1. Introduction
The incidence of sepsis, particularly severe sepsis and septic shock, is increasing among hospital transfer, with a mortality rate between 20 and 35% despite improved strategies of care [
1]. After the analysis of negative results from randomized clinical trials (RCTs) to reduce mortality after 28 days and/or 90 days [
2,
3,
4,
5], new aspects of sepsis have come to the front. The demonstration of systemic and tissue immuno-depression after a septic injury [
6] and the impact of co-morbidities [
7] both motivate a change in the sepsis syndrome paradigm [
8]. Research on the genetic predispositions associated with outcomes and polymorphisms for genes encoding mediators of inflammation, such as TNFα [
9] and IL-10 [
10], has been poorly replicated [
11], and negative results on crude mortality reduction were obtained when agents blocking TNFα were tested. Despite the large number of such studies, re-analyses of accumulated evidence do not definitely show any associated genes [
11]. As many other complex syndromes for which environmental and chronic disease risk factors are thought to interact with multiple genes, such analyses may benefit from recent genetic methodologies, such as genome-wide association studies (GWAS) [
12]. Instead of sepsis syndrome, the GWAS approach has provided important genetic and biological insight for other more specific infectious diseases, such as meningococcemia or malaria [
13,
14].
An article by Rautanen et al. [
15] presents the results from the first GWAS on survival from sepsis due to pneumonia, which was assessed in a multistage study, including four cohorts and tested almost 6 million single nucleotide polymorphisms. The authors identified genetic variants within the
FER gene showing consistent effects across the four cohorts studied. The rs4957796 C allele, with a near 20% frequency in European populations, was associated with reduced mortality in sepsis caused by pneumonia. Scherag et al. conducted another GWAS in patients with severe sepsis. They reported 14 loci with suggestive evidence of an association with 28-day mortality [
16]. Nevertheless, they did not find an association between rs4957796 and 28-day mortality and no evidence on an association between other loci identified by Rautanen et al. and 28-day mortality. They proposed that the focus on pneumonia-induced sepsis by Rautanen et al. may explain the observed discrepancies. In the same way, it was proposed that genetic studies should focus on specific traits related to severity and outcomes rather than on a broadly defined syndrome [
17].
We have collaborated to provide the first human GWAS on severe sepsis or septic shock using the randomized control trial PROWESS database [
18] to test the benefits of activated protein C (aPC) use on outcome. Because one arm of the trial received aPC treatment, the sepsis prognosis models have only been tested on the control arm. The prognosis was finally dominated by clinical variables with modest relation with the tested genetic markers. The second randomized control trial PROWESS SHOCK [
3] tested aPC exclusively on septic shock patients and failed to show a crude outcome benefit at day 28 or 90. This negative result of aPC treatment in septic shock prompted us to perform GWAS on outcome specific traits using the complete PROWESS dataset after selection of septic shock patients. The present study reports the first identification of genetic variants associated with the prognosis of septic shock when comorbidity levels and systemic inflammation intensities are integrated. Since the mechanisms for early death differ from those causing late death [
19,
20], we also investigated the association of genetic variants with both early and late death. We further focused on a non-coding region significantly associated with late death and showed its regulatory activity on one of the closest gene, the cytokine inducible SH2 containing protein
(CISH) gene. Interestingly, CISH is known to be a negative regulator of cytokine signaling.
3. Discussion
In this study, we assessed the association of SNPs with early and late mortality in septic shock patients at the genome level, and we looked for biological pathways that could be disrupted by genetic variation. We then annotated and prioritized the SNPs associated with mortality and the SNPs in linkage disequilibrium with them and characterized the functional significance of the best candidates.
This present GWAS follows a previous study using the same PROWESS cohort [
18], but designed after removal of the patients in the aPC arm. The next randomized clinical trial PROWESS SHOCK failed to show a benefit of aPC on mortality, motivating the immediate removal of aPC from the market. As a consequence, this aPC failure to reduce mortality in septic shock allowed to use the shocked patients from both placebo and aPC arms of the PROWESS cohort to perform the GWAS, a selection that differed from the previous GWAS [
18]. The treatment with aPC was then considered only as a covariate. The growing evidence for potential differing mechanisms for early versus late death [
19,
20] was then considered to test SNP associations. The early stimulation of inflammation processes appears to be rapidly followed by a downregulation of these processes through dominant anti-inflammatory patterns, which can be considered suitable for maintaining the tissue fitness and reducing the risk of death [
41]. If it persists over time, such acquired immunosuppression may be associated with higher risk of secondary infection [
42]. The present GWAS allowed us to identify SNPs associated with mortality in septic shock patients.
Based on an FDR of 5%, 32 and 107 SNPs were associated with early and late mortality, respectively. These associations reduced to 12 and 16 SNPs after the Bonferroni correction for early and late mortality, respectively. Individuals having four or more risk alleles had a 12-fold higher or a 123-fold higher risk of death than those without risk alleles for early and late death, respectively. For early death, the strongest associations between intra-genic SNPs were located within
CYP11B2, a gene encoding aldosterone synthase, a key enzyme of the aldosterone biosynthesis. Variants of such a gene have never been reported to be associated with human shock status and/or severe infection, but have been shown to be largely associated with hypertension and atrial fibrillation [
43]. The other important SNPs associated with early mortality were located within
PTPN11, which is also known as
SHP2. The proteins encoded by this gene are members of the protein tyrosine phosphatase (PTP) family that are known to be signaling molecules regulating a variety of cellular processes. This PTP family contains two
tandem Src homology-2 domains, which function as phospho-tyrosine binding domains and mediate the interaction of the PTP with their substrates. The protein encoded by
PTPN11 is implicated in reduced JAK/STAT signaling when it is elevated, which may reduce MHC expression induced by INFγ [
31]. SHP2 activation induced by human CMV infection is responsible for the downregulation of INFγ-induced STAT1 tyrosine phosphorylation [
44]. In addition, the PD1/PDL1 interaction has been shown to inhibit T cell receptor signaling by recruiting SHP1/2 phosphatases [
45]. For late mortality, it should be noticed that
FER reported to be associated with mortality in sepsis caused by pneumonia [
15] was associated with mortality in septic patients in our study on the basis of an FDR of 5%. FER that is a protein tyrosine kinase acting downstream of cell-surface receptors, has been shown to influence leucocyte recruitment in response to LPS [
46] to inhibit neutrophil chemotaxis [
47], and to alter the endothelial response to LPS stimulation [
48]. Furthermore, the strongest associations were found within cytokine-inducible SRC homology 2 (SH2) domain protein (
CISH) and
MAPKAPK3.
MAPKAPK3 is involved in the MAP Kinase pathway, which is known to regulate the activation of immune cells.
CISH is the first member of the suppressor of cytokine signaling (SOCS) family. An association has been shown between
CISH polymorphisms and susceptibility to infectious diseases including malaria, bacteremia or tuberculosis [
25]. In addition, rs414171-T allele that was associated with susceptibility to bacteremia, tuberculosis and malaria has been shown to reduce the promoter activity of
CISH and its expression in human PBMCs after stimulation by IL-2 [
25,
49]. Since CISH is known to suppress STAT5 in T cells, it has been proposed that decreased levels of
CISH lead to enhanced activation of STAT5 and enhanced activation of Treg lymphocytes, and as a consequence, a suppressed immune response against bacteria and other pathogens [
25].
Noticeably, the
CISH locus was highlighted by our bioinformatic analyses, which aimed to annotate and prioritize SNPs associated with mortality and the SNP in linkage disequilibrium with them. Since more than 95% of the SNPs associated with mortality or the SNPs in linkage disequilibrium with them were located in non-coding regions, we hypothesized that the vast majority of the causal genetic variants are regulatory variants. More generally, most of the GWAS variants are non-coding, emphasizing the potential role of regulatory variants in complex diseases [
50,
51]. Moreover, we investigated 1439 non-coding SNPs including SNPs associated with mortality and SNPs in linkage disequilibrium with them. Among those SNPs, rs143356980 was the best candidate using the IW-scoring method and was ranked in 13th position on the basis of the intergenic TAGOOS score. Interestingly, it is located near the
CISH gene, and is in strong linkage disequilibrium with four SNPs highly associated with late mortality in patients with septic shock; these includes rs2239751, which has been also associated with tuberculosis [
52,
53], and persistent hepatitis B virus infection [
54]. Furthermore, rs143356980 is located within a super-enhancer for monocytes and T lymphocytes, according to the database by Hnisz et al. [
33]. Using the CRISPR-Cas9 genome editing method, we showed that the sequence containing rs143356980 has an enhancer activity on the
CISH gene in unstimulated K562 cells and K562 cells stimulated with either LPS or IFNγ. Using the luciferase reporter assay, we showed the effect of rs143356980 on the enhancer activity in unstimulated K562 cells and K562 cells stimulated with IFNγ. More specifically, rs143356980-T decreased the enhancer activity compared to rs143356980-C allele. In all, these results suggest that genetic variation within the enhancer containing rs143356980 influences
CISH gene expression, Jak/Stat signal transduction, and the risk of death in septic shock patients. It is not excluded, nevertheless, that other SNPs within the same enhancer or other regulatory regions alter
CISH gene expression and susceptibility to sepsis. These include rs414171, which has been shown to reduce
CISH expression in human PBMCs after stimulation by IL-2 [
25,
49]. rs143356980 and other genetic variants in an enhancer may act through the same mechanisms, leading to susceptibility to sepsis.
Since the expression of
CISH is induced through the stimulation of other receptors, genetic variation altering
CISH expression may have functional consequences in other cells.
CISH expression is induced in response to EPO, IL-2, IL-3, IL-5, GM-CSF in hematopoietic cells, leading mostly to the activation of STAT5 [
55,
56]. In addition,
CISH is an inducible gene in NK cells stimulated by IL-15, and deletion of
CISH increased proliferation, IFNγ production and cytotoxicity against tumors [
57]. Since NK cells in septic patients have been shown to produce low levels of IFNγ and to have a decreased cytotoxicity activity [
58], low levels of
CISH may influence susceptibility to sepsis through an NK cell dependent mechanism.
GM-CSF expression is induced in macrophages infected by
M. tuberculosis, leading to
CISH expression and an increased replication of
M. tuberculosis [
59]. Moreover, LPS and IFNγ induce the expression of
CISH in human monocytes, as shown in a transcriptomic study [
60]. Similarly, we report in the present study an increase of the
CISH expression in K562 cells stimulated either by LPS or IFNγ. The functional effect of
CISH expression levels remains, however, to be investigated in monocytes or macrophages in septic patients.
Forty-five genes that encode proteins contained the SNPs associated with early or late mortality or the SNPs in linkage disequilibrium with the SNPs. Enrichment in biological pathways (Kyoto encyclopedia of genes and genomes-KEGG and BIOCARTA) was used to investigate the involved underlying biological functions. Since no significant enrichment based on the 45 genes was found, we mapped the proteins encoded by these 45 genes on a high-quality protein–protein interaction network [
30]. Thirty proteins out of the 45 proteins were included in the whole protein–protein network, leading to identify a sub-network that contains 27 proteins associated with mortality and their 298 direct interactors. For example, CISH and PTPN11 shared five direct interactors, whereas CISH and FER shared one interactor. This suggests that genetic variants altering the function or the expression of proteins belonging to the sub-network may act in combination to influence mortality in septic shock patients. Furthermore, this subnetwork was enriched for 79 significant pathways, including Toll-like receptor, IL-6, Jak-STAT, and T cell receptor signaling pathways as well as aldosterone-regulated sodium reabsorption. Thus, the dysregulation of the renin-angiotensin-aldosterone system and the dysregulation of the monocyte/macrophage activation or the T-cell activation may be involved in sepsis-induced associated organ failure. In the same way, sepsis-associated SNPs were enriched in the super-enhancers of adrenal gland that produces aldosterone; furthermore, sepsis-associated SNPs were highly enriched in the super-enhancers of monocytes and Th lymphocytes. Moreover, Davenport et al. recently reported that patients with higher early mortality had an increased expression of negative regulators of TLR signaling, and a downregulation of human leucocyte antigen class II genes and most genes implicated in T cell activation [
61]. The clinical relevance of these findings is strongly supported by the significant benefits of associating clinical traits with SNPs to predict early and late death [
62] (
Figure 1D,E).
Our results suggest that genetic variations in different genes including CISH alter the activation of immune cells and, in turn, increase the risk of both early and late mortality in patients with septic shock. To provide greater homogeneity to our GWAS study, only European patients with septic shock were selected. This allowed us to look for SNP in strong linkage disequilibrium with SNPs associated with mortality, and to annotate them for their molecular function. Furthermore, we looked for the potential functional significance of the identified SNPs using protein–protein interactions and bioinformatics tools predicting regulatory SNPs. Finally, we performed experimental studies confirming the regulatory effect of a bona fide candidate SNP.
In conclusion, this GWAS analysis identified new loci relevant for mortality in European patients with septic shock. Here, we provide evidence for (i) different covariates and SNPs that influence early or late mortality, supporting the concept to separate early and late mortality; (ii) different SNPs strongly associated either with early or with late mortality; (iii) a protein–protein sub-network highlighting biological pathways, such as the Jak/stat pathway; (iv) the combination of clinical traits and SNPs may better predict early and late mortality; (v) a regulatory effect of a sequence containing candidate SNPs on CISH expression, and a high effect of rs143356980 on the enhancer activity suggests an important role of this region on the immune response modulation in patients. However, independent GWAS testing the same SNPs or replication studies focused on the same phenotypes in septic shock patients are required to confirm our association results. Further studies depicting the effect of the transcription level of CISH on the intensity of the immune response in monocytes/macrophages, crucial during sepsis development are needed.