by Carol Beuchat PhD

In the 1920’s, understanding of genetics was progressing from mysterious forces to the notion of  “particles” of inheritance (we call them genes now) - actual, physical things that are passed from parent to offspring that somehow carry instructions sufficient to construct an entire animal and run it from the earliest embryo until death.  It was becoming clear that there was some predictability in this, and geneticists were working furiously trying to understand how this would work.

This was not just academic. Plant and animal breeders realized that this information had the potential to revolutionize animal breeding and agriculture, so there was much happening that was specifically directed at breeding livestock and crops.

Sewall Wright was an evolutionary geneticist more interested in theory than cattle and corn breeding, but in addition to various academic positions he was employed by the US Department of Agriculture, which wanted to learn how to apply “modern” genetics to animal breeding.  While he was at USDA, he wrote a short but seminal paper about inbreeding in animals called “Coefficients of Inbreeding and Relationship” (1921), in which he describes the derivation of his new coefficient of inbreeding (COI).

I think you will understand COI better if you know what problems Wright was trying to solve when he did this.  It was well known that inbreeding produced two effects (Wright 1921):

“First, a decline in all elements of vigor, as weight, fertility, vitality, etc., and second, an increase in uniformity within the inbred stock, correlated with which is an increase in prepotency in outside crosses... The best explanation of the decrease in vigor is dependent on the view that Mendelian factors unfavorable to vigor in any respect are more frequently recessive than dominant, a situation which is the logical consequence of the two propositions that mutations are more likely to injure than improve the complex adjustments within an organzism and that injurious dominant mutations will be relatively promptly weeded out, leaving the recessive ones to accumulate, especially if they happen to be linked with favorable dominant factors.  On this view, it may be readily shown that the decrease in vigor in starting inbreeding in a previously random-bred stock should be directly proportional to the increase in the percentage of homozygosity... As for the other effects of inbreeding, fixation of characters and increased prepotency, these are of course in direct proportion to the percentage of homozygosis.  Thus, if we can calculate the percentage of homozygosis which would follow on the average from a given system of mating, we can at once form the most natural coefficient of inbreeding.”

Wright was thinking that since we know inbreeding has two consequences that matter to a breeder, one positive (uniformity and prepotency) and one negative (loss of vigor and fertility), it would be useful to be able to calculate the degree of inbreeding of an animal because it would make these effects more predictable.  This was important because these two effects were at odds with each other; breeders could increase predictability and uniformity by inbreeding, but not without also having detrimental effects on an animal’s health and fertility.  Breeders couldn’t just keep inbreeding and inbreeding to get better and better cattle; with each incremental improvement in uniformity and homozygosity, there would be a price to pay in health, vitality, and reproductive performance, traits that are collectively referred to as “fitness”.  The short-term gain in consistency from inbreeding was paid for by a longer-term penalty in viability.  For those in the business of agriculture, it was critically important that the balance between the positive and negative be controlled.  Producing great cows today would be counterproductive if the ability to produce new stock in the future was diminished.

What Wright wanted was a way to predict the degree of homozygosity that would result from inbreeding, in which an allele originating from a single ancestor could be passed down on both sides of the pedigree and result in a descendant that is homozygous for that allele.  Knowing that there are two possible alleles for each gene, and that which of the two is inherited was random, Wright could assume that for each generation the probability of inheriting one allele or the other was 50%.  The sum of the probabilties for each generation would provide an estimate of homozygosity in a descendant.

“If an individual is inbred, his sire and dam are connected in the pedigree by lines of descent from a common ancestor or ancestors.  The coefficient of inbreeding is obtained by a summation of coefficients for every line by which the parents are connected, each line tracing back from the sire to a common ancestor and thence forward to the dam, and passing through no individual more than once.  The same ancestors may of course be involved in more than one line.”  (Wright 1921)

It’s not necessary for you to understand all the fancy math in his paper, but you can probably imagine how complicated the computations could get for tracing all possible paths from multiple common ancestors to the individual of interest.  You can do them by hand for a few generations, but past about 5 or so they are best left to a computer, and past about 10 the computer needs to be fast and efficient or the computing time can get quite long.  This is probably why most COIs on breed websites are reported for 5-10 generations of pedigree data; beyond that, the task of computing them becomes onerous.  But the number of generations used in the computation of COI profoundly affects the validity of the number you get (and you will always get a number, valid or not!), and it is critical for you to understand this and why it is a huge problem in interpreting data for dogs.

Complexity aside, the coefficient of inbreeding has become one of the most powerful tools a breeder can use, not only to predict the potential danger possible from a particular mating in terms of decreased fitness and increased incidence of genetic disease, but also to anticipate an animal’s possible value in the genetic improvement of the next generation.  Used without consideration of both positive and negative effects, inbreeding is invariably destructive.  But used carefully as part of a breeding plan that strategically balances its benefits and dangers, inbreeding can be a powerful tool.

Wright S, 1922.  Coefficients of inbreeding and relationship.  Am Nat 56: 330-338. (pdf)

By Carol Beuchat PhD

The development of ever better and cheaper tools for the analysis of DNA have driven the development of a genetic testing industry that allows breeders to identify potential genetic defects before breeding.  Through pedigree research and careful selection of tested animals, breeders can now produce a litter of puppies that is essentially guaranteed to be free of particular genetic disorders that plague the breed.

Breeders have adopted routine genetic testing with enthusiasm.  Nevertheless, purebred dogs are in a bad spot.  The list of genetic disorders is growing faster than the canine geneticists can identify the responsible genes, and the development of tests lags even further behind.  DNA tests are powerful tools - they can reliably identify the presence of particular mutations, but it sure doesn’t feel like we’re winning this war.  Is all of the effort and expense being invested in DNA analysis and testing actually improving the health of our dogs?

Every dog has many mutations, most of which we have no tests for - in fact, we don’t even know they exist unless they cause a problem in some dog.  For recessive mutations, this means they can be passed harmlessly from generation to generation like any other allele, and as long as they don’t meet up with a second copy (to become homozygous) in a dog they have essentially no effect.

The genes that we can test for are only a tiny fraction of all of the mutations in a population or even a single dog.  So the genes we test for are really just the bit of iceberg that sticks out above the water.  The much larger list of defects we don’t know about is out of sight, out of mind.  We can carefully navigate around the dangers we can see, but lurking below just out of sight is a much larger mass that can destroy the navigator who cuts it just a bit too close around the edge.

What do we actually accomplish with genetic testing?  Most of the tests currently available are for specific recessive mutations.  For these, breeding a carrier to a carrier has a 25% risk of producing an affected puppy.  (Get out your Punnett square if you need a refresher.)  So by testing, a breeder can completely eliminate the chance of  producing a puppy with a particular disorder.  The level of disease risk that a breeder is avoiding with this type of testing is 25% - one chance in four.

But what about all those other mutations we don’t know about?  The DNA tests we’ve done tell us nothing about them, and they work the same way as any other recessive mutation - a puppy that gets two copies of a mutation will be homozygous and therefore “affected”.  This is the part of the iceberg under water that we can’t see, and if we don’t properly navigate around it there could a serious consequence - a genetic disease.  How can we avoid these mutations if we don’t even know what they are?

We need to navigate in a way that will minimize the risk of a puppy inheriting two copies of the same recessive mutation.  Dogs that are closely related are more similar genetically than dogs that aren't, so the probability of producing offspring that inherit the same allele - good or bad - from both sides of the family is higher in dogs that are related.  We can estimate the liklihood of a puppy inheriting two copies of an allele from an ancestor on both sides of the pedigree with a statistic called the “coefficient of inbreeding” (COI).  A COI of 25% means that there is a 1-in-4 chance that a puppy will inherit identical copies of the same allele from both the sire and the dam.  This is the same as the risk we talked about earlier that a breeder avoids with DNA testing.

So think about this.  Breeders can reduce the possibility of puppies affected with a particular genetic disorder from 25% to 0% by means of a DNA test.  But if a breeder diligently runs all available DNA tests to prevent known genetic disorders, then does a breeding with a COI 25%, all they have done is swap the risk of known problems for the equally high risk of problems from those unknown, unseen mutations lurking beneath the surface.  DNA tests should definitely be used as a tool to prevent known genetic disorders in puppies.*  But it is pointless to be reducing the risk of one genetic disorder with one hand (and paying for the test to do it), only to be increasing the risk with the other hand by pairing two dogs with a high liklihood of sharing recessive mutations because of their shared ancestry.

This is the road we’re on.  Breeders run the available DNA tests to avoid known carriers, then line breed on that terrific dog a few generations back to double up on some of those genes - mutations included.  Or breed two dogs with lots of common ancestors to get the “look” of that line, without considering the probability that even if the dogs are “health tested” for known mutations, they nevertheless can both carry some of those unknown mutations, and the more ancestry they share the greater the risk.  It’s a self-defeating cycle - DNA testing to eliminate the possibility of homozygosity for a known mutation, do a breeding with a significant risk of producing homozygosity for unknown mutations, produce puppies that have a genetic disorder, commission a laboratory to identify the offending mutation and develop a test, then add this new silver bullet to the list of disorders you can avoid through testing in the next litter.  The future of the DNA testing companies is assured.

Responsible breeders can test for genetic disorders, but they negate their own efforts if a breeding is likely to pair up unknown mutations.  Breeding to eliminate one genetic problem while encouraging another isn’t the road to better health.  This is genetic whack-a-mole, and it is potentially endless. 

DNA tests will not improve the health of dogs if breeders don’t also adopt breeding strategies that will reduce the risk of producing a puppy that is homozygous for ANY mutation, not just the ones we know about.  We will never solve the problem of genetic disorders caused by recessive mutations in dogs until we understand why these mutations become a problem in the first place.  And when we understand that, it will be clear why DNA tests are not really necessary to avoid these problems.  If breeders reduce the risk of producing puppies that are homozygous for these unknown mutations - and the COI tells us the magnitude of that risk for a particular breeding - the risk of problems from ALL mutations, including those we now test for will also be reduced.  The lower the COI, the lower the risk.

What we need here isn’t genetic tests, it’s genetic management - breeding in a way that reduces the risk of all genetic disease to an acceptable level.  Breeders can do this by using the information in a pedigree database to compute COI.  (Again, we’re talking here only about recessive mutations, which cause disease only when homozygous.)  Apparently a 25% risk is unacceptably high, because breeders are paying for tests to avoid this.  Would 10% also be too high?  What about 5%, or 1%?  This might be different for every breeder, and perhaps even every breeding.  COI is only a statistical probability, but it is far better than simply crossing fingers after a breeding decision is made, and hoping that the stars are aligned, the karma is good, and that two months down the road you will be cuddling a litter of cute, healthy puppies.

The qualities of  “type” are fixed in a breed by reducing the variation in the population for the genes behind those traits.  For these, consistency is produced by homozygosity.  But homozygosity is also responsible for the high rates of genetic disorders in dogs caused by recessive mutations, so this is the crux of the problem.  What we really need is a way to increase homozygosity of the genes we want, without also increasing the homozygosity of the ones we don’t.  Is this possible?  Can we have both type and health?

The short answer is yes.  Breeders of other domestic animals have been doing this for decades.  In fact, the breeders of guide and service dogs have been quietly doing it as well.  With the right tools and appropriate expertise, breeders can improve type more efficiently than they are doing now, at the same time as they are also reducing genetic diseases.  Several European kennel clubs are beginning to implement these techniques, and it is only a matter of time before they are in widespread use by dog breeders.  But the longer it takes, the deeper the genetic hole we are digging for our dogs and the harder it will be to turn things around.  It will take cooperation among groups of breeders, oversight by clubs or breed organizations, some specialized expertise, the motivation to take the bull by the horns and just do it.  I am confident that, sooner or later, this will happen.

But in the meantime, breeders need to understand that DNA screening will prevent producing puppies afflicted with disorders that we can test for, but there is little gained by avoiding the part of the iceberg you can see only to be smashed by the larger, more dangerous part that you can’t.  You can navigate far more safely if you use the coefficient of inbreeding to assess the risk of having an unfortunate encounter with mutations that cause genetic disease, both the ones you know about and the ones you don’t.

You can learn more about the coefficient on inbreeding and how you can be using it to reduce the risks of genetic disorders in your dogs, especially those that could be caused by unknown mutations, ICB's 4 week online course "COI Bootcamp".

by Carol Beuchat PhD

Most genetic disorders in dogs are caused by single, autosomal recessive mutations. These are genes that somewhere along the line were copied improperly or damaged somehow, but have no ill effects as long as the animal has a copy of the normal gene.  These mutations can lurk in the genome for generations, passed on just like any other gene, and never causing a problem as long as they are bachelors.

But if an animal inherits two copies of the mutation, it won't have a copy of the normal gene so whatever function that gene had won't be done properly.  Perhaps it's something relatively trivial (e.g., ability to curl your tongue while whistling) that doesn't matter much to the animal's success at reproducing, but it can also be something serious, in which case it counts as a "genetic disease".

The list of these inherited diseases caused by single mutations seems to grow every day, and in fact you can keep track of the tally on a database called "Online Mendelian Inheritance in Animals".

The problem with these diseases caused by recessive mutations is that they sneak up on you.  Affected animals are homozygous for the allele, but there will be animals that carry the mutation, are heterozygous for the allele, that will have no symptoms.  These of course are the so-called "carriers", and carry is exactly what they do.  These animals sneaking around in the population are just one allele away from producing offspring with a particular genetic disorder, and the more carriers there are in a population the more likely one carrier is to bump into another and produce homozygous, affected offspring.

Genetic disorders in dogs caused by these recessive mutations usually seem to "suddenly" appear. The problem has never been seen in a breed, then a few dogs turn up affected but nobody gets too excited, then more dogs are are appearing, and finally breeders recognize that this particular disorder is becoming "a problem" in their breed.

Unfortunately, by the time alarm is raised, the mutation can be widespread in the breed.  We can use the Hardy-Weinberg equation to get a rough idea of how widespread a recessive mutation is in a population.  H-W tells us how the frequency of homozygous and heterozygous genotypes change with the frequency of one of the two alleles.  This is a graph of the estimated frequency of carriers of a recessive mutation based on the frequency of affected dogs that are observed. What you see here should be very worrisome.

When the frequency of affected dogs (aa) in a population is 1% - e.g., that's only three dogs in a population of three hundred - the frequency of carriers (Aa) is already almost 20%. Three dogs in 300 with a particular health issue would never be recognized by breeders as an impending crisis.  It's "just one of those things" or a bit of bad luck. Not knowing how the problem is inherited, breeders are unlikely to connect the dots, and in any case the number of affected cases would need to be much higher before a pedigree analysis would be very informative.

So, most recessive mutations are not detected until the number of affected animals is high enough to get everybody's attention. How high would that need to be? Well, if it's 5% affected, the fraction of carriers is already about 35%; if it's 1 in 10 dogs, the carrier frequency is over 40%.  Dealing with a nasty mutation that is carried by more than about 20% of a breed is extremely difficult and takes the cooperation of many breeders if the problem is going to be managed in the breed as a whole. (Of course, some breeders will hunker down and stay within their own lines, which they assume to be "safe", but of course unknown mutations are lurking in every animal and sooner or later one will become a problem.)  But if the problem isn't noticed until the carrier frequency is 40%, well then you've got yourselves a real problem.

EVERY dog has recessive mutations.  To cause a problem, a recessive mutation must find a mate, and the best place to find that is in a close relative that shares many of your genes.  The very reason breeders inbreed - to get two copies of the same allele for desirable traits - is the very reason the number of genetic disorders in dogs will continue to rise.  And as a breeder gets more and more consistency in the traits they want by increasing the amount of genetic homozygosity in their dogs, they also get a greater and greater risk of fixing the genes for some heretofore unknown genetic problem.  You can't increase homozygosity for the genes you want by breeding to related dogs without also increasing the homozygosity of the genes you don't.  This is a real bummer, but it's just math.  

Estimates of carrier rates using H-W aren't perfect. There are underlying assumptions that are definitely violated in purebred dog populations (such as "no selection"), and the dogs certainly are not bred randomly. But in general, the situation in dog breeds is likely to be worse than predicted by H-W because breeders tend to select dogs to breed to that have a HIGHER than average chance of sharing mutations (compared to a dog chosen at random) because they're related.

Go back and scan the last health survey of your breed. Notice any problems that are so uncommon that nobody is worrying about them?  The odd PRA here, a case or two of degenerative myelopathy, perhaps a few dogs with cerebellar ataxia or myasthenia gravis, but not in any dogs that you know.  Think about this the next time a breeder says of a genetic disorder "Oh, that's not been a problem in our breed."  The seeds of the next problem are already sown.