We tested the hypothesis that single nucleotide polymorphisms (SNPs) within genes of the NF-κB pathway are associated with altered clinical outcome of septic shock patients. We genotyped 59 SNPs in the NF-κB pathway in a discovery cohort of septic shock patients (St. Paul’s Hospital [SPH], N = 589), which identified the C allele of rs7222094 T/C within MAP3K14 (NF-κB inducing kinase; NIK) associated with increased 28-d mortality (uncorrected p = 0.00024, Bonferroni corrected p = 0.014). This result was replicated in a second cohort of septic shock patients (Vasopressin and Septic Shock Trial [VASST; N = 616]) in which the CC genotype of rs7222094 was associated with increased 28-d mortality (Cox regression: SPH cohort hazard ratio [HR], 1.35; 95% confidence interval [CI], 1.12–1.64; p = 0.002 Caucasian only; and VASST cohort HR, 1.24; 95% CI, 1.00–1.52; p = 0.048 Caucasian only). Patients having the CC genotype of rs7222094 in SPH experienced more renal and hematological dysfunction (p = 0.003 and p = 0.011), while patients of the VASST cohort with the rs7222094 CC genotype showed the same trend toward more renal dysfunction. In lymphoblastoid cell lines, we found the rs7222094 genotype most strongly associated with mRNA expression of CXCL10, a chemokine regulated by NF-κB. Accordingly, we measured CXCL10 protein levels and found that the CC genotype of rs7222094 was associated with significantly lower levels than those of the TT genotype in lymphoblastoid cell lines (p < 0.05) and in septic shock patients (p = 0.017). This suggests that the CC genotype of NIK rs7222094 is associated with increased mortality and organ dysfunction in septic shock patients, perhaps due to altered regulation of NF-κB pathway genes, including CXCL10.
Septic shock is an extreme manifestation of the host inflammatory response to severe infection, and the NF-κB signaling pathway is one of the most important signaling pathways involved in the pathogenesis of this pathological state (1). Septic shock is defined as sepsis accompanied by organ failure including cardiovascular failure (2). In response to infection, NF-κB signaling (among other pathways) leads to the transcription of a wealth of inflammatory mediators that contribute to the development of cardiovascular failure, including cytokines, chemokines, adhesion molecules, and reactive oxygen and nitrogen species, to name a few (3, 4). The induction of this response is necessary for the resolution of infection; however, this response, when extreme, leads to organ damage and mortality in many cases. A better understanding of the key pathways and critical molecules involved in the pathogenesis of septic shock is imperative.
NF-κB signaling is activated by a canonical (or classical) pathway and by a noncanonical (or alternative) pathway involving multiple genes (Supplemental Fig. 1) (5). Briefly, the canonical NF-κB pathway requires the IκB kinase (IKK) complex composed of IKKα/β/γ (3). Activation of the IKK complex in response to inflammatory stimuli results in phosphorylation and ubiquitin-dependent degradation of IκBα or IκBβ and translocation of p50-related dimers into the nucleus (3). In response to inflammation in the noncanonical pathway, NF-κB inducing kinase (NIK), a docking molecule, recruits IKKα to p100 and then activates IKKα (3, 6). IKKα phosphorylates p100, which is subjected to phosphorylation, ubiquitination, and proteosomal degradation, resulting in the release and nuclear translocation of p52 containing RelB heterodimers (3, 6).
Genetic variation in key inflammatory genes contributes to outcome in sepsis (7–12). Because NF-κB is a centrally important signaling pathway, we tested the hypothesis that genetic variation in genes involved in the NF-κB pathways would be associated with mortality in septic shock. To accomplish this, we genotyped 59 single nucleotide polymorphisms (SNPs) in 19 genes in a single center derivation cohort of septic shock patients. We next tested for replication of an arising SNP association in a second multicenter cohort of septic shock patients. We tested for similar association with organ dysfunction to understand the involved clinical phenotype in more detail. To test for biological plausibility of the observed SNP association, we next measured gene expression in genotyped lymphoblastoid cell lines and identified a candidate gene whose expression was associated with the candidate SNP. We then tested for association of this gene product with the candidate SNP both in vitro and in vivo.
Materials and Methods
St. Paul’s Hospital cohort (discovery cohort).
All patients admitted to the St. Paul’s Hospital (SPH) Intensive Care Unit (Vancouver, British Columbia, Canada) between July 2000 and January 2004 were screened. Using the current consensus definition (≥ 2 SIRS criteria, known or suspected infection, hypotension unresponsive to fluid resuscitation alone) 601 patients had septic shock on admission and had DNA available (12, 13). Twelve patients in this cohort had also been enrolled in the Vasopressin and Septic Shock clinical trial (14) and were therefore excluded. Thus 589 patients in total were included in this analysis. This study was approved by the Institutional Review Board at SPH and the University of British Columbia.
VASST cohort (replication cohort).
The Vasopressin and Septic Shock Trial (VASST) was a multicenter, randomized, double-blind, controlled trial evaluating the efficacy of vasopressin versus norepinephrine in 779 patients who were diagnosed with septic shock according to the current consensus definition (15). Clinical phenotyping has been described elsewhere (14). All patients were enrolled within 24 h of meeting the definition of septic shock, and DNA was available from 616 patients. The research ethics boards of all participating institutions approved this trial, and written informed consent was obtained from all patients or their authorized representatives. The research ethics board at the coordinating center (University of British Columbia) approved the genetic analysis.
SNP selection and genotyping of patient cohorts
We used Ingenuity IPA (version 8.6, build 93815, content 3003) to identify 19 cytosolic genes in the NF-κB canonical and noncanonical pathway that also had dense resequencing data publically available (Supplemental Fig. 1) (Seattle SNPs Program for Genomic Applications, http://pga.mbt.washington.edu/; Cardiogenomics, http://cardiogenomics.med.harvard.edu/home; Innate Immunity, http://www.pharmgat.org/IIPGA2/index_html, formerly http://innateimmunity.net/; Berkeley University PGA, http://pga.jgi-psf.org/; and SouthWestern PGA, http://pga.swmed.edu/). TagSNPs were identified for genotyping in patient cohorts using a linkage disequilibrium-based tag SNP selection method (16) and using an r2 threshold of 0.65 for SNPs with a minor-allele frequency >5% yielding 59 SNPs in 19 cytosolic genes; receptor and their ligands as well as downstream gene targets of NF-κB signaling were excluded (Table I). DNA was extracted from peripheral blood samples using a QIAamp DNA Blood Midi Kit (Qiagen, Mississauga, Ontario, Canada) and genotyped using the Illumina Golden Gate Assay at the UBC Centre for Molecular Medicine and Therapeutics genotyping core facility (Luminex Molecular Diagnostics, Toronto, Ontario, Canada). Primer probe sets are detailed in Supplemental Table IV.
The primary outcome was 28-d mortality. Secondary outcomes were days alive and free of organ dysfunction during the first 28 d calculated according to the Brussels criteria (15).
Biological plausibility experiments
Microarray mRNA expression analysis in vitro.
Lymphoblastoid DNA from the Coriell Institute was genotyped for rs7222094 in 85 CEPH population samples using Sanger sequencing of the region. Sequencing was performed at the McGill University and Génome Québec Innovation Centre (Montréal, Québec, Canada). Primers for sequencing the region surrounding rs7222094 are as follows: forward, 5′-GGGTTCCCTATGGAGGAGAG-3′; reverse, 5′-CTGTCCAGCTCTCCAGGTTC-3′. These 85 CEPH population lymphoblastoid cell lines of known genotype for NIK rs7222094 were cultured in RPMI 1640 and subsequently stimulated in triplicate by the addition of Cytomix (9, 17–19http://genomequebec.mcgill.ca/FlexArray/license.php). MIAME compliant microarray data are publically available at GEO (GSE25543; http://www.ncbi.nlm.nih.gov/geo/).
ELISA protein expression analysis in vitro.
ELISA protein expression analysis in vivo.
Briefly, whole-blood samples were drawn into chilled 7-ml EDTA Vacutainer tubes (BD, Mississauga, Ontario, Canada), put on ice immediately, then spun at 3000 rpm for 15 min at which point plasma was collected and stored at −70°C until further use (14
We assessed baseline characteristics using a χ2 test for categorical data and a Kruskal–Wallis test for continuous data and then reported the median and interquartile ranges. We then tested for association between SNP genotype and 28-d mortality in the SPH discovery cohort using an Armitage trend test, as is commonly used for initial discovery surveys in genome-wide association studies (Table I). One SNP emerged as statistically significant after a Bonferroni correction for multiple comparisons. We then tested for replication of this finding in the VASST cohort of septic shock patients. To correct for potentially confounding variables, including age, gender, ancestry, and surgical versus medical diagnostic category, we used Cox regression. We then tested for association between secondary outcome measures of days alive and free of organ failure using Kruskal–Wallis tests in both SPH and VASST cohorts. Vasopressin treatment by genotype (NIK rs7222094 CT) interaction was assessed using logistic regression analysis (interaction statistics): P(Death) ≈ vasopressin + genotype + vasopressin × genotype.
A Student t test was performed to test for differences between genetic groups (rs7222094 CC versus TT) of CXCL10 ELISA concentrations. Analyses used SPSS (version 16; SPSS, Chicago, IL), R statistical software package, and GraphPad Prism (version 5.02; GraphPad, La Jolla, CA).
Twenty-eight–day mortality in septic shock patients
Of the 59 tagSNPs in 19 genes in canonical and noncanonical NF-κB pathways, one SNP, rs7222094 in NIK, was significantly associated with 28-d mortality in the SPH discovery cohort (uncorrected p = 0.00024, Bonferroni corrected p = 0.014) (Supplemental Fig. 1, Table I).
Hardy–Weinberg equilibrium and minor allele frequencies of all SNPs genotyped are presented along with literature based minor allele frequencies in Supplemental Table I. Allele frequencies of rs7222094 differed between ethnic groups within the SPH and VASST cohorts (Supplemental Table II). Therefore, our primary analysis was limited to Caucasians only (SPH, n = 453, VASST, n = 517), and our secondary analysis of all patients included ethnicity as a covariate (SPH, N = 589; VASST, N = 616).
To consider and correct for potential confounding variables due to differences at baseline in septic shock patients, we used Cox regression to test an additive model in the SPH septic shock cohort and then used the same analysis in the VASST septic shock cohort. Patients in the SPH cohort who had the CC genotype of NIK rs7222094 had a significantly increased hazard of 28-d mortality compared with that of patients having the CT or TT genotypes of rs7222094 (hazard ratio [HR], 1.35; 95% confidence interval [CI], 1.12–1.64; p = 0.002 Caucasian only) (Table II). This finding was replicated in the VASST cohort (HR, 1.24; 95% CI, 1.00–1.52; p = 0.048 Caucasian only) (Table II). The results were similar for all patients with ethnicity included as a covariate in a Cox regression model (Fig. 1, Table III).
Similarly, in an unadjusted univariate analysis, the C allele of rs7222094 TC was associated with mortality in SPH (mortality, Caucasian only: TT, 33.8%; CT, 42.4%; CC, 53.7%; p = 0.005; mortality, all ethnicities: TT, 33.8%; CT, 43.2%; CC, 53.0%; p = 0.001), and a similar trend was observed in VASST (mortality, Caucasian only: TT, 26.3%; CT, 34.7%; CC, 38.4%; p = 0.09; mortality, all ethnicities: TT, 27.5%; CT, 33.2%; CC, 40.8%; p = 0.03). Allele frequencies of survivors versus nonsurvivors in both Caucasian and all ethnicities are reported in Table IV.
In the SPH Caucasian cohort, patients having the CC genotype of rs7222094 had greater baseline creatinine concentrations (p = 0.04) and significantly higher PaO2/FIO2 at baseline (p = 0.002) than patients having the CT or TT genotypes of rs7222094 (Table V). The only difference at baseline among VASST Caucasian patients was that patients having the CC genotype of rs7222094 had significantly lower platelet counts than those of patients having the CT or TT genotypes of rs7222094 (p = 0.02) (Table V). Because the VASST cohort was a clinical trial comparing efficacy of vasopressin versus norepinephrine in septic shock, in a secondary analysis we tested for an interaction by logistic regression between NIK rs7222094 and vasopressin treatment in Caucasian patients. We found no significant interaction (interaction statistic p = 0.462).
Organ failure in septic shock patients
Patients homozygous for the C allele of rs7222094 in both the SPH and VASST Caucasian cohorts had more organ dysfunction as defined by Brussels criteria compared with that in patients having the CT or TT genotypes of rs7222094 (15). SPH patients with rs7222094 CC genotype had significantly fewer days alive and free of renal dysfunction (p = 0.003) and fewer days alive and free of acute renal replacement therapy (p = 0.008) as well as fewer days alive and free of hematological dysfunction (p = 0.011) during the 28-d study period than those of patients having the CT or TT genotypes of rs7222094 (Table VI). Patients of the VASST cohort with the rs7222094 CC genotype show the same trend toward more renal dysfunction (Table VI). The number of patients affected by each type of organ dysfunction by genotypic group is outlined in Table VII.
In the SPH cohort, patients with the rs7222094 CC genotype also had significantly fewer days alive and free of cardiovascular dysfunction (p = 0.008), with correspondingly fewer days alive and free of vasopressor support (p = 0.01), which was also seen as a trend in the VASST cohort (Table VI). SPH cohort patients also experienced more hepatic and neurologic dysfunction (p = 0.007 and p = 0.006, respectively) (Table VI).
CXCL10 mRNA production by lymphoblastoid cell lines in vitro
The gene with the greatest difference (Δ) in fold change between major (TT) and minor (CC) genotypes was CXCL10 (Δ fold change, 0.67; uncorrected Student t test between groups, p = 0.055) suggesting lower mRNA expression of CXCL10 for the CC genotype compared with that of TT or TC (Supplemental Table III).
CXCL10 protein production by lymphoblastoid cell lines in vitro
Protein levels of CXCL10 were measured in 26 cell lines of known genotype for rs7222094. Cell lines homozygous for the C allele of rs7222094 produced less CXCL10 at both baseline and after inflammatory stimulation than that of cell lines homozygous for the T allele (p = 0.032 and p = 0.050, respectively) (Fig. 2).
CXCL10 protein production in VASST plasma samples
Baseline plasma specimens of a random sample of patients with septic shock from the VASST cohort were assayed in duplicate for CXCL10 by ELISA, and as was found in both the control and stimulated lymphoblastoid cell lines, patients of the CC genotype had significantly lower CXCL10 than that of the TT genotype (p = 0.017) (Fig. 3).
We found that patients of the CC genotype of NIK rs7222094 had significantly increased mortality compared with that of patients having the CT or TT genotypes of rs7222094 in two cohorts of patients who had septic shock. Specifically, Caucasian patients in the SPH cohort who had the CC genotype of NIK rs7222094 experienced a significant increase in the hazard of death over the 28 d (HR, 1.35; 95% CI, 1.12–1.64; p = 0.002). This effect was also found in the VASST Caucasian cohort (HR, 1.24; 95% CI, 1.00–1.52; p = 0.048). The results were similar for all patients with ethnicity included as a covariate in a Cox regression model. Also, patients of the CC genotype of rs7222094 of SPH experienced more renal and hematological dysfunction compared with that of patients having the CT or TT genotypes of rs7222094 (p = 0.003 and p = 0.011, respectively). Patients of the VASST cohort with the rs7222094 CC genotype showed the same trend toward more renal dysfunction as found in the SPH cohort.
NIK was first discovered in human B cells, and hence we chose lymphoblastoid cell lines for this phase of our study (20). CXCL10 is a chemokine transcribed in response to NF-κB during inflammation and is generally thought to signal in response to canonical pathway stimuli (21, 22). In a recent study by Zarnegar et al. (23), CXCL10 levels were dependent on NIK suggesting noncanonical signaling. CXCL10 is transcribed in response to NF-κB activation and has been shown to be downregulated upon inhibition of NIK (23). The CC genotype of NIK rs7222094 was associated with significantly decreased levels of CXCL10 in supernatant of lymphoblastoid cell lines at baseline and after Cytomix stimulation (p = 0.032 and p = 0.050). Similarly, CXCL10 levels were significantly lower in septic shock patients of the CC genotype (p = 0.017), suggesting a biologically plausible explanation for our observations.
It was interesting to find that the patients with the NIK rs7222094 CC genotype experienced increased mortality while supernatant from cell lines of the same genotype had decreased levels of CXCL10. This effect was replicated when CXCL10 concentration was measured in VASST patient samples. As mentioned previously, CXCL10 is a proinflammatory chemokine released during inflammatory states, such as allograph rejection and infection (24). It is plausible that proinflammatory molecules are necessary to mount an effective immune response during septic shock. CXCL10 is a chemokine that instigates chemotaxis of activated T cells and NK cells (24). We speculate that without enough CXCL10 to drive recruitment of inflammatory cells, it is possible that these patients (who had the CC genotype of NIK rs7222094) have an immunological disadvantage and so have increased mortality of septic shock.
The original characterization of NIK indicated that NIK was a powerful effector of canonical NF-κB signaling under TNF-α and IL-1β stimulation (20). Later studies were unable to confirm this mechanism, and instead the discovery of the noncanonical pathway emerged as subsequent research appeared to distill the two pathways into distinct and separate processes (6, 25–27). Evidence is building suggesting that the IKK–NIK axis is a pivot point for control of both noncanonical and canonical signaling (23, 28, 29). It is possible that the timing of experimental procedures is critical to our understanding of NIK, as many studies that separated the two pathways and excluded NIK from canonical signaling focused on early events (from minutes to less than 2 h) (25, 27). In contrast, many studies find that NIK plays a pivotal role in canonical signaling when evaluating effects at time points of several hours to days (20, 23, 28). The current understanding of the regulation of NIK is that NIK is constitutively transcribed, translated, and degraded via its interaction with TRAF3. However, after degradation of TRAF3 after appropriate stimuli, NIK recruits IKKα to p100, activating IKKα thereby initiating proteosomal degradation of p100 to p52 and consequently translocation of heterodimers to the nucleus (6, 30). It is conceivable that accumulation of NIK over time is a key facet to its mechanism.
Because NIK has been implicated in the host response to infection, we speculate that NIK could modulate the immune response during septic shock. NIK is important in the host response to numerous infections including respiratory syncytial virus [with evidence to suggest activation of both noncanonical and canonical signaling (31)], HIV (32), hepatitis B (33), Escherichia coli (34), as well as the response to LPS (35).
Several lines of evidence suggest why polymorphisms of NIK may be predictors of outcome in septic shock (1). First, NIK stimulates inflammation by upregulating the noncanonical (and possibly the canonical) pathway of NF-κB activation. Second, NIK is required for optimal IgG production by lymphocytes (36). Third, NIK may modulate blood pressure: NIK has a role in the mechanism of action of the calcium channel blocker nifedipine (37), and angiotensin II induces inflammation through NIK activation of the noncanonical NF-κB pathway (38). However, polymorphisms of NIK have not been widely studied. In a survey of 181 SNPs of 17 genes in the NF-κB pathway, polymorphisms of NIK were not associated with rheumatoid arthritis susceptibility (39). Notably, rs4792847 of NIK was found to be significantly associated with response to anti-TNF treatment in rheumatoid arthritis. Patients with GG genotype had the greatest improvement at the 6-mo mark in a discovery cohort; however, this effect was not found in a replication cohort (40). The GG genotype of rs4792847 (patients who may have had a more favorable response to anti-TNF) is in high linkage disequilibrium with the TT genotype of rs7222094 (r2 = 1.0). This is consistent with our observation of a protective effect of the TT genotype (i.e., lower mortality and higher CXCL10 production) seen in our study of patients who had septic shock and replicated in our in vitro experiments.
To our knowledge, NIK has not been implicated in septic shock to date. However, NIK is a drug target for other diseases (5). Inhibitors of NIK have been synthesized for diseases such as multiple myeloma and other cancers, as anti-inflammatory agents for inflammatory diseases (41), and as a vaccine adjuvant (42). Our data suggest that rs7222094 may be of interest in randomized controlled trials of therapies for septic shock by defining risk categories of patients and perhaps defining response to anti-inflammatory agents.
This study has several limitations. The analysis of the SPH and VASST cohorts was performed retrospectively. The association of rs7222094 with mortality, organ dysfunction, and CXCL10 levels does not prove a causal link. Furthermore, CXCL10 was used as a marker for differences in NIK-induced NF-κB signaling. We do not currently know of, nor did we test for, the influence of CXCL10 itself on organ dysfunction or mortality. In view of the large number of genes connected in some way to NF-κB signaling, we chose to limit our analysis to genes of the cytosolic members of the NF-κB pathway, excluding receptors and downstream targets. Therefore, this analysis does not include all potentially functional variants, in particular those recently published after the design of this study (43–45).
In conclusion, we found that NIK rs7222094 was consistently and significantly associated with mortality in two independent cohorts of patients who had septic shock. Patients homozygous for the CC genotype of rs7222094 had increased mortality and also experienced more renal and hematological failure in the SPH cohort of Caucasian patients with a similar trend in VASST Caucasian patients. Furthermore, we found that lymphoblastoid cell lines homozygous for CC of rs7222094 produced less CXCL10 in vitro at baseline and after Cytomix stimulation than that of cell lines having the TT genotype of NIK rs7222094 (p = 0.032 and p = 0.050, respectively). As well, patients who had septic shock in the VASST cohort who were NIK rs7222094 CC genotype had significantly lower plasma levels of CXCL10 than those of the TT genotype (p = 0.017). We speculate that polymorphisms of NIK could be used to predict risk of death from septic shock and to predict response to anti-inflammatory treatment, such as inhibitors of NIK.
S.A.T. has received grant support from MITACs made possible by an industry partnership with Sirius Genomics. J.A.R. holds stock in Sirius Genomics Inc., which has submitted patents owned by the University of British Columbia and licensed to Sirius Genomics that are related to the genetics of vasopressin, NIK, and protein C. The University of British Columbia has also submitted a patent related to the use of vasopressin in septic shock. J.A.R. is an inventor on these patents. J.A.R. has received consulting fees from Ferring, which manufactures vasopressin; from Astra Zeneca, which manufactures anti-TNF; and from Sirius Genomics Inc. J.A.R. has received grant support from Sirius Genomics, Novartis, Ferring, and Eli Lilly. J.A.R. has received speaking honoraria from Pfizer and Eli Lilly. K.R.W. holds stock in Sirius Genomics Inc., which has submitted patents owned by the University of British Columbia and licensed to Sirius Genomics that are related to the genetics of vasopressin, NIK, and protein C. The University of British Columbia has also submitted a patent related to the use of vasopressin in septic shock. K.R.W. is an inventor on these patents. K.R.W. has received grant support from Sirius Genomics. H.W. reports serving as director of research and holding shares at Sirius Genomics Inc., which has submitted a patent owned by the University of British Columbia and licensed to Sirius Genomics that is related to the genetics of LNPEP.
S.A.T. is the recipient of a Mathematics of Information Technology and Complex Systems fellowship. J.H.B. is the recipient of a Providence Health Care Research scholarship. T.-a.N. is the recipient of a Integrated and Mentored Pulmonary and Cardiovascular Training fellowship. K.R.W. is a Michael Smith Foundation for Health Research Distinguished Scholar. The Vasopressin and Septic Shock Trial is supported by Canadian Institutes of Health Research Grant MCT 44152.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- confidence interval
- hazard ratio
- IκB kinase
- NF-κB inducing kinase
- single nucleotide polymorphism
- St. Paul’s Hospital
- Vasopressin and Septic Shock Trial.
- Received September 2, 2010.
- Accepted December 7, 2010.
- Copyright © 2011 by The American Association of Immunologists, Inc.