Evolution of the Bovine TLR Gene Family and Member Associations with Mycobacterium avium Subspecies paratuberculosis Infection

Fisher, Colleen A.; Bhattarai, Eric K.; Osterstock, Jason B.; Dowd, Scot E.; Seabury, Paul M.; Vikram, Meenu; Whitlock, Robert H.; Schukken, Ynte H.; Schnabel, Robert D.; Taylor, Jeremy F.; Womack, James E.; Seabury, Christopher M.

To elucidate gene-specific departures from a strictly neutral model of molecular evolution, we used Tajima's frequency distribution test (D statistic) [43] , as applied to the discovery panel samples (all cattle from 31 breeds vs B. t. taurus), and evaluated the significance of the observed values (D) via coalescent simulation (Table 4 ). Departures from neutrality were detected for TLR3, TLR8, and TLR10. However, the direction of the deviation was not uniform across all three loci ( Table 4 ), suggesting that disparate modes of evolution (i.e., selection) may have influenced genetic diversity within these genes, and that there may be differences among cattle lineages (Table 4 , TLR10). For both TLR3 and TLR8, a significantly positive Tajima's D reflected an excess of moderate frequency alleles, whereas a large negative value for TLR10 (B. t. taurus) reflected an overabundance of rare, low frequency variants consistent with purifying selection [30] . Therefore, it is important to note that although a significant nonrandom trend toward benign or tolerated AA substitutions was detected across all investigated loci, the underlying reason for this functional and/or selective constraint appears to be fundamentally different between some gene family members (i.e., TLR3, TLR8 vs TLR10). Notably, we observed at least one moderate frequency AA substitution that was predicted to impact protein function in both TLR3 and TLR8 (Table 3) , whereas all AA substitutions predicted to impact protein function in TLR10 were detected at very low frequencies (Table 3) . To further investigate the overall magnitude and origin(s) of the most significant deviations from a strictly neutral model (Tajima's D; pyrosequencing discovery panel; Table 4 ), we used Fu's F S statistic [44] to estimate the probability of observing a number of haplotypes less than or equal to that predicted in our samples for TLR3 (B. t. taurus); TLR3-1 (B. t. taurus), and TLR8 (all cattle; B. t. taurus). For TLR3, we recognized that the inability to phase all individuals in the pyrosequencing discovery panel could lead to the absence of some low frequency alleles, thus potentially driving both Tajima's D and Fu's F S toward larger positive values. Consequently, we calculated Fu's F S and Tajima's D for TLR3 (B. t. taurus) and TLR3-1 (B. t. taurus) using the following approach: 1) Both test statistics were first calculated only for sires that could be phased with best-pairs probabilities$0.90, as depicted in Table 4 ; and 2) If a significant result was achieved in this analysis, we then added the taurine haplotypes with phase probabilities,0.90 into our analyses (D; F S ) by choosing the best haplotype pairs reconstructed for each sire. For Fu's F S , only TLR8 displayed unequivocal evidence for a departure from neutrality (All cattle F S = 10.2712, P,0.01; B. t. taurus F S = 10.296, P,0.01), with levels of significance that withstood conservative correction for multiple testing (correction = a/n locus-specific tests, 0.05/2 = Minimal P#0.025). For Tajima's D, inclusion of the best TLR3 haplotype pairs for sires with phase probabilities,0.90 resulted in very similar test statistics (TLR3 B. t. taurus D = 3.6034, P,0.001; TLR3-1 B. t. taurus D = 3.4895, P,0.002; Table 4 ), with levels of significance that endured correction for multiple testing (0.05/8 = Minimal P#0.00625).