By Carol Beuchat PhD

Heritability has a very specific meaning in genetics. The heritability of a trait is the fraction of the total variation in the trait among the animals in a population that can be accounted for by genetics. The variation that is not a consequence of genetic differences among individuals is lumped into the "non-genetic" category, which is called "environment". Values of heritability go from 0 to 1. If a trait is genetic but there is no variability in the population, the heritability is zero. If all of the variation from one individual to the next can be explained by differences in genetics, the heritability is 1; that is, 100% of the variation in the trait is accounted for by genetics. A trait that has no genetic influence, like hair color in a group of teens with neon spray-on paint, is not heritable. 

The heritability for most behavioral traits is rather low. This doesn't mean that genetics isn't important for behavior, but that our ability to infer the genetics of an animal from observations of behavior is limited because environmental factors produce so much variability that the fraction of total variation that can be attributed to genes is small.

The problem with the low heritability of behavioral traits in dogs is that behavior is usually evaluated very carefully when making breeding decisions - and the evaluator usually feels that they are making a useful assessment - yet all the studies and mounds of data generated over the last 50 years indicate that these assessments are very poor predictors of the genetic breeding value of the dog, as you will see.

This is well known among organizations that breed dogs for specific purposes, such as the military and guide dog organizations. A dog must meet certain standards of performance and behavior to be acceptable for work, and it is expensive and time-consuming to breed, raise, train, and evaluate a dog. Being able to identify at a young age the dogs that are most likely to make the grade is of huge importance, so the challenge becomes one of figuring out what can be assessed or measured in a puppy or young dog that will reflect its future behavior and performance. Just as important is the selection of animals to be bred, in which case you are hoping to choose the animals that are best suited genetically and therefore likely to pass their genetic value on to their offspring.

These organized breeding programs have a huge advantage over the average breeder in that they produce large numbers of puppies that are all raised under the same conditions, receive similar treatment, and are exposed to similar environments. Also, because many of these organizations have been systematically breeding dogs for many decades, they have accumulated lots of data and experience about evaluating and predicting behavior.

Let's look at a few representative studies that evaluate how well behavior assessments predict performance, and whether puppy tests are a useful way to predict behavior as an adult. The military generally keeps their data to themselves, but there is lots of information that has been produced by the guide dog organizations. 

ASSESSING PUPPIES

• Wilsson E & P-E Sundgren. 1998. Behaviour test for eight-week old puppies - heritabilities of tested behavior traits and its correspondence to later behaviour. App. Anim. Behav. Sci. 58: 151-162.

• Riemer S, C Muller, Z Viranyi, L Huber, & F Range. 2014. The predictive value of early behavioural assessments in pet dogs- a longitudinal study from neonates to adults. PLOS One 9(7): e101237.

How useful are those tests that evaluate temperament in puppies? With a quick Google search, you can come up with a bunch of examples, and many people swear by these. But it is really difficult to prove that they are useful.

The goal, of course, is to determine some traits in the puppy that will predict behavior and performance as an adult, but the tricky thing is that the tests you would use on a puppy aren't appropriate for an adult and vice versa. So right from the start, you know you are evaluating things that are "proxy traits" - that is, you expect them to tell you something useful about some entirely different behavior when the animal is older.

The two papers noted above are worth reading. The first one convincingly showed that evaluations of puppies were of little use in predicting behavior as an adult. They evaluated 630 German Shepherds at 8 weeks and again at 450-600 days (roughly 1.5-2 yrs). They found that heritability of some of the behavioral traits they looked at was modest, so selection for those would result in some improvement. But the "correspondence of puppy test results to performance at adult age was negligible and the puppy test was therefore not found useful in predicting adult suitability for service dog work".

The second study is recent so the information is up to date, and it does a good job of recognizing the potential pitfalls of studying puppies to predict adult behavior. But again, they "found little correspondence between individuals' behaviour in the neonate, puppy and adult test. Exploratory activity was the only behaviour that was significantly correlated between the puppy and the adult test. We conclude that the predictive validity of early tests for predicting specific behavioural traits in adult pet dogs in limited". 

An especially interesting outcome in this study is in the evaluation of fearfulness, which is the most common behavioral reason for a puppy to be rejected from a guide dog training program. They found that in evaluating fearfulness, "...some major changes were observed over time, with the initially most fearful individuals becoming most friendly to people or vice versa". That statement should give you pause. Could eliminating a puppy from a breeding program because of fearfulness be culling the dogs that will have the best temperaments as adults? 

People that have been breeding dogs for decades and have much experience working with puppies might feel that they have some skill at evaluating temperament and behavior, but the studies that have tried to confirm the value of these evaluations have uniformly failed to validate them. Mind you, these studies used hundreds or even thousands of puppies from dozens of litters; the testing conditions were uniform, the protocols carefully followed, and the data analyzed with appropriate statistics. In the absence of properly done studies that demonstrate otherwise, there is no evidence that puppy evaluations have any value at all.

So what about the breeders of guide dogs? Do they do temperament testing on young puppies? Analysis of over 10 years of data from one guide dog center showed that selecting breeding stock on the basis of puppy tests did improve the performance of puppies on the tests, but did not improve the success rate of dogs finishing the program (Scott & Bielfelt 1976; Goddard & Beilharz 1986). Consequently, they wait to evaluate the suitability of dogs for further training until they are older.

EVALUATING DOGS BEFORE TRAINING

• Duffy DL & JA Serpell 2012 Predictive validity of a method for evaluating temperament in young guide and service dogs. App. Anim. Behav. Sci. 138: 99-109.

Since assessments of very young puppies are not very predictive of adult behavior, at what age should they be assessed? This study tested a group of dogs at 6 and 12 months of age, before they entered their training period to become a guide dog. They noted which dogs were successful in graduating from the program and looked for associations with any of the measured traits. The test they used (which was developed by one of the authors, Serpell) was the C-BARQ, which has been widely used for behavioral assessment of dogs.  Note that they combined data from 5 different guide and service dog breeding programs, so they had a huge sample size - nearly 8,000 dogs. These were not anecdotal observations!

They found that dogs that successfully completed the guide dog training program "scored more favorably on 27 out of 36 C-BARQ traits at both 6 and 12 months of age compared to those that were released from the programs". Although they didn't find a trait that was very useful for predicting success, they were able to do a better job of rejecting dogs that were likely to be unsuccessful. In particular, owner-directed aggression was a very sensitive indicator of failing the program, and this at least is useful; eliminating dogs that are not likely to succeed spares the time and expense of training a dog that will ultimately wash out of the program. The other (rather unlikely) trait most predictive of failure in the guide dog program was "pulls excessively hard on a leash".

This study concluded that the C-BARQ test was useful in identifying the dogs that were least likely to succeed, but it was not a reliable indication of which dogs were most likely to be successful.


ASSESSING DIFFERENCES BETWEEN BREEDS AFTER TRAINING

• Ennik I, A-E Liinamo, E Leighton, & J van Arendonk. 2006. Suitability for field service in 4 breeds of guide dogs. J Vet. Behav. 1:67-74.

Historically, Golden Retrieivers, Labradors, and German Shepherds have been most commonly used as guide dogs, and more recently crosses such as Golden x Labrador have been added. If you're familiar with these breeds, you know that they can have very different temperaments, and in fact they were bred for very different purposes - the retrievers for hunting and the German Shepherd for herding and working. Certainly these breed-specific differences are genetic and are likely to be reflected in the success rates of the breeds in completion of training and becoming a guide dog.

This study looked at the success rates of these three breeds and the Golden x Labrador cross in becoming guide dogs, and it also looked for effects of sex and whether a dog received additional training (was "passed back") because there was no suitable match with a person when it finished training, there was a temporary medical issue, or it wasn't progressing adequately.

They found that there were significant differences among the breeds in the probability of becoming a successful guide dog. The Golden x Labrador cross dogs had the highest level of success in becoming guide dogs (59%), while the German Shepherd dogs were lowest (46%). The Goldens and Labradors were about the same (54% and 51%, respectively).

Not surprisingly, dogs that were passed back because of behavioral issues were the least likely to subsequently graduate (black bars), and the two most common reasons for being passed back were that the dog needed more work or lacked confidence (see the table below).

Overall, Labrador Retrievers and the Golden x Lab crosses were most likely to graduate after the normal training period, but success rates of Goldens and German Shepherds were higher if they were passed back and thus had a longer training period. Dogs that were passed back for behavioral reasons were least likely to be successful. There wasn't a consistent difference in success and failure rates by sex. The authors of this study noted that the crossbred dogs were more successful than any of the purebreds.


REASONS FOR FAILURE

The most common behavioral reasons for failing out of a guide dog training program are fearfulness, dog distraction, and excitability. Although the heritabilities of dog distraction and excitability are low and can differ between males and females, the heritability of fear is relatively high (0.46 over both males and females; Goddard & Beilharz 1982).

The relatively high heritability of fearfulness indicates that selection against it should be moderately successful, but in this not all breeds are equal. In a study that included four breeds as well as the crosses of these (Labrador retriever, Australin kelpie, boxer, and German shepherd dog), fearfulness was highest in German shepherds and lowest in Labradors, which also showed the greatest response to selection (Goddard & Beilharz 1986).

From these studies (and many more like them), we can conclude that evaluating behavior in dogs and using that information for selective breeding has proven to be extremely difficult, even in well-managed breeding programs in which the dogs are bred and raised under controlled conditions, receive similar treatment and experiences, and are assessed by trained handlers.

If we believe in genetics (which of course we do!), we would expect that dogs from selective breeding programs should have low rates of behavior problems, but still 70% of the dogs that fail to complete the program are removed because of behavior. Perhaps a better way to judge whether these specialized breeding programs are successful is to note that in the early days when dogs for guide dog training programs were donated by the public, the success rate was 9%; selective breeding improved the success rate to 90% in just a few generations (Pfaffenberger 1963). It is the remarkable early success of the guide dog breeding programs that provided the convincing evidence that broad differences in behavior among breeds are heritable, and that improvement within breeds is possible.

WHAT ABOUT TEMPERAMENT TESTING PUPPIES?

There are two very important messages to take away from this.

1) Be very skeptical of claims that temperament can be assessed through evaluation of young puppies. If there is somebody who can reliably do this, they should document their success in a publication, because it would revolutionize the breeding of dogs. Otherwise, the best way to predict the temperament and behavior of a dog as an adult is to assess it as an adult.

2) Nobody wants to produce puppies that will grow up to have behavior problems, but in breeds where low genetic diversity and small population sizes are resulting in high rates of genetic disorders and low fertility, excluding dogs from breeding based on evaluation of behavior, especially as puppies, should be done very carefully. Aggression and fearfulness have reasonable heritabilities and can be selected against if evaluated as adults, but for most other traits heritability is very low and therefore evaluations of behavior don't tell you much about a dog's genetic value for these traits.

One final point. The efficiency of selection can be improved by estimating breeding values statistically using information about not just the dog of interest but about parents, siblings, and progeny (i.e, estimated breeding values, EBV). The guide dog organizations have been using EBVs to select for the traits of interest to them (both behavioral and physical) for a couple of decades with good success (Leighton 2003), and the technique should be similarly useful to dog breeders if they are willing to collect the appropriate data to set up the program.

Just for fun:

If you have a dog, you can do your own C-BARQ assessment from their website at the University of Pennsylvania vet school:
http://vetapps.vet.upenn.edu/cbarq/

Duffy DL & JA Serpell 2012 Predictive validity of a method for evaluating temperament in young guide and service dogs. App. Anim. Behav. Sci. 138: 99-109.

Ennik I, A-E Liinamo, E Leighton, & J van Arendonk. 2006. Suitability for field service in 4 breeds of guide dogs. J Vet. Behav. 1:67-74.

Goddard ME and RG Beilharz. 1982. Genetic and environmental factors affecting the suitability of dogs as guide dogs for the blind. Theor. Appl. Genet. 62: 97-102.

Goddard ME and RG Beilharz. 1986. Early prediction of adult behaviour in potential guide dogs. App. Anim. Behav. Sci. 15: 247-260.

Leighton, EA. 2003. How to use estimated breeding values to genetically improve dog guides. Presented at a meeting of the "Original Group",  September 11-12, 2003.

Pfaffenberger C. 1963. The new knowledge of dog behavior.

Riemer S, C Muller, Z Viranyi, L Huber, & F Range. 2014. The predictive value of early behavioural assessments in pet dogs- a longitudinal study from neonates to adults. PLOS One 9(7): e101237.

Scott JP & SW Bielfelt. 1976. Analysis of the puppy testing program. Pp. 39-76, in CJ Pfaffenberger, JP Scott, JL Fuller, BE Ginsburg, and SW Bielfelt (Eds). Guide dogs for the blind: their selection, development and training.


Wilsson E & P-E Sundgren. 1998. Behaviour test for eight-week old puppies- heritabilities of tested behaviour traits and its correspondence to later behaviour. Appl Anim. Behav. Sci. 58: 151-162.

You can learn more about the genetics of dogs in ICB's online courses.
******************************

Visit our Facebook Groups

ICB Institute of Canine Biology
...the latest canine news and research

ICB Breeding for the Future
...the science of animal breeding 

By Carol Beuchat PhD

Immune system disorders seem to be fairly common in dogs- things like allergies, ear infections, vaccine reactions, food sensitivity, slow recovery from infection, etc. In addition, there are autoimmune diseases, in which the immune system mistakenly attacks the tissues of the body, and some of these can be quite serious. Many breeders, faced with a breeding decision involving a dog with an immune disorder, wonder if the gene for it is dominant or recessive. Most likely, the disorder isn't caused by a single gene, or at least not one that we can identify. In fact, the immune system is extremely - extremely - complex. Like a large, complicated piece of machinery, breaking one seemingly tiny thing can result in an entire system breaking down. Of course, fixing something like this is a challenge, and in the case of the genes of the immune system, the best plan is to make sure it doesn't get broken in the first place.

For example, a paper just out found that one of the DLA haplotypes that is part of the immune system of dogs is associated with symmetrical onychomadesis, an awful disorder in which the toe nails fall off over the course of a few months. At the same time, however, that particular DLA haplotype is protective for hypothyroidism. This one DLA haplotype is implicated in promoting one autoimmune disorder while at the same time it appears to prevent another. The dilemma for the breeder is obvious.

There is nothing like a little summary of how the immune system works to convince somebody that they should work hard not to break it, so to that end I have rounded up a few videos that explain things in the most basic (i.e., cartoon) terms. The first is extremely basic and actually cute. The next one is also basic, you will notice that they only address part of it, and about halfway through you will realize that what is being described is the equivalent of the US armed forces, with foot soldiers, air defenses, the submarine fleet, the intelligence people, all sorts of weaponry, and the central command, with complex defensive and offensive strategies all carefully orchestrated with the goal of defeating the enemy. Knock out the air support and there could be attacks on the ground. Lose a critical weapon system and the enemy can gain strength. It's hard to win, and really easy to lose. Like the military, the immune system wages some extremely complex battles. I think you'll be really impressed.

See, I said you'd be impressed, right? Remember, there are no DNA tests that will help you avoid these problems. We don't know which DLA haplotypes a particular dog should have and which might be causing problems. Perhaps one DLA haplotype doesn't function properly if some gene elsewhere isn't also present. Maybe there is a signaling system that is necessary for control. We really don't know.

To select against one gene, we remove a dog from the breeding population, and all of the good genes it might have go with it. You don't notice losing the immune system genes; they don't produce a coat color, or prick ears, or a better topline that you can see and evaluate. You will only preserve the genes of the immune system if you are preserving genetic diversity in general, by using a breed-wide strategy of genetic management. We can beat back bacterial infections with antibiotics if the immune system isn't up to the task, but the autoimmune disorders are quite another thing. Steroids will often help, but the side effects and collateral damage to the body are not benign. And it's not clear how you can "breed away" from something that's broken unless there are dogs with healthy immune systems around to breed to, and how can you know who they are?

Most people don't worry about the health of the immune system when they consider breeding options, or if they do it's down the list a ways from color, size, and other traits. But if we are not deliberately selecting for a healthy immune system, we are likely to break something that is going to be very, very hard to fix.

*** Visit our FACEBOOK pages ***

ICB Institute of Canine Biology
...the latest canine news and research

ICB Breeding for the Future
...the science of animal breeding 

By Carol Beuchat PhD

There's a paper just out about two autoimmune disorders, hypothyroidism and symmetrical onychomadesis, in Gordon and English Setters. Hypothyroidism is inadequate levels of thyroid hormone in the body, which can often be managed with appropriate treatment. Symmetrical onychomadesis (SO) results in the loss of the nails over the course of a few months. In some cases the nails grow back but in others euthanasia is the result. Obviously, both disorders are better avoided.

The study is easy enough to read and I've put a link to the pdf download below so you can read it yourself.

But it presents an opportunity to address the widely-believed notion that genetic disorders in dogs can be eliminated simply by getting rid of the offending genes. The argument is that inbreeding can be used as a tool to expose recessive mutations, and affected and carrier dogs can simply be eliminated from the population. It is true that breeding related dogs will unmask hidden recessives. Unfortunately, it is not true that breeding healthy dogs is a simple matter of eliminating all of the bad genes from the gene pool.

Just from an understanding of basic genetics, there are some obvious reasons why this won't work. First of all, you can't select for or against genes a la carte. Each dog comes pre-packaged with a collection of genes, and if you select against some specific gene it carries, you are also selecting against all of the genes it carries. And to complicate things, there is something called genetic "linkage", in which groups of genes tend to get inherited together as a block. If you select for a particular gene, the linked genes get selected for too. If some of the genes in a block are "good" genes, and some are "bad" genes, there is going to be some serious frustration. And if linkage wasn't itself a big enough problem, inbreeding tends to increase the size of the blocks of linked genes, something called "linkage disequilibrium". So while inbreeding might expose those recessive mutations, the resulting linkage disequilibrium is likely to frustrate your attempts to purge the mutations without collateral damage to the gene pool.

There is usually no dissuading the advocates of mutation-purging as the strategy for improving health in dogs. So, this study is the perfect illustration of the problem. They looked at the genes in the immune system, called "dog leukocyte antigens" (DLA), and found that a particular DLA haplotype was associated with both autoimmune diseases, hypothyroidism and SO. But most inconveniently, this DLA haplotype, which was associated with SO in Gordon Setters, was protective for hypothyroidism; a dog could be hypothyroid, or suffer from SO, but usually not both. Clearly, this DLA haplotype is associated with both of these disorders (one positively and one negatively), but there must be other genes involved that are yet to be identified. They might be closely linked genes, or genes that interact in a particular way, or who-knows-what? But the suggestion from this study is that selecting against the DLA haplotype that is associated with SO could increase the risk of hypothyroidism, trading one autoimmune disorder for another.

They conclude with advice to breeders:

"It is important to state that selection of breeding animals supported by DLA haplotypes/alleles could only be used if taking into account also potential associations to other autoimmune disorders in a dog breed as well as how it influence (sic) the genetic variation in the breed".

Of course, we know nearly nothing about associations of DLA haplotypes with specific autoimmune disorders, so fixing autoimmune disorders through selection of DLAs is likely to be a lot harder than putting Humpty Dumpty back together again.

So what are breeders to do? Good question. But for Gordon Setter breeders, trying to get rid of the offending gene is clearly not the way out of this one. If your breed is not already struggling with autoimmune disorders, keep it that way. Protect the genes in your gene pool that are playing well together, and manage the problem genes instead of trying to eliminate them. The immune system is extremely complex and we are far from understanding everything about how it works. If we break it, making it right again isn't going to be easy because we really don't know how to fix it.


Ziener ML, S Dahlgren, SI Thoresen, and F Lingaas. 2015. Genetics and epidemiology of hypothyroidism and symmetrical onychomadesis in the Gordon setter and English setter. Canine genetics and epidemiology 2:12. DOI 10.1186/s40575-015-0025-6. Download pdf.

By Carol Beuchat PhD

A paper just out about horses has the dog people all atwitter. Provocatively titled "Potential role of maternal lineage in the thoroughbred breeding strategy", it examines the relative genetic influence of the sire and dam in racing performance as documented in terms of lifetime earnings (1).

To determine the genetic influence of the sire and dam on the performance of a horse, they determined the heritability of performance (documented as lifetime winnings). The term "heritability" has a very specific meaning in genetics. Heritability is the fraction of the total variation observed in a trait that can be accounted for by genetics.

In this case, they have performance data for the sires, dams, and offspring in a population of racehorses, and they did statistical analyses that determined how much of the variation could be accounted for by the influence of the dam, how much depended on the sire, and how much was unaccounted for by genetics.

They found that 14% of the variation in performance among horses could be attributed to the genetic influence of the dam, but only 3.5% to the sire. Notice here the wording "variation in performance". This is NOT the same as saying that 14% of speed comes from the dam, which would be incorrect. It specifically references the variation in performance among the group of horses in the study, and attributes 14% of it to the genetics of the dam, while the 86% of variation that remains must be attributed to some other factors, which could include the paternal influence (which they found to be 3.5%) and all possible non-genetic factors that are collectively termed "environmental". The environmental factors could include any number of things - nutrition, exercise, stress level, environmental toxins, and even aspects of the maternal environment such as the dam's nutrition, and the interactions of foal and dam after birth.

So the quality of the dam does have an influence on the quality of the offspring, although note that since heritability can range from 0 and 100%, this influence (14%) is rather low.

How could the dam be contributing to the performance of her offspring in a way that the sire doesn't? After all, the sire and dam each contribute exactly half of the alleles on the autosomal chromosomes, which include all the chromosomes except the X and Y sex chromosomes. (For each gene on the autosomal chromosomes, there are two alleles - one that comes from the mother and one from the father.)

What the dam provides that the sire doesn't are the mitochondria, which are inherited solely from the mother (the mitochondria in the sperm are usually destroyed after fertilization). Mitochondria are organelles found in every cell of the body, and their job is to produce a molecule called ATP, which is the chemical "fuel" of the body. Without mitochondria you're dead. Literally.

The authors of this study speculate that variation in the maternally-inherited mitochondria could be contributing to the performance of offspring. Mitochondrial function has been linked to athletic performance (2), so presumably, inheritance of mitochondria that are more efficient at generating the ATP needed to fuel the body during a race provides the link between the performance of the dam and her offspring. The hypothesis that dams with more powerful mitochondria will pass the trait to their offspring would need to be tested by comparing the ATP production of mitochondria from elite versus poor performers.

So the importance of the quality of the dam to the performance of a Thoroughbred racehorse lies (potentially) in the energy-producing ability of the maternally-inherited mitochondria. But for the average domestic horse that doesn't race for a living, having exceptional mitochondria would not be especially useful.

Great mitochondria would not be a big benefit for the average dog either, except in extremely high-energy activities like racing or lure coursing. There is really nothing in this study that supports the notion that the "quality" of the dam is of greater importance than the sire in determining the quality of a litter of puppies. And as far as I'm aware, there are no data elsewhere that even address this in dogs. General statements in dog publications that the dam has some special importance to the quality of puppies are not supported by evidence. For example, one published years ago that is making the rounds again since publication of the horse paper, the author states that "the female or X chromosome is responsible for most of the highly desirable characteristics for which breeders strive" (3). The X chromosome carries only 5-10% of the genes, and these are responsible for a wide variety of things including production of specific enzymes, transcription factors, proteins for vision and muscle function, development of teeth, development of endocrine tissues and the nervous system, cell biochemistry, and many other things (4), all of which are essential and fundamental cellular functions but not "highly desirable characteristics" that would be under selection by dog breeders.

Good dams are probably more likely to produce good puppies than poor ones, and likewise for good sires, but the genetic contributions of the sire and dam to a puppy are essentially equal. So the "quality" of a puppy will be determined by which of the two alleles is inherited from each parent. If the dam really does has more to do with the quality of the puppies than the sire, it remains to be demonstrated by science.

1) Lin X, S Zhou, L Wen, A Davie, X Yao, W Liu, and Y Zhang. 2015. Potential role of maternal lineage in the thoroughbred breeding strategy. Reproduction, fertility and Development. http://dx.doi.org/10.1071/RD15063

2) Das J. 2006. The role of mitochondrial respiration in physiological and evolutionary adaptation. BioEssays 28: 890-901.

3) Andrews BJ. The genetic X factor.

4) National Institutes of Health. Genes on the X chromosome