Introduction

Sepsis is the leading cause of death in critically ill patients, with a hospital mortality ranging from 30% to 50% in patients with septic shock [1, 2]. Epidemiological studies on inflammatory mediators have suggested a strong genetic relationship on the susceptibility and outcome of sepsis. TNF-α-308A and TNF-β-252A polymorphisms have been reported to be associated with susceptibility to meningococcal disease [3] and a high risk for septic shock in patients admitted with community-acquired pneumonia [4], respectively. Regarding the risk of mortality Mira et al. [5] found that the TNF-α-308A allele had a 3.7-fold increase in death among septic shock patients whereas homozygosity for B2/B2 of TNF-β, which has been associated with a high tumor necrosis factor (TNF) response, increases the probability of severe posttraumatic sepsis in trauma by 3.4 [6]. With regard to the interleukin (IL) 1 pathway, the IL-1raA2 allele seems to be more prevalent among patients with severe sepsis than among healthy individuals [7].

Inflammation and coagulation are closely linked and most patients with severe sepsis and septic shock have coagulation abnormalities [8]. In this setting nonsurvivors present a particular hemostatic profile characterized by a marked activation of coagulation and a more intense inhibition of fibrinolysis [9, 10]. TNF-α and IL-1 increase the synthesis and release from endothelial cells of plasminogen activator inhibitor 1 (PAI-1), a key inhibitor of fibrinolysis and also decrease plasminogen activator synthesis [11, 12]. Thus, impaired fibrin degradation due to high circulating levels of PAI-1 may contribute to enhanced microvascular fibrin deposition, which favors the development of organ dysfunction [13]. PAI-1 may show a genomic variation in its promoter: a single base-pair deletion (4G) or insertion (5G). Homozygosity for the 4G allele is reported to have higher basal and inducible concentrations of PAI-1 and greater procoagulatory activity than one or two copies of the 5G allele [14]. This has been associated with an increased risk of death from meningococcal sepsis in children [15–17] and a poor survival after severe trauma [18]. On the other hand, the G4 variant of the PAI-1 gene seems to be associated with a weak increase to the risk of coronary disease [19], and G4/G4 genotype is associated with an increase in thromboembolic neurological complications after cardiac surgery [20]. Moreover, several reports claim for an involvement of the G4 variant in a miscellany of other cardiovascular [21], and systemic diseases [22–24], although these studies need confirmatory evidence.

The aim of the study was to evaluate the effect of the 4G/5G PAI-1 gene polymorphism in the development of organ failure and outcome in adult patients with septic shock. In addition, we evaluated the association of genomic variations in TNF-β and IL1-ra on the influencing the PAI-1 4G/4G gene effect on the primary endpoints.

Patients and methods

Patients

During the 12-month period from January 2003 to January 2004, 301 patients were admitted to the medical ICU of the Hospital Clinic of Barcelona. The study population consisted of 224 consecutively patients (mean age 62.3 ± 15.9 years, range 18–85; 63.8% men) with a minimum ICU stay of 48 h and without short-term irreversible diseases, as reported in the Electronic Supplementary Material (ESM; Fig. 1). The patients were included and followed until hospital discharge. Furthermore, 80 healthy blood donors recruited consecutively were used as the control population. All the subjects were of white descent and living in or around Barcelona. Informed consent was obtained from the patients or their relatives within 24 h after admission, and none refused to participate. The protocol was approved by the institutional review board of the Hospital Clinic of Barcelona.

Epidemiological data, the primary site of infection, and infection-related organisms were recorded (Table 1 and ESM). Clinical infections were defined according to the Centers for Diseases Control and Prevention criteria [25]. Severe sepsis and septic shock were defined according to SCCM/ESICM/ACCP/ATS/SIS consensus conference [26], as described in the ESM. Septic shock was considered nosocomial (late) when observed after 48 h after ICU entry. Severity indexes including Acute Physiology and Chronic Health Evaluation (APACHE) II, Simplified Acute Physiology Score (SAPS) II, and Sequential Organ Failure Assessment (SOFA) were calculated at ICU admission and thereafter on a daily basis. Within 48 h of ICU entry 88 patients exhibited septic shock and 26 severe sepsis (Fig. 1); of the remaining patients 51 had sepsis and 59 had noninfectious systemic inflammatory response syndrome (SIRS). Mean APACHE II, SAPS II, and SOFA scores at entry were 16.8 ± 6.3, 37.4 ± 12.3, and 8.6 ± 3.1 at ICU entry, respectively. Multiorgan failure was considered in case of acute progressive dysfunction of two or more organs systems, with a minimum failure score of 3 points for each organ. Acute respiratory distress syndrome was considered according to the European Consensus score [27], with patients showing a Murray et al. [28] score of 2.5 or higher. Blood samples and the information used in the study were coded and patient confidentiality was preserved. Biochemical studies and genotyping were performed in the patients after completion of follow-up. Patients with septic shock required surgery more frequently than patients with sepsis. Coagulation parameters at ICU entry did not differ among the groups of patients, except for greater fibrinogen levels in septic subjects than in those with noninfectious SIRS (Table 1).

Fig. 1
figure 1

Flowchart of the patient groups according to inclusion criteria

Full size image
Table 1 Epidemiological and clinical data of the patients on ICU admission (NA not applicable)
Full size table

Biochemical and genotype studies

Collection of venous blood samples and processing are reported in ESM. Prothrombin and activated partial thromboplastin times, fibrinogen, and D-dimer levels were measured [29]. Plasma antigen related to PAI-1 was measured by an enzyme-linked immunosorbent assay (Biopool, Umea, Sweden) based on a double-antibody principle [30]. Genomic DNA was extracted from 100 μl whole blood with a silica gel column method (QIAamp DNA blood mini kit, Qiagen, Hilden, Germany), and stored at −80 °C until further use. Genotyping was validated by direct sequencing of a group of random samples. We determined the PAI-1 4G/5G, TNF-β-252G B1/B2, and IL-1ra A2 polymorphisms [31–33], and methods for detection are shown in the ESM.

Statistical analysis

Results are given as means ± SD. Differences between groups were evaluated with the Fisher's exact and Student's t tests. Bonferroni's correction was applied to adjust the type I error in multiple comparisons. Pearson's correlation coefficients between continuous variables were calculated. Genotype distributions were tested for the Hardy-Weinberg equilibrium. To evaluate the variables independently associated with ICU and hospital mortality a logistic regression model (forward stepwise selection) was performed. In this model quantitative variables were introduced in its continuous form, and therefore the odds are referred to the natural unit of the predictor. All tests were performed with a bilateral significance level of p < 0.05. Statistical analysis was performed with SPSS statistical software (version 11.0.4; SPSS, Chicago, Ill., USA).

Results

Patient follow-up data

Data on the patients during ICU stay are summarized in Table 2. Patients with septic shock showed a significantly greater worsening in SAPS II score than those admitted because of sepsis or noninfectious SIRS. Thirty-two (14.3%) patients died in the ICU, mainly due to multiorgan failure. Mortality was higher in those with septic shock (27.3 vs. 5.8%, p < 0.001). Overall hospital mortality reached 20.1%, with higher post-ICU ward mortality in patients admitted for nonseptic SIRS (12.5 vs. 3.6%, p < 0.001). Logistic multivariate analysis showed that the clinical variables independently associated with ICU mortality in patients admitted with septic shock were the SAPS II score at admission [odds ratio (OR) 1.08, 95% confidence interval (CI) 1.02–1.15) and the worsening of this score during ICU stay (1.16, 1.09–1.24). Moreover, 14 patients (range between the groups 3.4–11.5%) developed nosocomial septic shock while in the ICU.

Table 2 Patients characteristics during Intensive Care Unit stay
Full size table

Genetic analysis

The distribution of the PAI-1 genotype was comparable in patients and controls and between the different subgroups of patients studied (Table 3). Likewise the overall allele frequency of 4G/5G was 0.49/0.51 in all patients, which did not differ from that observed in controls (0.54/0.46). Regarding the TNF-β genotype, homozygosity for the B2 allele was more prevalent in patients with septic shock than in controls (61.3 vs. 45.0%, p = 0.04). By contrast, the distribution of the IL1-ra genotype among the groups of patients did not differ from that of the healthy population. In patients with septic syndromes PAI-1 alleles distributed evenly with TNF-β and IL-1ra alleles (Table 4). Except for the higher prevalence of homozygosity for the B2 allele in patients admitted with septic shock, none of the genotypes analyzed was associated with diseases leading to ICU entry or the origin of infectious disease.

Table 3 Distribution of genotype polymorphisms among the groups of patients; no statistical differences were observed between the admission groups
Full size table
Table 4 Distribution of PAI-1 genotype polymorphisms according to TNF-β and IL1-ra genotype polymorphisms in patients with septic syndromes; no statistical differences were observed between the groups
Full size table

Coagulation abnormalities in organ dysfunction and septic shock

Plasma levels of the PAI-1 antigen on admission were correlated with the Δ SOFA score (r = 0.28, p = 0.02) and were greater in patients who died than in survivors (45.3 ± 23.8 vs. 33.8 ± 16.7 ng/ml, respectively, p = 0.02). The 4G/4G patients with septic shock had higher plasma concentrations of the PAI-1 antigen than 4G/5G and 5G/5G subjects (p < 0.001, both; Table 5). Coagulation data at ICU entry were not related to the PAI-1 genotype. However, 4G/4G patients showed a twice the D-dimer concentration of those with other PAI-1 genotypes (p < 0.05, both; Table 5). D-dimer levels at ICU entry were not correlated significantly with the SOFA score, but were also greater in patients with septic shock who died than in survivors (826 ± 919 vs. 513 ± 423 ng/ml, respectively, p = 0.03). A direct correlation between the PAI-1 antigen and D-dimer upon ICU admission was observed (r = 0.49, p < 0.001).

Table 5 Data of patients with septic shock according to the PAI-1 genotype along ICU stay
Full size table

Effect of TNF-β and IL-1ra genotypes on organ dysfunction and mortality

We found no differences in severity scores with respect to baseline values and changes during ICU stay among the groups of patients homozygous to the B2/B2 compared to other forms of the TNF-β gene (data not shown). However, the mortality of B2/B2 patients with septic shock was higher than that in patients carriers of the B1 allele (37.0 vs. 11.7%, respectively, p = 0.01). Mortality was not associated with the TNF-β polymorphism in the remaining groups of patients. On the other hand, neither was the IL1-ra genotype related to susceptibility to infection, clinical data or outcome in any group of the patients studied. Neither of these two genotypes was associated with the appearance of late (nosocomial) septic shock and survival.

Effect of the PAI-1 genotype on organ dysfunction and mortality

The PAI-1 genotype affected the clinical course of patients with septic shock, as 4G homozygotes exhibited a greater worsening in clinical status (i.e., Δ SOFA) than other PAI-1 genotypes (p < 0.05, all; Table 5). The presence of 4G/4G in patients with septic shock was associated with an increase of 1 point in the SOFA score (OR 2.83, 95% CI 1.04–7.67). However, nonsignificantly greater supportive measures (i.e., mechanical ventilation, vasopressors administration, parenteral nutrition) were needed by the 4G/4G patients. Clinical changes in the patients admitted for severe sepsis, sepsis, or noninfectious SIRS were independent of PAI-1 genotype.

ICU mortality in septic shock reached 44.0% in patients with 4G/4G, 23.4% with 4G/5G, and 12.5% in 5G/5G (p = 0.03 for 4G/4G vs. other genotypes; Fig. 2), mostly due to multiorgan failure. The 4G and 5G allele distribution was 0.50/0.50 in alive and 0.69/0.31 in nonsurvivors (p = 0.03). After adjusting for SAPS II at admission the genotypes independently associated with ICU mortality in septic shock were of TNF-B2/B2 (OR 2.83, 95% CI 1.04–7.67) and 4G/4G of PAI-1 (2.23, 1.02–4.85). Taking all the genotypes into consideration, we observed that 37.7% of the subjects with septic shock carrying 4G/4G or B2/B2 died, whereas the mortality decreased to 3.7% when both these polymorphisms were absent (p < 0.001). No additive effect on mortality was observed between the genotypes. The 4G allele of the PAI-1 gene was not related to the appearance of nosocomial septic shock and its mortality, although the number of patients evaluated was too small.

Fig. 2
figure 2

PAI-1 genotype distribution among patients with septic shock according to ICU survival. Patients homozygous for the 4G/4G polymorphism had higher mortality than those with other PAI-1 genotypes (p = 0.03)

Full size image

Discussion

In the present study we observed greater organ dysfunction and a twofold increase in the risk of mortality in patients with septic shock who were homozygous for the 4G-deletion polymorphism of the PAI-1 gene. In addition, similarly to previously published studies [6, 34], we observed that B2/B2 of the TNF-β gene confers a higher susceptibility to septic shock and these subjects have a worse outcome. Moreover, we confirmed the lack of association between the genomic variations in PAI-1, TNF-β, and IL-1ra with each other. The study was underpowered to identify a clinical significant difference in the supportive case among the groups of patients.

It is already known that virtually all patients with sepsis have coagulation abnormalities, with widespread microvascular fibrin deposition, which contributes to organ dysfunction and profuse bleeding from various sites [35]. Inhibition of the fibrinolytic system is a key element of the pathogenesis of fibrin deposition during severe inflammation. In sepsis-induced generalized activation of coagulation the generation of thrombin also initiates fibrinolysis through the release of tissue plasminogen activator. However, the activation of the fibrinolytic system is transient due to a strongly proinflammatory cytokine-induced PAI-1 expression and secretion by endothelial cells [36, 37]. Thus the net result is a proinflammatory and prothrombotic disorder with exhaustion of fibrinolysis and coagulation inhibitors [38]. In agreement with previous findings [39] we found that both plasma PAI-1 and D-dimer were elevated in patients with septic shock, mainly in those who died, and was correlated with the first to organ failure.

In addition to proinflammatory cytokines, the circulating plasma PAI-1 level is affected by a deletion/insertion (4G/5G) polymorphism in the promoter of the PAI-1 gene. Although both alleles bind a transcriptional activator, only the 5G allele binds a repressor protein at an overlapping site. Therefore homozygosity for the 4G allele renders this negative regulator unable to bind, resulting in greater transcription of the PAI-1 gene [9, 11, 14]. In this regard we documented that plasma PAI-1 concentrations in 4G homozygotes doubled than that observed in 5G carriers. As this increase in PAI-1 may show circadian changes [40], we determined the PAI-1 antigen from morning samples. Previous studies have shown that homozygosity for the 4G allele of the PAI-1 gene exerts a harmful effect on the prognosis of some subsets of critical patients, such as children with meningococcal sepsis, in whom the relative risk of dying from meningococcal sepsis increases from 2.0 to 2.4 compared with 5G carriers [15, 17]. In these patients logistical regression indicated a 40% reduction in the probability of death for 4G/5G patients and a 91% reduction for 5G/5G patients compared to 4G homozygotes [15]. Moreover, in patients with severe trauma, Menges et al. [18] found that the 4G allele not only has increased levels of PAI-1 during 14 days of observation after multiple traumas but also entailed a higher prevalence of sepsis and multiple organ failure. As a result 11 of 19 patients with the 4G/4G genotype did not survive, whereas only 8 of the 29 4G/5G patients, and 2 of 13 patients with 5G/5G of the PAI-1 gene died. No studies have yet determined the effect of the PAI-1 genotype in adult patients with septic shock of different origin. In the present study we observed that patients with septic shock without unsalvageable disease resulted in a hospital mortality of 52.0% for 4G/4G, 27.7% for 4G/5G and 18.8% for 5G/5G patients, mainly due to multiorgan failure. This association with PAI-1 gene polymorphism occurred independently of the causative infectious pathogen and was not observed in patients with sepsis and severe sepsis or in those with late (nosocomial) septic shock. However, this study has some limitations due to number of subjects and power of calculation, as well as derived from the complexity of the disease [41], and these initial positive findings needs to be confirmed in a large independent cohort in order to verify that the genetic background or environmental exposures did not affect the results. The G4 variant of the PAI-1 gene has also been implicated in many other diseases [42–49], as is reported in ESM.

On the other hand, in the present study we evaluated other genotypes that coded for the production of TNF-α and IL-1ra and found septic shock to be more prevalent in patients homozygous for the TNF-B2/B2 polymorphism. However, in contrast to the report by Fang et al. [7], the prevalence of IL-1raA2 among patients with severe sepsis and septic shock did not differ from that observed in healthy individuals. More importantly, TNF-B2/B2 was associated with a 2.8-fold increase in the risk of mortality in patients with septic shock. Other candidate genes which regulate the inflammatory response and which may be implicated in the appearance and end result of septic shock are those that code for the heat shock protein, IL-6, IL-10, CD-14, and Toll-like receptors [50], were not evaluated in the present study.

In summary, we observed that patients admitted to the ICU with septic shock had greater organ dysfunction and exhibited a higher mortality when homozygous to the 4G form of the PAI-1 gene. This observation may be related to a procoagulant state in sepsis, as a consequence of fibrinolysis inhibition by PAI-1. However, the evidence from this single association study needs to be replicated to provide insurance against errors and biases. In addition, further studies of combined gene effects and gene-gene interactions as well as therapeutic approaches based on genetic risk in patients with both sepsis and a derangement of coagulation are needed.