Dogs
Detailed postnatal records among the puppies born between 2011 and 2017 were used in the Labrador retriever breeding colony at the Hokkaido Guide Dogs for the Blind Association in Japan. Their pedigree data kept by the AGBN (Asia Guide Dog Breeding Network) was used for genetic analysis. It was excluded from this study for the dogs which could not be tracked with accurate postnatal records.
Inclusion criteria for swimmer puppy: Based on the inspection data by the veterinarian in charge of the Guide Dog Association, a swimmer puppy was defined as followed. Dogs that were seen to have all the following three symptoms at the same time between the 14th and 28th days after birth: 1) the forelimbs, hind limbs, or all four limbs are laterally deployed, 2) flat chest, 3) unable to stand. All individuals were inspected by the same veterinarian.
Data Analysis
Data were collected for incidence rate, prognosis, breed, sex, weight, and the litter size. The incidence rate of swimmer puppy was expressed as a percentage by dividing the number of affected individuals by all births. The prognosis survey was traced back with all postnatal data. The litter size, and body weight records at birth, 10 days, and 28 days after birth were used and compared between litters where swimmer puppy appeared (swimmer puppy group) and the litters without swimmer puppy (non-symptomatic group). The correlation between the litter size and the body weight at each period was statistically analyzed.
Inbreeding coefficient
The inbreeding coefficient was calculated based on the assumption that the onset of swimmer puppy is a qualitative trait and a hereditary disease due to a single recessive gene. An ancestor that frequently appears among the blood relatives of the affected individual was identified, and when that individual was designated as the common ancestor (A), the inbreeding coefficient derived from A was calculated for all individuals with records using the following formula.
where is the inbreeding coefficient of the individual X derived from A, is the number of individuals on the i-th pathway that connects one parent to the other parent via A, and FA is the common ancestor A's inbreeding coefficient. was calculated for all common ancestors appearing among the pedigree The values were estimated using the program inbh (http://www.obihiro.ac.jp/~suzukim/masuda/software_inbcoef.html), which applied the method of Meuwissen and Luo [12]. The program was modified for estimating the inbreeding coefficient originated by common ancestor (A). In order to search for the causative common ancestor, individuals with golden parents were excluded from consideration, and only Labradors and hybrids were included.
Heritability and genetic correlation
Assuming that the onset of swimmer puppy is a quantitative trait, the heritability was calculated. Heritability is a measure of how much an individual's phenotypic value is determined by additive genetic effects. A total of 436 records with 10946 individuals in pedigree records were used to estimate genetic parameters of Labradors. The onset trait of swimmer puppy in retriever species was estimated using the best linear unbiased prediction (BLUP) animal model below [13, 14].
Formula 1: ℓ = S + L + u + e
Formula 2: y = S + L + u + e
where ℓ is the liability records expressed by 0 (normal) or 1 (onset) for the onset of swimmer puppy, S is the fixed effect of gender (2 levels), and L is the covariate of the number of litters, u is the additive genetic effect, and e is the residual effect. The body weight on the 10th day was also regarded as a phenotype, and the heritability was calculated in the same manner (Formula 2). Where y is observation of body weight, and S, L, u and e are the same as in Formula 1. The genetic correlation between the above two traits was estimated according to the bivariate BLUP animal model analysis. The THRGIBBSf90 program (http://nce.ads.uga.edu/~ignacy), which could handle threshold traits, was used to estimate variance components and best linear unbiased estimators (BLUE) and BLUP [13]. Then Heritability h² was shown as additive genetic variance (A) / additive genetic variance + residual variance (A + R). To calculate the posterior means and standard deviations of (co)variances, heritabilities and genetic correlation, 400 000 samples after a burn-in of 100 000 iterations were used. Convergence was determined by using visual inspection of the plots of Gibbs samples.
Statistical analysis: EZR (EazyR: https://www.jichi.ac.jp/saitama-sct/SaitamaHP.files/statmed.html) was used for statistical analysis, and the chi-square test and Mann-Whitney U test were performed as statistical processing.