Inclusion criteria were adult patients diagnosed with sepsis, including severe sepsis and septic shock. There was no exclusion criterion. In total, 212 patients were enrolled from March 1, 2011 to October 31, 2012. All patients were >20 years of age [median 67.5 (range 29-95) years, M:F=149:63] and had been admitted to the ICU of a Asan Medical Center (Seoul, Korea). Sepsis, severe sepsis, and septic shock were defined using American College of Chest Physicians/Society of Critical Care Medicine.2829 All patients were managed according to therapeutic recommendations based on early goal-directed therapy and lung-protective ventilator strategy.2930 Survivors were defined as patients who had survived for 28 days after ICU admission. The study objectives and procedures were fully disclosed, and a case report form for this study was completed. All data were collected from the medical records and laboratory and radiographic findings in all patients. This study was approved by the Institutional Review Board (IRB) of the Asan Medical Center (2012-0878). Informed consent was confirmed by the IRB, and written informed consent was obtained from all study participants or their surrogates.

The following data were gathered from the medical records of patients: age, gender, the primary cause of sepsis on initial admission, underlying comorbidities, duration of mechanical ventilation, and lengths of stay in the ICU and hospital. All patients were categorized as sepsis, severe sepsis, and septic shock on ICU admission. Acute Physiology and Chronic Health Evaluation (APACHE) II and Sequential Organ Failure Assessment (SOFA) scores were calculated on the sampling day for this study.3132 We also identified the causative pathogen for sepsis in patients with positive blood culture, and classified them accordingly. In addition, we recorded laboratory data (complete blood count, lactate, C-reactive protein, procalcitonin) on sampling day, and surveyed for the presence of neutropenia (absolute neutrophil count <1500/mm³).

Blood samples were drawn within 24 hrs after ICU admission. Genomic DNA was isolated from 5 mL of ethylenediaminetetraacetic acid (EDTA)-anticoagulated venous blood by the standard method using proteinase K and phenol/chloroform extraction. SNP data for the HMGB1 gene [chromosome 13, position 29930000-29939000 (9 kb total)] was obtained from the HapMap data (version 2, release 21) for 45 unrelated Han Chinese individuals from Beijing, China (CHB) and 44 unrelated Japanese individuals from Tokyo, Japan (JPT) samples. From the database, a total of three SNPs with a minor allele frequency >0.05 (rs1045411, rs3742305, rs2249825) were identified in HMGB1 (excluding 5'- and 3'-flanking regions) and selected for genotyping; all are common SNPs with a minor allele frequency >0.05. Among the three SNPs, rs1045411 located in the 3'-untranslated region (Fig. 1) was chosen for further analyses, because this SNP seemed to show the most significant difference in allele frequencies between patients and normal healthy persons (χ² test, p<0.05) with the lowest p value and highest odds ratio, compared to the other two SNPs (data not shown). Genotypes for rs1045411 in all patients were determined by Sanger sequencing. The target region of HMGB1 was amplified using forward primer 5'-TGGAAGTGGGAGGCAAT TTA-3' (HMGB1_1045411_F) and reverse primer 5'-TGCTGT GCAAA GGTTGAGAG-3' (HMGB1_1045411_R). Amplification conditions were one cycle of 95℃ for 7 min, 30 cycle of 95℃ for 30 s, 56℃ for 30 s, 72℃ for 1 min, plus one cycle of 72℃ for 5 min. Amplification conditions were one cycle of 95℃ for 7 min, 30 cycles of 95℃ for 30 s, 56℃ for 30 s, and 72℃ for 1 min, plus one final cycle of 72℃ for 5 min. Amplified products of 303 bp were confirmed by agarose gel electrophoresis and purified by treating with Exo-SAP (10:1 U ratio mixture of exonuclease I and shrimp alkaline phosphatase, USB corp., Cleveland, OH, USA) at 37℃ for 15 min, 80℃ for 15 min, and a 4℃ hold. Sequencing was carried out using an ABI 3730XL sequencer by Cosmogenetech, Seoul, Korea. Briefly, sequencing amplifications were performed using the BigDye terminator (ver. 3.1) cycle sequencing kit (Applied Biosystems, Foster City, CA, USA). PCR products of 30-90 ng were used as templates and the cycling reaction consisted of one cycle of 95℃ for 90 s, 25 cycles of 95℃ for 30 s, 50℃ for 5 s, and 60℃ for 4 min, followed by a 4℃ hold with HMGB1_ 1045411_F and HMGB1_1045411_R. After purifying the cycle sequencing reaction products with magnetic beads (MagneSil GREEN, Promega, Madison, WI, USA), the sequencer instrument was run according to the manufacturer's protocol.

Blood samples for cytokine measurement were immediately centrifuged, and serum was frozen -80℃ until the assay could be performed. We measured serum HMGB1 levels and several inflammatory [interleukin (IL)-1β, IL-6, IL-17, and tumor necrosis factor (TNF)-α] and anti-inflammatory (IL-10) cytokines from 190 patients, because we did not have sufficient quantities of samples; we could not measure serum levels of HMGB1 or other cytokines in 22 patients. The serum HMGB1 level was measured using a sandwich enzyme-linked immunosorbent assay (HMGB1 ELISA; IBL International, Hamburg, Germany). Intra-assay and inter-assay coefficients of variation (CV) were 5.5-13.7% and 7.6-13.7%, respectively. Also, functional sensitivity was 0.2 ng/mL (lowest HMGB1 concentration with ≤20%). The serum levels of cytokines were determined by Luminex® Performance Assay multiplex kits (R&D Systems, Minneapolis, MN, USA). Analyses were performed in accordance with manufacturer's protocol.

The genotype frequencies of three SNPs were tested for Hardy-Weinberg equilibrium using Haploview (v4.2) (http://www.broadinstitute.org/haploview). No significant deviation from Hardy-Weinberg equilibrium was observed. Continuous variables are expressed as median with range. Student's t-test or the Mann-Whitney U-test, depending on the normality of distribution, were used to compare continuous variables between two groups, and the Kruskal-Wallis test was used for comparisons among three groups. Also, the χ² and Fisher's exact tests (for small numbers) were used to compare categorical variables. To evaluate whether these SNPs could influence clinical outcomes, univariate and multivariate Cox regression analyses were performed including all clinical data from medical records, with adjustment for age, gender. Survival curves were obtained with using the Kaplan-Meier method with the log-rank test. All statistical analyses were performed using the Statistical Package for the Social Sciences (ver. 19.0; IBM, Armonk, NY, USA). A two-tailed p value<0.05 was considered to indicate significant difference.

The clinical characteristics and comparisons of survivors and non-survivors are presented in Table 1. The median durations of stay in the ICU and hospital were six (range 1-104) and 20 (1-830) days, respectively. The ICU and hospital mortality rates were 29.2% and 36.3%, respectively. In total, 150 patients (70.8%) were diagnosed with septic shock; 143 (67.5%) received ventilator care. Also, 86 (40.6%) and 6 (2.8%) patients had renal failure and neutropenia on admission, respectively. The most common primary site of infection was respiratory tract (57.5%), and 72 patients (34.0%) had positive blood cultures (Table 1). The underlying diseases present in all the patients are shown in Table 2. Oncologic and chronic heart diseases were the most common.

When we performed rs1045411 genotyping, the proportion of patients groups by genotype was GG (71.2%), GA (26.4%), and AA (2.4%), respectively. There was no significant difference in other clinical characteristics and outcomes among these three groups (data not shown). When patients were categorized into those with GG (n=151) versus those with GA+AA (n=61), the proportions of A alleles among patients with septic shock, severe sepsis, and sepsis were 31.3%, 23.7%, and 20.8%, respectively. However, there were no significant differences in this allele according to sepsis severity (data not shown).

As underlying comorbidities or primary site of infection could influence clinical outcomes, we also evaluated whether the categories of SNPs (GG and GA+AA) had different implications according to specific comorbid conditions or primary site of infection. In patients with chronic lung diseases or diabetes, patients with the variant A allele had higher positive blood culture rates than those with GG genotype. This allele was also associated with a higher Gram-positive blood culture rate in male patients and Gram-negative culture rate in female and non-survivors. Patients with chronic renal disease had a higher incidence of shock. In cytokine analysis of patients with chronic lung diseases, patients with the variant A allele had significantly higher levels of IL-1β, IL-6, IL-10, IL-17, and TNF-α than those with GG genotype. In analysis of those with diabetes, patients with the variant A allele had higher IL-17 levels. In patients with sepsis caused by respiratory tract infection, patients with this allele had higher levels of IL-10 and IL-17 than those with GG genotype (Table 3).

A univariate Cox proportional hazards model was performed to evaluate the influence of the variant A allele of rs1045411 on survival in all patients, as well as some subgroups that showed significant findings. However, the A allele was not significantly associated with survival (Table 4), and thus, we could not clarify a significant Kaplan-Meier survival curve for this allele (data not shown).