Shar-Pei dogs have been companion animals for centuries within China where they were commissioned to guard and hunt, and to sometimes serve as fighting animals. At the beginning of the communist era dog ownership was highly taxed and the breed was brought close to extinction. A few Chinese Shar-Pei dogs were exported to the United States in the early 1970's and Shar-Pei descending from this limited number of animals have undergone strong selection for a wrinkled skin phenotype and heavily padded muzzle and are called the “meatmouth” type (Figure 1A–1C) and have now found global popularity. The ancestral Shar-Pei, referred to as the “traditional” type Shar-Pei, still occurs and it presents with a less accentuated skin condition (Figure 1D). The major constituent of the deposit in the thickened skin is hyaluronan or hyaluronic acid (HA). HA is a large, multifunctional, linear, negatively charged, non-sulfated glycosaminoglycan of the extracellular and pericellular matrices. It is composed of repeating disaccharides and is widely spread throughout epithelial, connective and neural tissues ^[1], ^[2]. The biological role of HA depends on its size, location and equilibrium between synthesis and degradation ^[1]–^[3]. Meatmouth Shar-Pei show two- to five-fold higher serum levels of HA compared to other breeds ^[4], allowing us to propose the term hyaluronanosis, a definition also used for a comparable human condition ^[5]. HA is synthesized at the plasma membrane by three HA synthases, HAS1, HAS2 and HAS3, with HAS2 being the rate limiting-enzyme ^[6]. HAS2 is overexpressed in dermal fibroblasts of Shar-Pei compared with other canine breeds ^[7] suggesting a regulatory mutation as causative for hyaluronanosis. HA is deposited throughout the skin of Shar-Pei, often in microscopic lakes and grossly evident vesicles, leading to the formation of thickened skin folds around the head and tibiotarsal (hock) joints (Figure 1E). Almost all Shar-Pei seem to be affected by hyaluronanosis, however the extent varies among individuals and adults exhibit less skin folds and hyaluronanosis than puppies. Strong selection by breeders for dogs who retained their skin folds into adulthood has altered the phenotype of the breed to the more commonly heavily wrinkled meatmouth type.

Meatmouth Shar-Pei also suffer a strong predisposition to an autoinflammatory disease, Familial Shar-Pei Fever (FSF), which clinically resembles some human hereditary periodic fever syndromes, such as Familial Mediterranean Fever (FMF) ^[8]. Both diseases are characterized by seemingly unprovoked episodes of fever and inflammation and both FMF and FSF present as short (12–48 hour) recurrent bouts of high fever, accompanied by localized inflammation usually involving major joints (especially the tibiotarsal joints). Patients with FMF or Shar-Pei with FSF can suffer episodes as often as every few weeks, but in the interim seem symptom free. However, since acute phase reactants may endure between episodes, a subclinical state and chronic autoinflammation may persist (Linda Tintle unpublished data). As a secondary complication, the chronic state puts human patients, as well as affected Shar-Pei dogs, at risk of developing reactive systemic AA amyloidosis and subsequent kidney or liver failure ^[8], ^[9]. In Shar-Pei, the fever episodes are typically more frequent during the first years of life and the percentage of affected dogs is very high, estimated to be 23% in the US in 1992 ^[9].

In order to find candidate loci for the breed-specific phenotype (hyaluronanosis), known to be under selective pressure, we screened the genome for signatures of selective sweeps. These sweeps can be recognized as long chromosomal segments with a low degree of heterozygosity within populations ^[10]. Using 50,000 single nucleotide polymorphisms (SNPs) distributed throughout the dog genome, the level of heterozygosity in windows of ten consecutive SNPs was compared between a set of Shar-Pei (n = 50, all from the US, Table S1) and the average of 24 other canine breeds (n = 230). On four chromosomes (Cfa 5, 6, 13 and X) the reduction in heterozygosity in Shar-Pei was greater than 4-fold the average of control breeds (Figure 2A). The strongest signal of reduced heterozygosity appeared within a 3.7 Mb stretch on chromosome 13 (CanFam 2.0 Chr13: 23,487,992–27,227,623) (http://genome.ucsc.edu/) near the HAS2 gene, where almost complete homozygosity was observed in Shar Pei (Figure 2C). Here the reduction in heterozygosity was greater than 10-fold in Shar-Pei and several smaller regions showed complete homozygosity. The same region was confirmed to show high levels of homozygosity when the analysis was repeated in 37 additional Shar-Pei dogs sampled from Spain (Table S1) and was overlapping a sweep region reported by others for this breed ^[11]. The strong signal, together with the known function of HAS2 and its aberrant expression pattern in Shar-Pei, made this region an obvious candidate for the mutation causing the wrinkled skin phenotype (hyaluronanosis).

In parallel, we performed a genome-wide association study to map the susceptibility locus for FSF, using Shar-Pei strictly classified as FSF affected (n = 24, classification code FSF+A and FSF+, described in Materials and Methods) and unaffected (n = 17, classification code H+, described in Materials and Methods). Five SNPs were significantly associated (best SNP p_raw = 7.0×10⁻⁷, p_genome = 0.005 based on 100,000 permutations; software package PLINK http://pngu.mgh.harvard.edu/~purcell/plink^[12]), all on chromosome 13 (CanFam 2.0 Chr13: 22.4–30.7 Mb, Figure 2B). After correcting for putative stratification, two outlier cases were removed (Figure S1) and the same SNPs, forming the same signal of association remained (best SNP p_raw = 2.3×10⁻⁶, p_genome = 0.01; Table S2) with a genomic inflation factor of 1.2. When the association signal and the sweep signal were compared they appeared interspersed, so that individual SNPs were either part of homozygous regions or showed association with FSF (Figure 2C). It was therefore difficult to determine exactly where the strongest association fell, as variation is required to detect association.

Targeted sequence capture technology was used to further investigate the sweep signal and to search for the hyaluronanosis causative mutation. We resequenced 1.5 Mb around and upstream of our candidate gene, HAS2 (CanFam 2.0 Chr13: 22,937,592–24,414,650) in four Shar-Pei (two meatmouth type with high serum HA levels and two traditional type) and three control dogs from other breeds. The obtained sequences were mapped to the boxer reference sequence providing at least 5X coverage for 96–98% of the resequenced region in each individual. The targeted region also included the large intergenic noncoding RNA, HAS2 antisense (HAS2as; Table S3) which has been proposed as a negative post-transcriptional regulator of HAS2 mRNA ^[13]. After masking repetitive sequences we identified ∼670 indels and ∼1,500 SNP in each dog (Table S4) as well as two overlapping duplications in the Shar-Pei (Figure 3A). Nine mutations (eight SNPs and one indel) located in conserved elements as well as two SNPs possibly regulating transcription, were selected for further investigation due to their unique pattern in the sequenced Shar-Pei dogs. Additional genotyping in Shar-Pei and dogs from other breeds (Tables S1, S5) showed these mutations were not specific to Shar-Pei and the variants were subsequently excluded as causative.

The two duplications were named after the Shar-Pei type in which they were first identified. The “meatmouth” duplication was the larger fragment, 16.1 Kb (CanFam 2.0 Chr13: 23,746,089–23,762,189) with breakpoints located in repeats (a SINE at the centromeric end and a LINE at the telomeric end) and individual copies separated by seven base pairs (Figure 3B). The “traditional” duplication was 14.3 Kb (CanFam 2.0 Chr13: 23,743,906–23,758,214) and was identified in the two Shar-Pei with a less accentuated skin phenotype (Figure 3B). We first examined the duplications via Southern blot with control breeds (n = 2), traditional (n = 2) and meatmouth Shar-Pei (n = 6) (Figure 3C). As the digest cut outside and within both duplications, we were able to observe the absence of the variants from control breeds and separate restriction patterns in traditional and meatmouth type Shar-Pei. Interestingly, one meatmouth dog contained both duplication types (Figure 3C lane 6 and confirmed by PCR across break points, data not shown). Two copy number assays were developed to quantify these elements. The first (CNV-E) measured only the meatmouth duplication whilst the second (CNV-748), detected both the traditional and meatmouth duplications. Copy number analysis was estimated as the relative fold enrichment (ΔΔCt) between an amplicon within the duplication and one outside the duplication in a housekeeping gene. Assay CNV-E was run on 90 Shar-Pei and 73 dogs from 24 other breeds (Table S1) and assay CNV-748 on a subset of 44 Shar-Pei and 14 dogs from other breeds. Assay CNV-748 demonstrated that both the traditional and meatmouth duplications are unique to the Shar-Pei breed (Figure 4 and Figure S2).

We used the results of both assays to search for a relationship between Familial Shar-Pei Fever (FSF) and either meatmouth copy number (Assay CNV-E), traditional copy number (the normalized difference between CNV-748 and CNV-E) or total traditional+meatmouth copy number (Assay CNV-748). Shar-Pei dogs were strictly classified as affected by FSF (n = 28, FSF+A and FSF+) or unaffected by FSF (n = 16, H+). The most significant association was found when only the meatmouth copy number was considered (p<0.0001, Figure 4) although a weaker association with total copy number (p<0.01) was also seen. The observed association between fever and meatmouth copy number, despite the very high homozygosity in this region, strongly suggests that a high copy number is not just a genetic marker for FSF but is causally related to the development of disease.

Of the 153 dogs analyzed with the meatmouth copy number assay, 31 Shar-Pei and 18 control animals also had serum measures of HA available. No clear association was detected between HA levels and copy number (Figure S3), however the mean HA level in Shar-Pei with ≥ six copies was 905±403 ug/L (n = 21), whilst Shar-Pei with fewer copies had a mean concentration of 770±494 ug/L (n = 12) and control breeds had HA serum levels of 206±145 ug/L (n = 19). Interestingly, the three traditional Shar-Pei dogs had serum HA levels between 73 and 266 ug/L, which fell within the normal range ^[4].

The link between copy number and the expression of HAS2 and HAS2as was examined on a smaller scale using dermal fibroblasts cultured from six separate meatmouth Shar-Pei. The expression of both genes was calibrated against the Shar-Pei with lowest copy number (CNV estimate  = 5) and both genes showed an increasing trend of expression with copy number (Figure 5). These data suggest that a regulatory element for HAS2 is located in the duplicated region, however the interpretation of the HAS2as result is less clear. A single study of a human osteosarcoma cell line demonstrated that the expression of two isoforms of HAS2as were able to reduce HAS2 expression, and so these mRNAs may act as regulators of HA production ^[13]. Our data could indicate that HAS2as expression is also influenced by a regulator element in the duplication, or that HAS2as is up-regulated in response to HAS2 levels. If either of these scenarios were true, it is possible that if RNA expression were measured at multiple time points we would see temporal HAS2 repression. It could also be that the interaction between canine fibroblast HAS2 and HAS2as does not mirror the human system and that the canine antisense mRNA is non-functional. At present our results must be considered as preliminary and it is clear that further exploration of the interaction between canine HAS2 and HAS2as is required.

Here we have identified a 16.1 Kb duplication located approximately 350 Kb upstream of HAS2. This is clearly a derived mutation since it occurs as a single copy sequence in other dog breeds. We postulate that this is a causative mutation associated with both hyaluronanosis and Shar-Pei fever, as the observed correlation between copy number and susceptibility to Shar-Pei fever was not expected if this was a linked, neutral polymorphism. We suggest that the unique region of the meatmouth type duplication identified in Shar-Pei contains one or more regulatory elements that alter the expression of HAS2. It appears possible that as the duplication copy number increases, so does the copy number of potential enhancer elements within the duplication, likely leading to a higher expression of HAS2 and elevated HA levels, and resulting in the development of hyaluronanosis in this breed. We propose a scenario whereby the traditional duplication arose de novo in the traditional type of Shar-Pei causing a milder skin phenotype. This event made the region unstable and allowed the second meatmouth duplication to occur. Breeders subsequently selected the meatmouth duplication as a higher copy number enhanced the phenotypic effect in appearance. However, it is not yet possible to say whether the meatmouth duplication first occurred at low frequency in the Chinese Shar-Pei population and quickly rose during breeding in America, or if the mutation occurred spontaneously during breed expansion in the West.

Tandem duplications are notoriously unstable and may show copy number variation due to unequal crossing-over, as is clearly illustrated by the copy number variation of a 450 Kb duplication associated with dominant white colour in pigs ^[14]. The meatmouth Shar-Pei duplication adds to the list of copy number variants (CNVs), which affect phenotypic traits in domestic animals (e.g. dominant white in pigs ^[14], gray color in horses ^[15], the hair-ridge in Rhodesian ridgeback dogs ^[16], and pea-comb in chicken ^[17]), several of which are linked not only to the desirable trait but also to disease. Interestingly, all of these except pea-comb, represent novel duplications derived from single copy sequences. This is in contrast to most reported CNVs in humans, which are mainly benign and represent expansions or contractions of duplicated sequences ^[18].

Although we failed to find a significant correlation between serum HA levels and copy number, this does not exclude our proposed hyaluronanosis scenario. Difficulties in correlating fluctuating serum levels of HA with other clinical and biomedical parameters have also been reported in many human studies, where no or only weak correlations were observed ^[19], ^[20]. We have shown that the 16.1 Kb duplication appears only in meatmouth Shar-Pei, a breed type that has elevated levels of HA compared to both traditional Shar-Pei and other breeds, and that copy number correlates with a breed-specific syndrome associated with excessive HA deposition and the over expression of a HA synthesizing gene. Because HA is primarily a component of the extracellular matrix, serum measurements may only broadly reflect total body HA.

Hyaluronan can bind to several cellular receptors (e.g. CD44, RHAMM and layilin), however it is the interaction between CD44 and HA which acts as a biological regulator, differentially modulating the cellular microenvironment in response to homeostatic versus inflammatory conditions ^[21]. Alterations in the balance between native high molecular weight HA versus fragmented HA may result in activation of innate immunity. HA has been linked to sterile inflammation as an endogenous response molecule to sterile tissue injury ^[21]. Shorter fragments of HA can be generated by environmental insults such as sterile trauma ^[22], reactive oxidative species (ROS) ^[23], or pathogenic hyaluronidases, and it is these low molecular weight fractions which can become pro-inflammatory danger associated molecular pattern (DAMP) molecules ^[22], ^[24] mimicking microbial surface molecules.

Using a mouse model, Yamasaki and colleagues ^[25] showed that HA can interact with the cell through two separate pathways that culminate in the release of IL-1β, which together with IL-6, is one of the main promoters of fever. In the first route, CD44 bound HA is degraded at the plasma membrane by hyaluronidase-2 (HYAL2) prior to endocytosis and further cleavage by lysosomal hyaluronidase-1 (HYAL1). The resultant small intracellular oligosaccharides of HA activate the NLRP3 inflammasome, a multiprotein complex consisting of the NLRP3 scaffold, the ASC adaptor and caspase-1 ^[26]. In the second arm, the CD44-HA complex activates toll like receptors 2 and 4 (TLR2 and 4), leading to intracellular IL-1β mRNA transcription and the formation of pro-IL-1β. Activation of the NLRP3 inflammasome by HA oligosaccharides allows cleavage of this pro-IL-1β by caspase-1 and subsequent release of IL-1β. The NLRP3 inflammasome is present in the cytosol of many cells including monocytes, macrophages and mast cells, and has been implicated in the pathogenesis of numerous autoinflammatory diseases in humans including the cryopyrin-associated periodic syndromes which result from mutations in NLRP3/CIAS1^[26].

The actual role of excessive HA in Shar-Pei needs to be investigated further. Shar-Pei may experience exogenous fragmentation of their over-abundant HA from sterile or pathogenic trauma. This, plus endogenous degradation of excessive native HA, may contribute to induction of recurrent episodes of fever and inflammation. Acute fever events in Shar-Pei respond rapidly to dipyrone, a potent antipyretic and analgesic pyrazolone, which has been demonstrated to inhibit IL-1β induced fever ^[27]–^{[29 and Linda Tintle unpublished data]}. It is therefore not surprising that the strong selection on the hyaluronanosis phenotype, with increased levels of cutaneous HA, may predispose Shar-Pei to autoinflammation, potentially contributing to other pathologies seen in this breed. One such example is renal medullary amyloidosis. Histopathologically, kidneys of Shar-Pei in renal failure have multifocal non-suppurative tubulointerstitial nephritis with fibrosis. Medullary amyloidosis predominates and glomerular deposition, although consistent, is highly variable in its extent ^[8], ^[30]. The renal medulla is naturally HA rich and enhanced renal interstitial HA accumulation can be coupled to inflammatory responses, such as ischemia-reperfusion injury, transplant-rejection, tubulointerstitial inflammation and diabetes ^[31]. In addition, Shar-Pei are prone to mast cell disease including mast cell tumors ^[32], ^[33]. The binding of HA to CD44 has been shown to play a critical role in regulation of murine cutaneous and connective tissue mast cell proliferation ^[34]. As the CD44-HA interaction may modulate local immune responses through regulation of mast cell functions ^[35], excessive HA and its subsequent damage and degradation may play a role also in the Shar-Pei breed’s predilection for allergic skin disease and other mast cell driven inflammation.

This study suggests that HAS2 dysregulation can trigger a periodic fever syndrome in dogs and therefore it will be relevant to examine the approximately 60% of human fever patients who currently have unexplained disease. Previously, the role of hyaluronan in sterile inflammation has focused on HA signaling and degradation; for example a deficiency of hyaluronidase causing mucopolysaccharidosis type IX in humans has some autoinflammatory features ^[36]. However by directly implicating HAS2 in inflammation, we suggest that a reexamination of genes further up the biosynthetic pathway, such as those involved in HA synthesis and polymerization is called for. In addition, the canine mutation appears regulatory in nature and therefore regulators of HA should be also be included in a broader scope pathway analysis of human patients with unexplained autoinflammatory disease.

Finally, this study illustrates how copy number variations can shape phenotypic traits and how strong artificial selection for certain phenotypic traits may not only affect the desired trait but also the health of the animal.

The authors KL-T, MO, and LT have filed a patent for development of a genetic test.

This work was supported by the Swedish Research Council; FORMAS; the Swedish Research Council for Environment, Agricultural Sciences, and Spatial Planning; the Swedish Foundation for Strategic Research; in part by the Intramural Research Program of the National Institute of Arthritis and Musculoskeletal and Skin Diseases and the National Human Genome Research Institute of the National Institutes of Health; the Chinese Shar-Pei Charitable Trust and the European Commission (FP7-LUPA, GA-201370). KL-T is the recipient of a EURYI award from the European Science Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

We thank all the dog owners, breeders, and breed clubs worldwide that have supported this study and contributed samples. We thank veterinarians and other colleagues for their help with samples including Jérôme Abadie, Laila Irene Baek, Kikka Posti, Hannes Lohi, Anne-lise Juncker, Barbara LaVere, Patricia and Harry Roach, and Jeff Vidt. We also thank Kathleen Long and Lazara Cuza for Shar-Pei photos. We thank Marie Lindersson and Kristina Larsson at the SNP Technology Platform in Uppsala (Sweden), the Broad Institute Genetic Analysis platform (US), as well as Freyja Imsland and Snaevar Sigurdsson for technical assistance with Illumina sequencing and bioinformatics.