019). Summarizing the history of canine cancer sentinel studies, Rabinowitz noted that studies of dogs as sentinels have untapped potential for detecting environmental carcinogens and improving human cancer prevention.

Lessons Learned, Looking Forward

Following the Flint, Michigan, water crisis that poisoned thousands of children with lead, it was found that local dogs also had higher levels of lead in their blood than unexposed controls (Langlois et al., 2017). Earlier studies in the 1970s had shown that dogs could be used to detect areas with problematic levels of lead paint, and thus address the environmental health needs of underserved populations, but this work was abandoned in the late 1970s, said Ruple. She added that if there had been an ongoing effort to identify lead levels in dogs this could have made a difference: “Imagine if we had not stopped … instead of waiting until more than 14,000 children were impacted in Flint, Michigan.”

The 1991 NASEM consensus report showcased a wealth of studies demonstrating the utility of using animals as sentinels for environmental health risks, concluding that “these findings indicate that well-designed epidemiological studies of spontaneous tumors in pet animals may provide insight into the role of environmental factors in human cancers and serve as a valuable sentinel model to identify environmental health hazards for humans” (NRC, 1991). But it also acknowledged that this approach had not gotten much traction. “One reason,” it offered, “might be the institutional inertia that accompanies integration of new scientific methods into the risk-assessment process and use of the results for risk management. Many government agencies do not recognize the importance of animal sentinels or agree on how to compare the findings obtained with them and the findings obtained with more traditional methods.” The report noted that research in animal sentinels was not a high priority for funding—however, it recognized that increasing interest in

TABLE 1 A Selection of Canine Cancer Sentinel Studies Showing Positive Association or No Association with Environmental Exposure

SOURCES: Adapted from Ruple and Rabinowitz presentations, December 1, 2021.

humane alternatives to lab animal experimentation might draw attention to the use of companion animals, as indeed has transpired.

Workshop participants considered how data on animal exposures could be used to advance public health. Ruple said that work needs to be done on both the human and animal sides to bridge the gap, and noted that veterinarians are often the first ones to sound the alarms on vector-borne disease. To bring the One Health²⁵ voice to the public health arena, she emphasized that veterinarians need to be included on public health boards, which is happening in some places. She added that updates are needed in the curriculum of training programs. Marcia Haigis (Harvard University) and Elaine Ostrander (National Human Genome Research Institute) emphasized the need to make the large canine ,-omic data that is now being accumulated freely accessible to investigators so they can use informatics approaches to generate hypotheses for cancer mechanisms, as is done for humans. Ostrander applauded the efforts of the NCI’s canine commons, but added that the problem of access will only be solved as a community. Rabinowitz envisioned a future when veterinarians could detect an environmental hazard in a companion animal and convey this information to the medical professionals caring for members of the household. “There needs to be a way to clinically get that information to people concerned with human health . . . I really hope we move in that direction,” he added.

How Diet Modulates the Tumor Microenvironment (TME)

Metabolites and the TME

Obesity is a risk factor for several types of human cancer (Lauby-Secretan et al., 2016), and this is becoming increasingly significant as our metabolic health demographics change, with nearly 50 percent of the U.S. population expected to be obese by 2030 (Ward et al., 2019), said Marcia Haigis (Harvard University). She described her efforts to understand the connection between diet, a type of environmental exposure, and cancer. Extending the central dogma to metabolites (DNA→RNA→protein→metabolite), Haigis noted that 25 percent of the proteome has a direct metabolite readout, and that this readout is directly affected by both the external environment and diet. Metabolites feed back to the earlier steps of the central dogma, controlling protein function through allosteric effects and posttranslational modification, and influencing DNA packaging, with consequences for epigenetics and gene expression.

___________________

²⁵ See https://www.cdc.gov/onehealth/index.html (accessed February 11, 2022).

Cancer cells in particular have tremendous energy demands and burn fuel by metabolizing glucose, fatty acids, and amino acids, generating profiles of metabolic byproducts distinct from those of healthy cells. For example, tumor cells from estrogen-receptor-positive (ER+) breast cancers accumulate large amounts of ammonia, which they secrete into the cellular environment as a metabolic waste product but then recycle and use for building amino acids, unlike healthy cells (Spinelli et al., 2017). Tumor metabolites can alter protein activity through mechanisms similar to those used by the metabolites of healthy cells. But tumor cells do not exist in isolation, said Haigis; their growth and survival are mediated by interactions with a host of cells and molecules within the tumor microenvironment (TME). Although killing by cytotoxic T lymphocytes (CTLs) is important for controlling tumors, tumors have evolved several mechanisms for evading CTLs, including by constraining the ability of CTLs to function in the TME, which can be accomplished by sequestering fuel away from the CTLs or by secreting metabolites—such as ammonia and lactate—that inhibit CTLs.

Diet and the TME

How does diet affect cancer risk? Haigis described highlights of a recently published study that investigated how a high-fat diet alters the function of CTLs and tumor cells, using implanted tumors to probe tumor-driven modification of the TME in mice (Ringel et al., 2020). Syngeneic²⁶ colorectal tumor cells were implanted into mice fed either a control or high-fat diet, and tumor growth and phenotypes were monitored over time. When mice were fed a high-fat diet, CTLs were suppressed in the TME and the incidence and growth of tumors increased. CD8⁺²⁷ CTLs in particular were fewer, less proliferative, and less cytotoxic in the TME of mice fed a high-fat (vs. control) diet. In addition, CD8⁺ CTLs from the TME of mice on the high-fat diet changed their choice of energy source, up-regulating fatty-acid metabolism and down-regulating glucose metabolism.

Drilling down to the molecular level, Haigis and colleagues found that expression of the prolyl-hydroxylase 3 (PHD3) protein was down-regulated in tumor cells exposed to a high-fat diet. PHD3 belongs to a family of enzymes that mediate the response to hypoxia stress. PHD3 represses fat metabolism (German et al., 2016), so down-regulation of PHD3 would be expected to increase fat burning and oxidation. The researchers found that, although the high-fat diet increased fatty acid metabolites (lipids) in circulation, as might

___________________

²⁶ Syngeneic is defined as genetically similar or identical.

²⁷ CD8-positive T cells are a critical subpopulation of major histocompatibility complex (MHC) class I-restricted T cell and are mediators of adaptive immunity.

be expected, it had the opposite effect in the TME, where it caused lipid accumulation to decrease, potentially because the tumor cells, having turned off their PHD3 gene, were taking up all the lipids. Suspecting this would have the effect of inhibiting CTLs by depriving them of energy from lipid metabolism, the researchers induced the tumor cells to overexpress PHD3. In these mice, lipid levels in the TME remained high even on a high-fat diet, CTL numbers increased, and tumor growth was reduced.

Taking this question to human studies, Haigis and colleagues found that PHD3 expression was similarly decreased in tumors from colorectal cancer patients with a high body mass index (BMI) compared to low-BMI patients, and these PHD3-low tumors were immunologically “cold,” with low levels of tumor-infiltrating lymphocytes, compared to PHD3-high tumors. Haigis concluded that obesity metabolically reprograms tumors to metabolize fat while changing the TME and impairing antitumor immunity—and this immunity may be restored by manipulating the expression of a single gene. Haigis emphasized the importance of studying the effect of environmental factors and aging on tumor growth on the cellular level. She also noted that most laboratory models of cancer use young animals, despite the importance of aging in cancer development.

Epigenetic Aging as a Target and Biomarker for Environmental Exposures

Environmental Exposures Alter the Epigenome

The epigenome, which consists of heritable chemical modifications of DNA, without changes to the sequence of bases, or chromatin, is a target for environmental exposures, said Cheryl Lyn Walker (Baylor College of Medicine). The primary type of epigenetic modification is methylation of CpG dinucleotides in the DNA sequence, and this can be inherited mitotically or meiotically. In contrast to DNA, which is protected from environmental influences by multiple mechanisms, the epigenome is inherently plastic, and it is well equipped to respond to the environment (Perera et al., 2020). One example would be epigenetic modifications in a fetus in response to starvation, which are adaptive for development in a nutrient-poor environment following birth, though they can lead to metabolic disease in an environment that is nutrient rich (Fleming et al., 2018). Epigenetic modifications may persist long after an exposure has occurred and can occur late in life. For example, breast tissue is developmentally plastic until the first full-term pregnancy (Walker and Ho, 2012).

Epigenetic modifications to DNA continue to change in species-specific patterns over time as the organism ages, and this predictable course of DNA

methylation has enabled the characterization of species-specific biological clocks (Field et al., 2018). The biological clock runs at different speeds in different individuals, revealing differences in “biological age” between humans of the same chronological age. These differences in biological clocks are excellent predictors of health, said Walker. Environmental perturbations can cause a change in the epigenetic clock, in which a “faster” epigenetic clock predicts decreased survival (Kochmanski et al., 2017), as was demonstrated over a 15-year time period for older males participating in the Veterans Health Administration Normative Aging Study (Belsky et al., 2020). Low socioeconomic status and high trauma were associated with faster clocks (Schmitz et al., 2021; Verhoeven et al., 2018). Epigenetic clocks can be compared across tissues and species. For example, although the maximum life span is 122 years for humans and 3.8 years for rats, recalibrating both species to the same scale shows 95 percent correlation in epigenetic aging between the two (Horvath et al., 2020). These correlations support study of the epigenome as a biosensor in companion animals, said Walker.

Epigenetic Modifications can Enhance Cancer Risk

Walker described her group’s work in the TaRGET-II consortium,²⁸ which is probing the ways in which multiple environmental perturbations alter the epigenome across multiple tissue types. When male mice were exposed in utero to the obesogen tributyltin (TBT), roughly half developed liver tumors by 10 months (Katz et al., 2020). In mice that developed tumors, researchers were surprised to discover that the transcriptional profile of the normal-appearing, uninvolved part of the liver looked nearly identical to that of the tumors. In contrast, the transcriptional profile of the normal liver from TBT-exposed animals that did not develop tumors resembled that of untreated control mice. Walker and colleagues hypothesized that TBT treatment had generated a distinct epigenetic “endotype” in the livers of half the exposed mice, which conferred a high risk for tumor development. Epigenetic analysis of these mice detected many differences in CpG²⁹ methylation in the uninvolved livers of TBT-treated mice that had developed tumors, compared to livers of untreated control mice, while TBT-treated mice without tumors resembled the controls. These methylation differences were observed in genes that normally underwent epigenetic modifications early in development, and they could be observed in the high-risk mice months before they developed any tumors. Walker surmised that TBT targeted this “plasticity of epigenetic aging,” and as a result, normal epigenetic aging failed to occur in the animals

___________________

²⁸ See https://target.wustl.edu/index.html (accessed December 15, 2021).

²⁹ A CpG is a C (cytosine) base followed immediately by a G (guanine) base.

that had been transformed to the high-risk type. Importantly, humans with liver cancer or non-alcoholic fatty liver disease displayed the same high-risk epigenetic signatures associated with TBT-induced reprogramming in mice, indicating that the same genes were altered in the human disease state. The epigenome is an incredible biosensor of the environment, said Walker, especially for analyzing the effects of complex exposures, and with the availability of two-species clocks, the faster epigenetic clock of companion animals presents a valuable tool for inferring the future health of the human epigenome and perhaps for predicting adverse outcomes due to environmental exposure; specific companion animal examples are provided in the next section.

Asked to consider differences in methylation patterns across different cell types, Walker replied that the TaRGET II consortium has not detected any correlation between epigenetic signatures in blood cells and epigenetic changes in target tissues. For example, blood cells failed to distinguish between individuals with different liver epigenome types following TBT treatment. Haigis cited her metabolic analyses, which detected a different abundance of fat metabolites in the plasma versus the tumor microenvironment. This is particularly relevant to epigenetics, she said, because the acetyl-CoA from fat oxidation is used in the acetylation of histones in chromatin, another epigenetic change that can alter gene expression, and there are many other metabolites that regulate enzymes. Different places in the body have different nutrient pools and react differently to changes in diet, said Haigis, and therefore, epigenetic differences between circulating lymphocytes and the cells in tissues should not come as a surprise.

Domestic Dogs as a System for Understanding Aging and Life Span

Development of Dual-Species Epigenetic Clocks

Historically, the epigenetic clocks in humans have not been strong predictors of age in other mammals, and this has presented a challenge to advancing dog studies of aging and cancer. To determine whether the epigenetic age of a dog could predict its life span (as it does in humans), Ostrander and colleagues characterized the methylomes (consisting of 50,000 CpG sites spanning the genome) of blood cells from 104 Labrador retrievers ranging in age from 2 months to 16 years (Wang et al., 2020). Young dogs showed high methylome similarities to young humans, and aged dogs to older humans, with a weaker correlation in midlife. Mapping the methylation patterns against age, the researchers detected a nonlinear relationship between dog and human age; following this formula: human age = 16(ln(dog age)) + 31 (see Figure 5). When life stages were plotted along this curve, the researchers found a strong agreement between the approximate times at which dogs and humans experience physiological mile-

stones in the transitions from infancy to juvenile, adolescent, mature, and senior life stages. The researchers also observed conservation of epigenetic progression between dogs and mice, suggesting that changes in methylation with age extend from humans to other mammalian genomes, said Ostrander.

Among the 50,000 methylation sites, Ostrander and colleagues identified 394 genes whose methylation patterns showed conserved time-dependent behavior across dogs, mice, and humans. When these genes were mapped onto the PCNet molecular interaction database,³⁰ which captures physical and functional relationships among genes and their products, they clustered into

___________________

³⁰ PCNet is a database of molecular interactions capturing physical and functional relationships among genes and their products.

five highly interconnected network modules involved in anatomical development, leukocyte differentiation and metabolism, neuroepithelial cell differentiation, and synapse assembly and regulation. Some of these genes are associated with cancer as well as aging. Unlike age predictions based on the 50,000-site whole methylome clock, which only worked within the same species used to create the clock, patterns of CpG methylations in these 394 conserved developmental gene modules could make accurate age predictions across multiple mammalian species. “This is allowing us to hone in on the most relevant genes … for building a really good epigenetic clock,” explained Ostrander.

Aging, Somatic Evolution, and Cancer—The Inexorable Link

Adaptive Oncogenesis: Why Does Cancer Increase with Age?

Ninety percent of human cancers develop after the age of 50, noted DeGregori, and this has traditionally been ascribed to the age-dependent accumulation of mutations. But he emphasized that the story is more complicated because different cancers are driven by different numbers of mutations, initiate in different stem cell pools, and have different age-dependent incidences. However, Rozhok and DeGregori (2019) found that when curves of cancer incidence are normalized based on their maximums, diverse cancers all demonstrate the same age-dependent increase in incidence over time, regardless of stem cell source or the number of driver mutations required for their development (see Figure 6). DeGregori noted that the likelihood of developing cancer and dying after age 50 is not linear; it gains speed over time, reflecting a dramatic drop-off in natural selection against cancer beyond the age where most individuals would have historically been dead by other causes. This derives, he said, from a variety of cellular and genetic mechanisms that provide resiliency to tissues throughout the reproductive years but decline thereafter, increasing vulnerability to a host of diseases, cancer among them.

The significance of these resiliency mechanisms in suppressing cancer is seen when humans age and the waning of tissue maintenance enables cancers to develop, through a process DeGregori termed “adaptive oncogenesis.” Adaptive oncogenesis predicts that an oncogenic mutation in one cell will reduce its fitness relative to the rest of the pool if that pool is healthy but will not incur a selective disadvantage if the pool is damaged (DeGregori, 2011). DeGregori and colleagues tested this theory by transplanting a small number of Ras-transformed stem cells into both young and old mice (Henry et al., 2015). The transformed cells expanded only in the old mice, resulting in leukemias; however, blocking inflammation prevented both expansion and leukemias. DeGregori observed that youth is intrinsically tumor suppressive, with healthy tissues exerting stabilizing selection against clonal expansion of transformed cells. As individuals age and their tissues become damaged, onco-

genic alterations that were maladaptive in a young environment are selected for in the older environment, and the incidence of cancer over time is shaped by the changing, age-dependent balance of genetic drift, stabilizing selection, and positive selection (for malignant growth).

How Improving the Function of Aging Tissue May Reduce Cancer Risk

In a computational study, Evans and DeGregori (2021) found that human tissues become “riddled” with cells carrying cancer-associated mutations as

they age, with over 100 billion cells in the body harboring at least one known oncogenic mutation by age 60. Roughly 40 percent of adults develop cancer, which arises from a single cell; this raises the question of why the other 100 billion cells with “oncogenic” mutations do not result in cancers. One clue may be found in a study of lung cancer, which demonstrated that, although smoking dramatically increases the likelihood of getting lung cancer at any age, lung cancer incidence shows a similar age-dependent increase in both current smokers and never-smokers (Samet et al., 2009).

To investigate how smoking status influences oncogenesis in the lung, DeGregori and colleagues examined normal (noncancerous) lungs in current, former, and never-smokers. Using duplex DNA sequencing, which enables detection of mutations with high sensitivity, the group found that even a small brushing of lung tissue from a 70-year-old contained hundreds of mutations, dozens of which were cancer associated, regardless of the person’s smoking history. However, smoking status influenced proliferation of the cells carrying oncogenic alleles. Cells with cancer-associated mutations demonstrated relatively little expansion in nonsmokers, compared to former and (even more) current smokers. Focusing on the p53 tumor suppressor gene, the highest frequency of cancer-associated alleles was seen in lungs of current smokers, more than former smokers. While these results clearly indicate that smoking promotes selection for oncogenic variants in the lung, they also suggest that quitting smoking may reverse this selection and thus explain why every year that passes after a person quits smoking reduces their odds of getting lung cancer (Tindle et al., 2018), said DeGregori. He added that they also support the goal of developing interventions to accelerate the restoration of more normal lung landscapes after quitting. In a broader sense, these observations indicate that efforts to prevent and treat cancer could converge with similar efforts to prevent other aging-associated diseases, he said.

Discussion: Animal Size, Reproductive Status, and Cancer

The Paradox of Large Animals and Cancer

Daniel Promislow (University of Washington) raised the question of Peto’s Paradox,³¹ which refers to the surprising fact that large animals like whales and elephants do not die of cancer when they are a few weeks old, given the number of cells that could potentially develop cancer compared to a mouse; rather, the larger species have longer lives. That large dog breeds

___________________

³¹ Sir Richard Peto noted that although mice have approximately 1,000 times fewer cells and > 30 times shorter life spans than humans, their risk of carcinogenesis is not markedly different—this is coined as Peto’s Paradox. For more information, see Peto (2015, 2016) and Callier (2019).

develop cancer more frequently than small breeds is the “paradox of the paradox,” said Promislow. DeGregori noted that, similar to dogs, big humans have higher cancer rates than smaller humans. The association of larger body size with increased cancer in dogs and humans is likely due to modulation of the insulin-like growth factor (IGF) pathway, which, when hypomorphic,³² consistently leads to a longer life span and less cancer in model organisms, said DeGregori. Selection for life span versus body size within a species therefore reflects differential investment, with natural selection pitting early reproduction and robustness against long-term maintenance and extended reproductive capacity. Ostrander cited data from her laboratory indicating that IGF alleles associated with small, longer-lived breeds were present heterozygously in wolves 50,000 years ago, roughly 30,000 years before domestication, but did not achieve homozygosity until about 10,000 years ago, when domestication and agrarian society were well established.

Reproductively Altered versus Intact Animals

Participants noted that reproductively altered animals might display different sensitivities to carcinogens than humans, owing to their different hormonal context. Ostrander noted that prostate cancer is a nearly ubiquitous disease of aging in men, except in those who lose their gonads early in life and never develop this type of cancer. In dogs, it is the opposite—prostate cancer is rare but is much more likely in neutered than intact dogs.

The question of longevity of altered versus non-altered companion animals is complicated. Ruple remarked that in the United States, intact animals typically live shorter lives, but they tend to be killed by cars and other traumas. Ostrander added that altered animals in the United States are more likely to be obese, which confounds the effect of the hormonal changes on cancer and mortality. However, spaying and neutering is not consistent worldwide, and research is being done in Hong Kong and other locations, where most dogs and cats are not altered but nonetheless live long lives, said Ruple.

Validation of Canine Models as Sentinels

A fundamental question, said Rod Page (Colorado State University), LeBlanc, and others, is how to validate a canine model as a sentinel. Dog cancer is not always directly comparable to human cancer, and many times it is not possible to obtain detailed diagnostic data from pets. But to what degree can

___________________

³² Hypomorphic is a type of mutation in which the altered gene product possesses a reduced level of activity.

the mechanisms be comparable when using data from dogs to evaluate the likely impact of an environmental exposure on human populations? Must they be identical, or is there a way to normalize the concept of a canine sentinel based on genetic, histopathological, or other biomarkers that may not be the same as those in humans? When thinking about these questions from the standpoint of the exposure, said Rabinowitz, researchers are trying to determine what in the environment is carcinogenic. The traditional method looks at animal experimental studies, which often produce a different type of tumor than what’s seen in humans, but nonetheless can reinforce the fact that a given substance or hazard is carcinogenic. The purpose is not to replace current methods but to introduce a whole new body of epidemiological evidence from companion animals to add to the epidemiological evidence from humans and the experimental evidence from animal exposure studies and in vitro work.

Walker commented that canaries did not go into coal mines because anyone thought they were little humans. It is useful to partition out hazard identification, exposure assessment, disease etiology/mechanism, and therapeutics. Each one of those endpoints would involve companion animals in a different way, said Walker.

METHODS AND CURRENT STUDIES

Speakers reviewed the current state of exposure science as it relates to methods, data, and analysis. Speaker presentations described examples of causal inference between exposures and cancer and aging, as well as potential data sources, including through the use of biosensors, needed to assess whether companion animals may serve as sentinels for human environmental exposures.

The Exposome and Health

The exposome idea, first introduced by Wild (2005), offers a way to consider the totality of an individual’s environmental exposures over the course of a lifetime, said Ellison (see Figure 7). Wild defined the exposome as all exposures from conception onward, including those from lifestyle, diet, and the environment. Although this enables consideration of multiple exposures, it carries a significant challenge in terms of managing the data and integrating it into models of disease risk, said Ellison. Gary Miller (Columbia University) and Dean Jones from Emory University refined Wild’s definition as “the cumulative measure of environmental influences and associated biological responses throughout the life span,” emphasizing that the biological signal left by a stressful exposure, such as an epigenetic mark, may persist long after the initial exposure is no longer present (Miller and Jones, 2014). Miller stressed the need to measure external influences on health in an unbiased way to avoid

making invalid associations and to create “a systematic, unbiased, and ,-omicscale examination of external factors contributing to disease or health status.”

Detecting the Metabolic Signatures of Biological Exposures

A broad range of physical and chemical exposures affects both humans and their animal companions (Figure 8), but in order to have an impact on health, an exposure must be converted to a biological signal inside the body, said Miller. Capturing the breadth of these exposures and signals will require a suite of tools, from wearable devices and biospecimens to computational imaging and satellites. In his research, Miller pairs high-resolution mass spectrometry (HRMS) with both liquid and gas chromatography to detect a broad range of volatile and soluble chemicals in biospecimens. This effort is complicated by the limited availability of authentic standards for many exogenous chemicals and their metabolites, requiring Miller and colleagues to compare individual compounds’ mass and retention times based on their peaks, which may be in the tens of thousands per measure, to PubChem and other databases containing nearly a hundred million compounds (Vermeulen et al., 2020). Miller added that this can be a “computational nightmare.” Another challenge is the under-

representation of biological metabolites of exogenous chemicals in the databases; these metabolites may comprise the only biological signature of exposure to many compounds, so it is important to know what they are, said Miller.

Miller and colleagues designed a method for identifying the biological metabolites of a vast range of agents, in which they exposed human liver fractions to known environmental chemicals and characterized the products based on their peaks following mass spectrometry and chromatography (Liu et al., 2021). They validated this method by treating mice with one of the tested chemicals, buproprion, detecting the same profile of metabolites in mouse blood that they had previously identified in the liver assay. These same metabolites were also detected in a human taking buproprion pharmaceutically. The researchers were surprised to discover a novel metabolite of caffeine, suggesting the existence of a large number of unidentified metabolites overall. Miller acknowledged that many analytical chemists would bristle at the identification of a compound in the absence of an authentic standard, but he noted that standards only exist for a few thousand compounds, while his group analyzes a hundred thousand peaks, and matching the peaks to those obtained from the metabolites of known compounds is quite effective.

Applications of Exposomics to Human Disease

Miller explained that this approach is beginning to yield valuable information linking exposure to disease. An exposome-wide association study (EWAS) found an association between environmental chemicals in plasma and two rare liver diseases, while a metabolome-wide association study (MWAS) uncovered disease-associated alterations in metabolic pathways in the same cohort (Walker et al., 2021). Analysis of 119 individuals from Richard Mayeux’s Washington Heights/Inwood Columbia Aging Project (WHICAP)³³ cohort distinguished between participants with and without Alzheimer’s disease (AD), as well as among different racial and ethnic groups, based on metabolites in their plasma (Vardarajan et al., 2020). Miller noted that the degree of separation was unlike anything Mayeux had observed in this same cohort over the course of 30 years and that a follow-up study was being performed on 8,000 metabolomes to determine what metabolic pathways were involved. The researchers began a parallel study in the Dominican Republic, where many members of Mayeux’s New York cohort originated. Miller noted that Mayeux has extensive environmental data from northern Manhattan for the past 50 years. Miller is also analyzing biobank plasma specimens from a cohort of individuals aged 20 to 80 years old who have been studied by cognitive testing and functional magnetic resonance imaging (fMRI) imaging (Stern et al., 2014).

___________________

³³ See https://dss.niagads.org/studies/sa000007 (accessed December 17, 2021).

Laboratory animal models may be used to test hypotheses generated by EWAS. For example, Miller and others described metabolites of dichlorodiphenyl-trichloroethane (DDT) that were elevated in humans with AD. Using an AD model strain of C. elegans containing a mutation in the tau gene, Kalia et al. (2021) found that tau mutant worms generated a distinct pattern of DDT metabolites compared to control worms. In a study of worms bearing a knockout of a Parkinson’s disease–associated gene, changes in gene expression were correlated with patterns of metabolite production (Mor et al., 2020).

The disease-related metabolic similarities between species as evolutionarily distant as worms and people suggest that the potential for cross-species EWAS/MWAS analysis in companion dogs and humans is very strong, said Miller. A collaboration with Mark Prausnitz produced a microneedle patch that can sample interstitial fluid from skin without a blood draw, which may be helpful for sampling pets at home (Samant et al., 2020). Miller cautioned that developmental windows matter, citing a study on couples trying to conceive that found significant variability in their exposomes despite their shared living environment (Chung et al., 2018). This highlights the importance of knowing which stages of the human and animal life spans were co-exposed when doing paired cross-species studies, said Miller.

As exposomics gains acceptance in the biomedical research community, where it has been used to study diverse organisms and can be performed on blood samples from any species without technical limitations, it is important to know what you are trying to achieve, said Miller. He favors untargeted HRMS due to its relative lack of “streetlamp bias,”³⁴ and because so many metabolites have not been identified. Targeted approaches have greater sensitivity and confidence in identification, but they are entirely dependent on this tough question: “Do you know what you are looking for?”

Biomonitoring of Chemical Exposure in Companion Animals and Humans

Evidence of Exposure to a Wide Variety of Agents in Cats and Dogs

Kannan began the discussion of biomonitoring by reviewing the literature on exposure to heavy metals in pets. The handful of papers published on this topic demonstrate that cadmium, mercury, and lead exposure can cause neurodegeneration in pet cats and dogs, said Kannan. High levels of mercury

___________________

³⁴ Streetlamp bias, or streetlight bias, is when people search for something in easier places instead of in the places that are most likely to yield the results they are seeking; i.e., looking under the lamppost.

were found in the brain, liver, and kidney tissues of the Minamata “dancing” cats (Eto et al., 2001; see section on Evidence of Exposure to a Wide Variety of Agents in Cats and Dogs). Toxic levels of heavy metals have been found in pet blood, hair, urine, and organ tissue, and lead concentrations in dogs and in children from the same household and socioeconomic status show an association, said Kannan. The persistent organic pollutants polychlorinated biphenyls (PCBs) and DDT and their metabolites have been widely found in cat and dog tissues, with PCB levels linked to diabetes and hypothyroidism, said Kannan. California cats had high serum concentrations of the flame retardant PBDE (polybrominated diphethyl ethers), which is linked to hyperthyroidism and house dust exposures (Guo et al., 2012, 2016). Thyroid, kidney, liver, and respiratory diseases have also been associated with high serum concentrations of poly- and perfluoroalkyl substances (PFAS) in cats, whose patterns of exposure to these substances are similar to those observed in humans (Bost et al., 2016).

Kannan and colleagues detected higher levels of metabolites of phthalates—plasticizers—in urine from cats and dogs compared to those reported in humans, with cats having the highest levels overall (Karthikraj et al., 2019), and used reverse dosimetry to estimate the daily exposure dose. Shelter animals had lower exposure doses than house pets, suggesting that some exposure was occurring in the home. Testing for additional compounds, the researchers found that of the three species (cats, dogs, and humans), dogs had the highest levels of antimicrobial parabens in their urine, less than 10 percent of which was acquired from their diet; much higher levels of parabens were found in dry foods than wet foods, and more in cat food than in dog food (Karthikraj et al., 2018b). The group also detected the kidney toxicant melamine and the herbicide glyphosate in dog and cat urine (Karthikraj et al., 2018a; Karthikraj and Kannan, 2019) and PFAS in pet feces (Ma et al., 2020). Many urine samples from pets and humans had levels of melamine exceeding those associated with chronic toxicity, and levels in children were higher than those in adults (Karthikraj et al., 2018a). Calculated PFAS exposures were one to three orders of magnitude above the Agency for Toxic Substances and Disease Registry (ATSDR) provisional minimal risk levels (Ma et al., 2020). Microplastic levels in cat and dog feces were higher than those in adult humans but similar to levels in 1-year-old children (Zhang et al., 2019), likely because infants and pets chew on textiles and plastics, conferring much higher exposures than diet alone. These observations and others from his laboratory indicate that biomonitoring methods can assess exposures and risks from a wide variety of contaminants, said Kannan. While pets often show higher levels of exposure than humans, patterns of exposure are similar, particularly with respect to indoor environmental contaminants,

and they do not primarily come from the diet. Kannan suggested a collaborative approach that combines clinical assessments of biomarkers in pets with paired human–pet biomonitoring studies to advance the use of pets as sentinels of environmental health.

Rabinowitz presented on unpublished research in which he and colleagues measured the levels of volatile organic compound (VOC) metabolites in dogs and humans living near natural gas wells, finding similar trends in exposure levels for both dogs and humans, with the dogs having higher levels of VOC metabolites than their human companions in 70 percent of cases, suggesting higher exposure in the dogs. As homes’ distance from a gas well increased, there was a greater distance-related reduction in VOC metabolites in dogs than in humans, indicating that the metabolites constituted a likely exposure biomarker in dogs more than in humans. Nicole Deziel (Yale School of Public Health) noted that childhood cancers are regrettably being used as a “sentinel” for adult cancers in populations exposed to fracking, because of the shorter latency, and that both companion and domestic animals would be ripe for study. This argues for routinely including animals whenever you build a longitudinal cohort for observational studies of potential hazards, such as those related to natural gas drilling, said Rabinowitz.

Assessing the Exposome using Wearable Sensors

Environmental exposures show dramatic variation in type, timing, and dose across populations, time, and space, and therefore require a variety of tools for their detection and analysis, said Cui, and the goal of the NIEHS is to use these tools to understand how exposome–genome interactions influence health outcomes (see Figure 9).

The NIEHS sponsors the development and refinement of a variety of wearable and field-deployable tools for characterizing personal exposures to air pollution and for monitoring physiology. Although these are being developed for use in humans, they can be adapted for companion animals, Cui explained. The NIEHS-supported sensor research includes development of hardware, software, miniaturization, improvements to battery life, and laboratory and field testing to improve accuracy and reliability. Cui described five examples of this work (see Box 2).

Cui mentioned that there are challenges associated with analyzing large volumes of sensor data, as well as privacy concerns. But she noted that the number of PubMed publications on wearable sensors has been increasing exponentially in recent years, indicating the great potential of this field.

Ongoing Canine Population Studies

The Domestic Dog as a Sentinel Species for Environmental Influence Associated with Bladder and Other Cancers

There are remarkable similarities between human and canine cancers, both genetically and from an anatomic pathology perspective, said Breen. Furthermore, health data and biological specimens from the 6 million dogs that are diagnosed with cancer each year provide an “enormous opportunity” for research, with millions of biopsy specimens from pet dogs stored around the United States. By cross-referencing data from dogs and humans for cancers that are considered comparable, Breen and colleagues have found many shared genetic features between human and dog tumors for a range of cancers (Angstadt et al., 2012; Mochizuki et al., 2015; Schiffman and Breen, 2015; Tawa et al., 2021). However, genetics are insufficient to explain why, among populations of dogs with the same genetic risk factors, some develop cancer and others do not. To understand how the genetics interact with environmen-

tal risks to produce cancer, “we need to understand a lot more about our dogs’ exposures, to understand more about our own exposures,” said Breen. To do this work, Breen and colleagues built a nationwide collaboration among dog owners, scientists, veterinarians, and breeders, with two fundamental goals: (1) to monitor changes in cancer incidence in dogs in different geographical regions, using the shorter latency of dog cancer to identify risks to humans, and (2) to engage dog owners in clinical trials that include exposure assessment. The best strategy, says Breen, is to do paired exposure studies of humans and dogs from the same household, where one of those species has been diagnosed with cancer.

In a collaboration between North Carolina State Veterinary School and Duke University, Breen and colleagues have begun enrolling dog patients in combined clinical–environmental studies, the first of which is looking for environmental associations with bladder cancer in dogs, who are typically diagnosed at age six or older. The researchers developed a ‘liquid biopsy’ that enabled them to detect very early bladder cancer based on gene signatures in the urine, allowing them to “wind the genomic clock back” and look for the earliest genetic changes. Working with the American Kennel Club (AKC), they recruited owners of 2,000 dogs aged 6–12 years from seven breeds with a high predisposition for bladder cancer. Their first project, involving use of silicone wrist bands and matched urine to investigate shared exposures, has revealed “striking . . . similarities between exposures of our dogs and ourselves,” said Breen (see section on Using Silicone Samplers to Assess Air Pollution in People and Pet Dogs).

Breen noted that the number one risk factor for bladder cancer in humans is smoking, but unlike lung cancer, bladder cancer has not declined with the

reduction in smoking over the last 30 years. This suggests that other environmental factors may be at play; for example, certain pockets of Ohio are bladder cancer hot spots. In a collaboration with the Cleveland Clinic, Breen’s group is enrolling human patients with bladder cancer and their dogs in a study to monitor exposures through water samples, urine, and wrist bands.

Bladder cancer differs between dogs and people and this complicates comparisons, suggested Beverley Koller (University of North Carolina). Eighty to 85 percent of dog urothelial carcinomas (bladder cancer) has the equivalent of a human V600E BRAF³⁵ mutation, which is present in only 1 to 2 percent of human bladder cancers, agreed Breen. However, he noted that there is regional variation in human bladder cancer, with the V600E mutation present in as many as 25 percent of human patients in certain cohorts. Mutational signatures need to be considered in terms of their geospatial distribution in order to explore the causes of regional variation, even within a state, said Breen.

A third study builds on an existing North Carolina cohort from the NIEHS Personalized Environment and Genes Study (PEGS).³⁶ Participants have contributed genotype, health, and environmental exposure data since 2002. Breen aims to obtain genetic and health information from participants’ dogs, overlay this onto the exposure maps that were developed for the human study, and determine whether exposures affect the health of dogs before that of humans. Lastly, as part of a large collaboration with Timothy Mousseau (University of South Carolina), Ostrander, Norman Kleiman (Columbia University), and others, Breen is studying the dog populations that have roamed the area surrounding the Chernobyl nuclear plant since the reactor melted down in 1986, as a model for the generational health impacts of chronic exposure to radiation, heavy metals, and other environmental toxins.

The Golden Retriever Lifetime Study

Rodney Page (Colorado State University) described the Golden Retriever Lifetime Study (GRLS), which emerged from an effort of the Morris Animal Foundation to make progress on canine cancer prevention that started 15 years ago (Box 3).

Participants were interested in the extent to which the longitudinal dog studies investigated drinking water. The GRLS surveys owners about water

___________________

³⁵ BRAF is a gene found on chromosome seven that encodes a protein also called BRAF. This protein plays a role in cell growth by sending signals inside the cell promoting, among other functions, cell division. The V600E BRAF mutation makes a protein that is involved in sending signals in cells and in cell growth

³⁶ See https://www.niehs.nih.gov/research/clinical/studies/pegs/index.cfm (accessed February 7, 2022).

source (bottled, tap, well) as well as filtration, type of pipes, and regional ground water contamination, said Page. Lauren Trepanier (University of Wisconsin) cautioned that handling of the water significantly impacts the ability to detect different chemicals. In two prospective studies, Trepanier explained

that she uses four different collection methods to test for five chemicals in the drinking water of dogs and their owners.

One participant remarked that the size of the GRLS could make it possible to identify genetic risk factors for cancers that occur less frequently in the golden retriever but are more common in other breeds, such as histiocytic

sarcoma. Page noted that it was too early to associate risk factors with tumors, but the ability to compare littermates and the availability of over 1,000 puppies could enable a study of the intergenerational effects of exposure, cost permitting.

The Dog Aging Project: A One Health Perspective on Healthy Aging

Daniel Promislow (University of Washington) described the Dog Aging Project (DAP) (Box 4), which was initiated 14 years ago as a collaboration with Kate Creevy at the University of Georgia, inspired by Ostrander’s work

and Page’s Golden Retriever study, and now involves a team of about 100 scientists in addition to many thousands of dog owners.

The DAP can be understood from a quantitative genetics perspective, said Promislow, where G = variation in genes, E = variation in the environment, P = variation in phenotype, and G + E → P. For genetic information, DAP uses whole-genome sequencing or owner-provided breed identification, which is less reliable. Environmental information is derived from both owner-completed surveys and other sources pegged to zip codes. Phenotypic information includes behavior, age-specific disease, frailty, comorbidities, mortality, and fecundity. To probe the intermediate mechanisms by which genes and environment trans-

late to phenotype, the DAP uses tools of systems biology to analyze clinical chemistry, metabolome, epigenome, microbiome, and transcriptome, as well as fluorescence-activated cell sorting, in the 1,000-dog Precision cohort.

Scientific questions being addressed by the DAP are split into four projects: (1) Aging, to create a measure of normative aging in dogs that can inform the care of dogs as they age; (2) Genetics of Aging, to understand the genetics of aging as well as gene–environment interactions in dogs; (3) Systems Biology of Aging, to understand the ’-omics of aging in dogs; and (4) TRIAD (Test of Rapamycin in Aging Dogs), the clinical trial. Other clinical trials, including tests of cancer treatments, can be envisioned in the future, said Promislow.

Promislow described several analyses that the DAP consortium planned to submit for publication in the coming months: three different measures of activity mapped a steady decline in activity with age of the dog; rural dogs were more active and for more time than urban and suburban dogs; and older owners rated their dogs’ activity levels differently than younger owners. In an analysis of meal frequency and health, eating just one meal daily, which is a type of intermittent fasting (Mattson et al., 2017), correlated with a lower rate of various diseases in dogs. Promislow noted that this in itself does not imply causation, as sicker dogs may eat more often, but diversity of eating patterns offers an opportunity to test the health effects of various forms of diet restriction on aging. The DAP collected a great deal of data on the frequency and types of cancers in various dog breeds and on the environment. Older owners rated dogs differently than younger owners in several areas, suggesting that owners’ answers to survey questions “will partly be a reflection of themselves and not their dogs,” said Promislow.

The DAP is an open science project, said Promislow; all data will be made available through the Terra platform at the Broad Institute of MIT and Harvard, which acts as a wrapper around the Google cloud, facilitating analysis or downloading of data. Promislow encouraged scientists to propose ancillary studies that can build on the DAP infrastructure and generate future avenues for research, making DAP a “forever study.”

Companion Dogs as Sentinels for Chemical Effects on Male Reproductive Function

Richard Lea (University of Nottingham) discussed research using companion dogs as sentinels for environmental effects on male reproductive function (Box 5).

One environmental factor that varies among countries is the available brands of dog food, and these differences can influence the microbiome, noted

Andrei Gudkov (Vaika Inc. and Roswell Park Cancer Institute). To determine exposures from food, Lea analyzed chemicals in a dozen different popular types of dog food in the United Kingdom; he found that the same chemicals elevated in semen and testes were also elevated in certain brands of food, suggesting that food is a major cause of exposure of these contaminants. While his hypothesis was that exposure to these chemicals during the developmental period was responsible for the observed symptoms of testicular dysgenesis syndrome (TDS), the fact that they have short-term effects on sperm motility in vitro suggests a potential effect at later life stages as well, regardless of exposure in utero, Lea added.

Discussion: Advancing Use of Companion Dogs as Sentinels

Addressing Causality and Establishing Best Practices

Participants discussed the difficulty in establishing causality of environmental exposures on health outcomes. This would be difficult using dogs alone, said Lea, who suggested a combination of approaches in multiple species. For example, once similar associations are identified in parallel human and dog studies, a rodent model could be used to test controlled exposures, singly or in combination. Another way to address causality is through highly exposed populations like those in Minamata, offered Kannan, who suggested identifying highly exposed communities and then looking for cancers and other diseases in the pets, followed by laboratory exposure studies in rodents to confirm causality. For instance, if a high incidence of cancers were found in pets in a given geographic region or location, studies focusing on etiology would help strengthen cause–effect linkages. Miller suggested drawing on human examples that have strong causation, like the effect of cigarette smoking on cancer or of air pollution on pulmonary outcomes (e.g., emergency room visits for asthma), and extending the data collection to include an assessment of dogs in the same environment. Jan Dye (EPA) considered the GRLS in the framework of DOHAD, the Developmental Origins of Health and Disease.³⁷ The ability to compare exposures in breeding females and monitor their pups prospectively raises the potential of studying the effects of gestational and early life exposures on disease, or even just to measure the kinetics of exposure to determine the extent of placental transmission of agents or the duration of their transfer through the milk, said Dye.

Breen reflected on instances where purebred littermates living in the same household develop the same cancer within a short time window, while their littermates in different households are unaffected. Once a region with a high incidence of cancer in a particular breed has been identified, littermates living in regions with no similar environmental exposure can serve as controls. The result is “almost like a twinning study,” said Breen, with sextuplets or octuplets, albeit not identical twins. This is where dog researchers will get pushback from the human research community because of “statistics, statistics, statistics,” said Ostrander. “Any scenario . . . that sounds fascinating, the first thing they’re going to say is, well, so how many? How many families of how many individuals into how many urban zones defined by zip code with what demographic, and then did you measure environmental exposures . . . exactly the way we want . . . and was it done the same all the way across the country?” In order to elevate the dog model as a sentinel, researchers need to design experiments together with bioinformaticians, genetic epidemiologists, and statistical geneticists “and figure out how to do this right,” she added.

___________________

³⁷ See https://dohadsoc.org (accessed January 13, 2022).

Unlike humans, only about 3 percent of pets have health insurance, which influences the choice of treatment, said Chad Johannes (Iowa State University). Participants noted that this also limits the amount of data that can be obtained. The number of dogs with cancer is underestimated because many never receive a definitive diagnosis or even see the vet because pet owners cannot afford the tests and treatments.

One participant suggested adding environmental screening diagnostics for pets to regular periodic veterinary care, for the purpose of a greater public good. This could be useful for human patients as well, replied Myrtle Davis (Bristol Myers Squibb) and Miller.

Participants also considered how to develop best practices for veterinary research. Promislow commented that historically there has not been much support for large-scale veterinary studies, and the veterinary community needs to have a seat at the table to ensure that researchers are asking the right questions.

Systematic Tracking of Environmental Exposures in Dogs

Participants considered current efforts to systematically track environmental exposures in canines. This would require a definition of the normal distribution of chemicals in dogs—a reference exposome that could be developed from the current populations in biobanks, said Miller. Promislow, who is asking targeted questions focused on a limited set of exposures in the DAP, noted that a wholesale exposomic study of his specimens would be ideal, albeit cost prohibitive.

The National Health and Nutrition Examination Survey (NHANES)³⁸ is a decades-long reference study run by the Centers for Disease Control and Prevention (CDC) that measures several hundred different chemicals in a representative population of humans across the United States, but it is a series of snapshots of different people over time, not a longitudinal study of the same people. A study that follows dogs and people in the same homes over time would be extremely valuable, said Promislow.

The need for standardization of the exposome platform was mentioned by several participants. Chemical analysis is currently performed on multiple platforms in different labs using different extraction protocols without a reference set. The only solution to reduce variability, said Miller, is for multiple labs to use the same equipment. Miller suggested using HHEAR network labs, which are already standardized, for exposome analysis of large canine studies. But Breen said that labs can harmonize their approaches and cross-reference data without using the same equipment and protocols. Miller responded that this

___________________

³⁸ See https://www.cdc.gov/nchs/nhanes/index.htm (accessed January 18, 2022).

works well for targeted searches, where national and international standards already exist, but not for untargeted searches.

Aging Dogs as a Model for Cognitive Dysfunction

The utility of dogs as a model for human cognitive disorders was briefly discussed. A DAP study recently demonstrated a positive correlation between amyloid beta levels in dog brains and their cognitive dysfunction scores prior to death (Urfer et al., 2021). Promislow noted that while large-breed dogs are on a quicker trajectory than small breeds regarding life span and cancer incidence, all breeds seem to have a similar age trajectory for cognitive dysfunction, which typically begins after age 12, so the smaller breeds tend to display it more because they live longer (Watowich et al., 2020). Miller said that he preferred the term cognitive decline, which can be measured across multiple species, rather than dysfunction. He noted that recent improvements in biomarkers, like phosphorylated tau protein, have made it more feasible to measure correlates of pathology in dogs. Finch added that, unlike rodents, dogs naturally accumulate plaques and tangles with aging, produce the same amyloid peptide as humans, and have the same distribution of Alzheimer’s-like pathology in the brain. Geriatrics is an important specialty that is currently missing from board certification in veterinary care, said Promislow.

Use of Electronic Medical Records in Research: Opportunities and Limitations

There is a need for standardized data models in canine population studies, said Anne Thessen (University of Colorado). Page and Promislow noted that electronic medical records (EMRs) systems in veterinary medicine are extremely variable and proprietary, and a useful standard for answering scientific questions would likely require a new build with standards compatible with those used for humans. The Integrated Canine Data Commons at the NIH may prove a useful foundation, said Page.

One goal of this research is to overlay human and canine electronic medical records (EMRs) with environmental exposure maps and geo databases, such as those being developed by the NIEHS in North Carolina, said Breen and DeGregori. This is expensive and requires coordination of human clinicians, veterinary medical professionals, and data scientists, noted Breen.

Raj Boya (private sector IT developer) asked whether current IT applications are meeting researchers’ needs for data collection, storage, retrieval, and model design. The DAP researchers are harmonizing a variety of data types into a single resource that can be efficiently stored and accessed, offered Promislow. In addition, the DAP has responded to an NIH request for proposals to improve the readiness of existing large, complex data sets for analysis by

artificial intelligence and machine learning (AI/ML). Page and Promislow noted that it is much easier to harmonize dog and human ’-omic data than clinical information. The first challenge, said Promislow, is to harmonize the data within a study, which may include thousands of variables. For the sake of reproducibility, it is important to document the methods of data collection—the metadata, said Cui.

RELEVANCE OF COMPANION ANIMAL EXPOSURES TO HUMAN CANCER AND AGING

Participants discussed the science connecting specific exposures with cancer and aging outcomes in companion animals. Speaker presentations considered the state of the science of environmental exposure monitoring in companion animal populations, the benefits and opportunities in conducting biomonitoring in companion animals for cancer and aging research, how companion animal data can be leveraged to inform human monitoring, and the challenges and barriers to implementing exposure monitoring studies in companion animals.

Exposures to Air Pollution, Smoking, and Lead

Ambient air pollution and cigarette smoking are responsible for the deaths of over 16 million people each year globally, said Finch (Trumble and Finch, 2019). This is twice the number of pre-COVID deaths from infectious disease and 10 times those from road injuries (WHO, 2018), and this loss of life expectancy is getting worse due to industrial use of fossil fuels. Roughly one-fifth to one-third of dementia in humans can also be attributed to air pollution or smoking (Barnes and Yaffe, 2011; Cacciottolo et al., 2017; Chen et al., 2017). Tobacco smoke and air pollution have similar effects on both carotid artery disease and cognitive decline in adults (Baldassarre, 2009; Gatto et al., 2014; Künzli et al., 2010), and the two interact synergistically to increase the risks of lung cancer, brain aging, metabolic disorder, and atherosclerosis (Forman and Finch, 2018).

Dogs and Cats as Sentinels for Air Pollution

There’s no question that dogs are a sentinel for air pollution, said Finch, citing postmortem studies that compared dogs in Mexico City, where the air contains high levels of particulates and polycyclic aromatic hydrocarbons, to those in coastal cities with far less pollution (Calderón-Garcidueñas et al., 2001a, 2001b; Calderón-Garcidueñas and Duyckaerts, 2017). The Mexico

City dogs had significant neuron loss in one region of the brain stem and their myocardium contained silica-like crystals. Two studies examined particulate exposures in Alaskan sled dogs while also testing new therapeutic approaches to mitigate the negative impacts of exposure, said Finch (Montrose et al., 2015; Witkop et al., 2021). Another study of secondhand smoke also provides an example of the utility of cats as a potential sentinel for young children, said Dye. While exposure to secondhand smoke (SHS) has been shown to cause lung cancer in humans, it may also put children at increased risk for lymphoma and leukemia (Rial-Berriel et al., 2020). Cats, like children, experience an increased risk of malignant lymphoma associated with secondhand smoke (Bertone et al., 2002), and this may be due to an additional exposure through their grooming, said Dye. In contrast, medium- or short-nosed dogs are at increased risk for lung cancer from secondhand smoke, while long-nosed breeds are at increased risk for nasal cancer (Reif et al., 1992, 1998).

Finch and colleagues proposed the concept of the “gero-exposome” to describe gene–environment interactions that produce aging phenotypes (Finch and Haghani, 2021). Many components of these exposomes apply equally as well to domestic animals as humans, he said. Finch raised the possibility of a connection between secondhand smoke and dementia in humans, which could be investigated in companion dogs because canine brain amyloid Ab42, which correlates with cognitive dysfunction, increases with aging (Urfer et al., 2021), He also noted that exposure to air pollution increases amyloid production in mice containing the ApoE4 risk allele for Alzheimer’s disease (Cacciottolo et al., 2017). Age, sex, and ApoE alleles all modulate the effect of pollution on the expression of a cluster of genes in mice that contribute to amyloid production (Haghani et al., 2020), providing a molecular basis for studying this effect. In the same study, mice that had been exposed to inhaled air pollution during gestation exhibited deficits in adult neurogenesis, behaviors, and fat metabolism that differed by sex and ApoE allele, with increased adult obesity and glucose intolerance.

Pets as Sentinels for Exposures Related to Low SES

The socioeconomic gradient has a huge influence on life span in humans, essentially shifting the entire survival curve by 15 years (Crimmins et al., 2009). This effect is larger than any of the genetic influences known to impact aging and is caused entirely by environmental factors, said Finch, with “not a shred of serious evidence that there are genetic differences [at play] except as gene–environment interactions.” Low socioeconomic status (SES) is associated with increased exposures to life-shortening environmental factors, including cigarette smoking, which exposes more than 35 percent of children in the United States to secondhand smoke during their development (Tsai et al., 2018). Low SES also impacts animal welfare, with significantly more pets sur-

rendered to shelters in areas with the highest poverty (Dyer and Milot, 2019; Ly et al., 2021; Monsalve et al., 2018).

Finch said that emerging literature is associating impulsivity in children with lead exposure, and there may be related animal behaviors, though they have not been examined (Dignam et al., 2019). Use of leaded gasoline caused an “immense public exposure, which is still being worked out” in studies on children’s development, he added, noting that stray dogs and cats in Naples, Italy, had high levels of lead and cadmium in their livers and kidneys (Esposito et al., 2019). Dogs in Flint, Michigan, however, showed a low rate of elevated blood-lead levels with no clinical-grade lead toxicity (Langlois et al., 2017), though behavioral impacts were uninvestigated. Paint is now responsible for the majority of children’s blood-lead elevation and there are very few case reports of lead-paint toxicity in pets, though it merits further investigation, said Finch.

Using Silicone Samplers to Assess Air Pollution in People and Pets

On average, people spend more than 90 percent of their time indoors, where they are exposed to a number of different chemicals, particularly organic contaminants, at higher levels than from exposure to chemicals outdoors, said Heather Stapleton (Duke University). Building materials and common household items that are chemically treated slowly emit these chemicals into the air over time, and they can accumulate indoors. Examples of such chemicals include flame retardants, synthetic dyes, and stain repellants in furniture and carpets; plasticizers, metals, and flame retardants in electronics; as well as components of insulation, drywall, and flooring. Thousands of different chemicals are detectable in indoor dust, though studies typically focus on only a dozen to a few hundred compounds.

Silicone Samplers for People

Contaminant exposures are typically measured by detection of chemicals in tissues such as blood, urine, and hair. Chemicals with half-lives of months to years are usually measured in blood, and those with shorter half-lives in urine. Industry has moved to using compounds with short half-lives, to reduce the use of persistent chemicals, which is leading to a greater reliance on urine to support biomonitoring studies (except for PFAS), said Stapleton. However, urine has several drawbacks. Metabolites measured in urine may be derived from multiple chemicals, making it difficult to trace a metabolite back to its parent chemical; and the concentrations of individual compounds in urine can show dramatic variability within a day and from one day to

the next. To overcome these limitations and get a fuller accounting of the exposome, which requires an assessment of cumulative exposures over time, Stapleton and colleagues have been testing wearable samplers in the form of silicone wristbands. These wristbands are made of polydimethylsiloxane (PDMS), which absorbs chemicals well and can be used to measure ambient exposure to a variety of compounds (O’Connell et al., 2014), particularly those that are inhaled or absorbed through the skin. Chemicals in the gas or particle phase adhere to the silicone well and desorb poorly, even in the shower, so that all but the most volatile organic compounds accumulate over time. Exhaustive solvent extractions are used to pull compounds off the samplers for analysis.

A pilot study with 40 participants found that the amount of a common flame retardant, TDCIPP,³⁹ collected from silicone wristbands worn for 5 days correlated well with the amount of a TDCIPP metabolite obtained in pooled urine specimens collected over the same time period (Hammel et al., 2016). In a separate study, she found that the wristbands were equal to or better than spot urine tests at predicting the total mass of several organophosphate ester metabolites excreted in the urine over the course of 5 days (Hoffman et al., 2021).

Silicone Samplers for Companion Animals

To test whether silicone samplers could characterize exposures in pet dogs and to compare the exposures of dogs with those of their owners, Stapleton, Breen, and colleagues studied 30 pairs of owners and dogs over the course of 5 days, with owners wearing silicone wristbands and dogs wearing silicone tags. High-resolution gas chromatography (GC) and mass spectrometry (MS) were used to support the analysis of over 150 target chemicals as well as to support “suspect screening,” which allows for identification of nontargeted chemicals that might be present in the specimens. Analysis of both organophosphate esters and pesticides showed significant correlations between dog tag and dog urine, and between wristband and human urine, with tighter correlations observed in dogs (Wise et al., 2020, 2022). The same study found a strong correlation between exposures measured on human wristbands and on companion dog tags for a range of chemicals, including legacy PCBs, a novel flame retardant, pesticides, and a component of flea/tick treatments. Silicone tags have also been successfully used to measure exposures to endocrine-disrupting compounds in cats (Poutasse et al., 2019).

___________________

³⁹ Tris(1,3-dichloroisopropyl)phosphate.

Silicone Samplers for Testing Exposure over Time

Stapleton described a recent, unpublished study of 110 pregnant women in New York City, in which the most abundant class of chemicals picked up by targeted screens using these wristbands consisted of phthalates, followed by organophosphate esters. But many additional chemicals were detected using a nontargeted approach. She highlighted a need to support more research on characterizing these mixtures—each individual’s “exposure fingerprint”—to support environmental health research and investigate associations between exposure to chemical mixtures and adverse health outcomes, said Stapleton.

Silicone samplers offer many advantages for companion animal research: they are noninvasive, easy to use at home and can be returned via the mail, easy to store, and amenable to both targeted and nontargeted screening approaches, said Stapleton. In addition, they could be used to support prospective research studies and to assess how our chemical exposures change over time. People move around and experience huge seasonal differences in exposures to many chemicals in the indoor environment, and these variations can be captured by silicone samplers. By providing a measure of the integrated average exposure over time, these samplers can help support prospective longitudinal studies. However, it should be noted that they do not capture dietary exposures or individual variation in metabolism of toxicants, said Stapleton. They do not trap metals well, but they might be modifiable to better accumulate particulates that contain metals. Stapleton and Breen have begun to use the tags to study gene–environment interactions associated with an increased risk of bladder cancer in Shelties carrying the BRAF mutation.⁴⁰ Citing the work of Kim Anderson’s group at Oregon State University, Stapleton noted that silicone tags have the potential to be used in a variety of model organisms, including horses.

Cats as Sentinels for Persistent Organic Pollutants Indoors

Cats have many disease syndromes that closely parallel those in humans, including cat versions of severe acute respiratory syndrome (SARS), acquired immunodeficiency syndrome (AIDS), and other viral diseases, as well as many respiratory and endocrine disorders, said Jan Dye (EPA). As was discussed for dogs, cats’ condensed life span, indoor lifestyle, veterinary records, and fully sequenced genome (Pontius et al., 2007) enable their use as sentinels to investigate human disease syndromes in terms of environmental exposures. Dye discussed studies of two persistent organic pollutants in cats: PBDEs and PFAS.

___________________

⁴⁰ See https://www.v.org/grants-awarded/to/matthew-breen-heather-stapleton (accessed on January 3, 2022).

PBDEs and Skinny (Hyperthyroid) Cats

The brominated flame retardants, PBDEs, are endocrine disruptors with structural similarity to thyroid hormones. The first reports of environmental contamination with PBDEs in 1979 coincided with the first reports of feline hyperthyroidism (FH), and increased numbers of FH cases were seen in hot spots for PBDE contamination. PBDEs bioaccumulate in the food chain and are found in fish, which cats eat; they were also used in many household products, including carpets, furniture, and electronics. FH is now the most common endocrine disorder in cats. It is characterized by excess T4 production, which causes weight loss in middle age, and if left untreated will cause heart problems, hypertension, and kidney damage. FH has been proposed as a model for toxic nodular goiter in humans (Peterson, 2014).

The biggest risk factor for FH is living indoors, but certain types of canned cat food are also associated with an increased risk. Dye and colleagues measured PBDE content in cat foods and found a strong correlation between the amount of PBDE that a cat consumed and its likelihood of developing FH (Dye et al., 2007; Martin et al. 2000). The researchers detected multiple different PBDE products accumulating in cat serum at levels 20- to 100-fold higher than median PBDE levels in U.S. adults (Dye et al., 2007; Schecter et al., 2005). PBDE levels were high across all cats and were not significantly different between FH and non-FH cats,⁴¹ though levels correlated with age. Noting that PBDE products tend to leach out of furniture and carpeting, with high levels of PBDEs ending up in house dust that can be ingested by cats during grooming, Dye hypothesized that the very high PBDE levels seen in cats were due in part to this exposure. Like cats, young children inhale, contact, and ingest greater quantities of house dust than adults, due to their increased floor contact time and mouthing behaviors; indeed, the amount of dust ingested by cats falls in the range of that estimated for toddlers⁴² and supports the use of cats as sentinels for toddlers, said Dye. Judging from the number of publications that have cited this work, there is widespread interest in this approach, she added.

PFAS and Fat Cats

Given the ubiquitous use of PFAS, which are stain-resistant agents and emerging environmental contaminants, Dye and colleagues performed a cross-sectional analysis of cats to determine whether exposure to PFAS was corre-

___________________

⁴¹ There was an increase in one specific PBDE, OctaPBDE, in the hyperthyroid cats (Dye et al., 2007).

⁴² See https://www.epa.gov/healthresearch/understanding-exposures-childrens-environments (accessed January 4, 2022).

lated with health issues. PFAS are found in many household products, including non-stick cookware, food packaging, furniture, clothes, carpets, paint, and cosmetics. The researchers measured three PFAS compounds (perfluorooctane sulfonic acid [PFOS], perfluorooctanoic acid [PFOA], and perfluorohexane sulfonate [PFHxS]) in house dust, finding them at low levels in most households but very high in some (Strynar and Lindstrom, 2008). Among 70 cats studied, several of the indoor cats (but no feral or outdoor cats) had “amazingly high” serum PFAS levels (Bost et al., 2016), similar to levels previously seen in a case report of human teenagers, whose analyte profile was consistent with the known content of a carpet treatment product (Beesoon et al., 2012). Cats were ranked in quartiles based on total serum PFAS concentrations. Those with the highest levels (quartile 4) were more likely to be overweight and to spend more time indoors (Figure 10). Interestingly, both cats and humans have experienced a similar trend toward increasing obesity since the 1970s,⁴³ and PFAS is a suspected “obesogen,” which describes compounds that interfere with insulin signaling and thyroid signaling and may affect metabolism, though Dye pointed out that the association of PFAS exposure with obesity in indoor cats does not imply causation.

In addition to its association with obesity, PFAS exposure in humans has been linked to kidney and testicular cancer, abnormal thyroid hormone levels, pregnancy-induced hypertension, and high cholesterol with associated liver

___________________

⁴³ See https://lanekenworthy.net/2012/05/31/why-the-surge-in-obesity and http://www.catalystcouncil.org/enewsletter/october2011 (both accessed January 4, 2022).

disease. When the researchers examined cats with liver disease, feline hyperthyroidism, respiratory effusions, or kidney disease, they found elevated levels of PFAS compounds in these sick cats compared to healthy cats (Bost et al., 2016). “You don’t have to look very far into the human literature to realize that obesity, liver, thyroid, and kidney disease are emerging issues that are increasingly being associated with these types of compounds,” said Dye, adding that there are many other compounds in the indoor environment. Noting that more than 50 percent of U.S. households have one or more pets, Dye challenged investigators to take collaborative, transdisciplinary approaches to address data gaps in this untapped population and take advantage of the newer technologies, with nontargeted analyses, to identify new associations between exposures and health outcomes.

Radon Exposures and Cancer in Pets

Johannes described research using dogs and cats as sentinels for lung cancer resulting from exposure to household radon. The reported incidence of lung cancer in these pets (where it is called primary pulmonary neoplasia, or PPN) is very low, at 4 per 10,000 dogs in the United States, with an average age at diagnosis of 11 years (Culp and Rebhun, 2020). Secondhand smoke and air pollution are potential environmental risk factors for PPN, and an experimental study in beagles that were exposed to plutonium indicated that radiation could also play a role (Tierney et al., 1996). When Johannes moved from North Carolina to Iowa, which has high levels of household radiation in the form of radon, he noticed an increased number of dogs and cats in his clinics who presented with PPN.

Radon and Lung Cancer in Dogs

Radon is a breakdown product of uranium in the soil that becomes gaseous and circulates in household air.⁴⁴ Inhaled radon enters the lung and creates reactive oxidative species that can cause double-stranded breaks in DNA and increase the risk of developing lung cancer (Bersimbaev et al., 2020). As was noted for other indoor contaminants, pet dogs and cats offer an opportunity to study the health effects of radon due to their shorter life span, significant amount of time spent indoors, and absence of compounding occupational exposures. To determine whether a correlation exists between radon and lung tumors in pets, Johannes and colleagues at 10 specialty clinics in universities around the United States calculated the incidence of PPN in dogs and cats living in regions with low, intermediate, and high levels of radon (Fowler et al., 2020). The incidence of PPN in dogs living in high radon zones was more than

___________________

⁴⁴ See https://theoarp.com/What-Is-Radon (accessed January 5, 2022).

twice that of dogs in intermediate or low radon zones, with similar findings in cats. Johannes noted that the reliance on hospital clinics for data skewed the patient population to those able to afford treatment at a specialty hospital, but he added that these same homes would be more likely to have radon mitigation so, if anything, this study might underestimate the true PPN risk. These results suggest related lines of inquiry into the potential role of radon in sinonasal tumors, of which Johannes sees a much higher incidence in pets than in people, as well as its role in dementia and Alzheimer’s. Tissue deposition analysis could indicate what regions of the lung or nasal cavity are most exposed to radon.

Johannes noted that veterinary oncology is a relatively new specialty, with the first specialists certified in 1990, and the actual incidence rates for cancers in pets remain uncertain, but he is working with the large diagnostic companies to improve data collection. Getting quality clinical data is another challenge, but the increasing corporate ownership of specialty veterinary practices may lead to greater standardization of medical records and thus provide a better pool of data, he added.⁴⁵ A next step would be to perform observational cohort studies that simultaneously study people with their pets. Johannes sees opportunities for public–private research partnerships, for example, by attaching radon monitoring devices to dog collars to signal exposure in the home.

DeGregori noted that, based on extensive studies in humans and mice, radiation exposure seems to primarily lead to selection for cell populations with specific mutations, and this raises the question of how to interpret mutational signatures associated with radon exposure. Birnbaum replied that mutational signatures can be developed for dog tumors, and indeed some are already available. However, to test whether a given chemical is a carcinogen, Birnbaum favors in vitro approaches, key characteristics, and rodent and fish studies; “to test that in the dog is not very helpful,” she said.

Heavy Metal Exposures in the Dogs of Chernobyl

The Chernobyl Nuclear Disaster

On April 26, 1986, the Chernobyl nuclear plant in Ukraine experienced an explosion and fires that spewed large quantities of radioactive iodine and cesium into the air and resulted in establishment of a 30-kilometer exclusion zone where no one is permitted to live to this day.⁴⁶ Helicopters

___________________

⁴⁵ See https://www.veterinarypracticenews.com/myvpnplus/the-corporatization-of-veterinary-practices-in-america-will-you-be-a-part-of-it (accessed January 5, 2022).

⁴⁶ See https://www.who.int/publications/m/item/1986-2016-chernobyl-at-30 (accessed January 5, 2022).

dumped 25 tons of lead powder on and around the reactor, and this lead still contaminates the environment, along with other heavy metals, pesticides, and industrial toxins. Today, thousands of workers visit Chernobyl daily to work on cleanup, remediation, and new construction. Researchers demonstrated a dose-dependent risk of radiation-induced cataracts in the workers who cleaned up Chernobyl in the years immediately following the explosion (Worgul et al., 2007) as well as an elevated frequency of cataracts in the local birds (Mousseau and Møller, 2013), said Norman Kleiman (Columbia University), who observed a dose response for radiation cataracts in Chernobyl’s voles.

Toxic Exposures in the Dogs of Chernobyl

Chernobyl is an environmental and ecological disaster as well as a radiological disaster, with multiple industrial waste sites and “contamination everywhere you turn,” said Kleiman. Numerous feral dogs populate Chernobyl—descendants of canine survivors after the meltdown—and Kleiman studies the dogs’ exposures to toxic and heavy metals, pesticides, and organics, with the hypothesis that the dogs could serve as surrogates for human exposures to the same compounds and provide a measure of the potential risks to human workers. The researchers obtained hair samples from dogs captured throughout the exclusion zone and from a town 40 kilometers outside the zone, as well as from pet dogs elsewhere in Ukraine and in New York City and used inductively coupled plasma mass spectroscopy (ICP-MS) to measure 26 different metals in the samples. They analyzed unwashed and washed hair samples, with the expectation that these measurements would provide information about both external contamination and body burdens of these compounds.

Of the 26 metals tested, 19 were significantly elevated in unwashed hair and 15 in washed hair from dogs living in the exclusion zone compared to unexposed controls, suggesting that exposure was a combination of ingestion/inhalation and external contamination, said Kleiman. Some of these elevations were dramatic—for example, beryllium was increased 66-fold in unwashed hair and 25-fold in washed hair from dogs living near the nuclear power plant compared to those outside the exclusion zone. Kleiman is currently working to correlate the highest exposure levels across the range of metals with a variety of health outcomes. These observations suggest that the heavy metals in the environment around the power plant could pose health hazards to both workers and animal populations and support the use of dog hair samples and other biomarkers as surrogate indicators for human risk at the power plant and other sites of industrial and radiological disasters, said Kleiman.

Exposures through Food

Dogs as Models for Human Dietary Exposures

Given the extent to which dogs recapitulate human exposures, they constitute the perfect model for testing nutritional interventions for humans, said Joseph Wakshlag (Cornell University), although he noted that the diet and eating schedules of humans differ greatly from those of companion dogs. Thanks to federal regulations, commercial dog and cat foods offer a complete spectrum of vitamins and minerals, but they also tend to be higher in fat and protein than the human diet, with more long-chain omega-3 fatty acids. Nonetheless, the diets of dogs, despite being contained in a dry pellet, resemble those of humans better than the casein-and-lard-based rodent chow fed to lab mice and rats. Birnbaum noted that the outcomes of nutrition studies in rodents were affected by the sources of the proteins, carbohydrates, and other nutrients, and Wakshlag said that the longitudinal dog studies will address some of these questions. He said that for canine cancer studies, it is very important to control the diet, noting that all dogs enrolled in the Vaika Aging Project,⁴⁷ which tests an antiaging intervention in retired sled dogs, receive the same complete and balanced kibble. Finch noted that the sources of protein in dog food are commodities driven and can vary widely with the season and international markets. The source of fiber is also significant, said Elizabeth Ryan (Colorado State University); beet pulp and rice bran can both contribute to the percent fiber on a label, but they contain different prebiotics and will impact the microbiome differently. Fiber also affects how much protein becomes bioavailable, while fat influences the bioavailability of lipid-soluble vitamins. It will be important to study nutrient availability in these diets, added Ryan.

Heterocyclic Amines and Acrylamide in Food

Heterocyclic amines, which form adducts that are mutagenic and carcinogenic (Wakabayashi et al., 1992), build up in the hair of human omnivores (but not vegetarians) and to varying degrees in dogs (Gu et al., 2012). The intake of heterocyclic amines by dogs was estimated to exceed that of humans by a factor of five, and when these compounds were extracted from commercial pet foods, they were found to have mutagenic activity in laboratory assays; of 25 pet foods tested, 24 were mutagenic (Knize et al., 2003). Commercial dog food provides a complete and balanced diet that avoids nutrient deficiencies, but dried food production is also likely to generate heterocyclic amines because pet foods are manufactured through a process of heating, extrusion,

___________________

⁴⁷ See https://www.vaika.org/the-project (accessed January 6, 2022).

and dehydration to produce a food that is dry and stable, said Wakshlag. Vegetarian dog foods are becoming increasingly popular, noted Wakshlag, with protein levels similar to meat-based foods, so comparing dogs on the two types of food could shed light on the effects of processing.

Commercial dog and cat foods also contain the neurotoxin acrylamide, as do fried potatoes that are commonly consumed by humans.⁴⁸ Acrylamide can be metabolized into the highly potent mutagen glycidamine and is listed as a carcinogen in the state of California, which sets the no-significant-risk level (NSRL) at 0.2 μg/day.⁴⁹ When Wakschlag and colleagues tested dog foods for the presence of acrylamide, which is more concentrated in dry than wet foods, the dry food with the highest level provided a dose of roughly 34 μg/day, far in excess of the NSRL but comparable to the exposure of a person eating French fries daily. Wakshlag said it will be interesting to compare diet-related cancer risks as fresher, minimally processed dog foods enter the market.

Heavy Metals in Food

The levels of heavy metals in pet food are thought to be relatively safe, said Wakshlag, though dry dog foods approach Conservative Toxicity Reference values for cadmium and arsenic (Macías-Montes et al., 2021). Levels are dependent on the protein source, with higher levels of mercury and arsenic in fish-based kibble than in meat-based or poultry-based kibble (Kim et al., 2018). He suggested developing a repository of data concerning the composition of commercial dog foods that includes not just nutrients but also heavy metals and heterocyclic amines.

Chemical Mixtures, Cancer, and Diet—The Example of Pesticides

Cancers are characterized by cellular hallmarks that reflect the transitions from tumor initiation to tissue invasion and metastasis (Goodson et al., 2015), said Elizabeth Ryan (Colorado State University). Noncarcinogenic pollutants in the environment can affect these hallmarks at low doses, leading to the hypothesis that the cumulative effects of combined exposures to multiple agents at low doses may promote tumor progression (see Figure 11). Pesticides exert harmful effects through oxidative stress, which causes DNA damage, and by modulating gene expression, possibly through epigenetic alterations, and thus pose many dangers to humans, including increasing the risk of cancer,

___________________

⁴⁸ See https://www.researchgate.net/publication/355392171_Beynen_AC_2021_Acrylamide_in_petfood (accessed March 29, 2022).

⁴⁹ See https://oehha.ca.gov/proposition-65/proposition-65-list (accessed January 5, 2022).

she said. Pesticides are often present in combination with other toxicants in the environment, where exposure occurs through contact with contaminated drinking water or food, said Ryan. Dogs drink large quantities of water both in the home and outside, exposing them to these mixtures (Aslan et al., 2020). Stapleton noted that although spot urine is variable for many chemicals, a lot of pesticides, phenolic compounds, and amines could be studied in dogs’ spot urine samples, in connection to their potential risks for adverse health outcomes later in life.

Metabolism May Modulate Exposure Risk

When studying the effects of complex, low-dose chemical exposures such as those from pesticides, it is important to take into account metabolic dif-

ferences among animals, which may influence exposure dose, immunity, and risks for developing cancer, said Ryan. For example, there are metabolic differences between small and large dogs. Among laboratory dogs that were housed together and fed identical diets, small dog breeds could be distinguished from large breeds based on distinct plasma metabolome signatures and distinct clinical signatures; there were also differences in the fecal microbiome (Middleton et al., 2017). Body condition also has an impact on metabolism, and potentially on exposure: When Ryan and colleagues fed an identical diet to 66 companion dogs that were healthy weight, overweight, or obese, the dogs’ plasma, fecal, and urine metabolomes differed based on body condition score (Forster et al., 2018). Plasma lipid composition was altered in the overweight and obese dogs, indicating that, as in humans, being obese/overweight affects metabolism in dogs and may impact levels of pesticides, said Ryan. Indeed, in a 1989 study on companion dogs that found an association between insecticide exposure and bladder cancer risk, this risk was enhanced in overweight or obese dogs (Glickman et al., 1989).

Diet May Modulate Exposure Risk

Diet also has the ability to modulate exposure, and this may derive in part from its effect on the microbiome. When Ryan and colleagues screened 21 healthy-weight dogs for 301 pesticides in their urine, they detected atrazine in all 21 dogs and combinations of multiple pesticides in 20 (Forster, et al., 2014). Changing to a 25 percent bean diet altered the dogs’ metabolomes (Forster et al., 2012, 2015) as well as their fecal microbiomes (Kerr et al., 2013). Ryan noted that chemicals from pesticides interact bidirectionally with the gut microbiota; on the one hand, microbial enzymes can alter these chemicals, potentially leading to activation or detoxification, and on the other hand, pesticides can contribute to dysbiosis (reduction in microbial diversity) (Giambò et al., 2021). Antibiotics can increase pesticide bioavailability (Zhan et al., 2018), indicating that disruptions to the gut microbiome may increase the risk from pesticides and highlighting the need to study the effect of medications on pesticide risk, said Ryan.

As a proof-of-concept for testing the ability of the microbiome to mediate the disease-causing effects of environmental chemicals, Ryan and colleagues examined the effect of diet change on the microbiota of dogs undergoing chemotherapy for cancer. Lymphoma alters the fecal microbiome of dogs (Gavazza et al., 2018), as does chemotherapy, she said. When dogs undergoing chemotherapy were fed a diet of 20 percent beans and 5 percent rice bran, some microbial diversity was restored to the gut and probiotic commensals increased, suggesting that it may be possible to modulate the effects of environmental exposures through dietary modifications, said Ryan. However, it remains unclear what constitutes a healthy microbiome

in dogs, or for that matter in humans, which makes it difficult to assess the impacts of dysbiosis. Nonetheless, there are similarities in human and dog gut microbiomes, and microbes are exchanged between humans, animals, and the environment (Trinh et al., 2018). Metabolomics offers a powerful tool for understanding the global landscape of human gut microbiota (Sung et al., 2017) and can be used for this purpose in dogs, added Ryan.

Chemicals in Flea-and-Tick-Control Products

The mechanisms of action of many pesticides used for flea and tick control are connected to pathways involved in cancer, said Ryan. Participants considered the relative benefits of controlling fleas and ticks in dogs, versus the risks of exposing dogs and humans to pesticides. “Dogs are constantly being bombarded with these pesticides” that carry a potential risk of cancer, said Wakshlag, to which Johannes replied that, from an infectious disease standpoint, keeping ticks and fleas off dogs has a tremendous quality-of-life benefit, and it may come down to weighing the trade-offs. Nontoxic alternatives were mentioned, but evidence of their efficacy is lacking.

The activities of topical flea and tick treatments could be altered by the skin microbiome, said Ryan, and participants considered studying the skin microbiome as a reflection of exposures. While dogs’ fur makes collection of skin microbes a challenge, Ryan remarked that techniques for collection of the skin microbiome have evolved, and it would be valuable to develop a standardized sampling scheme. Some animals also develop allergies to these chemicals, and age-related susceptibility to allergy may be an untapped indicator of environmental compounds acting on the immune system.

Discussion: Relevance of Companion Animal Exposures to Humans

Communicating Exposure Risks to the Public

Workshop participants considered the need and complexity of communicating exposure information to the public, particularly when researchers do not yet understand the relationship between a given exposure and disease because so much of the data from companion animals are correlative. The exposure–disease connection needs to be something the general public can understand, said Miller. Scientists are interested in molecular mechanisms but what really matters are the functional consequences, and these have to be actionable. Reducing plasticizers or changing diet are actionable goals, “and if

it doesn’t get to that translation point, we’re just wasting our time,” he added. This requires help from science communicators. For example, DNA testing companies put considerable thought into what to tell a person with a particular polymorphism.⁵⁰ “We really have to think about that message part at the end,” said Miller. Deziel added that communication with policy makers is also needed to drive action on public health. Stapleton said she reports measured exposures back to her study participants, and she stressed the importance of providing context. Her reporting template indicates whether a particular exposure is above or below the median, provides guidance for reducing exposure, and makes it clear that the researchers still do not know the risk associated with the exposure.

Deconstructing Complicated Environmental Exposures—Or Not

We live in a “chemical soup” together with our dogs, said Birnbaum. Rather than trying to deconstruct the mixtures, maybe we could look on a larger scale—diet, pesticides, air pollution—and not fixate on knowing everything in the mixture. Exhaustive efforts to identify carcinogens in cigarette smoke identified a group of “proven” and “possible” carcinogens, noted Finch. Yet, remarkably, cigarette smoke and air pollution do almost the same thing despite the huge range of variation, so singling out individual components would not be terribly useful. Approaching the problem on a macro level—for example, analyzing how diet and omega-3 content may protect against Alzheimer’s disease in relationship to environmental factors that are not yet defined—may prove the most rewarding, said Finch.

Dogs as Models for Interventional Studies

Dogs are a remarkable, and yet untapped, resource for interventional studies of cancer, said Wakshlag. Their nutrient intake is far easier to control than that of humans and ripe for studying the influence of the omnivore diet on both cancer prevention and progression. However, prevention studies are currently hamstrung by the fact that little is known about the composition of commercial dog food other than macronutrient content. To get the necessary information on heavy-metal exposure, pesticides, and so forth, when there are 250–300 companies, each with 5–40 brands on the market, is a “gargantuan

___________________

⁵⁰ See https://www.23andme.com (accessed January 13, 2022).

task,” but no manufacturer will fund this work, so it needs government support, said Wakshlag.

Johannes emphasized the need to partner with industry. He noted that state-of-the-art diagnostics are now available to veterinarians, making it possible to match environmental exposures to mutations and biomarkers. However, the therapeutics have lagged behind, so even when drugs exist for human cancers, veterinarians are often unable to use them because they have not been validated in dogs or cats. There is an opportunity to test mitigation strategies via parallel studies in dogs, said Page. This might include, for example, dietary measures to mitigate exposures to heavy metals or air pollution.

Prioritizing Exposures for Future Research

Given the broad range of exposures that could be measured and finite resources, prioritizing the questions to study will be crucial, said Promislow, who noted that studies vary in their aims. There are banked collections around the world and new collections being started (see section on Biobanks for Companion Animal Sentinel Studies). Some questions address widespread conditions and could potentially benefit many, whereas others address rarer conditions. How can the research be optimized? Stapleton envisioned focusing on ambient measurements that are cumulative over time, in an exposomic framework, while at the same time paying careful attention to the diet and microbiome. Ryan highlighted changes to the microbiota and immune evasion by tumors as two potential readouts of the effects of chemical mixtures.

The key value of these longitudinal studies is their tissue and data banks, said Wendy Shelton (Virtual Beast/Colorado State University), and it will be important for future researchers with fresh ideas and technologies to be able to access these resources. Ten years from now, when the results come in from prospective dog cohorts, we’ll be able to go back and look at specimens to determine the relationships between priority agents and disease, said Ryan. The question will then be whether to respond by designing an intervention, she added.

Wakshlag said that given the capability to measure “just about anything” in interstitial fluid and the ongoing longitudinal dog cohorts, now is the time to assemble expert panels to identify the important exposures to study, and to evaluate the risks of pesticide exposures based on available human and dog data. Considering his own data on heterocyclic amine and acrylamide exposure in dogs and humans and their potential long-term health risks, it is also time to reconsider the established safe upper limits for these compounds,

he added. Study design, statistical analysis, and readout also require consideration, said Ryan.

ACCELERATING CROSS-SPECIES COMPARISONS: OPPORTUNITIES AND CHALLENGES IN DATA SOURCES, COLLECTION, STORAGE, MODELING, AND SHARING

Participants discussed practical concerns for collecting and storing data and making it available to the wider research community, as well as methods for comparing data obtained from companion animals to that collected from humans. Speaker presentations considered strategies for standardizing, sharing, and aggregating health records and relevant metadata across species; the obstacles (e.g., scientific, infrastructure, ethical, and financial) to using companion animals as biomonitors; incentives needed to encourage data sharing and collaboration; and implications for expanded, systematic collection of these data.

Human Exposure Assessment

Widespread ambient environmental exposures, ranging from air pollution and drinking water contamination to nonchemical agents such as light pollution, may affect health even at low levels, said Rena Jones (NCI), who is building data resources to support cancer epidemiology studies of these exposures in humans. Jones combines individual and population exposure data, collected through environmental monitoring, biological sampling, and surveys, with secondary data derived from exposure assessments that were not specifically intended for epidemiological studies, such as regulatory monitoring measurements, census data, and satellite imagery; these data are then linked to individuals based on geographic location, using geographic information systems (GIS).

Use of Spatially Derived Exposure Assessments to Predict Individual Exposures

Jones described two studies that used spatially derived exposure assessments to predict individual exposures. In one study, Jones and colleagues probed the association between lung cancer and ultrafine particles (UFPs)—components of outdoors air pollution that are derived from combustion of diesel fuel and behave as a gas, passing through walls and cell membranes. For the Los Angeles Ultrafines Study, Jones and colleagues obtained health data from a cohort of 52,000 participants enrolled in an existing longitudinal

study.⁵¹ To determine UFP exposure, the researchers measured UFPs and other pollutants at 250 monitoring sites in the Los Angeles basin, detecting a range of exposures that were then analyzed by linear modeling called land use regression, which considered nearby traffic, proximity to airports, population density, and other location-based environmental characteristics. This generated an equation that could be used to predict UFPs at each participant’s location, and the predictions were validated against measurements taken at a subset of cohort residences (Jones et al., 2020).

In a second study, which assessed exposures to nitrate in residential well water, Jones and colleagues applied a different GIS approach—random forest modeling—to analyze nitrate measurements that had been obtained previously through community and personal well-water sampling in rural Iowa and North Carolina.⁵² Nitrate is an agricultural contaminant that is regulated in public drinking water but not in private wells. The researchers identified 58 predictive variables, which were validated against sampling with good accuracy and high specificity (Messier et al., 2019), enabling them to predict nitrate levels in private wells at levels below the regulatory limit.

Use of Regulatory Data to Predict Exposures

In ongoing work to test the feasibility of estimating exposure to PFAS from drinking water, Jones and colleagues are applying GIS analysis to a combination of regulatory data from the EPA that measures PFAS levels in the public water utilities⁵³ and participants’ responses to detailed surveys about their water intake, to identify determinants of serum PFAS in 1,200 women nested within a cohort in California.⁵⁴ Jones and colleagues also use regulatory data to estimate exposures to the nearly 800 chemicals, released from over 15,000 industrial point sources scattered around the country, that are reported to the EPA each year under the Toxics Release Inventory (TRI) program.⁵⁵ Half of these point sources emit known or probable carcinogens. To turn a TRI emissions estimate into an exposure estimate, the researchers employ inverse distance weighting, summing the emissions from all facilities within a given radius of a participant and weighting each one by the inverse of its distance to the participant, with additional fine-tuning (e.g., adjustment for prevailing wind direction). Jones noted that these estimates

___________________

⁵¹ See https://dietandhealth.cancer.gov/history.html (accessed January 18, 2022).

⁵² See https://aghealth.nih.gov (accessed January 18, 2022).

⁵³ See https://www.epa.gov/dwucmr (accessed January 18, 2022).

⁵⁴ See https://www.calteachersstudy.org (accessed January 18, 2022).

⁵⁵ See https://www.epa.gov/toxics-release-inventory-tri-program (accessed January 18, 2022).

are not surrogates for exposure, but they enable ranking of participants based on their relative exposures. Focusing on dioxin and PCB levels in serum obtained from the NHANES, the researchers validated these GIS-derived exposure estimates against actual burden as measured in serum. Adding satellite imagery allows for triangulation of location data, to (for example) determine the height of a smokestack, enabling the construction of actual pollution dispersion models.

Jones noted that these same types of exposure assessment approaches centered on residential locations can be applied to the study of companion animals as sentinels for humans, with the advantage that the shorter latencies and life spans of animals require a shorter monitoring period. Furthermore, although focusing on exposures in the home may be an oversimplification for humans, who move around, it may be sufficient for pets. Jones encouraged companion animal researchers to propose ancillary studies to human studies that are recruiting new cohorts, such as the intramural NCI Connect for Cancer Prevention study,⁵⁶ which includes infrastructure for GIS and location-based exposure assessments, with complete residence histories and drinking water intake data. She expressed the need to gather more data on disadvantaged populations, which tend to be underrepresented in many cancer cohorts.

Comparative Oncology: How Dogs are Helping Researchers Understand and Treat Cancer

Unlike laboratory models of disease, natural development of cancer in companion dogs offers “the really unique advantage of having a coevolving tumor, stroma, immune system, and microenvironment,” and dogs can be used to evaluate novel therapies that cannot be tested in treatment-naïve human disease, said LeBlanc (LeBlanc and Mazcko, 2020). Following a 2015 workshop on comparative oncology (NASEM, 2015), there has been a groundswell of research in this field, aided by a network of veterinary schools that conduct clinical trials through the NCI Comparative Oncology Trials Consortium (COTC).⁵⁷ This has driven a dramatic improvement in data to inform drug development and advance the translational relevance of canine cancer for humans, said LeBlanc, whose Comparative Oncology Program engages in laboratory research and COTC-run clinical trials and supports additional

___________________

⁵⁶ See https://www.cancer.gov/connect-prevention-study (accessed March 7, 2022).

⁵⁷ See https://ccr.cancer.gov/comparative-oncology-program/consortium (accessed January 18, 2022).

collaborations, such as the ICDC and a positron emission tomography–computed tomography (PET/CT) and Molecular Imaging Consortium.⁵⁸

The Companion Dog Model for Osteosarcoma

Osteosarcoma is diagnosed in about a thousand children or young adults per year in the United States, and this rarity makes it a challenge to study in humans. The 5-year survival rate for children with metastatic osteosarcoma is less than 30 percent, and treatments have not improved in 30 years, despite significant efforts. New drug targets are needed, said LeBlanc (Bishop et al., 2016). Companion dogs are good models for studying osteosarcoma because it is diagnosed in more than 10,000 dogs each year, she added. Most canine osteosarcomas are metastatic, with a survival time of about 1 year (Selmic et al., 2014). Pediatric and canine osteosarcomas look strikingly similar at every level: histopathological, molecular, and clinical, including the tendency of the tumor to metastasize to the lung (Gustafson et al., 2018). There are broader similarities as well, with osteosarcoma affecting big dogs and, in general, tall children. There may be a baseline risk that is modified through the onset of rapid scalable growth through puberty, with the effects of estrogen on bone exerting a protective effect on girls, who are affected less frequently than boys (Simpson et al., 2017), though this is harder to assess in dogs that are spayed or neutered. The clinical similarities in the human and dog diseases support hypotheses involving IGF pathways, micro stresses, micro fracture/trauma, or bone remodeling, said LeBlanc.

LeBlanc and colleagues are working to define a genomic landscape for canine osteosarcoma that can be integrated with human osteosarcoma data to identify new druggable targets for preclinical work in mice and then dogs, with the ultimate goal of benefitting both canine and human patients. Unlike humans who enroll in clinical trials of novel therapies, dogs with osteosarcoma provide a source of treatment-naïve tissue, enabling LeBlanc and colleagues to develop a biospecimen repository with primary tumors, matched normal tissue, blood, and other samples. The researchers are using these specimens to describe the molecular landscape of canine osteosarcoma, in a project called Deciphering the Osteosarcoma Genome in Dogs (dog²) (LeBlanc and Mazcko, 2020) (see Figure 12).

Using Dog Osteosarcoma Data to Inform Human Medicine

With the goal of using canine molecular and clinical data to inform development of therapies for osteosarcoma, researchers conducted clinical trials for

___________________

⁵⁸ See https://ccr.cancer.gov/comparative-oncology-program (accessed January 18, 2022).

dogs testing two different drug treatments and obtaining biospecimens from about 450 canine osteosarcoma patients. Their first published study, which found that adding rapamycin to dogs’ treatment did not improve outcomes, provides an example of the utility of this approach, said LeBlanc, informing the consideration of rapamycin and its analogues for treatment of children while providing an opportunity to ask why it failed to work in dogs on a molecular level (LeBlanc et al., 2021).

Within the dog² project, LeBlanc and colleagues are combining analysis of bulk genomic data from clinical specimens; AI/deep learning and computational modeling to identify biomarkers from imaging data; immune landscape profiling, using a new canine-specific immuno-oncology gene expression panel;⁵⁹ and analysis of the tumor microenvironment to interrogate tumor heterogeneity and progression, in order to define prognostic signatures and tumor subtypes that can form the basis for harmonized canine/human clinical trials. To support this initiative, the researchers developed a computational approach to identify cancer-specific gene modules and biological mechanisms that are shared between humans and dogs (Tawa et al., 2021). Perhaps combining genome-wide association studies (GWAS) and tumor genetics with activity tracking could elegantly address the question of whether osteosarcoma requires a certain number of load cycles on the limbs. If so, this might explain why it affects adolescent humans (ages 12–16) and geriatric dogs (ages 8–10), LeBlanc postulated.

Biobanks for Companion Animal Sentinel Studies

A biobank is “a facility for the collection, preservation, storage, and supply of biospecimens and associated data, which follows standardized operating procedures and provides material for scientific and clinical use” (Hewitt and Watson, 2013). Biobanks are intended to supply and distribute specimens for research, said Marta Castelhano (Cornell University), otherwise they become “biohoards” (Catchpoole, 2016). Nonetheless, there are considerable obstacles to the effective use of the material in many biobanks. These include lack of uniform systems for banking, data annotation, and bioinformatics, and insufficient or variable consent procedures (e.g., limiting analysis to a specific set of environmental agents). There is also often reluctance to share biospecimens and difficulty in accessing them from outside of the institution, and many existing biobanks lack the necessary information to link participants to exposure data.

___________________

⁵⁹ See https://www.nanostring.com/products/ncounter-assays-panels/oncology/canine-io (accessed January 18, 2022).

Standards for Biobanks

The first international standard for biobanks, ISO 20387, was published in 2018,⁶⁰ and two documents offer best practices to guide the collection, storage, and analysis of biobank specimens (Campbell et al., 2018; NCI, 2016). Castelhano emphasized the importance of conformity assessment (CA), which demonstrates that a biobank complies with specific requirements. CA can be performed at three levels of stringency: first party, where the biobank attests to its own conformity; second party, where a user reports on conformity; or third party, where an impartial external body provides certification or accreditation, with accreditation going the farthest by evaluating the technical competence of personnel (Mouttham et al., 2021). Biobanks should aim for the most stringent assessment practice with accreditation, said Castelhano, saying that ten billion dollars a year are lost to irreproducible research in the United States due to inadequate biological reagents and reference materials (Freedman et al., 2015). She emphasized that “standardization and auditing of biological materials . . . can enhance cumulative production of scientific knowledge by improving both availability and reliability of research inputs” (Freedman et al., 2015; Furman and Stern, 2011). In addition to assuring technical competence, said Castelhano, accreditation enables harmonization of procedures, allowing for interoperability among biobanks. This is especially important when multiple institutions are involved, as is often the case for exposure studies (Lermen et al., 2020).

Biobanks for Companion Animal Sentinel Studies in the United States

Established U.S. biobanks offer opportunities to advance research on the role of companion animals as sentinels for environmental exposure, said Castelhano. One of these is the National Institutes of Standards and Technology (NIST) biorepository in Charleston, South Carolina, which contains two environmental specimen collections: the National Biomonitoring Specimen Bank, established in 1979 to store human specimens, and the Marine Environmental Specimen Bank, established in 2002 to store diverse marine specimens (Becker and Wise, 2006). Both biobanks were designed for long-term storage and use over the next 50–100 years. The Centers for Research in Emerging Infectious Diseases (CREID)⁶¹ is creating an international, standardized biobanking infrastructure in several countries, specifically for pandemic pre-

___________________

⁶⁰ See https://www.iso.org/obp/ui/#iso:std:iso:20387:ed-1:v1:en (accessed January 18, 2022).

⁶¹ See https://creid-network.org (accessed January 18, 2022).

vention, which includes the collection of companion animal specimens, said Castelhano.

Castelhano coordinated the biobanking of high-quality, clinically annotated specimens from the Dog Aging Project, with the intent of making them available to the wider research community. As noted in the previous discussion of the Dog Aging Project, specimens are collected from thousands of dogs by veterinarians across the United States, sent to Texas A&M, and then shipped to Cornell, where they are catalogued in the Biobank Information Management System, processed, and stored. Whole blood specimens are processed into serum, plasma, and live peripheral blood mononuclear cells (PBMCs). Castelhano noted that while blood and urine are easy to obtain, they reflect recent exposure rather than long-term accumulation. Due to the relatively low levels of environmental contaminants in these fluids, large volumes are required to provide sufficient material for biobanking. In contrast, postmortem tissue offers greater potential for assessing bioaccumulation. Postmortem specimens from the Cornell Veterinary Biobank, collected by its rapid necropsy team with an average postmortem interval (time from death to specimen preservation) of 43 minutes, include multiple tissues “from brain to bone” as well as histopathology authentication. These specimens, along with clinical data, are available for use by the larger research community, said Castelhano.

To encourage specimen sharing and collaboration, biobanks need infrastructure support, including HEPA-filtered air in sample-preparation and freezer rooms; cryogenic homogenization; and 24-hour monitoring systems for freezers; as well as risk management plans to avoid loss of collections during a natural disaster, said Castelhano. Standardized protocols are crucial to ensure specimen integrity and stability during a long storage, and biospecimens should be routinely analyzed to assess stability. Standardized consent procedures are also essential. Researchers need a clear access policy, procedures for locating specimens and data, and standardized annotations, including geo-exposure data. Ideally, Castelhano said, she would like to see a national, centralized environmental specimen banking initiative for human and sentinel species. She suggested that researchers establishing biobanks engage three organizations with specimen and data locator capabilities: the Clinical and Translational Science Award One Health Alliance (COHA),⁶² the International Society for Biological and Environmental Repositories (ISBER),⁶³ and the NIST.⁶⁴ She also suggested using the model established by BBMRI-ERIC,⁶⁵

___________________

⁶² See https://www.ctsaonehealthalliance.org (accessed January 18, 2022).

⁶³ See https://www.isber.org/page/about (accessed January 18, 2022).

⁶⁴ See https://www.nist.gov (accessed January 18, 2022).

⁶⁵ BBMRI-ERIC is a European research infrastructure for biobanking. See https://www.bbmri-eric.eu (accessed January 18, 2022).

a distributed research infrastructure for biobanks in Europe that includes compliance audits, and added that the Cornell Veterinary Biobank accepts materials from smaller studies.

Using Ontologies to Unify Genomics and Phenomics across Species

The human genome has been sequenced, but researchers still do not know much about what it does, how the environment affects gene expression, or how these phenotypes manifest themselves, said Thessen. Adding more species to a data set improves the information landscape. For example, although only about 4,000 of the 19,000 known human genes have gene-to-phenotype annotations, assigning the annotations from five model organism species to their human orthologs boosts phenotypic coverage to 16,000 human genes. Noting that every species has unique phenotypes that optimize it for a different research purpose, from snails modeling spontaneous regeneration of injured neurons to armadillos acting as the only nonhuman host of M. leprae, Thessen draws on a wide range of organisms to construct two types of data infrastructure: ontologies and knowledge graphs.

Developing Ontologies across Species

An ontology is a formal, computational representation of knowledge in a particular domain, which uses defined terms with clearly defined hierarchical relationships (e.g., an efferent neuron is a type of neuron, which is a type of neural cell, which is a type of cell).⁶⁶ Each relationship between terms is accompanied by a series of formal logic rules readable by software “reasoners,” enabling logical inferences and sophisticated data queries. The challenge in developing a multispecies ontology is that each species’ genomic database employs a unique vocabulary to describe phenotypes (Thessen et al., 2020). For example, “ulcerated paws” in the mouse corresponds to “palmoplantar hyperkeratosis” in humans and “hairy dog feet” in dogs, so software could not connect the three. Instead, Thessen and colleagues break down complex phenotypes into basic, species-neutral elements, producing, for example, “increased keratinization of stratum corneum layer of the skin on the auto-pod,” which could describe this particular phenotype across all three species (Figure 13). The researchers then use a unifying ontology to combine multiple species in the same ontology and thereby generate testable hypotheses. For example, in the unifying phenotype ontology, UPheno,⁶⁷ “enlarged heart

___________________

⁶⁶ See bit.ly/ontology101 (accessed January 19, 2022).

⁶⁷ See https://obofoundry.org/ontology/upheno.html (accessed January 19, 2022).

in zebrafish” and “enlarged heart in human” are both subtypes of “enlarged heart,” so the human orthologues of genes that impact heart size in zebrafish might be hypothesized to influence heart size in humans, said Thessen.

Knowledge Graphs Map Data across Ontologies

As the data infrastructure grows more complex, it enters the territory of a knowledge graph, such as that of the Monarch Initiative (Shefchek et al., 2020).⁶⁸ The knowledge graph integrates genotype–phenotype data across species, mapping data from multiple sources to a small number of unifying ontologies (e.g., UPheno) and combining these unifying ontologies. The Monarch knowledge graph has helped diagnose rare diseases that manifest differently in each patient, using fuzzy matching over groups of phenotypes. It can also help prioritize candidate genes for disease diagnosis or treatment (Smedley et al., 2016).

The missing piece of the knowledge graph is exposures, said Thessen, who is working with exposure scientists and toxicologists to create a semantic data

___________________

⁶⁸ See https://monarchinitiative.org (accessed January 19, 2022).

model for representing exposure data. Each exposure can be modeled as an event that joins a stressor/stimulus and a receptor (individual, part, or group of organisms); the exposure has an outcome (phenotype or disease); and each of these elements—event, stressor, receptor, and outcome—contains its own set of metadata (Mattingly et al., 2012). The exposure ontology can then be integrated into the larger knowledge graph, connecting exposures and outcomes to genes and phenotypes across species. Thessen gave an example of a zebrafish exposure to a toxin, aldicarb, which promotes microcephaly, which is affected by the rpl11 gene (Figure 14).⁶⁹ The human rpl11 ortholog is connected to multiple phenotypes and diseases, raising the hypothesis that rpl11 may mediate disease caused by aldicarb exposure in humans. Dogs have a rpl11 ortholog, though standardized dog and cat phenotype and disease data sets are lacking; “we need a dog phenotype ontology,” said Thessen. The knowledge graph will offer the exciting opportunity to do “fuzzy matching” using exposures, genes, and phenotypes, bringing together heterogeneous data from multiple species and modalities. Thessen noted that while chemical exposures are straightforward to describe in machine-readable language, some exposures, including many related to socioeconomic status, are hard to define semantically, and individuals generally experience multiple exposures simultaneously.

Data, Samples, and Modeling

The Mars Petcare Science Engine

Mars Petcare has anonymized data on nearly a hundred million unique pets through its global network of brands in over 125 countries including a number of pet food brands,⁷⁰ more than 2,500 veterinary hospitals and diagnostic labs, DNA testing services,⁷¹ and pet activity trackers,⁷² said Angela Hughes (Mars Petcare). To enable robust analysis of the large amount of collected data, which includes granular information on genetics, health, nutrition, and behavior, researchers at Mars Petcare’s Waltham Petcare Science Institute built the Petcare Science Engine, a cloud-based platform.⁷³ The Petcare Science Engine inputs data from electronic medical records in both structured and unstructured formats and uses artificial intelligence to structure the unstructured text (medical notes) for analysis. Mars Petcare’s massive

___________________

⁶⁹ See https://ntp.niehs.nih.gov/whatwestudy/niceatm/test-method-evaluations/dev-tox/seazit/index.html (accessed February 11, 2022).

⁷⁰ See https://www.mars.com/made-by-mars/petcare (accessed March 8, 2022).

⁷¹ See https://www.wisdompanel.com/en-us (accessed January 20, 2022).

⁷² See https://www.whistle.com (accessed January 20, 2022).

⁷³ See https://www.youtube.com/watch?v=UanTIsbMY_8 (accessed January 20, 2022).

data resource can be used to improve pet health care, said Hughes, giving the example of anesthesia mortality, which is twice as high in cats as in dogs. To identify risk factors linked with mortality, researchers created over 300,000 structured anesthesia records from nearly 1 billion raw medical notes, which they are using to identify risk factors and build predictive algorithms that can provide real-time feedback to veterinarians to make anesthesia safer for cats, said Hughes.⁷⁴

With the aim of improving preventive health care, Mars Petcare’s Pet Insight Project uses the collar-mounted Whistle FIT activity tracker to monitor a wide range of behaviors in dogs such as eating, drinking, chewing, and scratching.⁷⁵ Synchronizing thousands of hours of video footage from citizen scientists with 11 million days of their dogs’ Whistle FIT data, researchers developed a deep-learning algorithm that can accurately classify dog behavior based on activity tracking alone (Chambers et al., 2021). Validation of the algorithm found that it correctly identified eating 94 percent of the time and

___________________

⁷⁴ See https://www.thetimes.co.uk/static/data-analytics-dogs-cats-ai-machine-learning-mars-petcare-raconteur (accessed January 20, 2022).

⁷⁵ See https://www.petinsight.com (accessed January 20, 2022).

drinking 98.8 percent of the time and could identify more nuanced behaviors, such as sniffing and licking, to a lesser extent. This activity tracker provides a new way to capture nuanced behaviors as a biomarker, said Hughes, enabling early detection of changes that could result from disease, such as increased scratching. Hughes also noted that Whistle GPS trackers could enable the study of social interactions in groups of dogs.

Hughes described several other applications of machine learning that were developed to analyze Mars Petcare’s large data sets, including automated evaluation of chest and abdominal radiographs in dogs and cats; reliable, automated cancer grading of biopsy slides using mitotic figure analysis; and early prediction of feline chronic kidney disease (CKD). CKD is the main cause of death in cats over 5 years old and has been difficult to diagnose in its early stages, said Hughes. Researchers at Mars Petcare developed a predictive tool that uses a combination of six common cat health measurements to predict CKD up to 2 years before a traditional diagnosis could have been made, enabling care to potentially slow the progression of disease, said Hughes (Bradley et al., 2019).

Embark: Using Dog Genetic Data to Drive Translational Discovery

The domestication of dogs from wolves to phenotypically differentiated closed breeding populations has provided a model for studying rapid evolution, genetic load, and personalized genomics, said Adam Boyko (Embark Veterinary, Inc./Cornell University). Genetic studies of dogs, accelerated by the sequencing of the dog genome (Lindblad-Toh et al., 2005), have identified hundreds of likely causal variants underlying Mendelian traits, most of which are potential models for human disease.⁷⁶ Performing genome-wide association studies on 4,200 dogs, Boyko and colleagues identified new genes associated with six complex diseases and two morphological traits (Hayward et al., 2016).

Village dogs offer a window into early dog evolution because they have been largely preserved from the type of artificial selection used to generate closed breeding populations, said Boyko. By comparing linkage disequilibrium in dogs around the world, Boyko mapped the likely origin of domesticated dogs to central Asia (Shannon et al., 2015).

Despite the progress in identifying Mendelian and large-effect loci in dogs, there has been slower progress in elucidating the genetics underlying behavior, longevity, and complex diseases such as cancer, which are big-data questions and will require large-scale studies on hundreds of thousands of dogs, said Boyko. Noting that direct-to-consumer genetic testing of humans has built data sets on the order of tens of millions of individuals, Boyko developed a similar

___________________

⁷⁶ See https://www.omia.org/home (accessed January 20, 2022).

model for dogs. Embark⁷⁷ was begun in 2015 and now has over 900,000 dogs in its database, each of which has been genotyped at over 230,000 markers, making it probably one of the biggest dog genetic databases in the world, said Boyko. The database consists of roughly 70 percent mixed-breed dogs, 30 percent purebred, with the remainder wolf dogs and village dogs. Boyko and colleagues are using the data for discovery and have identified novel loci associated with blue eyes in Siberian huskies (Deane-Coe et al., 2018) and with a mottled coat pattern (roaning) in Australian cattle dogs (Kawakami et al., 2021); the causal mutation in each case is a DNA duplication.

To obtain health data, Embark surveys owners annually, with over 180,000 surveys completed in 2021. Owners rate their dogs’ general health and report in detail on organ systems, symptoms, and diagnoses. The researchers found that inbreeding has a strong negative correlation with longevity, with outbred and mixed-breed dogs outliving inbred and purebred dogs at every body size (Yordy et al., 2020). Noting that big data is less clean than small data, Boyko aims to maximize statistical power by balancing sample size and the signal-to-noise ratio, using automated quality control and tracking. Boyko has been particularly interested in automating discovery, which enables the researchers to explore associations across the entire genome and then follow up with a targeted validation. Million-dog cohorts are too large for standard approaches like PLINK,⁷⁸ said Boyko. It is important to standardize not only data models but also how the significance of genetic variants is communicated to owners. This can vary with breed and is complicated for mixed breeds, he noted.

Boyko mentioned some limitations of using dogs as models for humans, including incomplete age data and differences in the degrees to which stature, reproductive history, smoking, and inbreeding influence health outcomes in the two species.

Data and Resource Sharing with the American Kennel Club

About eighteen percent of U.S. dogs are from litters registered with the AKC, said Mark Dunn (AKC). The AKC has 175,000 active breeders who volunteer 50,000 DNA specimens from pedigreed dogs with known familial history each year; the DNA collection now includes about 600,000 specimens going back over 20 years. The AKC also enrolls 70,000 spayed and neutered mixed-breed dogs each year. Dunn added that the AKC database was recently

___________________

⁷⁷ See https://embarkvet.com (accessed January 21, 2022).

⁷⁸ PLINK is a free, open-source whole genome association analysis toolset, designed to perform a range of basic, large-scale analyses in a computationally efficient manner. See https://www.cog-genomics.org/plink (accessed February 11, 2022).

augmented with third-party demographic profiles, which may be useful for correcting issues of low representation from underserved communities.

The AKC supports researchers through consulting on the U.S. dog population, particularly with respect to breed-specific population data (aided by its litter registry); this has included geospatial mapping to 10-digit zip codes and geolocation coordinates. The AKC uses its large social media presence and contact management capabilities to drive interest, awareness, and participation in research, directing dog owners toward clinical surveys or trials. The AKC recently transitioned its genotyping method from STRs (short tandem repeats) to SNPs (single nucleotide polymorphisms), increasing its analytical capabilities, and expects to double its DNA collections to 100,000/year. The AKC is also seeking to enhance its data collection on health outcomes and cause of death and is interested in collaborations to capture phenotypes and combine them with genotype and pedigree data. The AKC has a strong interest in the genetics of behavior and is working with James Serpell’s group at the University of Pennsylvania to make the Canine Behavioral Assessment & Research Questionnaire (C-BARQ)⁷⁹ canine behavioral evaluation tool available to all AKC members. Dunn reached out to the researchers for input in building infrastructure for sharing the AKC’s sophisticated data collection, noting that purebred AKC dogs have the advantage of known pedigrees.

Dunn and participants from the research community considered how best to identify the most important diseases in each breed for investigation of possible environmental contributions. Over the last 5 years, the AKC has worked with its national breed clubs (called “parent clubs”) to clarify each breed’s most significant health concerns. This has resulted in over 180 letters detailing breed-specific health screening recommendations, which are publicly available in the “health” section of each breed page on the AKC website, said Dunn.⁸⁰ In addition, breeders who participate in the Breeders of Merit or Bred with H.E.A.R.T. programs are required to perform all the health testing required by their club.

Discussion: Accelerating Cross-Species Comparisons

Integrating Dog and Human Studies

Participants considered the feasibility of obtaining parallel data on pets and their owners by either recruiting the pets of participants in human stud-

___________________

⁷⁹ See https://vetapps.vet.upenn.edu/cbarq (accessed January 24, 2022).

⁸⁰ See, for example, http://cdn.akc.org/Marketplace/Health-Statement/Collie.pdf (accessed January 24, 2022).

ies or vice versa. Freya Mowat (University of Wisconsin–Madison) noted the expense of contacting human research participants for participation in an add-on pet study, when half do not even own pets. She suggested, at the very least, that every human study participant be asked whether and what type of pet they own. “Do you have a pet?” is an important question for any human longitudinal health study, given the increasing understanding of the impact of the human–animal bond on people’s health, said Hughes. Promislow, who has addressed this question with leadership at the Baltimore Longitudinal Study of Aging⁸¹ and the Framingham Heart Study,⁸² commented that there is a high bar to adding even one more question to the surveys. He suggested bringing together leaders of human and pet prospective studies in a roundtable to figure out how to work collaboratively and support one another’s research. Members of the dog community who can best identify the important non-monogenic breed-specific diseases need to sit down with their human-disease counterparts to pinpoint areas for collaboration, said Mowat. On the dog side this could involve the AKC parent clubs, said Dunn.

About 17 veterinary schools are currently affiliated with clinical and translational science awards (CTSA) to medical schools, which are designed to train clinicians in translational research, and the NIH has several initiatives aimed at bringing MDs and DVMs together around shared diseases, said Trepanier. Many of the affiliated medical schools have prospective longitudinal studies in humans, and work is under way to link veterinary studies to those human studies. However, in only six of these schools are the veterinary and medical schools housed on the same campus.

Participants were asked how they envision pairing human and pet data. Boyko suggested doing paired GWAS of dogs and their owners, to see which phenotypes map to the dog genome and which map to the human genome. Castelhano is currently integrating data from human and dog biobanks, with bioinformaticians working to standardize the platforms. The ICDC is funded by the NCI’s Cancer Moonshot Initiative,⁸³ which supports data sharing and harmonization, noted LeBlanc. Once the Moonshot-funded Center for Cancer Data Harmonization⁸⁴ finishes creating its harmonized data model, ICDC data will be combined with human data in the Data Commons, amplifying the importance of uploading data to the ICDC, added Thessen. Although the ICDC was initially conceived as a repository of canine cancer genome data,

___________________

⁸¹ See https://www.blsa.nih.gov (accessed January 24, 2022).

⁸² See https://framinghamheartstudy.org (accessed January 24, 2022).

⁸³ See https://www.cancer.gov/research/key-initiatives/moonshot-cancer-initiative (accessed January 25, 2022).

⁸⁴ See https://datascience.cancer.gov/data-commons/center-cancer-data-harmonization-ccdh (accessed January 25, 2022).

it has broadened to include various types of data related to cancer development that have a rationale for comparison to humans, and this offers a huge opportunity for inclusion of environmental exposures, added LeBlanc. The ICDC has been opened for 6 months and contains data on bladder cancer, osteosarcoma, and glioma, including genomics, transcriptomics, and imaging, said LeBlanc, who encouraged researchers to use it.

Citizen Science: Opportunities and Concerns

This is a “golden age” for citizen science, said Boyko, highlighting Darwin’s Ark⁸⁵ and Loyal for Dogs⁸⁶ in addition to Embark, the Golden Retriever Lifetime Study, and the Dog Aging Project. There is currently no comprehensive list of these projects, but the American Veterinary Medical Association (AVMA) has a clinical trial database akin to clinicaltrials.gov, said LeBlanc, and it would be straightforward to register studies on that website. Hughes suggested that the AKC publish a list.

Citizen science raises multiple concerns regarding data health, noted Dunn, with owners frequently misreporting phenotypes and breed status. Ostrander and LeBlanc noted the importance of verifying breed status genetically. Health outcome data need to be collected in a way that ensures it is all correct, or that enables drilling down to the useful data, said Dunn.

Dunn has also seen big differences in reported household demographics between subjective and third-party (objective) reports. There is also selection bias; the people most likely to work with researchers are the ones already in the database. Dunn also noted ethical concerns surrounding informed consent, privacy, and commercialization of data (see section on Ethical Considerations of Using Companion Animals as Sentinels).

Sharing Commercially Held and Biobanked Data

Shelton asked about open access for academic and other researchers to the huge databases of Embark and Mars Petcare. Boyko described three ways in which Embark makes its data available: Raw data for published papers are submitted to public archives, researchers can contact Embark directly to obtain consent for the data they need, or researchers can obtain data directly from pet owners. This last method worked “spectacularly well,” said Ostrander, when Embark contacted owners of pets with a particular disease and invited them to get in touch with Ostrander. Mars Petcare’s Science Engine integrates data

___________________

⁸⁵ See https://darwinsark.org (accessed January 24, 2022).

⁸⁶ See https://loyalfordogs.com/about (accessed January 24, 2022).

from across the business, said Hughes, and researchers are invited to partner with its Data and Analytics team. In addition to various institutional partners who work with its data, Mars Petcare collaborates on clinical trials with VCA Animal Hospitals,⁸⁷ BluePearl Pet Hospitals,⁸⁸ and others, added Hughes.

The AKC works closely with the Orthopedic Foundation for Animals (OFA),⁸⁹ whose Canine Health Information Center (CHIC) program⁹⁰ database is an actively collecting biobank that includes scored X-rays, health testing, and health outcomes for thousands of dogs, said Dunn. Many breeders who do Embark testing submit these data to the CHIC, added Boyko, which could enable cross-referencing of hip X-rays with genotype data. Castelhano noted that the CHIC is among a network of biobanks around the country; researchers can contact any biobank in the network and will be referred to whichever one contains the specimens they need.

Purebred versus Crossbred Dogs as Models for Environmentally Induced Cancers

Participants considered the relative advantages and disadvantages of using purebred versus mixed-breed dogs to study the influence of environmental agents on the development of cancer. Dye raised the question of whether, in a breed at high risk for a particular cancer, the hardwired genetic risk might potentially overwhelm the ability to detect the influence of an environmental exposure—or on the other hand, in cases where gene–environment interaction is critical, would that breed demonstrate particular sensitivity to the environmental agent? That depends on the genetic architecture underlying the risk, said Boyko. As noted by Page and Trepanier, the importance of the environment to cancer incidence in susceptible purebred dogs has been demonstrated: For example, pesticide exposure increases the risk of bladder cancer in Scottish Terriers, a high-risk breed, whereas ingestion of vegetables reduces the risk of bladder cancer in both people and dogs and improves survival time in patients with the disease (Glickman et al., 2004; Raghavan et al., 2005; Tang et al., 2010). Page also added the GRLS finding that 50 percent of cancer deaths were due to hemangiosarcoma, which is uncommon in dogs outside the United States and thus suggests the presence of an environmental influence. However, if the cancer susceptibility is fixed within a breed, as is suspected for the Scottish Terrier, it would not be possible to elucidate the underlying genetics by studying that breed alone; dogs outcrossed from the sensitive breed would be used to

___________________

⁸⁷ See https://vcahospitals.com (accessed January 24, 2022).

⁸⁸ See https://bluepearlvet.com (accessed January 24, 2022).

⁸⁹ See https://www.ofa.org (accessed January 24, 2022).

⁹⁰ See https://www.ofa.org/about/chic-program (accessed January 24, 2022).

identify the region(s) of the genome responsible for the predisposition to risk, said Boyko. Chromosomal analysis of outbred dogs narrows down the region of interest and allows for a much finer GWAS, added Ostrander.

High-cancer breeds may have certain additional factors that sensitize them to an environmental exposure, said Boyko, citing a study from the Ostrander lab demonstrating the importance of genetic background in determining whether a cancer-predisposing mutation in the Kit ligand gene was expressed in poodles. Although the Kit ligand mutation was found in both black and white poodles, only poodles carrying a particular allele in another gene, which was associated with black coat color, expressed it (Karyadi et al., 2013). Trepanier added that GWAS studies in dogs have failed to identify many cancer drivers, suggesting that most of these tumors require a gene–environment interaction. Boyko, Trepanier, Dye, and Ostrander concurred that purebred and mixed-breed dogs would reveal different things depending on the genetic architecture of risk, and it was important to study both.

Incorporating Environmental Measures into the Database

Noting that chronic noncommunicable diseases in humans are associated with environmental factors acting on multiple genes, with each gene contributing a relatively small amount to the health effect, Birnbaum wondered how environmental measures would be incorporated into these databases. The environment, said Boyko, can be thought of as just another phenotype, albeit a highly complicated one. This involves collecting data over a time series—not just for exposures but for phenotypes like obesity, which is not a single data point, and which can be combined with environmental inputs like nutrition and exercise over time. Time-series data may be available from various sources—for example, the genetics committee of the American College of Veterinary Ophthalmologists (ACVO) reviews annual screens for potentially heritable disease in dog eyes, noted Mowat.

Jones encouraged researchers to leverage existing, geographically linked exposure data as an additional resource to supplement the collection of environmental and biospecimens, noting that “those data exist and have been mined and cleaned and resourced for lots of different purposes” and could function as a hazard identification tool, particularly for investigating new exposures. LeBlanc has considered geospatially mapping the residences of dogs enrolled in osteosarcoma drug trials. The veterinary schools where these dogs are treated are geographically dispersed across the United States and Canada, and LeBlanc has observed differences in outcome by location. This could reflect differences in care or environment, she noted, cautioning that with only 350 dogs in the cohort, these finding were not statistically significant.

Birnbaum suggested that researchers explore the NIH All of Us research

program,⁹¹ which aims to recruit a million volunteers around the United States and is now in its second phase and collecting basic environmental data, possibly including a question about pet ownership. This will be a huge study with follow-up and opportunities to access the data, she added. A participant noted that the 2000 U.S. Census included a question about whether there was a dog in the household, but this was eliminated from the 2010 and 2020 questionnaires. Restoring that question could help address the denominator problem, which is the lack of knowledge regarding how many animals reside in a given location, said Jones. The pet industry and the AKC have both lobbied to restore the question, added Dunn. The AVMA could join that conversation, suggested LeBlanc.

EQUITY, ETHICS, AND POLICY

Participants discussed issues of ethics and equity raised by the use of companion animals in research. Speaker presentations considered the opportunities to proactively ensure that these research methods do not perpetuate structural inequities or discriminate against disadvantaged populations; the appropriate role of pet owners in these studies; and the collaboration opportunities (e.g., multi-sectoral collaboration mechanisms, community engagement strategies, educational opportunities) needed to effectively implement the core capacities and interventions of One Health principles within communities.

Ethical Considerations of Using Companion Animals as Sentinels: Research Subject Protections, Citizen Science Issues, and Shared Health

Embedding ethical considerations into research strengthens study design and validity, and is the right thing to do, said Lisa Moses (Harvard Medical School), who is one of two ethicists consulting on the Dog Aging Project. “Everyday ethics,” said Moses, requires examining projects to identify issues like cognitive bias or hidden clashes of values, “those common issues that trigger vague feelings of discomfort or red flags . . . should we pay enrollees, or why is the project having trouble enrolling diverse participants?” Applying basic principles of ethical analysis when planning projects, such as listing everyone who may be affected by a project’s activity or outcome, can help clarify these issues, she said. When planning partnerships with prospective commercial partners or private funding sources, each prospective partner’s objectives should be compared to prevent future conflicts “before the challenges emerge.”

___________________

⁹¹ See https://allofus.nih.gov (accessed January 25, 2022).

Though significant human–animal relationships have existed for millennia, the meaning of these relationships has changed as humans have urbanized and become increasingly likely to live alone and to delay marriage and childbearing, said Moses. Pets now fill holes in the support network for millions of people. By and large, ethicists consider companion animals as having a special, protected moral status compared to other animals (Yeates and Savulescu, 2017). In Moses’s view, the primary argument supporting this special status is relational, where the complexity of human–pet relations resembles those between humans, and thus lends them moral weight (Ashall and Hobson-West, 2017).

Though this workshop mostly covered observational studies, which carry less potential for harm than interventional studies, Moses highlighted three basic ethical issues that remain relevant:

1. Informed consent does not exist for animals. Whereas human subjects give informed consent to protect their own autonomy, consent by a pet owner reflects their ability to preserve the value (economic, emotional, etc.) the animal has for them and does not require prioritizing the best interest of the animal (Ashall et al., 2018; Gray et al., 2018).
2. Pets are not protected by laboratory animal research regulations. Although dogs and cats are protected species under the Animal Welfare Act, most of what is assessed by an Institutional Animal Care and Use Committee (IACUC) does not apply to animals housed outside of labs (Walker and Fisher, 2018). Moses urged researchers to seek additional review of their study plans by a Veterinary Clinical Trial Ethical Review Committee (Bertout et al., 2021).
3. Payment or other compensation for enrollment can be coercive when pets are the subject, due to inequities in access to veterinary care in the United States. If the compensation is veterinary care, that might be the only care the pet receives, in which case it incurs a moral obligation on the investigators. Another concern is therapeutic misconception, where owners believe mistakenly that the research will directly benefit their pet.

While citizen science has tremendous potential to democratize science, it carries its own ethical concerns, added Moses. It is essential to incorporate the principles of reciprocity, where participants benefit from their involvement, and reflexivity, where investigators examine the influence of their own beliefs and practices on the research. Other useful strategies include diversifying participants, which improves data quality, and building in learning objectives to enhance participants’ scientific literacy. There are well-developed tools for

engaging in these partnerships, including a NASEM consensus study report, said Moses (NASEM, 2018; Soleri et al., 2016).

The Potential for Coercion in Interventional Studies

The perception of coercion in interventional studies was of particular concern to participants. Nicole Erhart (Colorado State University) noted the concept that compensation could be coercive when it covers standard-of-care treatment. To offset the perception among pet owners that this is the only way they can afford to care for their pets, researchers could ensure that participants have another way to obtain care, which may involve steering them toward other resources, said Moses. Another strategy is for a social worker to privately discuss access to pet care with the owner during the original consent process. Trepanier raised the scenario of a cancer clinical trial that enrolls patients through a free veterinary clinic, with compensation in the form of specialized care, such as free tumor biopsy, genetic analysis, and therapy. This presents an ethical issue, said Moses, if the only way people can obtain that level of care is by participating in the study. “The potential for therapeutic misconception and misidentification is huge when people are desperate,” she added. When they cannot obtain the same level of care elsewhere, it is essential that both owners and researchers are aware of this and that owners are clearly informed of their obligation to their pet and their ability to withdraw at any time. For complex situations like these, Moses suggested writing consent documents that address all the essential ethical concerns on the first page, with more details in other parts of the document. A clinic offers the opportunity for face-to-face communication, which goes a long way toward finding the participant’s source of discomfort or conflict. “When it becomes the unspoken thing, that is when things get dangerous, and people do not feel comfortable speaking up when they are not sure . . . what the implications of their decisions are,” she added.

Moses also expressed concern with the question of euthanasia, particularly in oncology studies, where owners want to contribute to the science on the one hand but also want to decide when to end their pet’s suffering on the other, and recommended involving social workers in that discussion.

Veterinary cancer patients resemble pediatric cancer patients whose parents give consent for their participation in clinical trials, suggested Page. In both cases, problems arise from a lack of full medical coverage and the potential coercion embedded in free care. The main difference between veterinarians and pediatricians, said Moses, is that veterinary patients are a different species with different needs than humans, and veterinarians understand this in a way that owners often do not. Parents can usually be assumed to know their child best and advocate for his or her best interest. In contrast, veterinarians are called on to advocate for and protect their patients in a different way, said Moses.

One Health Approaches in Arctic Indigenous Communities

Persistent Organic Pollutants: Contamination of Arctic Wildlife and People

The rise in synthetic chemical pollution is outpacing all other measured agents of global change, including habitat loss and the increase in atmospheric levels of CO₂, but synthetic chemicals remain the least studied of all these agents (Bernhardt et al., 2017). The heat of equatorial climates causes light- to intermediate-weight pollutants to evaporate and disperse in the earth’s atmosphere, where they condense and are deposited in colder regions. Over time, through cycles of volatilization and deposition, persistent organic pollutants (POPs) work their way to the highest latitudes, and this has made the Arctic the most polluted place on earth for these compounds, said Frank A. von Hippel (University of Arizona). POPs bioaccumulate in animals and magnify in food webs, with long-lived predators such as polar bears and whales registering the highest concentrations of POPs on earth. These chemicals are further concentrated in the Arctic indigenous people who subsist on these animals, said von Hippel, adding that mining, oil extraction, and about 600 former defense sites contribute to the pollution in Alaska (von Hippel et al., 2016).⁹²

Cancer is the main cause of death among Alaskan natives, with significantly higher rates of lung cancer and colorectal cancer in Alaskan native men and women compared to the general U.S. population.⁹³ Additional health patterns among indigenous Alaskans that may relate to POP exposure include thyroid disease, diabetes, heart disease, birth defects, and other reproductive health problems.

Sled Dogs as Sentinels for Environmental Exposures

Dogs are a major part of Alaskan culture, where they are relied on for transportation, subsistence hunting, safety on the ice, and protection from predators, in addition to sport, said von Hippel, noting that dogsled races form the centerpiece of cultural events and that the Iditarod race commemorating the 1925 diphtheria antitoxin run to Nome celebrates “one of the first One Health projects.” Many Alaskans also rely on their dogs for emotional health and sobriety. Arctic dogs are more like workers than pets, with paral-

___________________

⁹² See https://www.niehs.nih.gov/research/supported/translational/peph/assets/docs/the_use_of_sentinel_species_in_health_disparities_research_508.pdf (accessed January 27, 2022).

⁹³ See http://anthctoday.org/epicenter/publications/Cancer_50year_Report/antr_fifty_year_report_web.pdf (accessed January 27, 2022).

lels to military dogs or rescue dogs, which may raise different ethical issues, noted Birnbaum.

Alaska presents unique issues for studying companion dogs as sentinels of environmental exposure, said von Hippel. Most dogs live outdoors, so they are not a suitable model for studying household exposures. Arctic communities are very remote, and transportation is mostly by small plane, and there are no facilities or housing for researchers. These regions have high levels of poverty and virtually no veterinary care, which amplifies ethical concerns. The Arctic has a broad host spectrum for rabies, and rabid dogs pose a major public health risk. Dogs can also be a source of conflict when they get caught in trap lines. On the other hand, Alaskan dogs and people eat the same food, making dogs a good sentinel for POPs exposure, which primarily occurs through the diet; sled dogs have the same profile of contaminants as their owners and suffer similar health effects, said von Hippel. The research enterprise itself can bring improvements to the community in terms of increased access to veterinary care and jobs, improved infrastructure, improvements to human health through treatment of zoonotic diseases, and conflict reduction through community engagement. Although many communities are legally dry, drug and alcohol addiction is a major problem throughout the Arctic, and community engagement in research can help address this problem by giving people agency and improving resilience, he added. The remoteness of these communities limits confounding variables, and there is excellent scientific literature on the use of sled dogs as sentinels throughout the Arctic.

Community-Engaged Research to Study Exposures in Alaskan Natives and Dogs

Von Hippel’s research always originates with a request for help from the community, generally out of concern for contamination of their villages and subsistence foods. For example, a health aide on St. Lawrence Island, concerned about whether the high rates of cancer in her community might be caused in part by exposure to thousands of barrels of unknown contaminants that had been left behind by the U.S. military, contacted a nongovernmental organization (NGO) that recruited scientists to work on the problem. Another tribe asked for von Hippel’s help dealing with pollution from World War II defense sites in the Aleutian Islands. The community originates the question, and a working group forms between the scientists and community members to strategize a way to get an answer. Community-engaged research is essential to avoid perpetuating inequities or discriminating against disadvantaged populations, and ideally takes the form of community-based participatory research (CBPR), said von Hippel (Wallerstein and Duran, 2010). When working with indigenous popula-

tions, it is essential to obtain consent from the community and tribe in addition to the dog’s owner, and the community should benefit from the work, he said. The tribe owns von Hippel’s data and he reports results to them before publishing. More researchers are needed in the Arctic and there are opportunities for collaboration, said von Hippel, though he cautioned that this work proceeds on a long timeline, which requires that scientists educate funding agencies and build a trusting relationship with the community before writing the grant proposal.

There are important parallels between climate change and synthetic chemical pollution in terms of their effect on the Arctic, added von Hippel. Like POPs, the effects of climate change are most pronounced in the Arctic, which is warming at twice the rate of the rest of the planet. Both POPs contamination and climate change are cases of environmental injustice, exerting their greatest effects on populations that did not create or use these compounds. Declines in polar bears and other Arctic wildlife are due more to POPs exposure than to climate change, and the effects of the two exacerbate one another: Climate change increases the geographic spread of disease while POPs compromise immune function and amplify disease susceptibility. Noting that the attention received by synthetic chemical pollution is disproportionately low relative to the increasing urgency of the problem, von Hippel cited a recent paper advocating for an intergovernmental panel to tackle global chemicals and waste (Wang et al., 2021).

Aligning Health Care for a Bonded Family Society

Two-thirds of U.S. households have pets—considerably more than have children—and as many as 95 percent of these consider their pets to be family members.⁹⁴ Many families with strong human–animal bonds, or “bonded families,” have limited assets; these include individuals employed in the service of others or retired on limited incomes, with a disproportionate representation of minorities and women, said Michael Blackwell (University of Tennessee, Knoxville), citing a study by the Access to Veterinary Care Coalition (AVCC).⁹⁵ Those with the least often demonstrate the strongest human–animal bonds, and this includes people who are unsheltered—in some cases because available housing options exclude their pet. Twenty-eight percent of pet owners in the AVCC study reported a barrier to veterinary care in the pre-

___________________

⁹⁴ See https://www.prnewswire.com/news-releases/more-than-ever-pets-are-members-ofthe-family-300114501.html (accessed May 23, 2022).

⁹⁵ See https://pphe.utk.edu/wp-content/uploads/2020/09/avcc-report.pdf (accessed January 27, 2022).

vious 2 years, said Blackwell. These barriers to care were not just financial but included practical hurdles like transportation. Pets exist throughout society because they benefit humans, both physiologically and psychologically, and this results in reduced morbidity and mortality (Matchock, 2015); therefore, rather than trying to remove pets from owners who cannot afford veterinary care, Blackwell wants to change the system to help these families.

Societal support for pet care is also becoming a public health imperative, argued Blackwell, as climate change broadens the range of disease vectors and leads to an increase in zoonotic diseases. The incidences of both ehrlichiosis and Rocky Mountain spotted fever have risen dramatically in recent years.⁹⁶ Ignoring the health care needs of bonded families would adversely impact the health of the nation overall, said Blackwell. But addressing this requires revisiting public policies, such as those that concern housing, and factoring in the needs of bonded families. This requires a paradigm shift to One Health, which recognizes that efforts to improve health outcomes for humans must take the health of animals and the environment into account, said Blackwell.

There was also discussion about the particular needs and potential pitfalls of conducting research in which bonded families participate. “In my experience, pets very often will reflect the reality of the family,” said Blackwell. For example, low-income families are more likely to feed their pet the same food they eat rather than buying pet food. This can lead to similar health conditions in people and their pets and it could be beneficial to study conditions like diabetes or obesity in the companion animals as well as the people. Treating the bonded family as a unit would enable researchers to engage the companion animal in studies of family dynamics around mental and emotional health in addition to infectious disease.

Discussion: Equity, Ethics, and Policy

Different Uses of Companion Animals Lead to Different Shared Exposures

Shelton was intrigued by the shared diet between sled dogs and humans in their community and its ability to shed light on the physiological effects of contaminants. Von Hippel agreed, citing research in Greenland, Canada, and Alaska on persistent organic pollutants and associated disease states, which compared sled dogs with Arctic foxes, polar bears, and their human owners; “the parallels are really striking,” he said.

___________________

⁹⁶ See https://www.cdc.gov/ehrlichiosis/stats/index.html and https://www.cdc.gov/rmsf/stats/index.html (both accessed January 27, 2022).

Unlike Arctic sled dogs, military dogs live and sleep with their individual human comrades-in-arms, so exposures to stressors may be even more linked in these parings than with pets, noted Paula Genik (Colorado State University). Responding to Blackwell’s comment that the emotional, psychological, and mental health of both dog owners and their pet dogs are deeply affected by their common lived experience, Genik suggested that military dogs are excellent sentinels of combatant mental health, and that perhaps Arctic dogs could be able to provide insight into the mental health of their owners. Von Hippel agreed that relationships between owners and working dogs are fundamentally different from those between owners and pets and suggested that researchers explore how to better incorporate working dogs into the One Health framework.

Research, Service, and Education: Balancing Priorities

Sue VandeWoude (Colorado State University) expressed concern about how to prioritize the need to provide underserved communities with human and/or veterinary medical care, as well as research needs. Disease distribution and trends among companion animals in these communities is poorly understood, said Blackwell, so that has to be studied in order to determine what kind of medical care to prioritize. Based on its survey of pet owners and veterinarians, the AVCC report recommended development of new service delivery models, including methods for controlling costs, to find the most economical way to provide veterinary services when income is limited.

Promislow asked if there is an ethical approach to using compensation to attract a more diverse participant pool to observational studies, particularly in communities that might not have easy access to veterinary care. In the case of a disadvantaged community, you can find more creative solutions, said Moses, such as working with local care providers to offer a free or low-cost clinic event that supports the community as a whole, while building trust in the research project. The key with citizen science is building relationships within the community and asking participants about what help they need, she added; a project that addresses systemic barriers to care in the community would significantly help with participant diversity and equity.

Families need help understanding the needs of their nonhuman family members, and with so many people turning to the internet for advice, pet owners need guidance on how to use research tools to obtain reliable information, said Blackwell. There are many ways to incorporate education into citizen science, said Moses, and it is important to help participants increase their scientific literacy. This might include training in data collection methods, which is important for study validity, as well as directed training for adult learners. For large-scale projects that involve many communities, Moses sug-

gested partnering with local schoolteachers, who are always looking for ways to engage their students with science and incorporating a trained educator into the project design team. “[It] is really about maximizing very explicit and direct educational opportunities for participants,” said Moses.

Cultural Competence and Establishing Partnerships in Underserved Communities

When working with underserved communities, communication can be a challenge, said Tracy Webb (Colorado State University), noting that not everyone has reliable access to email or participates in social media, and there may be barriers to what individuals are willing to report. Citing von Hippel’s emphasis on both top-down and bottom-up communication, with the community directing the research to address the questions most important to them, Webb asked how to actualize these concepts and establish effective two-way communication in underserved communities.

In Arctic communities, and indigenous communities in general, “there is no substitute for being there,” said von Hippel, so “the [funding] agencies have to be willing to fund travel to get to these remote places.” Von Hippel’s group makes many trips and does community presentations, which involve local leadership, and these have to be done in person because most areas do not have broadband internet. In Australia, where he works with an aboriginal community, von Hippel reaches the community through a local radio show. “We talk to the radio host about the research and what we are finding. So it’s a matter of finding the right means within each local community that reaches people.”

It is critically important to gain cultural competencies around the cultures one is studying, said Blackwell. Historical experiences, such as the Tuskegee Syphilis study,⁹⁷ have molded perceptions of research within these communities. Often the only connection they have with outsiders around their pets is through animal control, which translates as policing, so we need to be attentive to those perceptions, said Blackwell. One example that works, said Moses, is a program called Pets for Life,⁹⁸ whose workers go door to door, ask whether there are pets in the home, ask what the residents need to maintain the pet, and provide it. Partnering with an organization like this, which has established trusting relationships in the neighborhood, is the most direct way to begin a research project, suggested Moses.

___________________

⁹⁷ See https://www.cdc.gov/tuskegee/timeline.htm (accessed February 27, 2022).

⁹⁸ See https://humanepro.org/programs/pets-for-life (accessed February 11, 2022).

Research with a long timeline can be challenging, said Blackwell, but underserved and underresourced communities may require a long period of engagement to establish trust with researchers. Noting that Promislow took 3 years to establish infrastructure for the Dog Aging Project, Don Ostrow (educator, NYC) asked how to establish research models that acknowledge the necessary time to establish trust in the community, challenge basic assumptions within the research community, and remove obstacles to providing genuine two-way communication.

Von Hippel writes memoranda of understanding (MOUs) between the university and the tribe that clearly state the relationship, data sovereignty, and other issues important for community participation. “I think that protection for the tribe is critical so they feel confident that they are not being exploited and that the work is really to benefit the community, rather than just researchers coming in, doing their work, and never hearing from them again, which is the way it used to be that was so problematic,” he added.

Moses suggested involving an expert in research ethics in the initial planning phase, which can speed up the process. Blackwell suggested engaging individuals with leadership roles, which may vary according to the community. For example, within the Black community this has traditionally included ministers and teachers, and Blackwell suggested starting by building relationships with these individuals, adding that “. . . if one is serious about this, from the standpoint of respecting me as a minority and recognizing that there is some history that I embody and has been handed down, that literally defines the culture—if you can’t come in respecting my culture, you will be obvious from the beginning by what you say, what you don’t say, and how you go about doing it.”

IDENTIFYING RESEARCH GAPS AND SETTING A RESEARCH AGENDA: EXPLORING NEXT STEPS FOR THE PATH FORWARD

Planning committee members summarized facilitated discussion tables in which workshop participants considered lessons learned from the workshop and strategies for charting a path forward. This was followed by open discussion with suggestions to move the field forward from planning committee members, invited speakers, and audience members.

Obtaining Data on Pets and Exposures

Demographic information is an important denominator for some of this research and is lacking for all populations, including purebred dogs, mixed breeds, and cats. The AKC, AVMA, and industry can help obtain this infor-

mation, said William Farland (Colorado State University, emeritus); adding a question about pets to the 2030 Census would also be useful. One approach to improving the denominator, suggested Promislow, would be if a team of human demographers and geospatial scientists worked with the databases from Mars Petcare’s veterinary hospitals and clinics to develop models that could derive estimates of the denominators based on existing information such as breed, age, and size. This is important for understanding risk factors and incidence of disease—particularly for rare diseases, said Promislow, which may not be so rare in certain populations.

Danielle Carlin (NIEHS) suggested developing studies that would allow physicians and veterinarians to collect biospecimens from humans and pets simultaneously—for example, through existing mobile clinics. Promislow noted that dog owners are often more willing to do things for their dog than for themselves, which could extend to the collection of dust and other samples in the home. The fact that pet research would be included with the human research might motivate owners to engage more in this kind of work, and it underscores the value of piggybacking dog studies onto longitudinal human studies, he added, which would benefit both studies.

Interdisciplinary Training of Researchers, Physicians, and Veterinarians for One Health

Johannes discussed the need for physicians and veterinarians to be trained in public health, particularly as veterinarians become more specialized and 80 percent go into private practice, where their focus is on caring for patients. Most veterinary residents are trained at universities, and it is the universities’ role to convey the importance of participating in One Health, he suggested. Ryan suggested encouraging veterinary students to take public health courses and, in some cases, to complete dual DVM and MPH degrees. The fields of oncology, internal medicine, and nutrition would dovetail well with public health, said Wakshlag, and DVM programs could be tailored to include an extra year to accommodate this training. The problem seems to be funding, rather than interest, he added.

Greater support is also needed for the teaching of comparative medicine and research in medical schools because this trained medical workforce will be crucial for collaborating on research with practicing veterinarians, said Carlin. These companion animal studies can be used to train interdisciplinary personnel in public health, medical, and veterinary schools, as well as professionals in the social sciences who will be needed to bridge cultures and work closely with participants in community-based participatory research, added Farland.

Focusing on the animal side, Trepanier noted that veterinary students whose main interest is in primary care receive no training in research. To

remedy this, Trepanier plans to pilot a veterinary student training program with CTSA One Health Alliance, to build interest in participating in citizen science in the primary care setting, and to view this participation as important for both the students’ professional development and the growth of the field.

The primary care veterinary community plays a critical role in helping to gather data for companion animal studies, noted Promislow, but primary care veterinarians often do not have the bandwidth to participate in studies, and there is no funding for their participation. Furthermore, because many primary care veterinarians do not encounter the scientific literature in their training, they may not appreciate the value of doing this extra work for science. This presents an education challenge for researchers, who need to convince the veterinarians that it is worth their time to participate in a scientific study and to help them guide their clients through informed decision making about their own participation.

Practicing veterinarians could be engaged through continuing education (CE), particularly if it is provided in an American Association of Veterinary State Boards’ (AAVSB’s) Registry of Approved Continuing Education (RACE)⁹⁹ board-approved format that meets CE requirements for licensure, said Marguerite Pappaioanou (University of Washington). Community research education would probably be a very popular topic, added Page, who suggested building it into conferences and CE expectations for licensure. Veterinarians on the Dog Aging Project team are creating CE content as a way of “giving back” to the many practicing veterinarians helping with the rapamycin trial (TRIAD), said Promislow. Any veterinarian who participates in the DAP is offered CE, and once RACE approval is obtained, CE will be offered as distance learning on-demand, added Ruple.

Veterinarians are among the most respected sources of information to the general public, noted Farland, and it is important for researchers to collaborate and utilize veterinarians as a trusted resource. The research community could reach veterinarians through respected sources, such as the AKC and similar groups, the Toxicology Education Foundation,¹⁰⁰ and presentations at society meetings, he added. The American College of Veterinary Preventive Medicine would be a valuable partner, added Pappaioanou.

Translating the Science between Pets and People

How exposure data might be translated between humans and animals remains to be determined, said Carlin. She offered the example of physi-

___________________

⁹⁹ See https://www.aavsb.org/ce-services/race/program-overview (accessed February 2, 2022).

¹⁰⁰ See https://toxedfoundation.org (accessed Feburary 9, 2022).

ologically based pharmacokinetic (PBPK) modeling (Jones and Rowland-Yeo, 2013), which can help predict tissue-specific exposures to a drug in one population based on extrapolation from another population along with relevant physiological parameters. In this case, the pharmacokinetic (PK) for each species needs to be determined before effective comparisons can be made between species, said Carlin.

“This is about more than just genes and environment,” added Farland, noting the observed roles of metabolism, receptors, the microbiome, and other aspects of the organism that modulate its response to environmental stimuli. Models could incorporate these other types of information, perhaps employing the toxicological concept of adverse outcome pathways, said Farland.

Data Integration

What does it mean to create compatible, harmonized data across species for so many different data types? There are many definitions of data and many components, noted Page. This may include questionnaire data and EMR data, but also sample data, which could take numerous forms, such as histopathology, which would require digitization. In addition, he asked, “Who is going to pay for that?”

Carlin emphasized support for a publicly available database and biobanks. Promislow envisioned the creation of a DNA Data Commons for dogs, where all the studies collecting DNA data would share information. For example, if a participant consents to having their dog enter an oncology clinic for treatment or a research study, the principal investigator (PI) or clinician would be able to access that dog’s genetic data and analyze it for high-risk markers. Sharing DNA data brings challenges regarding consent, privacy, and consistency, with different platforms providing different information, but there may be easy ways to harmonize it, said Promislow. Genik cited a life sciences data archive maintained by NASA, which holds samples and ’-omics data collected from astronauts and biological organisms both flown in space and on the ground and suggested that a similar structure might be useful to the NIH and the NIEHS. “I think what that comment underscores,” said Promislow, “is that there are a lot of people who are not at this meeting who don’t think about the questions that we’re thinking about but who have solved some of the problems that we’re thinking about. And one of our action items should be to think broadly and deeply about who else is out there solving different problems but with the kinds of solutions that we also might need to implement.”

“We should not have to reinvent the wheel,” said Page, suggesting instead that researchers expand the NCI’s ICDC to contain exposure data and other types of data that would be generated from studies on a broader scale across species, with the vision of a common exposure or common genomic analysis

arrangement. This will be expensive, added Page. There needs to be active work to promote data accuracy, collation, integration, curation, and accessibility; and that will require targeted funding, said Breen. The Dog Aging Project and Golden Retriever Lifetime Study cannot fund all the analytics needed to study the data they produce; rather, “they are an amazing resource for other people to adopt and fund,” said Breen.

Data Collection: Long-Term Investment

There is an opportunity to build data infrastructure from scratch, but existing systems, such as human cancer registries or reportable illnesses in humans and pets, can also be leveraged for this effort, said Deziel. COVID19 provides a case study of how a health data resource can be built quickly, given the will and the funding, he added. An EMR system for pets would help significantly, but the importance of this—and how it would benefit veterinarians and their patients—needs to be communicated to veterinarians, who are already overworked, said Deziel. Nonetheless, Promislow noted, there are also challenges getting data out of EMRs, as well as challenges getting veterinarians to share data.

Sample collection is expensive, making it a challenge to fund, particularly for larger-scale studies, said Promislow. Small clinical trials are straightforward, but for human studies and larger dog studies, with thousands or tens of thousands of specimens, the cost can be prohibitive. The research community needs to develop funding strategies, said Carlin, noting tremendous variability in access to resources.

Data Collection: Tapping into Existing Cohorts (Short Term)

There is a need to better understand exposure and the environment that companion animals live in, and this requires augmenting the information that veterinarians typically collect, said Farland, who reviewed methods covered at the meeting, including GIS, proximity to sources, and both active and passive sampling. He suggested leveraging existing environmental data and exposure information that has been collected by the NIEHS, the EPA (through its research and development program), and state and local health departments. Plenty of specimens have already been carefully collected and stored, noted Farland; the question is how to make the best use of these and surmount institutional barriers to making them available for study.

Deziel noted many opportunities for researchers to tap into existing cohorts, both human and animal. One example is the NCI-CONNECT cohort. A helpful resource would be a repository of existing cohorts that are amenable to additions. There are also underutilized data sets. For example, Poi-

son Control¹⁰¹ collects both human and animal data, as well as some location information. The research community needs to figure out how to collaborate with and access the data from Mars Petcare and other large data sources, added Carlin. This may include data from diverse sources, such as the Multi-Ethnic Study of Atherosclerosis (MESA).¹⁰² NHANES collected data on water and dust samples and has access to Social Security and Medicaid data, as well as a significant amount of longitudinal data, said Carlin. Available GIS data could also be leveraged for studying companion animals, Carlin added.

Trepanier and Birnbaum discussed the possibility that the NIEHS and the NCI might provide supplemental funding to add pet biomonitoring and outcome data to existing longitudinal studies in humans. Studies initiated in disadvantaged communities might fit into other NIH institutes, including the National Institute on Minority Health and Health Disparities, added Birnbaum.

Continuing and Expanding the Conversation

Page suggested a follow-up meeting to consider what “data” really means and how to define it, and to create recommendations for building more robust data systems. Data convergence, pulling the dog and human data together, is another important conversation, said Shelton. Breen suggested establishment of focus groups “to develop the right questions to ask of these data.”

“Where are sentinel animals poised to have the most impact?” asked Carlin, suggesting studies on physical health, exercise, and obesity, as well as mental, emotional, and social health and well-being. She also suggested setting up working groups to identify focus areas based on existing collaborative networks.

Shelton suggested that a communication committee could be tasked with getting information out to relevant stakeholders. Hughes also stressed the importance of having a central body to manage communications from the research community, with a consistent message. Breen said, “This is going to take a massive coordinated effort,” which will require a partnership between scientists, veterinarians, companion animal owners, and others, as well as education of a new generation of leaders in research.

Community Research—Engagement and Equity

As researchers continue to grow the infrastructure for community science, they need to focus on making the science more accessible and understandable

___________________

¹⁰¹ See https://www.poison.org (accessed February 11, 2022).

¹⁰² See https://www.mesa-nhlbi.org (accessed February 11, 2022).

to the public, said Megan Dillon (North Carolina State University). This will enhance the value of the science and also generate support for funding, she added.

Science literacy and public awareness about these projects need to be increased, said Farland. The research community could work more with nonprofit organizations to reach out to communities, which could include working with animal shelters and rescue groups, and organizations that support people who are homeless, said Carlin, adding that researchers could engage marketing and community specialists to return results directly to the pet owners.

As this research community continues to engage in citizen science, it needs to put into practice community-based participatory research and find opportunities for two-way communication, said Farland. Promislow suggested changing the wording from “citizen science” to “community science,” noting that some target participants who are not U.S. citizens may be put off by the former, whereas “community” embraces everyone. Promislow also cautioned that an equity issue arises in research where participants fund some of the cost, as only people with money can participate in these studies.

Noting that current leaders in companion animal research have been working with stakeholders for many years and have developed a strong level of trust, Breen suggested “a kind of legacy planning” to train the next generation of both researchers and community members.

CONCLUDING REMARKS AND POTENTIAL NEXT STEPS

The history of using animals as sentinels extends long before the canary in the coal mine to ancient times, said Birnbaum, when animals were used to determine whether a particular food was safe. Animals can be effective sentinels because of greater susceptibility, greater exposure, and shorter latency for adverse effects of environmental factors. Furthermore, domesticated dogs have been exposed to some of the same environmental selection pressures as people, and dog orthologous genes more closely resemble those of humans than do rodent genes.

This is a golden age of cancer research, said Birnbaum, with increasing awareness of the pressing relevance of environmental exposures. However, the latency of effects in humans makes it difficult to identify environmental carcinogens and gerontogens.¹⁰³ Cross-sectional studies are valuable for generating hypotheses, but longitudinal prospective studies are needed to establish strong associations and causality, and these studies would benefit from pairing

___________________

¹⁰³ Gerontogens are environmental agents that can accelerate aging.

humans with companion animals. Birnbaum suggested looking in both directions: adding pets to human studies and adding pet owners to longitudinal dog studies. She also noted the great potential of including sensors for personal monitoring.

These large studies will generate Big Data, and it is only going to get bigger and more complicated. Big cohort studies with tissue banks and data banks have real long-term benefits, said Birnbaum, who urged all players to work together: “We need to have everyone at the table,” including public–private partnerships, and to use the entire arsenal of cross-disciplinary approaches. One Health should include public health, but it has been a challenge to involve the public health and medical communities. “It’s bigger than we can address here, but we should be aware of it,” she said. Untargeted analysis, suspect screening, and targeted analysis are all needed to identify environmental agents, she added. Practices and procedures in both data handling and biobanking need to be harmonized, and standardization of veterinary EMRs would help this effort.

To refine these studies, “we need to define what we mean by environment,” said Birnbaum, noting that in this workshop, “environment” seemed to include all nongenetic factors while incorporating windows of susceptibility, effects of early life exposure, transgenerational effects, mixtures, interactions, and cumulative exposures. The genome, epigenome, microbiome, and metabolome act as both targets and mediators of environmental exposures. Considering the vastness of the exposome, Birnbaum wondered how necessary it was to analyze each exposure separately. “I think our companion animals are living in the same soup that we are living in, and rather than trying to break down what that soup is, maybe we need to start looking at the totality of the soup and how that affects the health of the dogs and cats and people,” she suggested.

Our companion animals can inform us not only about cancer and aging but many other human health conditions, said Birnbaum, including those involving the immune system and the microbiome. Although several presentations emphasized the similarities of cancers in dogs and people, Birnbaum cautioned that neutered dogs may not be the best models for all diseases because the changed endocrine milieu may have a dramatic effect on the outcomes of some cancers.

Researchers also need to make more concerted efforts to reach out to disadvantaged communities. “We can’t just build it and they will come,” said Birnbaum, noting that dogs as well as people experience health disparities, and that greater diversification of participants will produce better-quality data. “The animal–human bond really defines our human society,” added Birnbaum, but “dogs are not always household buddies,” and working dogs may offer the opportunity to ask different questions. Although there is overlap

between citizen science and CBPR, these are not the same thing. Academic or corporate researchers cannot helicopter into communities, get what they want, and leave. Developing a trusting relationship with communities can take time, which can be a problem for funders, but it is the reality of dealing with different communities, said Birnbaum. It is essential to communicate results back to the participants, even if the researchers do not know what the results mean. There also needs to be careful consideration of informed consent, privacy, and commercialization of data for both veterinary and human studies. “We have to be sure the work is not exploitative.” Planning is key, and researchers should aim to involve an ethicist and an educator in the initial planning process, she added.

Reflecting on the presentations as a whole, Birnbaum emphasized that everyone comes to this from their own perspective. Clinicians, veterinarians, veterinary researchers, human health researchers, and exposure scientists all approached the work from a somewhat different angle. She reiterated a theme that had surfaced several times in the context of studying the health impacts of environmental exposures: “What is the question you are asking?”

Breen said, “The challenge is what to analyze, how to analyze, and how to merge the data. . . . It’s the data management that’s going to cause us a bit of a roadblock. Gathering the data is one thing, harnessing and using it is another.” He observed that exposure research is very much bidirectional and he hoped that it will accelerate as a consequence of the workshop. “We do absolutely share exposures with our pets, and . . . the keys to unlocking some of the puzzles about environmental exposures have been walking alongside us for many years.”

References

London, L. G. Barber, C. E. Balkman, A. L. McCleary-Wheeler, S. E. Suter, O. Martin, A. Borgatti, K. Burgess, M. O. Childress, J. L. Fidel, S. D. Allstadt, D. L. Gustafson, L. E. Selmic, C. Khanna, and T. M. Fan. 2021. Adjuvant sirolimus does not improve outcome in pet dogs receiving standard-of-care therapy for appendicular osteosarcoma: A prospective, randomized trial of 324 dogs. Clinical Cancer Research: An Official Journal of the American Association for Cancer Research 27(11):3005–3016.

Le Moal, J., S. Goria, A. Guillet, A. Rigou, and J. Chesneau. 2021. Time and spatial trends of operated cryptorchidism in France and environmental hypotheses: A nationwide study from 2002 to 2014. Human Reproduction 36(5):1383–1394.

Lermen, D., F. Gwinner, M. Bartel-Steinbach, S. C. Mueller, J. K. Habermann, M.-B. Balwir, E. Smits, A. Virgolino, U. Fiddicke, M. Berglund, A. Åkesson, A. Bergstrom, K. Leander, M. Horvat, J. Snoj Tratnik, M. Posada de la Paz, A. Castaño Calvo, M. Esteban López, H. von Briesen, H. Zimmermann, and M. Kolossa-Gehring. 2020. Towards harmonized biobanking for biomonitoring: A comparison of human biomonitoring-related and clinical biorepositories. Biopreservation and Biobanking 18(2):122–135.

Levine, H., N. Jørgensen, A. Martino-Andrade, J. Mendiola, D. Weksler-Derri, I. Mindlis, R. Pinotti, and S. H. Swan. 2017. Temporal trends in sperm count: A systematic review and meta-regression analysis. Human Reproduction Update 23(6):646–659.

Lindblad-Toh, K., C. M. Wade, T. S. Mikkelsen, E. K. Karlsson, D. B. Jaffe, M. Kamal, M. Clamp, J. L. Chang, E. J. Kulbokas, M. C. Zody, E. Mauceli, X. Xie, M. Breen, R. K. Wayne, E. A. Ostrander, C. P. Ponting, F. Galibert, D. R. Smith, P. J. DeJong, E. Kirkness, P. Alvarez, T. Biagi, W. Brockman, J. Butler, C.-W. Chin, A. Cook, J. Cuff, M. J. Daly, D. DeCaprio, S. Gnerre, M. Grabherr, M. Kellis, M. Kleber, C. Bardeleben, L. Goodstadt, A. Heger, C. Hitte, L. Kim, K.-P. Koepfli, H. G. Parker, J. P. Pollinger, S. M. J. Searle, N. B. Sutter, R. Thomas, C. Webber, J. Baldwin, A. Abebe, A. Abouelleil, L. Aftuck, M. Ait-Zahra, T. Aldredge, N. Allen, P. An, S. Anderson, C. Antoine, H. Arachchi, A. Aslam, L. Ayotte, P. Bachantsang, A. Barry, T. Bayul, M. Benamara, A. Berlin, D. Bessette, B. Blitshteyn, T. Bloom, J. Blye, L. Boguslavskiy, C. Bonnet, B. Boukhgalter, A. Brown, P. Cahill, N. Calixte, J. Camarata, Y. Cheshatsang, J. Chu, M. Citroen, A. Collymore, P. Cooke, T. Dawoe, R. Daza, K. Decktor, S. DeGray, N. Dhargay, K. Dooley, K. Dooley, P. Dorje, K. Dorjee, L. Dorris, N. Duffey, A. Dupes, O. Egbiremolen, R. Elong, J. Falk, A. Farina, S. Faro, D. Ferguson, P. Ferreira, S. Fisher, M. FitzGerald, K. Foley, C. Foley, A. Franke, D. Friedrich, D. Gage, M. Garber, G. Gearin, G. Giannoukos, T. Goode, A. Goyette, J. Graham, E. Grandbois, K. Gyaltsen, N. Hafez, D. Hagopian, B. Hagos, J. Hall, C. Healy, R. Hegarty, T. Honan, A. Horn, N. Houde, L. Hughes, L. Hunnicutt, M. Husby, B. Jester, C. Jones, A. Kamat, B. Kanga, C. Kells, D. Khazanovich, A. C. Kieu, P. Kisner, M. Kumar, K. Lance, T. Landers, M. Lara, W. Lee, J.-P. Leger, N. Lennon, L. Leuper, S. LeVine, J. Liu, X. Liu, Y. Lokyitsang, T. Lokyitsang, A. Lui, J. Macdonald, J. Major, R. Marabella, K. Maru, C. Matthews, S. McDonough, T. Mehta, J. Meldrim, A. Melnikov, L. Meneus, A. Mihalev, T. Mihova, K. Miller, R. Mittelman, V. Mlenga, L. Mulrain, G. Munson, A. Navidi, J. Naylor, T. Nguyen, N. Nguyen, C. Nguyen, T. Nguyen, R. Nicol, N. Norbu, C. Norbu, N. Novod, T. Nyima, P. Olandt, B. O’Neill, K. O’Neill, S. Osman, L. Oyono, C. Patti, D. Perrin, P. Phunkhang, F. Pierre, M. Priest, A. Rachupka, S. Raghuraman, R. Rameau, V. Ray, C. Raymond, F. Rege, C. Rise, J. Rogers, P. Rogov, J. Sahalie, S. Settipalli, T. Sharpe, T. Shea, M. Sheehan, N. Sherpa, J. Shi, D. Shih,

J. Sloan, C. Smith, T. Sparrow, J. Stalker, N. Stange-Thomann, S. Stavropoulos, C. Stone, S. Stone, S. Sykes, P. Tchuinga, P. Tenzing, S. Tesfaye, D. Thoulutsang, Y. Thoulutsang, K. Topham, I. Topping, T. Tsamla, H. Vassiliev, V. Venkataraman, A. Vo, T. Wangchuk, T. Wangdi, M. Weiand, J. Wilkinson, A. Wilson, S. Yadav, S. Yang, X. Yang, G. Young, Q. Yu, J. Zainoun, L. Zembek, A. Zimmer, and E. S. Lander. 2005. Genome sequence, comparative analysis and haplotype structure of the domestic dog. Nature 438(7069):803–819.

Liu, K. H., C. M. Lee, G. Singer, P. Bais, F. Castellanos, M. H. Woodworth, T. R. Ziegler, C. S. Kraft, G. W. Miller, S. Li, Y. M. Go, E. T. Morgan, and D. P. Jones. 2021. Large scale enzyme based xenobiotic identification for exposomics. Nature Communications 12(1):5418.

Ly, L. H., E. Gordon, and A. Protopopova. 2021. Exploring the relationship between human social deprivation and animal surrender to shelters in British Columbia, Canada. Frontiers in Veterinary Science 8:656597.

Ma, J., H. Zhu, and K. Kannan. 2020. Fecal excretion of perfluoroalkyl and polyfluoroalkyl substances in pets from New York State, United States. Environmental Science & Technology Letters 7(3):135–142.

Macías-Montes, A., M. Zumbado, O. P. Luzardo, Á. Rodríguez-Hernández, A. Acosta-Dacal, C. Rial-Berriel, L. D. Boada, and L. A. Henríquez-Hernández. 2021. Nutritional evaluation and risk assessment of the exposure to essential and toxic elements in dogs and cats through the consumption of pelleted dry food: How important is the quality of the feed? Toxics 9(6):133.

Martin, K. M., M. A. Rossing, L. M. Ryland, R. F. DiGiacomo, and W. A. Freitag. 2000. Evaluation of dietary and environmental risk factors for hyperthyroidism in cats. Journal of the American Veterinary Medical Association 217(6):853–856.

Martincorena, I., K. M. Raine, M. Gerstung, K. J. Dawson, K. Haase, P. Van Loo, H. Davies, M. R. Stratton, and P. J. Campbell. 2017. Universal patterns of selection in cancer and somatic tissues. Cell 171(5):1029–1041.e21. Erratum in: Cell 173(7):1823.

Matchock, R. L. 2015. Pet ownership and physical health. Current Opinion in Psychiatry 28(5):386–392.

Mattingly, C. J., T. E. McKone, M. A. Callahan, J. A. Blake, and E. A. C. Hubal. 2012. Providing the missing link: The exposure science ontology ExO. Environmental Science & Technology 46(6):3046–3053.

Mattson, M. P., V. D. Longo, and M. Harvie. 2017. Impact of intermittent fasting on health and disease processes. Ageing Research Reviews 39:46–58.

McHale, C. M., G. Osborne, R. Morello-Frosch, A. G. Salmon, M. S. Sandy, G. Solomon, L. Zhang, M. T. Smith, and L. Zeise. 2018. Assessing health risks from multiple environmental stressors: Moving from G×E to I×E. Mutation Research. Reviews in Mutation Research 775:11–20.

Messier, K. P., D. C. Wheeler, A. R. Flory, R. R. Jones, D. Patel, B. T. Nolan, and M. H. Ward. 2019. Modeling groundwater nitrate exposure in private wells of North Carolina for the Agricultural Health Study. The Science of the Total Environment 655:512–519.

Middleton, R. P., S. Lacroix, M.-P. Scott-Boyer, N. Dordevic, A. D. Kennedy, A. R. Slusky, J. Carayol, C. Petzinger-Germain, A. Beloshapka, and J. Kaput. 2017. Metabolic differences between dogs of different body sizes. Journal of Nutrition and Metabolism 2017:4535710.

J. O. B. Jacobsen, J. A. McMurry, T. Groza, S. Köhler, D. Smedley, P. N. Robinson, C. J. Mungall, M. A. Haendel, M. C. Munoz-Torres, and D. Osumi-Sutherland. 2020. The Monarch Initiative in 2019: An integrative data and analytic platform connecting phenotypes to genotypes across species. Nucleic Acids Research 48(D1):D704–D715.

Simpson, S., M. D. Dunning, S. de Brot, L. Grau-Roma, N. P. Mongan, and C. S. Rutland. 2017. Comparative review of human and canine osteosarcoma: Morphology, epidemiology, prognosis, treatment and genetics. Acta Veterinaria Scandinavica 59(1):71.

Skinner, C. G., J. D. Thomas, and J. D. Osterloh. 2010. Melamine toxicity. Journal of Medical Toxicology: Official Journal of the American College of Medical Toxicology 6(1):50–55.

Smedley, D., M. Schubach, J. O. B. Jacobsen, S. Köhler, T. Zemojtel, M. Spielmann, M. Jäger, H. Hochheiser, N. L. Washington, J. A. McMurry, M. A. Haendel, C. J. Mungall, S. E. Lewis, T. Groza, G. Valentini, and P. N. Robinson. 2016. A whole-genome analysis framework for effective identification of pathogenic regulatory variants in Mendelian disease. American Journal of Human Genetics 99(3):595–606.

Smith, M. T., K. Z. Guyton, C. F. Gibbons, J. M. Fritz, C. J. Portier, I. Rusyn, D. M. DeMarini, J. C. Caldwell, R. J. Kavlock, P. F. Lambert, S. S. Hecht, J. R. Bucher, B. W. Stewart, R. A. Baan, V. J. Cogliano, and K. Straif. 2016. Key characteristics of carcinogens as a basis for organizing data on mechanisms of carcinogenesis. Environmental Health Perspectives 124(6):713–721.

Smith, N., K. R. Luethcke, K. Craun, and L. Trepanier. 2021. Risk of bladder cancer and lymphoma in dogs is associated with pollution indices by county of residence. Veterinary and Comparative Oncology 20(1):246–255.

Soleri, D., J. Long, M. D. Ramirez-Andreotta, R. Eitemiller, and R. Pandya. 2016. Finding pathways to more equitable and meaningful public-scientist partnerships. Citizen Science: Theory and Practice 1(1):9.

Spinelli, J. B., H. Yoon, A. E. Ringel, S. Jeanfavre, C. B. Clish, and M. C. Haigis. 2017. Metabolic recycling of ammonia via glutamate dehydrogenase supports breast cancer biomass. Science 358(6365):941–946.

Stern, Y., C. Habeck, J. Steffener, D. Barulli, Y. Gazes, Q. Razlighi, D. Shaked, and T. Salthouse. 2014. The Reference Ability Neural Network Study: Motivation, design, and initial feasibility analyses. NeuroImage 103:139–151.

Stoffel, E. M., and C. C. Murphy. 2020. Epidemiology and mechanisms of the increasing incidence of colon and rectal cancers in young adults. Gastroenterology 158(2):341–353.

Strynar, M. J., and A. B. Lindstrom. 2008. Perfluorinated compounds in house dust from Ohio and North Carolina, USA. Environmental Science & Technology 42(10):3751–3756.

Sumner, R. N., M. Tomlinson, J. Craigon, G. C. W. England, and R. G. Lea. 2019. Independent and combined effects of diethylhexyl phthalate and polychlorinated biphenyl 153 on sperm quality in the human and dog. Scientific Reports 9(1):3409.

Sumner, R. N., I. T. Harris, M. Van der Mescht, A. Byers, G. C. W. England, and R. G. Lea. 2020. The dog as a sentinel species for environmental effects on human fertility. Reproduction 159(6):R265–R276.

Sung, J., S. Kim, J. J. T. Cabatbat, S. Jang, Y.-S. Jin, G. Y. Jung, N. Chia, and P.-J. Kim. 2017. Global metabolic interaction network of the human gut microbiota for context-specific community-scale analysis. Nature Communications 8:15393.