The Effects of FGF4 Retrogenes on Canine Morphology

Two FGF4 retrogenes (FGF4L1 on chromosome 18 and FGF4L2 on chromosome 12) have been identified to cause dwarfism across many dog breeds. Some breeds are nearly homozygous for both retrogenes (e.g., Dachshunds) and others are homozygous for just one (e.g., Beagles and Scottish Terriers). Since most breeds do not segregate both of these retrogenes, it is challenging to evaluate their individual effects on long bone length and body size. We identified two dog breeds selected for hunting ability, the Alpine Dachsbracke and the Schweizer Niederlaufhund, that segregate both of these retrogenes. Using individual measurements of height at the shoulder, back length, head width, thorax depth and width, and thoracic limb measurements, we evaluated the combined effects of FGF4 retrogenes within these breeds. We applied multivariable linear regression analysis to determine the effects of retrogene copy numbers on the measurements. Copy numbers of both retrogenes had significant effects reducing height at the shoulders and antebrachial length, with FGF4L1 having a much greater effect than FGF4L2. FGF4L1 alone influenced the degree of carpal valgus and FGF4L2 alone increased head width. Neither retrogene had an effect on thorax width or depth. Selectively breeding dogs with FGF4L1 and without FGF4L2 would likely lead to a reduction in the FGF4L2-related risk of intervertebral disc herniation while maintaining the reduction in leg length resulting from FGF4L1.


Introduction
Dog breeds come in a wide variety of shapes and sizes. Dwarfism or skeletal dysplasia characterizes many dog breeds and was selected to slow the speed of hunting dogs, allow entry into small burrows to pursue game and to reduce overall size for companion dogs (www.fci.be and www.akc.org (accessed on 25 January 2021)). Two different FGF4 retrogenes have been associated with dwarfism across dog breeds [1,2]. Interestingly, the most common form of human dwarfism is caused by reoccurring activating mutations in the FGFR3 gene, which is one of the receptors for FGF4 [3]. Retrogenes are functional, active intronless copies of the parent gene that integrate into chromosomal DNA through the action of endogenous LINE 1 retrotransposons [4]. The exact mechanism linking FGF4 retrogene expression and skeletal dysplasia is not defined; limited data are available relating to temporal and spatial expression of FGF4 retrogenes and although FGF4 protein overexpression acting through FGFR3 is an intriguing comparative model, FGF ligands are Genes 2022, 13, 325 2 of 8 known to bind to multiple receptors [1,2,5]. We identified five additional FGF4 retrogenes in canids leading to the corrected nomenclature of the retrogenes as L for "like", followed by a number representing the order in which they were discovered [6].
One of the known canine FGF4 retrogenes is located on chromosome 18 (FGF4L1) and is associated with a phenotype termed chondrodysplasia [2]. The other FGF4 retrogene is located on chromosome 12 (FGF4L2) and has been associated with a phenotype previously described as chondrodystrophy [7]. Chondrodystrophy is characterized by both short legs and abnormal intervertebral discs [7]. Chondrodysplasia was defined as dogs having very short legs and is required by many breed standards [2]. Both loci have been identified in across breed association studies for height at the shoulders [8][9][10]. However, the direct effect of each retrogene on height or individual bone segments has not been evaluated.
The breed distribution of the two FGF4 retrogenes suggests that the effects of the retrogenes are additive, since the breeds with the shortest legs, for example the Dachshunds and Corgis, have both retrogenes at high allele frequencies [1,11]. Since the majority of dog breed standards are very prescriptive about height, most breeds are unlikely to segregate alleles for height. This has limited the ability to detect the effects of these variants on height within breeds. Two breeds, the Alpine Dachsbracke (AD) and the Schweizer Niederlaufhund (SN), segregate different copy numbers of both FGF4 retrogenes within each breed [11], likely reflecting breed standards, emphasizing hunting ability rather than a specifically defined absolute height. Height at the shoulders in the SN varies from 35-43 cm with an additional tolerance of 2 cm in males and in the AD 34-42 cm (http://www.fci.be (accessed on 25 January 2021)). We directly evaluated genotype phenotype correlations with height at the shoulder in both breeds and demonstrated that the effects of the two FGF4 retrogenes additively decrease height. In the SN, we also evaluated the effect on additional specific morphologic measurements obtained from individual dogs.

Sample and Data Collection
DNA samples, sex and height at the shoulder measurements for 111 dogs of the AD breed (58 males) were obtained from dog owners. DNA samples were submitted for genetic testing to the Vetsuisse Biobank at the University of Bern. DNA and morphological measurements from 36 SN (11 males) dogs were obtained by the investigators at gatherings of the official breed club in Switzerland. Procedures were approved by the "Cantonal Committee for Animal Experiments" (Canton of Bern; permits 48/13, 75/16 and 71/19) and the UC Davis Institutional Animal Care and Use Committee (Protocol #21190).

Measurements
Height at the shoulders was reported by owners of AD (Supplemental Table S1). All other measurements were collected by one author, DB, from SN over 1 year of age (Supplemental Table S2). Height at the shoulders was measured in cm with a level and movable arm. Aluminum body calipers were used to measure width of skull at the widest point, width of the thorax at the widest point and depth of the thorax at the greatest depth, length of the antebrachium from the olecranon process of the ulna to the palmar surface of the manus with the elbow and carpus held at a 90 • angle, and length of the manus from the cranial surface of the antebrachium to the distal aspect of the third phalanx of digits 3 and 4, with the elbow and carpus held at a 90 • angle. A tape measure was used to measure the length of the humerus (proximal aspect of the humeral greater tubercle to the olecranon process of the ulna), scapula (dorsal border to the proximal aspect of the humeral greater tubercle), back length (distal edge of the scapula to the proximal aspect of the ilium). A goniometer was used to measure carpal valgus, which was recorded as the angle of deviation [11]. Reproducibility of these measures was determined by comparing repeated measures taken of the same three dogs on three different days by the same observer. The coefficient of variation was calculated for each measurement for each dog and the mean percentage reported (Supplemental Table S3).

Genotyping
Genotyping for FGF4L1 and FGF4L2 was performed using a PCR based assay as previously described [1]. FGF4L1 is the FGF4 retrogene referred to in this manuscript as chromosome 18 FGF4 and FGFL2 is the FGF4 retrogene referred to in this manuscript as chromosome 12 FGF4. In brief, three primer PCR is used to flank the novel insertion sites for FGF4L1 and FGF4L2 in separate PCR reactions. The third prime is an FGF4 internal primer. Different product sizes are generated for the absence and presence of the retrogene insertion.

Statistical Analysis
One dog had an injury that prevented thoracic limb measurements of the right limb. Other limb measurements and back length measurements from both the left and right were used across all genotype classes. Multivariable linear regression analysis was used to identify contributions of genotype to measurement using stats package in R version 4.1.1 [12]. Sex, FGF4L1 genotype (zero copies, one copy, two copies) and FGF4L2 genotype (zero copies, one copy, two copies) were independent variables. Reference categories for categorical variables were zero copies for retrogene genotypes and female sex. Univariable analyses were performed initially, and all independent variables were tested for inclusion in the multivariable model if either retrogene had Wald p < 0.2. Results were reported as differences in mean length cm or percent change and surrounding 95% confidence intervals (CI) (Supplemental Tables S4 and S5). p values were reported for the inclusion of each variable in the model. Carpal valgus was analyzed by rank ordering the measurements first.

Evaluation of Height at the Shoulder
Two breeds of dogs which segregated different copy numbers of FGF4 retrogenes were chosen to study the effects of the retrogenes on morphology. The Alpine Dachsbracke dogs (N = 111) had height at the shoulders reported by the owners and the Schweizer Niederlaufhund (N = 36) were measured during this study. Based on the possible combinations of the two retrogenes, there were nine possible genotype classes. However, not all genotype classes were represented in each breed ( Figure 1). Based on comparing the mean height of each genotype class, there was an additive effect of copy number for each retrogene and between FGF4L1 and FGF4L2 on reducing height at the shoulder. In order to investigate this further, univariable and multivariable linear regression were performed to determine the effect of retrogene copy number and gender on height (Tables 1 and 2). Copy numbers of both retrogenes and gender had significant effects on height at the shoulder. In both breeds, there was a high coefficient of multiple correlation between height and the number of copies of the FGF4 retrogenes (r 2 = 0.75 in SN and 0.71 in AD). The magnitude of the effect on height reduction was much greater for FGF4L1 (15.9 to 27.9%) than for FGF4L2 (3.2 to 10.5%) in both breeds.

Evaluation of Specific Morphologic Parameters in the Schweizer Niederlaufhund
Measurements of thoracic limb segments, thorax depth and width, back length, head width and amplitude of carpal valgus were determined. The mean coefficient of variation ranged from 2.1% to 4.1% for the caliper measurements, 3.0-7.2% for the tape measurements and 2.3% for the shoulder height. Carpal valgus measured with a goniometer had a mean coefficient of variation of 14.5% (Supplemental Table S3). Univariable and multivariable linear regression were performed for all measurements to determine the effect of retrogene copy number and sex (Supplemental Tables S4 and S5). FGF4L1 copy number was found to have a significant effect on increasing carpal valgus (p < 2 × 10 −16 ) and reducing scapular length (p = 1.1 × 10 −7 ). FGF4L2 copy number was found to have a significant effect increasing head width (p = 0.015). Effects of copy number of both retrogenes were statistically significant, reducing humerus length (FGF4L1 p = 8.3 × 10 −9 , FGF4L2 p = 0.008), antebrachial length (FGF4L1 p = 2.0 × 10 −16 , FGF4L2 p = 1.6 × 10 −16 ), and manus length (FGF4L1 p = 0.0002, FGF4L2 p = 0.03). The largest effect was on antebrachial length, where two copies of FGF4L1 reduced length by 28% and two copies of FGF4L2 reduced the length by 10%. Neither retrogene had a significant effect on thorax depth or width. A paradoxical effect for both retrogenes was seen on back length, where one copy of the retrogene increased back length and two copies decreased back length (FGF4L1 p = 0.001, FGF4L2 p = 0.0007). A summary of the significant results is shown in Figure 2. Sex had a significant effect on manus length (p = 0.005), scapular length (p = 0.04), and head width (p = 0.003).

Evaluation of Specific Morphologic Parameters in the Schweizer Niederlaufhund
Measurements of thoracic limb segments, thorax depth and width, back length, h width and amplitude of carpal valgus were determined. The mean coefficient of varia ranged from 2.1% to 4.1% for the caliper measurements, 3.0-7.2% for the tape meas ments and 2.3% for the shoulder height. Carpal valgus measured with a goniometer a mean coefficient of variation of 14.5% (Supplemental Table S3). Univariable and m variable linear regression were performed for all measurements to determine the effec retrogene copy number and sex (Supplemental Tables S4 and S5). FGF4L1 copy num   was found to have a significant effect on increasing carpal valgus (p < 2 × 10 −16 ) and reducing scapular length (p = 1.1 × 10 −7 ). FGF4L2 copy number was found to have a significant effect increasing head width (p = 0.015). Effects of copy number of both retrogenes were statistically significant, reducing humerus length (FGF4L1 p = 8.3 × 10 −9 , FGF4L2 p = 0.008), antebrachial length (FGF4L1 p = 2.0 × 10 −16 , FGF4L2 p = 1.6 × 10 −16 ), and manus length (FGF4L1 p = 0.0002, FGF4L2 p = 0.03). The largest effect was on antebrachial length, where two copies of FGF4L1 reduced length by 28% and two copies of FGF4L2 reduced the length by 10%. Neither retrogene had a significant effect on thorax depth or width. A paradoxical effect for both retrogenes was seen on back length, where one copy of the retrogene increased back length and two copies decreased back length (FGF4L1 p = 0.001, FGF4L2 p = 0.0007). A summary of the significant results is shown in Figure 2. Sex had a significant effect on manus length (p = 0.005), scapular length (p = 0.04), and head width (p = 0.003).

Discussion
FGF4 retrogenes had an additive effect reducing height at the shoulders in two breeds of dog. Individual thoracic limb segments performed in one breed were all affected by FGF4L1, but only the more distal segments showed significant effects of FGF4L2. Neither FGF4 retrogene had an effect on thorax depth or width. FGF4L1 alone affected carpal valgus, which is consistent with other breed phenotypes. For example, Beagles and American Cocker spaniels that only have FGF4L2 do not typically have carpal valgus as a breedrelated phenotype, while Basset hounds and Dachshunds do [13]. Interestingly, FGF4L2 caused increasing skull width, which has not been reported previously but is also consistent with some breed traits, for example in French Bulldogs and Beagles [13].

Discussion
FGF4 retrogenes had an additive effect reducing height at the shoulders in two breeds of dog. Individual thoracic limb segments performed in one breed were all affected by FGF4L1, but only the more distal segments showed significant effects of FGF4L2. Neither FGF4 retrogene had an effect on thorax depth or width. FGF4L1 alone affected carpal valgus, which is consistent with other breed phenotypes. For example, Beagles and American Cocker spaniels that only have FGF4L2 do not typically have carpal valgus as a breedrelated phenotype, while Basset hounds and Dachshunds do [13]. Interestingly, FGF4L2 caused increasing skull width, which has not been reported previously but is also consistent with some breed traits, for example in French Bulldogs and Beagles [13].
Limitations of this study include using owner reported heights in the AD without the ability to assess the reproducibility of the measurements. In addition, the limited number of dogs of both sexes in each genotype class for the SN could have masked more subtle effects of both retrogenes on the measurements. Sex was always a variable used for the multivariable models and was significant for height, skull width, scapula, humeral and manus length.
Several other studies evaluated the effects of height-related alleles segregating within breeds. In the Labrador retriever, a COL11A2 missense variant conferred a 10% reduction in height at the shoulders [14]. A variant in ADAMTS17 confers a risk of glaucoma in Petit Basset Griffon Vendéen (PBGV) and Shar Pei but was also demonstrated to have an effect on height. In the Shar Pei, a 6% reduction in height was observed within homozygous animals and in the PBGV, a 4% reduction was observed [15]. The PBGV is a dwarf breed homozygous for FGF4L1. One study evaluated the effect of FGF4L1 on height within the Havanese breed and identified that dogs homozygous for FGF4L1 were 4.9 cm shorter than heterozygotes, with a 16% decrease in height [16]. In that study, only FGF4L1 was evaluated, however FGF4L2 also segregates in the breed [17].
Dwarfism in dogs has not been characterized relative to its effects on different long bone segments. Based on the data presented in this manuscript, the antebrachium is the most affected by FGF4L1, however all segments of the thoracic limb had a significant decrease in length. FGF4L2 affected all segments except the scapula but to a lesser degree than FGF4L1. Both retrogenes had additive effects on the thoracic limb. The two described FGF4 retrogene insertions lead to semi-dominant gain of function alleles [1,2]. The resulting phenotypic differences are most likely due to the specific expression control based on the insertion sites [1,2,6]. Further evidence for this is seen in the distinct phenotype differences seen in dogs with different FGF4L1 and FGF4L2 genotype combinations [16]. One interesting finding was the increase in head width significantly associated with FGF4L2, but not FGF4L1. Previous volumetric evaluation of canine skull based on CT scans identified FGF4L1 as increasing the neurocranium in dogs [18]. These studies did not identify any association with FGF4L2, however our measurement was for the width of the skull rather than its volume. This phenotype could also be a driving selection for the FGF4L2 insertion allele.
We identified a paradoxical relationship with both retrogenes and back length, where heterozygotes had longer backs and homozygotes had shorter backs than dogs without the retrogenes. The method that was used to measure back length was a tape from the distal scapula border to the proximal ilium. This paradoxical effect may have been caused by the presence of lordosis in dogs with two FGF4 retrogene copies but not in dogs with one copy. There could also be additional genetic or environmental factors segregating in these dogs. Studies aimed at precisely measuring vertebral body length might be able to address this question more appropriately.
Breed height standards constrain the population of animals that may be chosen for breeding, which may maintain FGF4L1 in the homozygous state since it has such a strong effect on height. Batcher et al. [11] identified eight breeds homozygous for FGF4L1 and only three breeds homozygous for FGF4L2. FGF4L2 has a more modest effect on height, particularly in the context of breeds homozygous for FGF4L1, allowing it to segregate in more breeds (37 vs. 24) [11]. One of the motivations for this study was to determine how strong the effect of FGF4L2 is on the phenotypes of dogs, since FGF4L2 is undesirable due to its association with Hansen type I intervertebral disc disease [11]. In the AD, being homozygous for FGF4L2 reduces the shoulder height by only 6% and, in this breed, those dogs were right in the middle of the height standard. Reducing or eliminating FGF4L2 from breeds due to its associated significant health concerns and its modest effect on height should be possible, particularly in breeds that segregate FGF4L2.  Informed Consent Statement: The dogs' owners gave informed consent to use samples and data for this study.

Data Availability Statement:
All data is available within the manuscript.