Labrador retrievers are popular for use as companion and working dogs worldwide. Like other large sized breeds; this breed is also prone to lower back diseases like lumbosacral stenosis. In this dog version of human lumbar spinal stenosis (leading cause of spinal surgery in Americans over 60), abnormal narrowing of the spinal canal in the lower back region causes several neurologic complications. Symptoms of lumbosacral stenosis are not always obvious and/or externally visible making detection difficult. If detected early, the condition is manageable by therapy, but by the time symptoms usually appear therapy becomes less effective. The problem is even more pronounced in working dogs that can get retired early – a major concern for military and police organizations that invest a lot of resources in recruiting and training working dogs. Expensive imaging methods such as computed tomography (CT) can confirm the diagnosis, but are not routinely performed without obvious symptoms. Diagnostic methods capable of detecting lumbosacral stenosis early are needed, preferably ones that are inexpensive, accurate and rapid (like blood/saliva tests). However it is important to understand the genetics of a disease to design such tests. Since we could not find any genetic studies of lumbosacral stenosis in Labrador retrievers, we explored said genetics in this preliminary study. We randomly selected 40 young Labrador retriever military working dogs with no outward signs of lumbosacral stenosis and assessed them for presence or absence of lumbosacral stenosis based on CT image analyses. We identified 4 dogs affected and 4 dogs unaffected by lumbosacral stenosis and sequenced their DNA to identify genetic differences between normal and diseased dogs. One mutation was identified in a gene that could be associated with lumbosacral stenosis – transthyretin (protein associated with human lumbar spinal stenosis). Other genes identified with mutations could also be important but we could not find evidence of this in the published literature, so more research is needed before they can be ruled out definitively. Once we can identify gene/s responsible for this condition, we will be that much closer to designing improved early detection methods for early detection of lumbosacral stenosis.

Working dogs are high-performance athletes that assist human team-members in supporting public service, national security and military missions in the U.S. and around the world [1]. Working dogs perform a variety of tasks such as sentry-and-patrol; search and rescue; mobility support for disabled persons; and detection of explosives, arson accelerants, and illegal drugs. Labrador retrievers are one of the most popular working dog breeds [2, 3]. As of 2016, the U.S. Military had an estimated 2300–2500 military working dogs (MWDs) across all branches of the Armed Forces [4, 5]. Each year the U.S. Military invests major financial and personnel resources in procuring, training and maintaining these MWDs in peak physical condition. According to a 2011 U.S. Government Pentagon memo, typical purchasing and training costs for a high quality MWD can go as high as $40,000 U.S. dollars [6]. Since 2001, the U.S. Department of Defense has spent more than $941,000 U.S. dollars towards this objective [4]. The demand for high quality working dogs (especially detection dogs) has been increasing in recent years [3]. But the demand often exceeds the supply [2], especially due to the high cost of breeding, raising and subsequent training of MWDs [3]. Ideally these trained working dogs are expected to maintain functionality for at least 10–12 years [7]. Early retirement of such trained MWDs means both a functional loss in the productivity of the team that depends on the trained dog as well as a financial loss for the Military.

Spinal diseases are a leading cause for early retirement in MWDs [8]. Lumbosacral stenosis (LS) is a common pathological condition affecting the canine lumbosacral spine, especially in large breeds like German shepherds and Labrador retrievers [9–18]. Military working dogs are also affected by LS [19], but there have been no breed-specific studies in Labrador retrievers.

Canine LS is defined as an abnormal narrowing of the lumbosacral canal, vertebral canal, and/or the intervertebral foramina [20]. This morphologic problem can be a risk factor for disability, often due to compression of the underlying neural and/or vascular tissues leading to clinical conditions like cauda equina syndrome (CES) [21]. Low back pain (LBP) on palpation of the lumbosacral spine is considered to be the primary clinical sign of LS [22]. However, there are drawbacks of using LBP status alone in diagnosis of LS in dogs: (i) symptoms can be intermittent with appearance only after hard physical exertion; (ii) symptoms can mimic those of other spinal diseases like intervertebral disc degeneration [23], degenerative sacroiliac joint disease [24], foraminal stenosis [25], and Schmorl’s nodes [26]; (iii) stoic dogs may not consistently vocalize pain; and (iv) dogs are specifically bred to have high-drive (desirable in MWDs, but can delay the detection of sub-clinical conditions like LS). A dog can be structurally LS positive, while remaining clinically asymptomatic until the condition worsens to such an extent that therapeutic and surgical options become unviable, and the only remaining course of action is retirement (in some severe cases, even euthanasia). Therefore improved methods for early detection of LS are critical for minimizing the risk of early retirement in these valuable canines.

Computed tomography (CT) is an established non-invasive diagnostic imaging technique for clinical diagnosis of canine LS [27, 28]. Morphological abnormalities within the canine lumbosacral canal that can lead to narrowing of the vertebral canal/intervertebral foramen (based on CT observations in surgically confirmed cases of LS) have been previously reported [17]. Based on these previously published CT diagnostic criteria, a recent study found the quantitative CT diagnosis of LS made by fat area ratio (FAR) measurements to be comparable with the qualitative CT diagnosis of LS made by a board-certified veterinary radiologist [29]. However, CT diagnoses of LS are not always possible especially since expensive imaging studies are not routinely performed unless a dog presents with some clinical signs of LS. This allows the condition to go undetected longer until the damage becomes irreversible and effectiveness of therapeutic interventions is significantly reduced. Previous studies have suggested LS therapy to be more effective in younger MWDs with mild clinical signs, with an increased chance of a successful return to active duty [19, 30]. Therefore, there is a need for methods capable of detecting LS early.

Similar to human lumbar spinal stenosis (LSS), canine LS has two distinct etiologies: congenital (idiopathic and developmental) or the more common acquired (degenerative and post-traumatic) [31]. Studies have reported association between genetic polymorphisms and degenerative LSS [32, 33]. Developmental LSS is usually observed in individuals with achondroplasia [34], and has known genetic causes [35]. Studies have also suggested that genetic factors could be influencing early manifestation of canine LS [36, 37]. This genetic predisposition of canine LS has been predominantly accounted for by lumbosacral transitional vertebra (LTV) – a congenital structural anomaly where the vertebra forms abnormally, usually between the last lumbar and first sacral vertebra [38, 39]. Presence of LTV can in turn cause CES due to abnormally narrow spinal canal (i.e. LS) [40]. An improved understanding of the currently unknown genomic mechanisms underlying LS in Labrador retrievers would be beneficial in designing improved diagnostic tests.

Whole exome sequencing (WES) method can explore the underlying genetic mechanisms of both Mendelian disorders and complex multi-factorial diseases [41–43]. Since coding regions constitute only about 1% of the whole genome, WES is an efficient, cost-effective and sensitive method for exploring the genomics of complex disorders (like cancer [44]; autism [45, 46]; hereditary myopathy in respiratory failure [47]; and osteogenesis imperfecta and Marfan’s syndrome [48]). Therefore WES approach was used for this initial genomic exploration study of the complex trait of LS in young Labrador retriever military working dogs without any outward signs of the condition. The exomes of a group of these dogs representing extremes of the LS phenotype (based on CT image analyses) were sequenced and annotated to identify possible exonic variants between LS negative and LS positive dogs. The findings from this exploratory study were intended to provide background data for future in-depth genetic association studies of LS in this vulnerable population of a predisposed breed.

To the authors’ best knowledge there are no previous studies that have investigated the genomics of lumbosacral stenosis (LS) in Labrador retrievers. The current study is the first attempt at understanding the genomic mechanisms underlying LS in the Labrador retriever breed with the search focused on the protein-coding exome. The genes identified here with exonic variants could serve as candidate genes for future in-depth genetic studies of canine LS in this breed. Preliminary findings suggest that one of these variants warrants further investigation as a possible genetic risk factor for lumbosacral stenosis in Labrador retriever MWDs – a missense mutation (Serine ➔ Glycine) in the TTR (transthyretin) gene.

Genome-wide association studies (GWAS) have been the prevalent approach to identify variants associated with complex traits [65, 66]. But GWAS can be a difficult approach for initial exploratory studies – they usually require large sample sizes [67], and usually are able to only detect the common small-effect variants [43, 68], instead of many rare variants with combined larger effects [69, 70]. Comparatively, whole exome sequencing (WES) studies are possible in smaller sample sizes, and are unbiased in selection of variants. Whole exome sequencing studies are also preferred due to the relative ease of availability and inexpensiveness of the commercially available exome-capture platforms [71]. Therefore WES approach was selected for the initial exploration of the genomics underlying LS in this population of MWDs of the Labrador retriever breed.

Sampling the extremes of quantitative traits (or unusual samples of qualitative traits) can further reduce the required sample size of WES studies without sacrificing statistical power [72, 73]. This sampling strategy usually involves 2 phases – initial broad discovery phase where extreme sampling in small sample size identifies significant variants/genes; followed by large scale replication/validation phase where the significant findings from the discovery phase are replicated for verification in larger sample sizes. Therefore the current study can be considered to be the initial “discovery phase” of the genomic exploration of LS in Labrador retriever military working dogs using the extreme-phenotype sampling strategy. The MWD population of this breed was ideal for sampling the “unusual” qualitative LS phenotype since the breed is at high-risk for LS and working dogs usually display an earlier than normal incidence of LS. An additional level of this sampling strategy was carried out in the study by selecting dogs with the “extremes” of the quantitative LS phenotype (based on FAR). Thus attempts were made to maximize the statistical power of the study.

A physically strenuous lifestyle (like that led by human athletes) has previously been associated with premature degeneration of the musculoskeletal system [74]. Degenerative lumbar spinal stenosis (DLSS) is a common spinal degenerative disease [75]. Even though there are studies that have reported reluctance to work as a probable clinical sign of LS in working dogs [9, 19, 30, 76], there is a lack of scientific evidence testing the contribution of working tasks themselves on the disease pathology of LS. We hypothesize that even though canine LS is most often associated with aging related degenerative changes, the physically strenuous life led by military working dogs might be a contributing factor for the early incidence of a degenerative condition like LS in young athletic dogs. Our study focused on the Labrador retriever breed alone, but we propose a similar mechanism could be at play with LS in young MWDs of other breeds as well. Gene-environment interaction could also have had an effect on the phenotypic expression of a complex trait like LS, which in turn could have resulted in misleading genotype-phenotype associations. To determine the true effect of genetics alone, the effect of external variables needed to be minimized [66]. This was a benefit of using a single breed to explore the genetics underlying LS since selecting dogs from multiple breeds could have introduced genetic variation unrelated to LS disease pathology. Military working dogs housed at the same military base share a highly uniform life (extremely similar diet, exercise/training regimen, climate, healthcare access etc.). Therefore the advantage of using MWDs as the study subjects was the removal of various external/environmental sources of variation that could have lead to incorrect genotype-phenotype association in LS.

Previous studies of canine LS have reported certain biases with regards to the relationship between qualitative LS status and multiple demographic factors such as age, gender and body weight [9, 77–79]. Lumbosacral stenosis usually affects older dogs more than younger ones (aging related degenerative LS is more common form of LS), males more than females (faster growth rate, larger body weight and more popular choice as working dogs), and heavier dogs more than dogs that weigh less (male dogs tend to be heavier than females, so the gender and body weight trends coincide with each other). The current study can be split into two parts: 1st part – comprising the entire study population (N = 40, 7 LS negative, 33 LS positive); and 2nd part – sub-study of the dogs chosen for exome sequencing (N = 8, 4 LS negative, 4 LS positive). In 1st part of the study, the association between demographic factors and LS status mostly agreed with previous reports except with the gender ratio. There were roughly equal number of males and females in both LS negative and LS positive groups. And though the affected dogs were on average older than unaffected dogs, the difference was very small. In comparison, the association between demographic factors and LS statuses of the dogs in second part of the study consistently agreed with findings in previous studies. The gender bias was obvious since all 4 LS negative dogs were females and all 4 LS positive dogs were males [9, 77–79]. Dog coat color displayed no obvious trends in either parts of the study. Even though there was a slight overrepresentation of black dogs in the LS positive group in 1st part, this trend was not observed in 2nd part (WES). The average age as well as the average body weight of the affected dogs was more than that of the unaffected dogs in both parts of the study (in agreement with previous reports). Similar to a previous CT phenotyping study of LS [29], no correlation could be detected between LS status and low back pain status. Small size of each group (in both parts of the study) prevented statistical testing of the associations between demographic factors and LS statuses of the dogs.

Three of the 4 LS negative dogs selected for exome sequencing were related to each other – a dam and 2 offsprings from the same litter. The familial relationship reduced the essential sample size of the LS negative or control group in the exome sequencing study from 4 to 2, i.e. the 3 genetically related dogs sharing the same bloodline representing “dog 1” and the unrelated dog as “dog 2”. This could be a possible source of bias in the results. However, variant selection criterion for LS negative dogs (homozygous for the “wild” allele i.e. “a/a”) should have ensured that the reduced sample size of the LS negative group did not affect the variant sequence identity. Genetically related study subjects are a common issue in genetic studies where sampling is done in genetically isolated populations [80]. Analyses factoring in this effect would have been desirable, but they were beyond the scope of the current study since complete pedigree information of all 40 dogs in the study was not available (due to procurement of dogs from multiple vendors).

Clinical severity of LS was not graded in the current study. The definition of clinical severity is subjective, and based on qualitative and clinical assessment it can vary. Therefore for the purposes of the current study, the dogs were just given a binary assignment for LS – negative or positive, at each of the 8 vertebral locations assessed. The subjective assessment of LS severity made by practicing veterinary radiologists remains the gold standard for clinical applications. However the fat area ratio (FAR) variable is a quantifiable measure of LS [29] that could potentially be used for quantitatively grading the severity of LS. This potential LS severity grading system using FAR could also be applied to future quantitative trait loci (QTL) mapping studies, but the novel grading system needs further development and validation before it can be implemented in research setting.

Out of 110,980 total variants called in all 8 samples, only 252 met the study selection criteria (LS negative: homozygous for wild allele; LS positive: heterozygous/homozygous for variant allele). Even though Ensembl’s VEP could not identify any genes with high impact variants, it was able to identify 10 genes with moderate impact variants, and 24 genes with low impact variants (total = 34 genes). The manual method identified 82 exonic variants in a total of 33 protein-coding genes. Thus VEP was able to identify variants in one additional gene than the manual method. Three exonic variants could have possible associations with LS in Labrador retrievers based on findings in previous studies – a moderate impact variant in Transthyretin (TTR); and two low impact variants, one each in Folate Receptor 2 (FOLR2) and Ubiquitin Specific Peptidase 9 X-linked (USP9X).

A missense mutation (preserved length but with different amino acid) was identified in the TTR gene (Serine ➔ Glycine). Transthyretin is one of at least 30 different types of amyloid proteins that can be deposited in the human body [81]. Recent studies have reported association between transthyretin-derived amyloidosis (ATTR) and human senile systemic amyloidosis (SSA) – a group of disorders involving the localized deposition of amyloid proteins in a variety of tissues [82]. A 2011 study found an association between SSA and human lumbar spinal stenosis (LSS) in the Japanese population [83]. The study tested ligamentum flavum specimens from 36 individuals with LSS. Amyloid protein deposits were found in 19 of the 36 samples, with 16 of those 19 testing positive for ATTR. Another study has reported an association between ATTR and LSS within the Swedish population [84]. This study found ATTR deposits in 25 out of 26 samples (resected bone fragments, pieces of ligament and other connective tissues from patients undergoing LSS surgery). Amyloid protein deposits in the brain has been well investigated for several years because of the role it plays in Alzheimer’s disease and Down’s syndrome [85]. Amyloid deposits are also found in connective tissues like ligaments, tendons and cartilages especially in the joints (knee, hip, vertebrae) of elderly individuals [86–88]; but the biological significance of this type of amyloidosis is unknown [84]. Previous reports suggest that similar to LSS, ATTR is an aging-related degenerative condition usually observed in individuals older than sixty, and more commonly in males than in females. The ATTR condition is usually also accompanied by cardiomyopathy and carpal tunnel syndrome. Therefore, studies indicate that TTR could be a possible candidate gene for human LSS; and that TTR could also be a possible candidate gene for canine LS in Labrador retrievers.

A low-impact variant found in FOLR2 could also be a possible candidate gene for both human LSS and canine LS in Labrador retrievers. This gene was previously detected in macrophages in the synovial fluid of patients suffering from osteoarthritis [89], a condition well documented to be present concurrently with both human LSS and canine LS. Another low-impact variant that could be associated with LS was found in USP9X gene – a X-linked gene that escapes X-inactivation in mammalian females [90], resulting in females having twice the dose of the gene product than males. The sex-specific trend of canine LS (males affected almost twice as much as females) could be due to LS being a X-linked condition, with USP9X gene product playing a protective role in LS disease pathology. However, it should be noted that both FOLR2 and USP9X variants were reported by VEP as low impact, i.e. most likely “harmless” to protein functionality. Therefore the associations of these genes with LS could be circumstantial only, and insignificant to LS disease pathology. An important finding in this study was the absence of any exonic variants in the human LSS candidate genes. The authors acknowledge that the sample size of this study was too low to completely rule out the involvement of human LSS candidate genes in canine LS. Another possible reason for not completely ruling out the role of human LSS candidate genes in canine LS is that these genes could be associated with degenerative canine LS instead of the (suspected) idiopathic canine LS observed in current study population. The authors suggest that the young Labrador retriever MWDs in this study could be affected by idiopathic LS instead of degenerative LS since they are too young for normal degenerative changes commonly associated with aging to set in.

A possible limitation of this study is that the findings might not be generalizable to other breeds of military (or other) working dogs, or even non-working dogs of the same breed (Labrador retrievers). Future genetic studies in varying canine populations (other breeds of MWDs, other working groups of Labradors and other breeds, non-working Labradors and other breeds) accounting for known covariates (like age, sex and diet) in their study designs need to be carried out to understand the true role of genetic factors on LS. Future studies investigating the role of genetically related individual dogs in presentation of LS and the etiology of LS presented within the young MWD population would be of interest to the military as well as other agencies that procure and/or breed working dogs. Therefore it would also be desirable to have future LS genetic studies in genetically related dogs that include and factor in complete pedigree information in their analyses. Future studies with large sample sizes investigating the association between LS and demographic factors such as age, gender and body weight also need to be carried out to identify any possible relationships. Sequencing studies remain more informative than traditional GWAS for study of complex traits. Whole genome sequencing (WGS) are even more informative than whole exome sequencing (WES) studies since WES only explores the exome and many non-protein coding regions have previously been implicated to be biologically significant in complex traits [91]. However until WGS becomes cheaper, WES remains the cost-effective option for exploratory sequencing association studies of complex traits. Even though extreme-phenotype sampling gives higher power compared to random sampling in smaller sample sizes, the strategy is limited by its inability of being generalizable to the general population [71]. Additionally if the complex trait is influenced by multiple loci, the power of detecting small-effects loci is decreased [92]. Therefore large scale WES studies would be desirable that can investigate the genetics of LS without the extreme-phenotype sampling strategy. Association of variants identified in sequencing studies with the complex trait of interest is not definitive, so follow-up replication studies (comparison with GWAS, targeted genotyping of candidate genes, large scale unbiased re-sequencing) in larger samples are needed to verify significance of the variants identified in the previous discovery phase [71]. But similar to discovery phase studies, replication studies also cannot signal causality. To understand the actual role of the variants on the disease mechanism, and/or to identify causation, further molecular/cellular experiments (like expression QTL and in-vitro protein assays) need to be carried out.

In conclusion, lumbosacral stenosis is a major concern for the U.S. Military since it is a leading risk factor for early retirement in MWDs, especially due to delayed detection. Better diagnostic methods are needed that can detect LS early, so that therapeutic interventions can be effective. As current study findings suggest, LS is not limited to older dogs. Thereby suggesting the presence of idiopathic instead of degenerative etiology of LS in at least this population of dogs (young Labrador retriever MWDs). The current study is the first breed-specific genetic study of LS. The exomes of 8 young MWDs (4 LS negative and 4 LS positive) of this breed were investigated. No high-impact variants were identified, but 17 moderate and 70 low impact variants were. One moderate-impact variant was detected in a gene that could be a possible “candidate” gene for future in-depth genetic studies of LS since it codes for a protein that has previously been associated with human LSS – TTR (transthyretin). No exonic variants were detected in any of the previously reported human LSS candidate genes. However, the findings of the current study should be interpreted with caution and no genes should be ruled out yet. These preliminary findings are a result of an exploratory study that was based on a very specific population of dogs – young military working dogs of Labrador retriever breed. Further investigations are necessary to validate the findings of this study in larger samples of young Labrador retriever MWDs. Further studies are also needed to test whether the current study findings could be generalized to other breeds of dogs – both working as well as non-working dogs. Findings of this study are solely intended to provide a starting-point for future in-depth genetic studies of LS in dogs.