Bioinformatics analysis of the structural and evolutionary characteristics for toll-like receptor 15

Phylogenetic analysis

The coding sequences of vertebrate TLR genes were retrieved from NCBI (GenBank) and Ensembl databases (Benson et al., 2013; Cunningham et al., 2015). All the partial sequences (<1,800 bp) and pseudogene sequences were excluded. A multiple sequence alignment of the TLR proteins was constructed with MAFFT (FFT-NS-i, BLOSUM62) (Katoh & Standley, 2013), and a phylogenetic tree was calculated from it with PhyML using the LG substitution model and four substitution rate categories (Guindon et al., 2010; Le & Gascuel, 2008). Branch support was calculated with an approximate likelihood ratio tests (aLRT SH-like) (Anisimova & Gascuel, 2006). The phylogenetic tree was visualized with MEGA 6 (Tamura et al., 2013).

Structural elements analysis

SignalP, SMART, and TMHMM were used to identify the signal peptide, ectodomain, transmembrane region, and intracellular TIR domain of TLR15 (Krogh et al., 2001; Letunic, Doerks & Bork, 2015; Petersen et al., 2011). The delimitation of each LRR module in the TLR15 ectodomain was determined with the LRRfinder software (Offord, Coffey & Werling, 2010). The delimitations of chicken TLR15 were used as the reference for the remaining species.

Structural modeling

I-TASSER was used to model the structure of the ectodomain and TIR domain of TLR15 based on a threading approach (Yang et al., 2015). I-TASSER is a hierarchical method for protein structure prediction. Structural templates were first identified from the PDB by the multiple-threading program LOMETS; then, full-length models were constructed by iterative template fragment assembly simulations. The modeled structure was displayed by PyMol (Schrödinger, LLC).

Protein–protein docking analysis

The two modeled structures of chicken TLR15 TIR domains were submitted as target 1 and target 2, respectively, to PRISM protein–protein docking server (Baspinar et al., 2014; Tuncbag et al., 2011). PRISM predicts possible interactions, and how the interaction partners connect structurally, based on geometrical comparisons of the template structures and the target structures.

Residue conservation analysis

A multiple sequence alignment of the TLR15 proteins was constructed with MAFFT. The evolutionary conservations of amino acid residue positions in the TLR15 sequences was estimated by using ConSurf algorithm (Ashkenazy et al., 2010). The JTT substitution matrix was used and computation was based on the empirical Bayesian paradigm. The conservation scale was defined from the most variable amino acid positions (grade 1) which were considered as rapidly evolving, to the most conservative amino acid positions (grade 9) which were considered as slowly evolving. The sequence and modeling structure of chicken TLR15 were used to show the nine-color conservation grades.

Codon-based analyses of positive selection

A multiple sequence alignment of the nucleic acid sequences of the TLR15 genes, based on their codons, was established with TranslatorX (Abascal, Zardoya & Telford, 2010). The ratio of the number of nonsynonymous substitutions per nonsynonymous site (dN) to the number of synonymous substitutions per synonymous site (dS), dN/dS, an indicator of the selective pressure acting on a protein-coding gene, at the TLR15 locus and the corresponding 95% confidence interval were calculated with the Datamonkey web server (Delport et al., 2010).

The SLAC, FEL, REL, and FUBAR methods were implemented in Datamonkey to explore the evidence of positive selection acting on the individual codon of the TLR15 sequences (Kosakovsky Pond & Frost, 2005; Murrell et al., 2013). To minimize the overestimation of the positively selected codons, codons with p values < 0.1 for SLAC and FEL, with Bayes Factor > 50 for REL, and with posterior probabilities > 0.9 for FUBAR were considered as candidates to be under positive selection.

TLR15 is phylogenetically closely related to family 1

We obtained all the known vertebrate full-length TLR gene sequences and constructed their phylogenetic relationships based on maximum likelihood method (Fig. 1A). TLR15 clusters into an individual clade and is closely related to the TLRs in family 1 (TLR1, TLR2, TLR6, TLR10, TLR14, TLR18, TLR25 and TLR27) on the phylogenetic tree. This is consistent with the early finding (Roach et al., 2005).

We identified 57 completed TLR15 genes through the phylogenetic analysis (Fig. 1B, Table S1). These TLR15 sequences are derived from 54 avian and 3 reptilian species (Burmese python, Chinese alligator, and American alligator). TLR15 exists in these species as a single gene copy in each species’ genome. Also, the phylogenetic tree showed that TLR15 has evolved following the phylogeny of species. These results indicate that TLR15 is possibly subjected to strong selection constraints.

TLR15 ectodomain possesses an intact asparagine ladder

TLR15 generally include 860∼880 amino acids. Motif prediction showed that TLR15 included the signal peptide, ectodomain, transmembrane region, and intracellular TIR domain. The structure of chicken TLR15 ectodomain was modeled with threading approach (Fig. 2A). The estimated TM-score (0.72 ± 0.11) for this modeled structure showed that it was acceptable. TLR15 ectodomain contains a LRRNT, a LRRCT, and 19 LRR modules, which is identical to the number of LRR modules of the known TLRs in family 1. The LRR3 module of TLR15 is quite long, for example, the LRR3 of chicken TLR15 has 99 amino acids that far exceed the lengths of common LRR modules (∼24 amino acids).

The LRR module is characterized by a conserved “LxxLxLxxNxL” motif on the concave surface. Among them, the conserved asparagine plays important role in maintaining the overall shape of TLR ectodomain by forming a continuous hydrogen-bond network (called as asparagine ladder) with backbone carbonyl oxygens of neighboring LRR modules (Kang & Lee, 2011). The asparagines can be substituted by other amino acids that are able to donating hydrogens, such as threonine, serine, and cysteine. We analyzed the asparagine ladder in the ectodomain of TLR15. The results showed that the asparagine ladder was intact throughout the TLR15 ectodomain, suggesting that TLR15 possesses an intact and continuous hydrogen-bond network (Fig. 2A).

Although both TLR15 and the TLRs in family 1 contain 19 LRR modules, the known crystal structures indicate that the asparagine ladders of the ectodomains of TLRs in family 1 are broken in the middle (Fig. 2B). TLR15 ectodomain more approximates to those TLRs with intact asparagine ladders in the ectodomains, for example, TLR3, TLR5, TLR7, and TLR21 (Table S2). Therefore, the ectodomain of TLR15 is obviously structurally different from the TLRs in family 1.

The convex surface of TLR15 ectodomain has a highly evolutionarily conserved region

We calculated the evolutionary conservation scores of all residue positions of TLR15 based on the phylogenetic relationships between homologous sequences (Fig. 3). The conservation scale was defined from the most variable amino acid positions, which were considered as rapidly evolving, to the most conservative amino acid positions, which were considered as slowly evolving. Further, the mean evolutionary conservation score of each module in TLR15 was also calculated (Fig. 4).

The results showed that the majority of amino acid positions of intracellular TIR domain were highly evolutionarily conserved. However, the average evolutionary conservation of different LRR module in the ectodomain had large differences. Most of LRR3 module was composed of highly variable positions. The loop of convex surface in LRR9 module also includes a number of highly variable positions. The average evolutionary conservations of LRR18, LRR19, and LRRCT modules are very high, this could be adapted to their function participating in the dimerization of C-terminal region of TLR15 ectodomain. The average evolutionary conservations of LRR10-13 modules are also very high, indicating that they are possibly related to the function of the ectodomain sensing pathogens.

We further displayed the evolutionary conservations of amino acid residue positions in the ectodomain of TLR15 using the modeled structure (Fig. 5). The asparagine ladder with high evolutionary conservation is located on the ascending lateral surface, which contains loops connecting the β-strand of the concave surface to the convex surface and participates in the dimerization of TLRs (Botos, Segal & Davies, 2011). Compared to the ascending lateral surface, its opposite side, the descending lateral surface, includes fewer evolutionarily conserved positions. Interestingly, we found a highly conserved region on the convex surface of TLR15 ectodomain, which is exactly located on the LRR11 module. For chicken TLR15, the highly conserved region on the convex surface of its LRR11 module ranges from 397th to 406th amino acid residues (SIVELPEWFA). The high conservation of this region across species might suggest that it is involved in ligand recognition.

TLR15 TIR domains are able to form homodimers

We further studied TLR15 intracellular TIR domain that is responsible for the signal transduction. The structure of TIR domain (from Pro706 to Thr848) of chicken TLR15 was modeled with threading approach. This modeled structure mainly contains a central three-stranded parallel β-sheet surrounded by five α-helices (Fig. 6A).

Protein–protein docking analysis showed that TLR15 TIR domain was able to form the homodimer with itself (Fig. 6A). The dimeric interface is mainly formed by the residues from the BB-loops and αC-helixes. The fifteen residues of TLR15 TIR domain were predicted to participate in its homodimerization, the five of them are located on the BB-loops and six of them are located on the αC-helixes (Fig. 6B). Cys777 in αC-helix at one monomer (chTLR15) probably formed disulfide bond with the same residue from the second molecule (chTLR15’). The amino acid residue positions involved in the interaction of TIR homodimer are highly evolutionarily conserved among 57 TLR15 sequences (Fig. 6B). The fourteen of fifteen residue positions involved in the homodimeric interaction have an evolutionary conservation score greater than or equal to the grade 8.

We compared the predicted dimerization of TLR15 with those of the TLRs in family 1. The multiple sequence alignment showed that there were 30% identical amino acid residues among the TIR domains of chicken TLR15, human TLR1, human TLR2, human TLR6, and human TLR10 (Fig. 7). The known five functionally critical residues in TLR2 TIR domain are conserved among different TLRs (Tao et al., 2002), which correspond to Phe743, Pro745, Cys777, Leu781, and Lys812 in chicken TLR15. The homodimer interface of TLR6 TIR domains is considered to be formed mainly by the residues from αC-helixes (Jang & Park, 2014), the six of its eight interaction residues was also found in the interaction of TLR15 TIR domain homodimer. The homodimer interface of TLR10 TIR domains was reported to mainly contain residues from the BB-loops and αC-helixes (Nyman et al., 2008). The nine of its eleven interaction residues was also detected in the interaction of TLR15 TIR domain homodimer. Comparatively speaking, the interaction mode of TLR15 TIR homodimer is more similar to that of TLR10.

Twenty-seven sites under positive selection were detected for TLR15

The estimated dN/dS value for the TLR15 locus is 0.318 (95% CI [0.308–0.328]), suggesting a clear excess of synonymous over nonsynonymous substitutions at TLR15 locus. However, numerous sites under positive selection were found in TLR15 with the codon-based maximum likelihood methods (Table 1). The SLAC, FEL, REL, and FUBAR methods identified statistically significant positive selection in 24, 29, 22, 14 amino acid sites across TLR15s, respectively. To identify robust candidate sites under positive selection, only 27 codons with evidence of positive selection identified by at least two methods were considered, which account for 3.1% of the total sites in TLR15.

These 27 sites under positive selection belong to the most variable positions identified by evolutionary conservation analysis (Fig. 3). The 24 of 27 sites under positive selection are located on TLR15 ectodomain. Among them, the LRR3 module contains seven sites under positive selection. The proline-rich loop on the convex surface of LRR9 module includes two sites under positive selection. There is also a site under positive selection (407th) closely neighboring to the highly evolutionarily conserved region on the convex surface of LRR11 module.