When comparing dog breeds and mutts, distinctive physical traits are not highly correlated to inheritable behavior traits

A genetic study of 2,155 purebred and mixed-breed dogs combined with 18,385 owner surveys has challenged existing notions about dog breed stereotypes and personality types. The study identified 11 locations along the canine genome that were strongly associated with behavior, none of which were specific for breed, suggesting that these personality traits predate modern canine breeding by humans.  

Noah Snyder-Mackler

Canine behavioral disorders are often proposed as a natural model for human neuropsychiatric disorders. Compulsive disorders, for instance, are often observed to manifest similarly in both humans and dogs. Thanks to the power of current DNA sequencing technology and the close relationship between pet and owner, canine genome-wide association studies (GWAS) have the potential to identify unique genetic areas in the dog genome that could lead to new insights into similar genes in humans. Karlsson and colleagues show that large-scale GWAS in dogs can yield genetic loci associated with behavioral traits.

“Although friendliness is the trait we commonly associate with golden retrievers, what we found is that the defining criteria of a golden retriever — what makes a golden retriever a golden retriever — are its physical characteristics, the shape of its ears, the color and quality of its fur, its size; not whether it is friendly,” said Karlsson, associate professor of molecular medicine at UMass Chan Medical School.

“While genetics plays a role in the personality of any individual dog, the specific dog breed is not a good predictor of those traits,” explained Karlsson, who is also the director of the Vertebrate Genomics Group at the Broad Institute of the Massachusetts Institute of Technology and Harvard University. “A dog’s personality and behavior are shaped by many genes as well as their life experiences. This makes them difficult traits to select for through breeding. For the most part, pure breeds are only subtly different from other dogs. A golden retriever is only marginally more likely to be more friendly than a mixed-breed or another purebred dog, such as a dachshund.”

A dog's tale

The story of how modern-day dog breeds emerged is a relatively short one in evolutionary terms, contrasted against the history of dog domestication from prehistoric wolves. Genetic research pegs the change from wolf to dog at about 10,000 to 15,000 years ago. Humans didn’t begin intentionally breeding dogs until roughly 2,000 years ago, when they were being selected for work roles such as hunting, guarding and herding. It wasn’t until the Victorian era in the 1800s that humans began selecting dogs consistently for the physical appearance and aesthetic traits that today we commonly associate with modern breeds.

Yet modern dog breeds are often credited with characteristics and temperaments (bold, affectionate, friendly, trainable) that correlate to their ancestral function (herding, guarding or hunting). Likewise, the breed ancestry of dogs is assumed to be predictive of temperament and behavior. DNA tests are even marketed as tools for dog owners to learn about a pet’s individual personality. However, there is a lack of genetic studies linking behavioral tendencies to ancestry or other genetic, heritable factors.

By pairing genome-wide association mapping technologies with pet owner surveys obtained through Darwin’s Ark, an open-source database of owner-reported canine traits and behaviors, Karlsson and first author Kathleen Morrill, a PhD student in the Morningside Graduate School of Biomedical Sciences at UMass Chan, explored the complicated relationship between modern canine breeds and behavioral characteristics.  

Genome-wide association studies (GWAS) are an approach used in genetics research to try to associate specific areas of variation in the human genome that align with certain phenotypes. Taking whole genome sequences from hundreds of thousands of people with the same disease, for instance, researchers look for common genetic variations among these people to pinpoint broad areas in the genome that may be predictive or causative for specific diseases.

Karlsson, Morrill and their colleagues applied this same strategy to correlate areas of the canine genome with certain behavioral traits or characteristics. Pet owners who participate in Darwin’s Dogs provide scientists with saliva samples from their dogs. Researchers run whole genome sequencing on these samples to generate a robust genetic data set for investigation. Additionally, for this study, owners filled out 12 short surveys totaling 117 questions about their pet’s behaviors and physical traits. Combined, this data provides the basis for scientists to associate genetics with owner-reported behavior. 

“Given a large enough sample size, GWAS are a really powerful tool for learning about genetics,” said Morrill. “We only get that size by looking at all dogs – not just purebred dogs, but mixed-breed dogs too. We compare all these DNA sequences computationally, using complex algorithms, to identify areas of differences and commonalities that might be of interest.”

Karlsson and Morrill collected more than 2,000 canine genomes and 200,000 survey answers through Darwin’s Dogs. Because of existing stereotypes about dog behavior and breeds, Karlsson and Morrill designed the study to account for possible owner bias, in part, by establishing standard definitions for reporting and rating canine behavioral traits such as biddability (a dog response to human direction), dog-human sociability (a dog’s comfort with people, including strangers) and toy-directed motor patterns (a dog’s interest and interaction with toys), among others. Physical and aesthetic trait standards were pulled from those published by the American Kennel Club.

Genetic research pegs the change from wolf to dog at about 10,000 to 15,000 years ago. Photo courtesy of Noah Snyder-Mackler

More to behavior than just breed

Behavioral data was analyzed across owner-reported breeds and genetically detected breed ancestries. The results of these tests, which included data from 78 breeds, showed that while breed explained some minor variation in behavior, its contributions were relatively small (9%). For certain behavioral traits, such as toy-directed motor patterns, age was a better predictor of behavior: Younger dogs were more likely to score higher in this category. For specific survey items, such as “lifts leg to urinate,” a dog’s sex was the best predictor of behavior. Physical traits like coat color were more than five times more likely to be predicted by breed than behavioral traits.

Additionally, investigators failed to find behaviors that were exclusive to any one breed. Even in Labrador retrievers, which had the lowest propensity for howling, 8% of owners reported their Labradors sometimes howl. Likewise, while 90% of greyhound owners reported that their dogs never bury their toys, three owners described greyhound dogs as frequent buriers.

Complementing the survey analysis of breeds to measure breed-behavioral propensities, the researchers also leveraged the genetic ancestry of highly mixed-breed dogs to test whether behavior is heritable in a breed-dependent manner. In some cases, heritable behavioral traits like biddability are somewhat more likely to correlate with breed, even if mixed a few generations back. In the case of purebred dogs, ancestry can make behavioral predictions somewhat more accurate. For less heritable, less breed-differentiated traits, like agonistics threshold, which measures how easily a dog is provoked by frightening, uncomfortable or annoying stimuli, breed is almost useless as a predictor of behavior.

A comparison of dog genomes was performed to identify genetic variations tracking along breed, as well as along individual physical and behavioral traits. Karlsson and Morrill identified 11 locations on the dog genome strongly associated with behavioral differences, none of which were specific for breed, and another 136 suggestively associated with behavior. The genetic differences between breeds such as golden retrievers, Chihuahuas, Labrador retrievers, German shepherd dogs and others, primarily affected genes that control coat color, fur length and other physical traits — far more than breed differences affected behavioral genes.

Overall, Karlsson and Morrill found that behavioral characteristics were influenced by multiple factors, including environment and individual genetics, but that modern breed classification played a modest role in the outcome.

“The majority of behaviors that we think of as characteristics of specific modern dog breeds have most likely come about from thousands of years of evolution from wolf to wild canine to domesticated dog, and finally to modern breeds,” said Karlsson. “These heritable traits predate our concept of modern dog breeds by thousands of years. Each breed inherited the genetic variation carried by those ancient dogs, but not always at exactly the same frequencies. Today, those differences show up as differences in personality and behavior seen in some, but not all, dogs from a breed.”

Next up: Understanding 'dog years' 

For ASU's Snyder-Mackler, the study represents a significant opening salvo demonstrating the power of using modern genomic tools to study companion animals to gain new insights into the causes and consequences of variation in the social environment, from the molecular to the organismal levels.

"This study would not have been possible without the Herculean efforts to generate a massive behavioral dataset and combine that with whole genome sequencing of almost 2,000 dogs – without spending an arm and a leg. This was achieved by using some really cool computational approaches to try and gain as much information from each dog with as little sequencing as possible, which we found worked better than the most common approach to genotyping — microarrays," said Snyder-Mackler. His lab focuses on studies that allow him to probe questions central to human health, aging and evolutionary biology, often using companion animals or field animal research.

Next up is another huge citizen science canine effort now underway called the Dog Aging Project, where, according to Snyder-Mackler, "We are trying to understand what makes dogs tick (and age). Our lab uses molecular tools to try and identify how age and the environment interact to alter the dog immune system. So this new study has provided a really powerful roadmap to carrying out future studies in larger cohorts, which Dr. Karlsson is leading as part of the Dog Aging Project." 

The project promises to explore what exactly are "dog years," other than the conventional wisdom of one dog year equal to about seven human years. What is known is that big dogs typically age more rapidly, an estimated 10 times faster than humans. By contrast, little dogs have a longer life span and can often live up to 20 years old. 

Given that dogs share the human environment and have a sophisticated health care system but are much shorter-lived than people, they offer a unique opportunity to identify the genetic, environmental and lifestyle factors associated with healthy life span.

Snyder-Mackler and collaborators outlined the goals of their efforts in a recent Nature paper. The Dog Aging Project is an interdisciplinary, open-data, community science project that consists of a team of more than 100 staff, students, faculty and veterinarians from more than 20 academic institutions, along with over 30,000 canine participants and their owners. 

The study investigators also include Kate E. Creevy, Joshua M. Akey, Matt Kaeberlein and Daniel E. L. Promislow, and afford great opportunities for Snyder-Mackler lab ASU graduate students, such as Brianah McCoy and Layla Brassington, who are helping to lead efforts to better understand the role of the environment, diet, drugs and the epigenome on dog aging. 

"I wanted to work at the interface of the public and academic science, and the Dog Aging Project is a great way to do that. So I hopped on the project," said Mccoy, who is a nontraditional student and among the first ASU Online students to pursue a PhD, having spent some time at the National Institute of Aging and Harvard Medical School prior to joining the Snyder-Mackler lab. Mccoy is particularly interested in diet and longevity, and is studying a subset of participating dogs that will be selected to be part of a new clinical study to explore the potential of the drug rapamycin to improve health.

Brassington is a master's degree student now in the molecular and cellular biology program at ASU who hopes to graduate in winter 2022.

"I read about the Dog Aging Project, and I was super excited to find a project that looked at health-related changes due to the environment," she said. She will be examining the environmental air quality and pollution levels of ozone, carbon monoxide and others nationwide. 

Both are now working hard with the citizen science effort, generating experiments, and hope to publish their first data from the Dog Aging Project later this year. 

The Dog Aging Project is actively looking for more participants. If you are a dog owner wanting to get involved, you can learn more at dogagingproject.org.

Written by Jim  Fessenden, UMass Chan Medical School Communications, with contributions from Joe Caspermeyer, ASU.

Top photo: Kristoff, one of Sparky's Service Dogs, keeps watch at the feet of his handler Taylor Randle on Hayden Lawn on Sept. 8, 2016, as they hang out with other puppies being trained by students. Photo by Charlie Leight/ASU News