Inflammatory diseases such as mastitis, inflammatory bowel disease, metritis and laminitis are economically important diseases for the dairy industry. Mastitis is the most common amongst them, alone accounting for losses in excess of $2 billion to the US dairy industry [1]. A number of therapeutic, prophylactic and management strategies have been proposed to minimize this complex disease. However, a widely accepted complimentary strategy is based on improving the host genetics through selective breeding. Since it is difficult to use a direct index to measure the mastitis phenotype, milk somatic cell score (SCS) is often used as an indirect index to select animals for breeding [2]. Milk SCS is a log₂ score of the milk somatic cell count and has been genetically correlated with clinical mastitis (r = 0.7) [3]. It has been demonstrated in sheep that intramammary infections occur less frequently after one generation of breeding for low SCS[4]. Breeding strategies in dairy cattle are similarly based on selection for low estimated breeding values (EBVs) for SCS [5].

The etiology of mammary inflammatory disorders is diverse and may include infections caused by bacteria, parasites, viruses and fungi. In order for an effective host immune response to occur against a wide variety of pathogens, the host must possess receptors that recognize conserved molecular patterns associated with different classes of pathogens. These are referred to as pathogen associated molecular patterns (PAMPs). Receptors that recognize PAMPs are collectively referred to as pattern recognition molecules (PRMs) [6]. The PRMs are expressed widely by different cells including monocytes, granulocytes, dendritic cells and epithelial cells [7]. PAMP recognition leads to the secretion of cytokines and chemokines by the epithelial cells that recruit effector cells (neutrophils and monocytes) of the immune system to the site of infection where they contribute to the host inflammatory immune response and subsequent acquired immune response.

Phagosomal toll-like receptor-2 (TLR2) is involved in recognizing PAMPs associated with Gram-positive bacteria. Murine studies for example, have demonstrated that TLR2 is in involved in the early recruitment of neutrophils in response to intraperitoneal challenge with either Staphylococcus aureus, or peptidoglycan (PGN) from S. aureus [8]. Ruminant studies have demonstrated that TLR2 gene (designated as 'TLR2') expression occurs in dermal and gut-associated tissues [9], and is highly induced during mastitis caused by S. aureus [10]. Bovine TLR2 was recently radiation hybrid mapped to Bos taurus autosome (Bta) 17 [11].

Caspase Recruitment Domain 15 (CARD15), also known as Nucleotide Oligomerization Domain 2 (NOD2), is a cytosolic protein capable of initiating inflammation following PAMP recognition. The gene encoding bovine CARD15 (designated as 'CARD15') is located on Bta18. It was previously categorized as a member of the CATERPILLER family, but was reassigned to the phylogenetically conserved NLR (NACHT-LRR) protein family [12]. CARD15 shares a common tripartite domain structure with other members of this family. The tripartite domain consists of a carboxy (C)-terminal leucine rich repeat (LRR) domain; a central NACHT (NAIP CIITA HET-E and TP1) domain, and an amino (N) – terminal domain that is composed of two CARD domains. The LRR domain is involved primarily in the recognition of bacterial peptidoglycans (PGN), whereas the central NACHT domain facilitates self-oligomerization and has ATPase activity. The CARD domains are known to interact with CARD containing serine/threonine kinase Rip2 (known as RICK), via homophilic CARD-CARD interaction; this leads to the activation of NF-κB [13].

The major portion of the PGN recognition system in mammals is constituted by CARD15 along with NOD1 and TLR2 [14]. CARD15 is involved in intracellular recognition of muramyl dipeptide (MDP), the minimal bioactive structure of PGN, which is common to the cell wall of both Gram-positive and gram-negative bacteria. Therefore, CARD15 acts as a general sensor of bacterial infection [15, 16]. A recent study suggests that NOD2 might be involved in sensing of PGN motifs of S. aureus, after its phagocytosis [17]. There are several reports of interactions between CARD15 and other PRMs, especially TLR2 [18–20]. This demonstrates the importance of CARD15 in signalling events associated with recognition of different PAMPs. Polymorphisms, particularly single nucleotide polymorphisms (SNPs), within the genes coding for different receptor proteins may impair the ability of certain individuals to respond properly to infections [21]. In the case of TLR2 and CARD15, they have been identified as susceptibility loci for different inflammatory bowel diseases in humans [22–25]. It is possible that SNPs within these genes influence other inflammatory diseases such as mastitis in cattle. Therefore, understanding the genetic variation underlying PRMs might help livestock breeders to identify and select animals with enhanced resistance to mastitis for breeding programs. The purpose of this study was to identify SNPs in the bovine TLR2 and CARD15 and to evaluate the association of these SNPs with SCS and other production traits in a population of Canadian Holstein bulls.

TLR2 and CARD15 play a role in the initiation of inflammatory and immune responses to bacterial infections. A number of studies have reported that SNPs in PRMs of different species play an important role in contributing towards disease susceptibility. Therefore, the main objective of this study was to identify SNPs in the TLR2 and CARD15 and to estimate the extent of associations between these SNPs and SCS, a trait directly related to udder inflammation. In the current experiment, the DNA pool used for the purpose of SNP detection was comprised of DNA samples from 40 animals with extreme EBVs for SCS. The exons, flanking intronic sequences, and promoter region were targeted for SNP detection. Therefore it is unlikely, that all SNPs in the population were identified.

Although no SNPs were identified in TLR2, four SNPs were identified in CARD15; only one SNP was found in the coding region (c.3020A>T) of this gene. This SNP, located in exon 12, is a non-synonymous SNP coding the terminal LRR domain of CARD15 receptor. While allele 'A' at this position codes for glutamine, allele 'T' codes for leucine. The association of the T allele with increased SCS and decreased udder depth, which predisposes animals to mammary infections, indicates that changes in the composition of the LRR domain of CARD15 may contribute to disease susceptibility. The terminal LRR domains, as in CARD15, are common to different PRMs and are responsible for PAMP recognition. There are several reports of polymorphisms in the human and mouse LRR coding gene segments that contribute to differential binding to several bacterial components [15, 24, 27]. In humans, CARD15 variants have been hypothesized to alter bacterial component recognition by altering the structure of LRR domain or the adjacent region [23]. It is possible that SNP c.3020A>T may compromise protein functionality by altering the conformation of the binding site in a similar fashion. This could contribute to the development of inflammatory disorders and warrants further investigation. The association between this SNP and udder depth might be an indirect result of the genetic correlation that exists between SCS and udder depth (rg= -0.26) [28].

Strong associations were also observed between the analyzed CARD15 SNPs and production traits. Such associations may be a result of linkage between these SNPs and other genes on the same chromosome having a significant effect on these production traits. Significant QTLs have been found on Bta18 for all the traits used in the analysis. While QTLs for SCS and udder composite index exist close to the location of this gene (about 230 kbs apart)[29, 30] on the chromosome, QTLs for milk yield, protein yield and fat yield were situated farther away (>1,500 kbs apart) [30–32].

Haplotype reconstruction revealed all four possible CARD15 haplotypes to be segregating in the sampled bulls. The second most frequent haplotype (Hap21) differed significantly from the most common haplotype (Hap12) with respect to its effect on SCS EBVs. In agreement with the results of the allele substitution analyses, the linear effect of Hap21 carrying the allele 'T' at SNP position c.3020A>T was significantly different from Hap12. However, we did not see any difference between the linear effects of Hap22 and Hap12 for SCS. Strong associations were observed between Hap22 and milk and protein yields (p < 0.0001 and p = 0.0007 respectively). Since, Hap22 carries alleles 'T' and 'C', both of which are associated with increased milk and protein yields, these results were in agreement with the allele substitution analysis. It is interesting to note that the most common haplotype, Hap12, carries alleles 'A' and 'C' at positions c.3020A>T and c.4500A>C respectively, and is beneficial not only for reducing SCS but also for increasing production. Thus selection for Hap12 in the population seems promising.

In conclusion, four SNPs in CARD15 in a sample of Canadian Holstein bulls were discovered. Statistical analyses revealed that SNP c.3020A>T was associated with EBVs for SCS and udder depth and milk and protein yields, while SNP c.4500A>C was only associated with milk, fat and protein yields. The most common haplotypes for these two SNPs in the population differed significantly for their effect on SCS. Moreover, the most common haplotype carried alleles at both loci that are favourable for reducing SCS and increasing production EBVs. This implies that these two SNP, together with other gene polymorphisms, may be potentially used for genetic selection for mastitis resistance and production. The findings of this study indicate that SNP c.3020A>T is a candidate for further detailed studies on its functional mechanisms.