Very few canine polymorphisms can be found in databases compared to human polymorphism data. These few polymorphisms in databases are usually predicted by in silico methods. For this reason, first a bibliographic review was carried out for SNP candidates with high allele frequency and having a putative miRNA binding site. Previous studies also follow this study design when analyising dog polymorphisms being miRNA binding sites 29. It is of importance to confirm the existence of in silico predicted SNPs in databases to be able to create a comprehensive database for dogs similar to other organisms.
It has recently been demonstrated that miRNAs play a crucial role in the regulation of protein synthesis, and polymorphisms located in the 3’UTR might alter the binding efficiency of miRNAs contributing to the fine tuning of this regulation. Several miRNA studies have been performed in the last couple of years in humans, mostly in clinical subjects 30–32 but there are some available data with behavioral aspects 33–35 as well. Behavioral traits and their connection with miRNAs were also investigated in our laboratory previously 16,36,37.
Although some miRNAs were proven to have a binding site in the 3’UTR of different canine genes 22, the functional effect of these miRNAs are still mostly unidentified, and data are available practically only in clinical settings 38–40. The miRBase database (http://mirbase.org/) provides the sequence of 502 dog miRNAs, which are annotated in the dog genome (ftp://mirbase.org/pub/mirbase/CURRENT/genomes/cfa.gff3). A publication of an improved canine genome build, canFam3.1 41 offers a chance to improve this knowledge: previously annotated dog miRNAs were validated and new miRNA sites were identified based on canFam3.1 genome assembly and RNA sequencing data 42. According to our best knowledge, we are the first to report the functional effect of a SNP influencing the efficiency of microRNA binding in canine WFS1 gene associated with behavioral traits.
The 3’UTR of the WFS1 gene containing the rs852850348 SNP was investigated by in vitro luciferase reporter system. We observed that the expression of the reporter protein was significantly lower in the presence of the “A” allele compared to the “G” allele at the rs852850348 locus. This finding was of especial interest as the studied SNP was not located in the predicted seed sequence of the cfa-miR-8834a miRNA. This result confirms that although nucleotides at position 2 to 8 are crucial in miRNA–mRNA interaction, further bases of the mature miRNA can also influence this effect 43.
Please note, that the rs852850348 SNP site flanked the seed region. “A” allele is the major allele of rs852850348 SNP (http://www.ensembl.org/). Therefore, carrying the mutated alelle (“G”) does not result in high mRNA-miRNA interaction. Our results confirm the hypothesis that miRNAs usually negatively regulate protein expression in case of carrying wild type genotype 44. However, the construct containing “G” allele showed also reduced luciferase activity compared to the control. This result suggests that one base pair change does not cause complete mRNA-miRNA dissociation. A similar finding was published by our laboratory previously 45. The rs1046322 SNP in the 3’UTR of the human WFS1 gene was also examined as putative miRNA binding site polymorphism. Its functional effect was demonstrated in a luciferase reporter system: the minor “A” allele showed lower repression compared to the major “G” allele, if co-expressed with miR-668 16, similar to our results. Nothwithstanding, the rs852850348 SNP might be in the seed region of another, unknown miRNA. Therefore, the underpining molecular mechanism needs further studies.
Although in silico sequence alignment suggested the putative role of cfa-miR-1838 as well, it could not be confirmed by our luciferase reporter system. Neither significant decrease when applying this miRNA, nor any difference between the effect of the two allelic variants could be detected. These results suggest that this miRNA does not have a role in regulating WFS1 expression.
Our association study suggests that the food possessivity behaviour might have genetic background in some breeds. Border collie dogs carrying GG genotype were more food possessive than dogs with AA genotypes. Food possessivity is often linked to possessive aggression, involving growling, baring the teeth, snapping, or biting when the dog possesses an object (food, bone, toy) and someone (family, stranger, animal) approaches and/or attempts to take it away 46,47. However, food possessivity was unrelated to aggression in our test, as only a few dogs growled or attempted to bite. The lack of association of bone possessivity to possessive agression in our study could be due to the low sample size, the unfamiliarity of the situation for the dogs, or the specific characteristics of the Border collie breed.
Wolframin is an endoplasmic reticulum membrane protein. Its function is not completely understood yet, however, its malfunction induces endoplasmic reticulum stress 18. Endoplasmic reticulum stress activates the unfolded protein response and mediates the pathogenesis of psychiatric diseases when the damage occurs in the hippocampus, the amygdala and striatum 48, where wolframin is greatly present 17. Therefore, food possession in dogs could be partially due to the polymorphism of the WFS1 gene.