Canine Hip Dysplasia Can Now Be Predicted At Birth By DNA

http://www.flickr.com/photos/dimas_sc/3100685891/

A few breeders and I occasionally have a chuckle over the fact that pet buyers seem to be more concerned with knowing hip scores than are most other breeders.  Perhaps the other breeders have already done their homework on the OFA database or our websites; it certainly appears that the puppy buyers have done their homework regarding finding a reputable breeder. Good for them; that’s the kind of home I like my puppies to go to.

Canine Hip Dysplasia is a physically debilitating disorder that is both emotionally and financial draining for dog owners.  With no cure other than pain management or hip replacement surgery, diagnosis of hip dysplasia is the breeder’s as well as the pet owner’s heartbreak.  Stocky breeds such as the Bulldog and those weighing over 60 pounds such as the St. Bernard experience hip dysplasia rates of up to 72% and 46% respectively.  It is estimated by Cornell University that 22% of canine hospitalizations in the USA are hip dysplasia related. Interestingly, the disease’s occurrence is minimal in sight hounds such as the whippet, saluki, borzoi and greyhound with an average occurrence of just 1-2%.  Border collies fall in the middle range of dysplasia occurrence at around 11%.

For over 40 years, breeders have attempted to minimize the occurrence of canine hip dysplasia (CHD) through the use of phenotypical screening methods.   As a predictive method for removing poor breeding candidates from their breeding programs, responsible breeders have submitted x-rays of mature dogs for hip scoring to institutions such as the OFA, BVA, FCI and University of Pennsylvannia seeking to determine the laxity of the hip joint, and therefore the likelihood of the animal developing CHD as a trait that might be passed on to its progeny.

Various studies of the efficacy of these screening programs have met with mixed results.  Some studies show little to no improvement in the occurrence of CHD. Even in countries, such as Finland, where an enforced threshold of hip scores was applied to registrations, results were varied across breeds, with some breeds demonstrating a lower incidence of CHD and others showing an actual increase in CHD.

Canine hip dysplasia is understood to be influenced by both genetic and environmental factors.  As a polygenetic mode of inheritance has long been suspected, attempts to isolate the genetics of the disease have long been frustrated.  Breeders seeking to improve their lines and their breed have been marginally successful as a whole, despite their best efforts.  Forty years of breeders’ best efforts may call into question the scoring methodologies of the various institutions. 

OFA uses a single radiographic view, the hip-extended view, to determine laxity of the hip joint (a predictive measure of the likelihood of developing CHD).  By contrast, the PennHIP method uses 3 radiographic views (the distraction view, the compression view and the hip-extended view) to determine the laxity of the joint.  A study released in 2010 by the University of Pennsylvania (http://webcanine.com/2010/canine-hip-dysplasia-may-be-underreported/), compared results from dogs scored by both OFA and PennHIP to determine the likelihood of CHD onset. Their findings were disturbing. According to Penn researchers, “Even if breeders were to selectively breed only those dogs having OFA-rated “excellent” hips — the highest ranking but in some breeds, a very small gene pool, the study suggests that 52-100 percent of the progeny, depending on breed, would be susceptible to hip dysplasia based on the Penn Vet scoring method.”  In fact 52% of the dogs with OFA “excellent” scores, 82% of the OFA “good” scores and 95% of the OFA “fair” scores fell below the PennHIP threshold of.3, making them by PennHIP standards susceptible to CHD. (It should be noted that the University of Pennsylvania holds the patent for the PennHIP test, and a financial incentive could exist in their findings).  Could the PennHIP method provide the key to a significant reduction in incidence of CHD?

PennHIP allows for early testing of puppies; they may be tested as early as sixteen weeks, which is still well beyond the typical puppy placement age of around 8 weeks. By contrast, OFA scores may not be obtained until the dog has reached 2 years of age.  As 3 x-ray views are required for the PennHIP testing method, depending on your area of the country, the cost can be substantially greater than OFA testing.  Additionally, the PennHIP method does not currently have the same recognition international as the OFA methodology.  What if a genetic test existed that could predict the likelihood of developing CHD?

The canine genome project, begun in June 2003, determined that the canine genome contains approximately 2.5 billion DNA pairs, much like that of humans and other mammals.  In 2004, a preliminary set of about 600,000 single nucleotide polymorphisms (SNPs), which amounts to a SNP roughly every 5,000 DNA base pairs, was released to researchers to aid in the study of genetically transmitted diseases and disorders. Armed with this massive amount of data, researchers at the Department of Animal Science, China Agricultural University,Beijing, China developed an algorithm that accurately predicts CHD from genomic data.  This means that puppies can be tested prior to weaning, in fact as soon as DNA is available, allowing for poor breeding prospects to be petted-out well before maturity. 

The Beijing researchers’ model was based on 359 dogs in the first sample and 38 dogs in the second sample.  The dogs whose results were used in the study came from the Baker Institute, The Guiding Eyes for the Blind and Cornell University Hospital for Animals (as the 38 dog control group).

They began by developing an Estimated Breeding Value (EBV) of each of the dogs based on 4 criteria: the dog’s Norberg Angle (NA), OFA score, the distraction index (DI) and the dorsolateral subluxation score (DLS). The former two are evaluated from the extended hip projection and are phenotypically as well as genetically correlated while the latter two are evaluated on different x-ray views and are also phenotypically and genetically correlated. It is interesting to note that the researchers believed that no measure alone completely represents hip morphology. In fact, their methodology was the similar to that of the PennHIP with the OFA score thrown in for good measure.  In the final analysis, the Norberg angle (NA) correlated to the OFA score and as most dogs had NAs measured, NA was chosen for this study to determine the EBV. The researchers then used 22,000 SNP’s across the genome and the EBV of the sample dogs to produce the predictive formula, an algorithm, termed the Genomic Breeding Value (GBV). 

The researchers performed two types of validations consisting of a cross validation and an independent validation. For the cross validation they used a Jackknife Cross Validation.  This consisted of hiding the dog’s EBV and removing its GBV from the dataset used to create the original predictive formula. The formula was then recalculated by using the EBV and genotype from the other 358 dogs in the study. This process was repeated for each of the 359 dogs. The cross validation performed showed a strong correlation (R>0.7) above 70% between the EBV (phenotype) and the GBV (genetic).

To develop the independent validation, the GBV formula was applied to 38 Labrador retriever dogs with no pedigree relationship to the original 359 dogs from which the GBV formula was derived. Correlations between the predicted GBV and the actual EBV (the physical conformation of the dog) were compared.

When 5,000 SNPs out of the original 22,000 SNPs remained in the formula, the correlation was remarkable, above 0.98 or 98%. The researchers then identified the 100-500 most influential SNPs that had provided the most information to the GBV formula. The reduced number of SNPs showed that genomic prediction could remain effective with the most influential 100-500 SNPs chosen from the original 22,000SNPs.  However, as the number of SNPs in the reference panel dropped below 50, the formula became significantly less reliable.

While the majority of the 359 dogs used in the study were Labrador retrievers and Lab crosses with Greyhounds, other minor breeds’ hip morphologies were also well predicted by the GBV formula. Therefore, we can expect that multiple breeds will be able to be integrated together into a single genetic test, even though they are from diversified phenotypic and genotypic populations.

The possibility of a genetic test for predicting CHD in young puppies raises some interesting ethical questions.  Puppies genetically predisposed to CHD should be removed from breeding programs, but where do they go?  What are the ethics of breeders presenting puppies to pet puppy buyers knowing that they have a strong likelihood (70% or greater) of developing CHD?  Additionally, if the genetic statistics correspond with those of the University of Pennsylvania survey, we are left with a very small gene pool in many breeds.  Add to that other genetic disorders being researched, such as cancer and breed specific issues, and the gene pool of good breeding candidates is likely to grow even smaller.  Relying on a substantially reduced gene pool could have unintended consequences.  Recessive, genetic “trash” could manifest into new, more terrifying health issues. 

These are all questions that are sure to be discussed heatedly as understanding of the canine genome progresses. Once these tests are available to the public, they will have to be financially feasible in order to effectively make an impact on breeding programs.  As we approach this new era in breeding prospect selectivity, perhaps a stepped approach to improving breeding prospects is in order.  Under this approach, initially breeders would breed one dog with a relatively high EBV and GBV to one with a moderate EBV and GBV, and seek to improve the GBV threshold for the sire or dam on each successive generation.  This would be a measured approach, however canine generations evolve very quickly – every 2 years compared with 20 years in humans.  Your thoughts and comments are welcomed!

photo credit: Dimas_ via photopincc