Association of common genetic variation in the protein C pathway genes with clinical outcomes in acute respiratory distress syndrome

Acute respiratory distress syndrome (ARDS) is a common cause of respiratory failure characterized by acute pulmonary edema and lung inflammation [1]. ARDS occurs in both adults and children and has an incidence of approximately 200,000 patients per year in the United States. Estimates of mortality range from 18 % to 58 % [2, 3]. The majority of deaths among patients with ARDS are attributed to multiorgan failure [1, 4].

A number of experimental and human studies suggest that excessive activation of coagulation is associated with increased mortality in patients with ARDS [5‐11]. Activated protein C (PC) is an endogenous regulator of coagulation that has both anticoagulant and anti-inflammatory effects [12, 13]. Protein C is activated by thrombin in the presence of thrombomodulin (TM), and membrane-bound endothelial protein C receptor (EPCR) potentiates this activation [14]. Human lung epithelial cells express protein C, EPCR, and TM, and the lung epithelium can actively modulate the protein C pathway [15]. Alterations in plasma levels of protein C, TM, and soluble EPCR are associated with increased mortality and greater severity of illness among patients with ARDS [13, 16‐18]. It is unclear if these alterations and their associations with clinical outcomes are determined exclusively by environmental factors that precipitate ARDS, such as the virulence of the infection, the extent of aspiration, and the severity of shock, or whether they are also influenced by genetic variation.

Common genetic variation (e.g., polymorphisms with minor allele frequency >5 %) in the genes encoding for PC, EPCR, and TM has been well-characterized and is associated with adverse clinical outcomes in disorders such as sepsis and cardiovascular disease [19‐29]. Since the genetics of diseases such as ARDS are likely complex and therefore likely to involve several low-penetrance loci [30, 31], analyzing multiple variants of genes encoding components of the same physiological cascade may prove to be a more powerful approach than studies of single candidates [32‐35]. We hypothesized that common genetic variations in the genes encoding for protein C, EPCR, and TM are associated with adverse clinical outcomes in patients with ARDS. We also hypothesized that variants in each individual gene may each have a small effect and that a combination of these variants would be associated with adverse clinical outcomes.

The baseline characteristics of the self-identified white patients of European ancestry enrolled in the FACTT trial stratified by the 320 for whom DNA was available and the 321 for whom DNA was not available are depicted in Table 1. The baseline characteristics of the study population are similar to the white patients of European ancestry in the FACTT trial for whom DNA was not available. Genotype frequencies of the assayed SNPs in the protein C, EPCR, and TM genes and the frequencies stratified by mortality at 60 days are depicted in Tables 2, 3 and 4.

In the EPCR gene, the GC/GG genotypes of the rs9574 SNP were associated with higher mortality (24 % vs. 11 %) compared with the CC genotype (p = 0.04 after correction for multiple SNPs tested in the EPCR gene). Two other SNPs in the EPCR gene—rs2069952 and rs2069948—that were highly correlated with rs9574 (r ² = 0.95), were also associated with mortality at 60 days; however, when tested in a regression model that included these two SNPs and rs9574, only rs9574 retained the association with mortality, suggesting that the effect was not independent of the effect of the rs9574 SNP. The association of the rs9574 SNP with mortality was independent of age, severity of illness, sepsis as the primary cause of ARDS, and the fluid management arm (Table 5).

Two SNPs in the TM gene were independently associated with mortality at 60 days. The CC genotype of the rs3176123 SNP was associated with higher mortality (57 % vs. 20 %) compared with the AC/CC genotypes (p = 0.002 corrected for multiple SNPs tested). The GG genotype of the rs1042580 SNP was also associated with higher mortality (35 % vs. 19 %) compared with the AG/AA genotypes (p = 0.02 corrected for multiple SNPs tested). These two SNPs had limited correlation with each other (r ² = 0.14), and both SNPs were independently associated with mortality when analyzed in a joint logistic regression model including the two SNPS and age, presence of non-pulmonary sepsis, APACHE score, and allocation to fluid management strategy (Table 5). The association of the rs3176123 and rs1042580 SNPs with mortality was independent of age, severity of illness, sepsis, and the fluid management arm (Table 5). Of the protein C SNPs analyzed, the AA genotype of the rs1799810 SNP showed a trend toward increased mortality (26 % vs. 19 %) compared with the TT/AT genotypes, but this trend was not statistically significant (p = 0.18).

We tested the effect of a combination of these unfavorable SNPs on mortality at 60 days in a joint logistic regression model. On the basis of their biological interaction in a common pathway, we hypothesized that multiple unfavorable SNPs would have a more significant impact on the protein C pathway and therefore would be associated with increased mortality. The CC genotype of the rs3176123 SNP, the GG genotype of the rs1042580 SNP in the TM gene, and the GC/CC genotypes of rs9574 SNP in the EPCR gene were independently associated with mortality at 60 days. This association was independent of age, severity of illness, sepsis, and the fluid management arm (Table 5). We also examined mortality at 60 days among patients stratified by the number of high-risk genotypes possessed by each individual. There was an increase in mortality at 60 days with the presence of each additional high-risk genotype: 5.8 % [1.2–16.2], 20 % [15‐26], and 41 % [27–57] (p < 0.001) (Fig. 1). The association was independent of age, severity of illness, sepsis, and the fluid management arm.

The number of VFDs was also compared among patients stratified by the number of high-risk genotypes carried by each individual. There was a decrease in the number of VFDs with the presence of each additional high-risk genotype (p = 0.01) (Fig. 2). This association was also independent of age, severity of illness, sepsis, and the fluid management arm.

The number of coagulation, renal, cardiovascular, and central nervous system organ failure-free days was compared among patients stratified by the number of high-risk genotypes carried by each individual. There was also a decrease in the number of coagulation, renal, cardiovascular, and central nervous system organ failure-free days with the presence of each additional high-risk genotype (Fig. 3).

Given our a priori hypothesis that genetic heterogeneity in all three genes (protein C, EPCR, and TM) in this pathway would have an effect on mortality in patients with ARDS, we conducted an additional analysis with four SNPs in the model: the three SNPs that are independently associated with mortality at 60 days and a fourth, the rs1799810 SNP in the protein C gene that had a trend toward increased mortality which did not achieve statistical significance. Even though the protein C SNP rs1799810 individually did not have a statistically significant association with mortality, we found that the mortality increased from 10.3 % [2.2–27.3] for those with no high-risk SNPs vs. 15.2 % [9.9–21.7], 28.5 % [19.8–38.5], and 63.6 % [30.8–89.1] for those with one, two, and three high-risk SNPs, respectively. Those with greater numbers of high-risk SNPs had fewer VFDs and fewer organ failure-free days (see Additional file 1). We did not find any joint haplotype effect on either the primary or secondary outcomes in any of these three genes.

The results of this study indicate that common genetic variations in the protein C pathway are associated with adverse clinical outcomes in adult patients with ARDS. The GC/CC genotypes of the rs9574 SNP in the EPCR gene, as well as the GG genotype of the rs1042580 SNP and the CC genotype of the rs3176123 SNP, both in the TM gene, were independently associated with mortality in ARDS. We also found that a combination of these high-risk genotypes had an additive effect on mortality, ventilator-free days, and organ failure-free days in ARDS. However, the effect of SNPs was not mediated via their effect on the levels of soluble TM or EPCR in plasma, suggesting the SNPs may have their effect through other mechanisms related to altered functioning of these molecules. This finding has major implications for current understanding of the pathogenesis of ARDS. Abnormalities of the coagulation pathway and its regulatory proteins and the association of these abnormalities with clinical outcomes in patients with ARDS have been described previously [6‐11]. However, these abnormalities were thought to be due largely to environmental factors (e.g., severity of illness, virulence of organisms, severity of lung injury). The present findings suggest that genetic susceptibility may contribute to this dysregulated coagulation and the poor clinical outcomes in patients with ARDS.

Both intraalveolar and systemic coagulation are activated in patients with ARDS [8‐10, 44]. Although intraalveolar fibrin deposition may have beneficial effects on gas exchange by sealing leakage sites and compartmentalizing infection, excessive fibrin deposition can be harmful because it can activate neutrophils and fibroblasts, compromise endothelial integrity, contribute to a loss of surfactant activity, decrease alveolar fluid clearance, and induce thrombotic obstruction of the microcirculation [6, 7]. The injury to the pulmonary microcirculation via inflammatory and thrombotic mechanisms may contribute to the increase in the pulmonary dead space fraction that is an independent predictor of mortality in ARDS [45]. In addition, systemic activation of coagulation may also contribute to hypercoagulability and the development of multiorgan failure with widespread microvascular thrombus formation [1, 4, 17].

Activated protein C is an endogenous regulator of coagulation that has both anticoagulant and anti-inflammatory effects. Alterations in plasma levels of protein C, EPCR, or TM that may contribute to decreased availability of activated protein C are associated with adverse clinical outcomes in patients with ARDS [16, 17]. In the present study, we examined the association of genetic variation in these protein C pathway genes with clinical outcomes in patients with ARDS. The results therefore add to the significance of the previously reported associations of alterations in these protein biomarker levels with protein C pathway proteins and adverse clinical outcomes, and they suggest that genetic predisposition may play a role in these previously reported associations of abnormalities in protein C pathway proteins and clinical outcomes in ARDS.

Although we chose a hypothesis-free approach within the candidate genes, the SNPs that were associated with adverse clinical outcomes in this study have all previously been reported (directly or indirectly via a tightly linked SNP) to have an association with protein levels and/or clinical outcomes in other conditions. The rs3176123 SNP in the TM gene is in tight LD with the rs1042579 SNP (r ² = 1), a coding region nonsynonymous SNP. The minor allele of this SNP is associated with increased incidence of venous thrombosis [46, 47]. In the present study, the minor allele homozygotes had increased mortality. The GG genotype of the rs1042580 SNP in the TM gene, which was associated with increased mortality in this study, has been associated with increased cardiovascular disease in females in combination with factor V Leiden in previous studies [32]. In our present study, the rs1042580 SNP was associated with variation in the soluble TM levels in plasma, but that explained only a minor part of the overall effect of the SNP on mortality, suggesting that the SNP may have its effect of through other mechanisms, including the possibility of altering the functional activity of TM. Finally, in a previous study, the haplotype tagged by the C allele of the rs9574 SNP in the EPCR gene was associated with increased levels of activated protein C levels and reduced risk of venous thromboembolism [21, 22]. In the present study, we found that the G allele at this locus is associated with increased mortality. The rs9574 SNP was associated with variation in the soluble EPCR levels in plasma, but that variation explained only a part of the overall effect of the SNP on mortality, suggesting that the SNP may have its effect through other mechanisms, including the possibility that the SNP may have an additional effect on the functional activity of EPCR.

A novel feature of this study is the combined analysis of the three genes. In a complex illness such as ARDS, the impact of genetic factors is likely to be determined by several variants of small effect size and their possible interactions. When acting together, these gene variants may affect the disease outcomes more profoundly than do the single predisposing variants [32]. An analysis comprising several genes that belong to the same pathway may reveal cumulative allelic effects (additive or multiplicative) as compared with single SNPs, which individually may have only a modest effect or no measurable impact on clinical outcomes. The combined analysis of these three proteins is based on the hypothesis that the three proteins of the protein C pathway act together to generate the final product (i.e., activated protein C, which might be the biologically active product responsible for the clinical effect). Given our a priori hypothesis that all three genes in this pathway would have an effect on mortality in patients with ARDS, we carried out an additional exploratory analysis using the rs1799810 SNP from the protein C gene based on the biological role of protein C in the common pathway, even though this SNP individually had only showed a trend toward association with increased mortality. Interestingly, the combination of high-risk genotypes from all three genes in the pathway was associated with mortality and adverse clinical outcomes upon addition of the protein C SNP (which was not independently associated with mortality) to the model.

There are several strengths of our study. These include the well-characterized clinical phenotype and the multicenter cohort of patients from a large, well-designed clinical trial. The previously reported association of clinical outcomes with protein levels and of the coagulation genes chosen for this study provides the biological plausibility for the candidate genes chosen for this study. All the SNPs associated with poor outcomes in ARDS in our study (or a tightly linked SNP) have been previously reported to affect protein levels and clinical outcomes in other populations [19‐21, 26, 32, 48]. This finding further supports the biological plausibility and increases the prior probability for the reported associations. This point is important because genetic association studies with high prior probability are likely to have a low probability of reporting false-positive associations [49]. Finally, we found not only an association of genotypes with mortality but also a strong association with multiple-organ failure-free days. This association is significant because multiorgan failure is the most common attributable cause of mortality among patients with ARDS [1, 4]. This could be related to the fact that multiorgan failure and death are associated with each other; alternatively, the association of the high-risk genotypes with the number of organ failure-free days may suggest that multiorgan failure may be an intermediate phenotype, which in turn leads to the higher mortality associated with the high-risk genotypes.

This study also has some limitations. First, the results were obtained only in self-identified white patients of European ancestry. Second, we did not correct for study-wide multiple comparisons. Since our hypothesis was based on the biological plausibility of the three genes in the protein C pathway and previously reported associations with protein levels, we corrected for multiple comparisons for multiple SNPs tested within each gene rather than for study-wide multiple comparisons as is the case with hypothesis-free genome-wide association studies. Therefore, it is unlikely that these associations were due to chance alone. However, as is true for all genetic epidemiology studies, these findings need to be tested and validated in other patient cohorts.

On the basis of this cohort, the GC/GG genotypes of the rs9574 SNP in the EPCR gene, as well as the CC genotype of rs3716123 and the GG genotype of the rs124580 SNPs in the TM gene, were associated with increased mortality in adults with ARDS. A combination of these genotypes had an additive effect on mortality, ventilator-free days, and organ failure-free days in patients with ARDS. These findings suggest that genetic differences may be at least partially responsible for the observed associations between dysregulated coagulation and poor outcomes in patients with ARDS. If confirmed, these findings may support the potential value of testing targeted therapies for ARDS in genetically predisposed patients.