The low-dose IV LPS infusion model has been used extensively in equine research to study the pathophysiology of early gram negative septicemia and test the efficacy of potential drug candidates.8,9,13,24−30 A strong inflammatory response to low-dose IV LPS is typically associated with clinical signs of systemic inflammation as well as increased TNFα serum concentrations by ≥ 50%.13,24,25 However, up to 20% of horses fail to respond to this model.9 One possible explanation is endotoxin tolerance, which is characterized by a diminished pro-inflammatory response to endotoxin that is recognized in animals and humans.31–33 However, endotoxin tolerance typically subsides within 3 weeks of endotoxin exposure in horses,34 making it an unlikely explanation in clinically normal horses. This study aimed to investigate TLR4 and MD2 variations among horses with different responses to LPS.
DNA sequencing of TLR4 and MD2 from horses representing multiple breeds with different TNFα dynamics after experimental IV LPS infusion revealed four haplotypes and two AA sequences for each gene. There were no associations between any haplotypes or single variant and TNFα response categories (high, moderate, low) for either gene. Similarly, when TNFα baseline concentrations were categorized as high, moderate, or low, there were no associations with haplotypes for either gene. The only significant association was between the individual baseline TNFα concentration and a missense MD2 variant. However, this result was based on a single horse (H1) which had the highest baseline TNFα concentration and was the only horse to carry the MD2 missense variant (the MD2-2 AA sequence). H1 was a Paint, and while one other Paint horse (H3) also had a higher baseline TNFα concentration, the third Paint (H9) in the study had an undetectable baseline TNFα concentration; indeed, H9 had the lowest baseline value in the study. Given the small sample size, conclusions regarding the significance of this association cannot be made.
Each haplotype combination was modeled as a 3D structure, using the known human TLR4 and MD2 structures as templates. This allowed visualization of each AA-changing variant’s position in relation to the LPS binding site, the TLR4-MD2 interface, and the dimer-dimer interfaces, with the aim of elucidating whether any of the AA changes might alter function and explain the variable TNFα values. There are six lipid chains in each LPS molecule, five of which are buried inside the hydrophobic pocket. The remaining chain is exposed to the surface of MD2 and forms hydrophobic interactions with TLR4.20 Since LPS binds both TLR4 and MD2 proteins, variations in their critical LPS-binding domains or in overall protein structural integrity could lead to a significant effect. Complexes 1, 2, and 4 were not associated with any significant differences in TNFα values (baseline or response), although some of the AA changes could potentially result in alteration to attraction/repulsion at the quaternary protein structure level. For example, the variant Lys325 residues in Complex 1, which has two identical TLR4-1 subunits, are located just 5.25Å apart. Lysine is a positively charged residue, and this small distance could create charge repulsion not observed in the wild type proteins. Similarly, also in Complex 1, the variant Arg106 residues in MD2-1 is close to TLR4’s Asp209, which likely creates a strong ionic attraction between the positively charged Arg106 and the negatively charged Asp209. Such small changes, though not statistically significant in the present work, may still contribute to overall protein stability.
Complex 3 contained one copy of the MD2-2 haplotype and therefore the MD2 missense variant, which was the only variant significantly associated with TNFα baseline concentration. In the dimer-dimer complex, one MD2 protein had Arg106 and the other had Thr106. One monomer (consisting of one TLR4 subunit and one MD2 subunit) having Arg106 is predicted to be stabilized by this residue interacting with TLR4’s Asp209 (as shown in supplementary Fig. 5C), while the other monomer having Thr106 is predicted to have no substantial interactions with the neighboring TLR4 protein (Fig. 5B). Threonine, a polar, uncharged AA is smaller than arginine, which is a basic AA residue, and the closest TLR4 residue (Phe263) was 4.42Å distant. It is unclear how this single missense change translates into the high baseline TNFα concentration in this single horse (H1). It is possible the significant association is spurious and this missense change is well-tolerated. In fact, the European Variant Archive (https://www.ebi.ac.uk/eva/), under accession datasets PRJEB28306 and PRJEB9799, has whole genome sequence from 94 horses representing over two dozen breeds; nine of these horses are heterozygous for the same MD2 missense variant (minor allele frequency = 0.049). This demonstrates that the variant is not overly rare among horses, although the impact of this variant on MD2 protein function, response to LPS, and resulting TNFα concentration response, are still unknown. Ultimately, it is important to remember that H1 was clinically normal at baseline for all other parameters, yet very high for baseline TNFα concentration; the present protein modeling work has unfortunately not provided any meaningful additional explanation for this finding.
Baseline TNFα concentrations varied widely among the horses in this study, despite their normal clinical examinations, CBCs, and biochemical profiles at the time of TNFα measurements. While a Kendall’s tau test comparing baseline TNFα concentrations with response TNFα concentrations (each categorized as high, moderate, or low) suggested a significant negative correlation (P = 0.033), the data set was very small. Further, all horses (n = 3) that had a high baseline TNFα concentration were categorized as low responders, and every horse (n = 4) that was categorized as a high responder started with a low baseline TNFα concentration. It is possible that subclinical inflammation was present in horses with “moderate” or “high” TNFα concentrations, but to date, large-scale screening studies of TNFα concentrations in healthy horses have not been performed, and there is a paucity in the literature of correlations between TNFα concentrations and clinical status in this species.35 It is also possible that initial high TNFα concentrations might prevent a further measurable response, rendering the response automatically to the “low” TNFα response category. Unexpectedly, two horses (H10 and H12) had decreased TNFα concentrations one hour after LPS infusion, the meaning of which is uncertain. Until more is known, it seems reasonable to continue to screen horses for TNFα concentrations, but perhaps not to use this metric as either inclusion or exclusion criteria.
The major limitation to this study is the small sample size, with only 15 and 11 horses sequenced for TLR4 and MD2, respectively. While horses are an excellent clinical and research model for endotoxemia, they remain an expensive animal to house, decreasing the available sample size. Another limitation is the lack of measurement of other cytokines; TNFα concentrations may not be the best marker for LPS efficacy. Another limitation is the absence of generated sequences in the study horses for introns or untranslated regions of TLR4 or MD2. Many non-coding variants can still impact gene and protein expression; these were not assessed in the present study.
The overall goal of this study was to determine the relationship between gene variants and TNFα profile. Future work should examine a larger cohort of horses and investigate associations between numerous inflammatory cytokines (e.g. TNFα, IL1β, IL6, and IL8) and TLR4 and MD2 variation. It is also possible that variants in other related genes (e.g., LPS binding protein, CD14) or other genetic predispositions are involved. Comparison among different breeds of horse might also show varying TLR4-MD2 complexes that could affect differences in LPS binding. This work combines for the first time TLR4 and MD2 genotyping and protein structural predictions with equine TNFα phenotypes and informs future work to further elucidate these molecular relationships.