A Genetic Variation of Lipopolysaccharide Binding Protein Affects the Inflammatory Response and Is Associated with Improved Outcome during Sepsis

The LPS binding protein (LBP) is an acute-phase protein primarily secreted from the liver with substantial concentrations also released from pulmonary and gut epithelial cells (1–4). It belongs to a larger family of phospholipid transfer proteins (PLTPs), which transfer lipid derivates and lipopeptides throughout the body influencing lipid homeostasis (5, 6). These functions are associated with common diseases like vascular and heart disease but are also known to be involved in innate immunity (7, 8). It has recently been shown that LBP particularly facilitates the transfer of multimers of LPS to its sensing receptor consisting of CD14, TLR4, and MD2, initiating the inflammatory response (9–11). Furthermore, it plays a role in detoxification of LPS by transferring LPS into lipoproteins (4, 12–15). In addition, LBP is also able to bind lipopeptides originating from both Gram-negative and Gram-positive bacteria and to mediate their proinflammatory effects (16–18). Recently, it has been postulated to be associated with bacterial translocation in the gut, potentially adding function to this molecule by sensing/scavenging bacterial material entering the body through a deranged gut wall (19). Finally, it has been shown that LBP, in a similar manner as other members of the genetically related family of PLTPs, such as cholesterol ester transfer protein and PLTP, can bind and transfer phospholipids. This mechanism most likely is important for LPS transfer from micelles to HDL particles and into membranes but potentially may represent a general lipid transport system (20, 21).

In critically ill patients, LBP levels are markedly increased severalfold (22). LBP effects within the host depend strongly on its concentrations in serum, in which lower concentrations are responsible for effective initiation of LPS sensing. However, high acute-phase concentrations of LBP exert a rather inhibiting effect on the immune response (23–25). Underlying mechanisms of this phenomenon are not yet completely understood but may involve facilitation of LPS internalization brought about by LBP (26).

Hereditary factors have been studied in the context of sepsis, and consequently, a growing number of genetic factors was identified that contribute to risk and the course of severe forms of sepsis (27, 28). Single nucleotide polymorphisms (SNPs) of the innate immune system have been described that influence individual responses to invading pathogens (29).

For the LBP gene on chromosome 20, at least two studies were able to show association of mutations with altered inflammatory responses (30, 31), and haplotype studies were able to associate genetic alteration to outcome (32, 33). In a very recent comprehensive review of this year, it was concluded that genetic variations of LBP alter the risk for inflammatory complications (34). For the purposes of this study, nonsynonymous LBP SNPs were deemed as appropriate candidates to assess structural changes in the molecule in combination with functional and clinical data. Other LBP SNPs identified with regard to infections and sepsis were mostly promotor variants or synonymous SNPs (34). In this study, we compared patients carrying different variants of LBP regarding a coding SNP (rs2232618) at position 436, leading to an amino acid exchange from phenylalanine (Phe) to leucine in the LBP molecule (31, 35, 36). We previously studied a different SNP (rs2232613) leading to a change at amino acid position 333 in a comparable experimental setting with patient data from the same cohort also partly included in this study. These results showed an altered molecular response and an influence on clinical outcome (37). Patients carrying this SNP were excluded from the primary analysis. The SNP further evaluated in this study has been shown by others to be associated with altered sepsis prevalence and mortality in Chinese trauma victims (38, 39).

The SNP was initially thought to cause an exchange from leucine to Phe, but recent sequence analysis revealed that the cystine to thymidine exchange is more frequent in Western European inhabitants. Therefore, the c1341t genotype (nucleotide sequence: TTC leading to Phe) was considered to be the common variant. As a continuation of a previous study, we show in this study how functional properties of the LBP molecule are altered by this SNP (37). We assessed binding of several ligands to the variants of LBP. As molecular patterns originating from bacteria, we used LPS and lipopeptides. Phosphatidylethanolamine was used as a nonbacterial ligand to assess transfer ability to LBP as has been done in previous experiments (21). Furthermore, we used the murine RAW 246.7 cell line and stimulated the cells with the ligands in the presence and absence of serum from patients carrying the rare variant of LBP. We used this as a macrophage model for LPS recognition and measured the in vitro release of a proinflammatory cytokine (TNF). In this way, we were able to simulate the structural and biochemical influence of this SNP on the basis of the recently discovered molecular structure of the whole LBP molecule using different bioinformatic tools (40). We also examined this functionally relevant SNP in a cohort of 363 surgical intensive care unit (ICU) patients from European ancestry and associated it with the susceptibility to postoperative infections and outcome. We hypothesized that the variant of LBP studied in this study would affect its ability to confer its main biological function. We sought to determine whether structural changes in the molecule could explain this behavior. Finally, we tried to determine whether these findings of altered structure and function would influence the clinical course of patients with postoperative infections carrying this SNP.

For binding assays, either LPS (Escherichia coli O111:B4 “smooth” LPS, Sigma-Aldrich, Taufkirchen, Germany), Pam₂Cys [Pam₂Cys-Ser-(Lys)₄ × 3 trifluoroacetic acid, EMC Microcollections, Tübingen, Germany], Pam₃Cys (Pam₂Cys-Ser-Lys₄, EMC Microcollections), or 3-sn-PE (3-sn-phosphatidylethanolamine, Sigma-Aldrich) were placed on an ELISA plate in a concentration of 30 µg/ml each in carbonate buffer (pH 8.2). Plates were incubated with sera containing LBP of individuals carrying either the common or the rare genotype in increasing concentrations (0.2, 0.39, 0.78, 1.56, 3.13, 6.25, 12.5, and 25%, respectively). These sera were diluted 1:2 from a maximal concentration of 25%. We used sera from different individuals with a mean LBP concentration of 17 ng/ml for common variant sera and 15.8 ng/ml for sera of heterozygous individuals. Following incubation, the photometric reaction was stopped with 1 M H₂SO₄. OD of the bound ligands was photometrically measured at 450 nm in an ELISA reader (Photometer Spectra Fluor Plus, Tecan, Crailsheim, Germany).

Murine macrophage cells (RAW 264.7, Leibniz-Institut DSMZ, Braunschweig, Germany) were stimulated with 1, 10, and 100 ng/ml of two different LPS types (smooth LPS [E. coli 0111:B4] and “rough” LPS [Salmonella minnesota Re595, Sigma-Aldrich]) and two different lipopeptides (Pam₂Cys [Pam₂Cys-Ser-(Lys)₄ × 3 TFA, EMC Microcollections] and Pam₃Cys [Pam₃Cys-Ser-(Lys)₄, EMC Microcollections]) in addition to serum of individuals with different genotypes (concentrations of either 1, 2, and 5%). Concentrations of TNF were determined at 4 h using ELISA (DPC Biermann, Bad Nauheim, Germany). All measurements were performed in duplicate, resulting in four values per experiment.

The sequence information about human LBP SNP rs2232618 was retrieved from dbSNP (http://www.ncbi.nlm.nih.gov/SNP/). MutPred was used to determine the effect of the amino acid substitutions (41). To investigate the secondary structures of common and rare variants, PSIPRED was used (42). I-Mutant 3.0 (http://gpcr2.biocomp.unibo.it/cgi/predictors/I-Mutant3.0/I-Mutant3.0.cgi) was applied for the determination of free energy change (ΔΔG) and protein stability changes owing to single-site mutations (43). The crystal structure of the LBP protein was obtained from Protein Data Bank (PDB identifier [ID]: 4M4D) (44). This was used as a template for generating the three-dimensional structure of common and rare variant human LBP. The three-dimensional structures were generated and energy minimized with SWISS-MODEL and Swiss-PdbViewer, respectively (45, 46). For the purpose of validating these structures, PROSA, ProQ, and SAVES (https://saves.mbi.ucla.edu/) were used (47, 48). Finally, the modeled structures of human LBP common and rare variant types were visualized with PyMOL. Docking of LPS with variants of human LBP were carried out using PatchDock and Firedock software (49, 50). The software DUET (http://biosig.unimelb.edu.au/duet/) was applied to study the role of mutation on the protein structure, using a combination of support vector machine and machine learning algorithms. The implication of mutation on the domain core stability was predicted using ELASPIC (51). ProFunc was used to figure out the alterations between common and rare variants of the proteins regarding functional clefts and cavities (52). ConSurf estimated the conservation profile of the key residues related to functional regions in the structures (53). The differences in the conserved residues of the major functional clefts were studied. CASTp was employed to examine the prevalence of functional pockets in the modeled structures (54). The solvent-accessible area and changes because of mutations were determined with GETAREA (55).

We also subjected the sequence information from the L variant of LBP SNP rs2232613 described in (37) to MutPred, PSIPRED, and I-Mutant 3.0, respectively, and analyzed the modeled structures of this variant with DUET, ELASPIC, ProFunc, ConSurf, CASTp, and GETAREA.

The frequency of the LBP SNP F436L (rs2232618) was determined by real-time PCR assays with subsequent melting curve analysis using the LightCycler 1.5 (Roche Diagnostics, Mannheim, Germany). Sequenced controls representing different genotypes were included in each reaction. Oligonucleotides used for genotyping were as follows: primers, forward: 5′-TTTGCTTTTCCCAAGCGTT-3′ and reverse: 5′-GAGCCCTGTTTTCCAAGTCC-3′; and probes, sensor: 5′-CTATTACATCCTTAACACCCTCTAC-FL-3′ and anchor: red 5′-640-CCAAGTTCAATGGTAAGAATCACTGTGG-3′. One reaction volume of 20 μl contained 2 μl 10× PCR Buffer, 2 mM MgCl₂, 125 μM NTPs, 5 U Taq polymerase, 3 μg BSA, primers at 0.5 μM (LBP forward/reverse), fluorescence probes at 0.2 μM each, and 5–20 ng DNA. On the LightCycler 1.5 platform, PCR parameters were as follows: initial denaturation at 95°C for 4 min, 40 cycles of denaturation (95°C for 1 s), annealing (56°C for 10 s), and extension (72°C for 8 s) with subsequent melting curve analysis: 1 cycle at 95°C for 10 s, 40°C for 30 s, followed by an increase of temperature to 80°C at a slope of 0.1°C/s. All oligonucleotides were manufactured by TIB MOLBIOL (Berlin, Germany). PCR reagents were obtained from Rapidozym (Berlin, Germany).

The local ethics committee of the Charité – Universitätsmedizin Berlin approved this clinical study (AA3/03/45). DNA testing was permitted by a signed broad written consent including DNA testing before surgery. All steps were performed according to the Helsinki declaration. Statistical analysis was carried out after anonymization of the patients’ data. Definition of sepsis (systemic inflammatory response syndrome, sepsis, severe sepsis, and septic shock) was based on published criteria (56). We did not reclassify the patients according to new sepsis definitions because recent publications show that matching is not reliable (57, 58).

The patient cohort from this study was previously described (59). Infections were defined as described by the clinical classification for nosocomial infections of the National Institutes of Health (60). A total number of 363 patients fulfilled the inclusion criteria. Severity of disease was assessed by the Simplified Acute Physiology Score on admission (61). Patients were followed up until discharge from the hospital. Patient DNA was extracted from blood or tissue specimens collected prior to surgery and were examined for the LBP SNPs.

Then, 675 individuals from a group of 692 volunteers (healthy individuals and blood donors) served as controls for frequencies of these SNPs. Characteristics of these individuals were published recently (62). All individuals consented to genetic testing. Blood donors were anonymized, and the healthy individuals gave written informed consent. Either blood or oral swabs were used for DNA extraction using QIAamp DNA Mini Kit (QIAGEN, Hilden, Germany).

Contingency tables were statistically tested with the χ² test or Fisher exact test where appropriate for differences in frequencies. Odds ratios (OR) were determined using the χ² test. Differences in numerical data were compared with unpaired t test or Mann–Whitney U test. For statistical analysis, the IBM SPSS Statistics 20.0 software package (IBM) and the Prism 8 Software package (GraphPad Software) were used. A two-tailed p < 0.05 was considered statistically significant.

We hypothesized that the F436L variant of LBP was associated with altered binding of bacterial ligands. To study this, we incubated plates containing different bacterial ligands to the binding of progressively diluted serum of individuals with different genotypes. For all tested ligands, we were able to record reduced binding in carriers of the F436L mutation of LBP as shown in (Fig. 1. Very low concentrations showed only little influence that was not statistically significant. When reaching increased concentrations (3.13% up to 25% of serum), carriers of the mutation bound ligands to a lesser degree.

This was detectable for ligands originating from Gram-negative bacteria (LPS 0111:B4) for serum concentrations of 6.25 and 12.5%, in which binding of LPS differed significantly between common and rare variant LBP (6.25: 0.507 versus 0.354, p = 0.048 and 12.5: 0.734 versus 0.526, p = 0.010) (Fig. 1A). This was as detectable in ligands from Gram-positive bacteria, although to a lesser degree. In this study, binding of the lipopeptide Pam₂Cys was markedly reduced in all experiments (0.78%: 0.169 versus 0.155, p = 0.026; 1.56%: 0.231 versus 0.179, p = 0.002; 3.125%: 0.248 versus 0.187, p < 0.001; 6.25%: 0.254 versus 0.187, p < 0.001; 12.5%: 0.250 versus 0.196, p < 0.001; and 25%: 0.234 versus 0.194, p < 0.001) (Fig. 1B). Similar results were seen in binding of the lipopeptide Pam₃Cys, although only for higher concentrations of serum (6.25%: 0.291 versus 0.226, p = 0.047; 12.5%: 0.327 versus 0.245, p = 0.034; and 25%: 0.311 versus 0.21, p = 0.003). (Fig. 1C). Binding of the artificial ligand 3-sn-phosphatidylethalonamine was also markedly reduced in patients carrying the genetic variation (0.78%: 0.175 versus 0.136, p = 0.004; 1.56%: 0.274 versus 0.198, p = 0.006; 3.125%: 0.361 versus 0.251, p < 0.001; 6.25%: 0.378 versus 0.263, p < 0.001; 12.5%: 0.426 versus 0.297, p = 0.002; and 25%: 0.491 versus 0.354, p = 0.003) (Fig. 1D). An additional finding was a saturation effect seen with increasing serum concentrations. This effect was detectable for all ligands, confirming previous observations (12). We saw no influence of the genotype with this regard.

After we found an effect on binding of ligands induced by the protein changes brought about by the genetic variations of LBP, we investigated whether this would affect their ability to facilitate the LPS transfer to TLR4 and thus the release of the proinflammatory cytokine TNF. We incubated RAW cells with LPS and other ligands in the absence and presence of sera obtained from individuals differing in their LBP genotype. As expected, we found the LPS-induced release of TNF to depend on the addition of LBP-containing serum. Mean concentrations of TNF induced by the smooth LPS 0111:B4 (containing carbohydrates in addition to the lipid A core) increased markedly stronger as compared with the rough variant of LPS Re595 (see Supplemental Table I). As is shown in (Fig. 1E, the addition of LBP-containing serum induced a severalfold stronger release of TNF as compared with LPS 0111:B4 alone. In detail, 100 ng/ml LPS 0111:B4 with 1% serum resulted in a mean ± SEM of 6514 ± 667.0 ng/ml TNF with serum of rare variant LBP compared with 8655 ± 538.6 ng/ml TNF with added common variant serum (p value: 0.012; Mann–Whitney U test).

In the experiments with lower concentrations of LPS 0111:B4 (1 and 10ng/ml) with 1% serum either from common variant or rare variant, the results in individuals was lower TNF values with a similar trend between the genotypes just failing to reach statistical significance. Mean ± SEM of TNF was 4248 ± 1006 ng/ml for the rare variant and 2327 ± 626.8 ng/ml for common variant serum (p = 0.16, unpaired t test). Interestingly, adding 2 or 5% serum to the above-mentioned doses resulted in a less pronounced difference (LPS 0111:B4 10 ng/ml + 5% serum, p = 0.98; LPS B4 100 ng/ml + 5% serum, p = 0.61, unpaired t test) (Fig. 1E, Supplemental Table I). Overall, the “response curve” was flatter, with individuals carrying the rare variant.

Adding rough LPS (Re595) to the LBP-containing sera resulted in an overall lower induction of TNF and no statistically significant differences between the genotypes. Other experiments with lipopeptides (Pam₂Cys or Pam₃Cys) and LBP-containing sera with different genotypes showed an increase in TNF concentrations, but there was no statistical difference between the genotype groups (see Supplemental Table I). Of note, the number of repetitions in these experiments was low because of sparse amounts of human sera available for the different genotypes.

We assessed a potential impact of the F436L LBP genotype on protein function on a structural level by computer modeling. Three-dimensional structural models of the common and rare variant human LBP protein were generated by homology modeling based on the crystal structure of LBP recently published (37, 63). The LBP molecule is boomerang shaped with three functionally distinct parts. The first is the N-terminal part, which is believed to primarily interact with LPS (9). The C-terminal portion is potentially interacting with CD14, and this subsequently leads to TLR4 interaction and thus may initiate signaling. The groove in human LBP is a unique region differing from related proteins such as bactericidal/permeability increasing protein (21). Position 436 is buried and located inside the C-terminal domain of LBP between the A′ loop and the Phe core at the tip of the molecule. In both models, F436/L436 was positioned within an α-helix (helix B) lining the inner surface of the C-terminal phospholipid binding pocket formed by helix A and B and the β2 and β6 sheets (23). (Fig. 2A shows the three-dimensional structure of the LBP variant and the location of the F436L exchange in the groove region.

We compared hydrophobicity, surface charge, and the intactness of LBP-specific structural features such as the Phe core and the LPS binding groove formed by the A′ loop (21). The close ups and top view portrayed the nature of the interior surface of common and rare variant protein in (Fig. 2B–E.

When analyzing the inner surface of this pocket, we noted that one continuous hydrophobic channel is formed in the common variant model, similar to the murine LBP crystal structure in which both phospholipid acyl chains that were seen in the crystal could be accommodated in the channel (23). In the L436 model, however, the leucine side chain protruded into this channel, thus effectively blocking the channel and reducing the accessible size of the pocket by 50%. This means in the L436 structure, only one acyl chain could be accommodated as evident from superpositions with phospholipid molecules. We also found that the rare variant differed in size of the first major cleft from the common variant (2845.0 versus 2416.08 ų). These clefts are known to harbor functional residues (64). The highlighted alterations in surface topography may therefore cause changes in molecular function.

Molecular docking simulations show the interaction between human LBP and LPS. (Fig. 2 H–I demonstrates the differences in docking residues of human LBP with LPS in both variants in the groove region. In the common variant, Ile251 and Tyr284 were the supporting residues for a potential interaction with LPS. Interestingly, Ser224 and Arg254 residues can interact with LPS in the rare variant in addition to other supporting residues. This difference in docking residues in the common variant and rare variant could cause structural conformational rearrangements influencing molecule function. These results are supported by the observations from MutPred, I-Mutant 3.0, DUET, and ELASPIC, which indicate a loss of stability of the rare variant LBP. This is further supported by solvent accessibility analysis showing a change in polar energy, surface atoms, and buried atoms.

The results are presented in Tables I and II. An additional comparison with the rs2232613 rare variant described in a previous study is presented in Supplemental Table III.

To evaluate if the tested genotypes would potentially influence susceptibility to and course of clinical infections, we examined a cohort of patients following surgery. In this cohort, 290 of 363 patients carried the more frequent variant. In 73 patients, the LBP SNP F436L (rs2232618) was present. Of these, 69 patients were heterozygous and 4 were homozygous carriers of the mutated alleles. We found a similar distribution of frequency in the control group consisting of 675 healthy volunteers: 555 were carriers of the common variant alleles only, 112 were heterozygous, and 8 were homozygous. The resulting allele frequencies were 0.091 and 0.095, respectively, as shown in Table III. The SNP was in complete Hardy–Weinberg equilibrium in both groups (patients: χ²<0.002, p = 0.96; control group: χ²= 0.75, p = 0.39). The distribution of the SNP is in accordance with available data (http://www.ensembl.org/Homo_sapiens/Transcript/ Haplotypes?db=core;g=ENSG00000129988;r=20:38346482-3837 7013;t=ENST00000217407).

Comparison of clinical characteristics between the genotype groups revealed no differences with regard to age, gender distribution, or preexisting conditions. Overall, preadmission disease severity scores (American Society of Anesthesiologists classification and Simplified Acute Physiology Score) were equal, as shown in Supplemental Table II. After performing risk calculation, none of these conditions influenced susceptibility to infections or outcome (data not shown). Of the 363 studied patients, 202 (55.7%) developed infections in the ICU. In terms of sepsis susceptibility, no influence of the studied genotype could be found (common variant: 163/290 [56.2%]; F436L: 39/73 [53.4%]), which was also the case for the type of infection. These results are presented in Table IV. The most frequent infection in three groups was pneumonia followed by abdominal and wound infections. The genotypes had no influence on the prevalence of infections nor on the causing pathogens. The distribution of pathogens was equal in all groups. In ∼10% of the patients, there was more than one pathogen identified, and in a similar number of patients, no pathogen could be retrieved by microbiological examination. Both facts were not different between the groups and did not influence further outcome (data not shown).

As compared with the prevalence of infections, we found an association with severity of infection between the genotype groups. The F436L genotype group showed a lower risk for septic shock (OR 0.34; 95% confidence interval [CI] 0.11–1.02; p < 0.05) as compared with the common variant group. Furthermore, patients carrying the F436L allele had a lower mortality associated with overall infectious complications (2.6 versus 10.4%) or septic shock (25.0 versus 39.0%) as compared with common variant patients. The latter fact was not statistically significant (OR 0.23; 95% CI 0.03–1.75; p = 0.21 for infectious complications; OR 0.47; 95% CI 0.05–4.92; p = 0.64 for septic shock, respectively). We found no difference in the clinical course of homozygous compared with heterozygous patients.

Interestingly, we found that the previously studied P333L SNP showed almost opposite effects in most of the studied aspects (37). With the additional genotyping data for the study cohort, we were able to directly compare both rare variants. We found mortality was significantly higher in the P333L group than in the F436L group (21.2%: 7/33 versus 2.4%: 1/39; p = 0.02 tested with Fisher exact test). However, we noted differences in the rate of preexisting conditions between the genotype groups that did not reach statistical significance, except for diabetes. More details for the comparison of these variants are provided in the supplemental material (Supplemental Table IV). Overall, the genotype groups did not differ with regards to risk factors for worse outcome.

This work shows a series of phenotypic effects of a genetic variation in the LBP gene on either functional properties of the molecule as well as clinical effects. We found that ligand binding to LPS was reduced when the serum of carriers of the rare variant was added as compared with serum of individuals carrying the common variant. This might be explained by a physically impaired adherence of these molecules to the ligand binding domain of LBP. To effectively exert its function, LBP must bind to its ligands to induce CD14 and thus TLR binding and consecutively induce intracellular signal transduction (65). The experimental concept employed in this study aimed at simulating macrophage function in the presence of serum proteins that are also part of the ligand transfer like MD-2 or CD14. In our cell stimulation experiments, we found lower cytokine concentrations induced in macrophages by bacterial ligands associated with the rare variant of LBP. This was particularly the case with high concentrations of LPS and the serum of individuals carrying the rare variant. Taken together, it appears that reduced binding of LBP ligands associated with the rare variant leads to lesser pronounced induction of proinflammatory cytokines. In the early response to bacterial infection, the release of high concentrations of proinflammatory cytokines might be associated with a more pronounced inflammatory reaction. Recent sepsis definitions articulate a dysregulated or uncontrolled immune reaction as a major factor for morbidity and mortality in this setting (66). However, the host ability to combat bacteria might also be related to adequate amounts of cytokines released (67).

The function of LBP is not only restricted to the transfer from LPS or other ligands to its innate immune receptors but also has immune inhibiting properties, particularly in higher acute-phase concentrations, by scavenging LPS (23). The observed genetic changes might affect this function of the molecule as well. In our experiments, higher concentrations of LBP did not result in increased binding of the molecule. However, we also could confirm previous findings, which show that high concentrations of serum tend to scavenge LPS, but this effect was reduced in the rare variant (Fig. 1A). This effect was also seen with binding to bacterial lipopeptides, although the effect was less pronounced. Interestingly, binding to a nonbacterial ligand did not show this effect.

In an in silico approach, we tried to explain the results through molecular modeling of the rare variant as compared with the common variant. In the F436L variant, the changed amino acid is situated close to the groove of the molecule near the center. This means that in the L436 structure, only one acyl chain could be accommodated, as evident from superpositions with phospholipid molecules. Although this has not been formally proven, from a structural perspective, it appears plausible that the two hydrophobic channels in LBP contribute to LPS or lipopeptide binding. Therefore, the F436L variant might alter the ability to bind these hydrophobic substances. Analysis of MutPred implied change in function of the protein because of destabilization.

Examination of the three-dimensional structures revealed some structural changes because of amino acid change from F436L. This is further supported by the destabilization occurring owing to free energy change based on I-Mutant 3.0, DUET, and ELASPIC analysis. Additionally, docking analysis revealed a shift in the amino acids residues binding to LPS in the rare variant compared with the common variant. Topographical and conformational changes associated with the amino acid change from F436 to L436 is evident from the altered binding of the rare variant to LPS. An increase in size of cleft 1 in the rare variant of LBP demonstrated that it had a higher number of active sites and was more receptive for ligand binding interactions owing to extra structural rearrangements. Finally, functional pockets and the overall number of highly conserved residues from two major functional clefts of the rare variant were higher in number as compared with the common variant. Comparatively higher number of pockets in the rare variant point to the fact that it may be slightly more flexible than the common variant.

From the clinical data of our study, we show a lower risk of septic shock in the presence of the rare LBP variant. It was not associated with sepsis prevalence and susceptibility to infection. Although not statistically significant, we also found an association with reduced mortality. Clinical effects of LBP variants have been observed in previous studies showing an influence on susceptibility to and also severity of infections or other inflammatory diseases (34). As we have shown in a previous study, another functional LBP SNP (P333L and rs2232613) was associated with a more severe course of sepsis (37). By direct comparison of these SNPs in our clinical cohort, we observed a marked difference in the clinical course of infections, suggesting an opposite effect.

In other studies, LBP variants like an LBP haplotype variation were associated with an increased rate of Gram-negative infections in recipients of homolog bone marrow for hematologic malignancies (30). In other patients, LBP haplotype variation was a risk factor for ventilator-associated pneumonia and sepsis in pediatric patients (32, 33). The LBP haplotypes variations did include wildtype and mutated variants. Therefore, no conclusive picture regarding risk alleles could be drawn from these data. Interestingly, the F436L SNP examined in this study was associated with a higher prevalence of sepsis in a cohort of trauma victims of Han Chinese origin with a different demographic composition (38, 39).

In conclusion, this study shows a functional relevant SNP in the LBP gene that is associated with changes in its properties regarding binding capability and function with regard to stimulation of macrophages. In silico analysis revealed molecular alterations that could be associated with these functional changes. We found a favorable outcome following severe infections in patients after surgery. We propose several potential ways of how a loss of function or a gain of function could be explained by this variant and potentially contribute to patient outcome.

There are potential limitations of our study. First, in our experimental approach, we used a xenogenic model, including mouse macrophages and human serum. In our view, this is justified by the fact that human macrophages (i.e., derived from transformed THP-1 cells) are not readily available and might not display full functional capacity. Human serum used in this study, in addition, contains functional molecules that are needed for sufficient signal transduction in contrast to recombinant protein. The retrospective analysis of only two variants of LBP, furthermore, is a limitation of this study. The association found should be studied prospectively in a broader approach, including other nonsynonymous SNPs.

To further delineate functional consequences regarding concentration-dependent effects of this SNP, however, more experiments may be necessary. Elucidating the complex cascade of events leading from recognition of pathogens to systemic inflammation and disease may in the future lead to novel intervention strategies currently needed to improve the outcome of sepsis. Furthermore, a genetic risk stratification may allow for better prevention of clinical complications in patients at risk for infectious diseases.

We thank Fränzi Creutzburg, Diana Woellner, and Ina Wendler (Institute of Microbiology, Charité, Berlin, Germany) for excellent technical assistance throughout this project. Michael Kabesch (Children‘s University Hospital Regensburg, Department of Pediatric Pneumology and Allergy, Campus St. Hedwig, Regensburg, Germany) is acknowledged for providing sera containing mutant LBP.

The authors have no financial conflicts of interest.