Genetic variants in the TIRAP gene are associated with increased risk of sepsis-associated acute lung injury

The present study was reviewed and approved by the Ethics Study Board of Zhongshan Hospital, Fudan University, Shanghai, China (No: 2006-23). Informed written consent was obtained from all subjects or from their legal surrogates before enrollment. Definitions of sepsis and ALI/ARDS were in accordance with the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference [20] and the American-European Consensus Conference statements (AECC) (Additional file 1: Supplemental Table S1) [21]. All sepsis subjects enrolled had either severe sepsis or septic shock. All patients were selected from the Emergency and Respiratory ICUs at Zhongshan Hospital (Shanghai, China), and were treated according to the Surviving Sepsis Campaign guidelines [22]. Exclusion criteria included age below 18 years, severe chronic respiratory disease, severe chronic liver disease (defined as a Child-Pugh score of > 10), using of high-dose immunosuppressive therapy and AIDS patients. All sepsis patients were screened daily for ALI/ARDS development and those who fulfilled the AECC criteria for ALI/ARDS were considered as ALI cases, which included ALI and ARDS patients; whereas those patients who did not develop ALI/ARDS during hospital stay were considered as sepsis alone patients. Baseline characteristics, source of infection, and Acute Physiology and Chronic Health Evaluation (APACHE) II scores of all patients were obtained during ICU stay. Sex- and age-matched controls were selected from healthy blood donors. Questionnaires including smoking, chronic illness and the history of ALI or sepsis were obtained from all control subjects. Healthy controls were defined as individuals without any recent acute illness, any chronic illness and a history of ALI or sepsis. To reduce the potential confounding from ethnic backgrounds, we only enrolled people with self-reported origin of central Han Chinese, including indigenous people from Zhejiang Province, Jiangsu Province, Anhui Province and Shanghai [23, 24].

Tag SNPs were selected based on the data of Han Chinese in Beijing (CHB) from the HapMap project phase II [25]. Three tag SNPs (rs595209, rs3802813 and rs8177375) for the 11.85-kb region encompassing the entire TIRAP gene were identified by Tagger within Haploview using the following tagging criteria: pairwise tagging of the HapMap CHB Population with r² ≥ 0.8 and a minor allele frequency (MAF) ≥ 5%. Additionally, two coding SNPs (rs8177374 predicting Ser180Leu, and rs7932766 predicting Ala186Ala) were also genotyped in this study as they have showed evidence of association with other inflammatory diseases [15‐19]. These were not genotyped as part of the HapMap project.

Genomic DNA was extracted from whole blood using FlexiGene DNA Kit (Qiagen Hilden, Germany) following the manufacturer's protocol. Genotyping was performed by direct sequencing at the Chinese National Human Genome Center in Shanghai, China. Using Primer 3 software, we designed two primers to completely incorporate the five SNPs. A 693 bp fragment harboring rs595209 and rs3802813 was amplified using the following primers: forward, 5'-TGGTGAAACCCCGTCTCTAC-3' and reverse, 5'-TGGCACAGCTCGGACACTAT-3'. Another 519 bp fragment harboring rs7932766, rs8177374 and rs8177375 was amplified using the following primers: forward, 5'-GCCAGGCACTGAGCAGTAGT-3' and reverse, 5'-ATGTTCTGAGCCCTTCGTGT-3'. The PCR cycling conditions for both fragments involved a denaturation step at 95°C for 5 min and then 35 cycles of 95°C for 30 s, 57°C for 40 s and 72°C for 45 s, with a final elongation cycle of 72°C for 8 min.

The sequencing reactions were performed using Applied Biosystems BigDye (version 3.1) chemistry (Applied Biosystem, Foster City, CA, USA), and the sequences were resolved using an ABI 3730 Genetic Analyzer. Analyses of the sequence traces were performed using the Staden package and double scored by a second operator [26]. A duplicate were added to each 96-well sample plate for quality assurance and quality control validation of inter-plate discordance, and we placed an extra 10 duplicates into our sample set in order to test for experiment-wide discordance. The data completion rate was 99%.

The demographic variables between different groups were compared by χ² test for categorical variables and by one-way ANOVA or Student's t-test for continuous variables. Genotype distributions were evaluated for departures from Hardy-Weinberg equilibrium by the Haploview v4.1 software [27]. The differences in allele and genotype distributions between groups were examined for statistical significance with χ² test or with Fisher's exact test when appropriate. P values for genotypic distributions were calculated using the global genotype test. Allele counts in cases and controls were used to calculate the OR and the 95% CI. Multiple logistic regression was used to evaluate if each SNP was independently associated with ALI when adjusted for the potential confounding effects of important clinical variables. When comparing ALI patients to sepsis alone patients, age, gender, body mass index (BMI), history of smoking, diabetes, liver cirrhosis and APACHE II score were included in the multivariate models because of their established association with ALI [28‐30]. When comparing ALI patients to healthy controls, age, gender, BMI and history of smoking were included in the multivariate models. A two tailed P-value of < 0.05 was considered statistically significant, whereas a value of corrected P < (0.05/number of tests), was considered significant after Bonferroni correction. The linkage disequilibrium (LD) between SNPs was calculated in terms of r² values by the Haploview v4.1 software [31]. Two-locus haplotypes were estimated using EM algorithm by Haploview v4.1. The haplotype-specific association was analyzed by Haploview v4.1 and the omnibus test were was analyzed by PLINK [32]. The case/control omnibus test is an H-1 degree of freedom test, where H is the number of different haplotypes. Each haplotype was compared with all other haplotypes as the reference in calculating the OR. The SNP-SNP interaction (epistasis) was also investigated among the SNPs using PLINK [32]. A P-value less than 0.05 was considered statistically significant. The software used for statistical calculations was the SPSS 15.0 (SPSS Inc., Chicago, IL, USA) unless specified.

From February 2006 to August 2009, a total of 278 sepsis-associated ALI (103 ALI, 175 ARDS) and 288 sepsis alone patients were enrolled in this study. An additional population of 298 ethnic-matched healthy blood donors was recruited for comparison. The baseline characteristics of the study population are shown in table 1. The primary source of infection was the lungs, involving 87.1% of the combined sample of sepsis alone and ALI patients. There was no significant difference in age, gender, BMI, diabetes, liver cirrhosis and history of smoking between ALI patients and sepsis alone patients (P > 0.05). Sepsis-associated ALI patients had average a higher APACHE II scores and mortality ratio than sepsis alone patients (P < 0.001), although they were comparable in infection sites.

Of the three tag SNPs, the MAFs of rs595209, rs3802813 and rs8177375 in our sample were 32.1%, 16.3% and 11.6% respectively. The MAFs of the two reported SNPs (rs7932766 and rs8177374) were lower than 5% in our data (1.1% and 2.5%, respectively). The genotype and allele frequencies of the five polymorphisms in all studied subjects are summarized in Additional file 1: Supplemental Table S2. The genotyping success rates ranged from 98% to 99% and did not diverge from Hardy-Weinberg equilibrium (P > 0.05) (Additional file 1: Supplemental Table S2).

Single locus analysis showed two SNPs (rs595209 and rs8177375) were associated with ALI risk, whereas other three SNPs (rs3802813, rs7932766 and rs8177374) showed no association (Table 2). The alleles of rs595209A and rs8177375G occurred significantly more frequently in the ALI group than in both the healthy control group (OR = 1.47, P = 0.0027; OR = 1.97, P = 0.0001, respectively) and the sepsis alone group (OR = 1.44, P = 0.0041; OR = 1.82, P = 0.00079, respectively), which remained significantly after Bonferroni correction. Moreover, in multivariate analyses after adjustment for covariates, both SNPs were still significantly associated with the development of ALI when compared with healthy control group (OR_adj = 1.43, P_adj = 0.0036; OR_adj = 1.81, P_adj = 0.00031, respectively) and sepsis alone group (OR_adj = 1.36, P_adj = 0.0082; OR_adj = 1.51, P_adj = 0.0041, respectively). The genotypes frequencies of rs595209 and rs8177375 in the ALI group were also significantly different from that in the healthy control group (P = 0.0034; P = 0.00047, respectively) and the sepsis alone group (P = 0.0084; P = 0.0026, respectively), the significance remained present in a multivariate analysis controlling for covariates and after Bonferroni correction (Table 2). However, the difference of the allele and genotype frequencies of rs595209 and rs8177375 between subjects with sepsis alone and healthy controls was not statistically significant (P > 0.05) (Additional file 1: Supplemental Table S3). No significant interaction was found between rs595209 and rs8177375 (P > 0.05). The low r² value between the rs595209 and rs8177375 (r² = 0.17) indicated the two signals were independent not simply due to LD (Figure 1).

We next carried out haplotype analysis on the basis of these two associated SNPs and found that a two-allele model of risk provided the strongest predictor of risk, which was highly significant (P = 0.00006 in ALI group vs control group and P = 0.00003 in ALI group vs sepsis alone group). The two SNPs (rs595209 C/A and rs8177375 A/G) generated three common haplotypes with frequency >5%: CA with frequency of 66.4%, AA with frequency of 22.1% and AG with frequency of 9.8%. A global test showed significant association with omnibus p value of 0.015 in ALI group vs control group and 0.011 in ALI group vs sepsis alone group. In haplotype-specific analysis, the frequency of haplotype AG (rs595209A, rs8177375G) in the ALI samples was significantly higher than that in the healthy control group (OR = 2.13, 95% CI: 1.46-3.09, P = 0.00006) and the sepsis alone group (OR = 2.24, 95% CI: 1.52-3.29, P = 0.00003), suggesting its role as an ALI risk haplotype. After adjustment for covariates, the haplotype AG still associated with increased risk of ALI (Table 3). Another haplotype CA (rs595209C, rs8177375A) appeared protective, as it was found to be more frequent in both healthy control and sepsis alone group than in the ALI group. Carriers of the CA haplotype had a lower risk for ALI compared with healthy control (OR = 0.69, 95% CI: 0.54-0.88, P = 0.0003) and sepsis alone group (OR = 0.71, 95% CI: 0.55-0.91, P = 0.0006). The protective effect of haplotype CA remained significant after adjustment for covariates (Table 3). No association was found between haplotype AA and ALI susceptibility. Similar to the results from individual SNP analysis, the difference in the three haplotypes frequencies between sepsis alone patients and healthy controls was not statistically significant (P > 0.05) (Additional file 1: Supplemental Table S4).