Association of TNF2, a TNF-α Promoter Polymorphism, With Septic Shock Susceptibility and Mortality: Title and subTitle BreakA Multicenter Study FREE

Septic shock is the primary cause of death in intensive care units (ICUs). With a mortality rate in excess of 50%, it results in more than 100,000 deaths a year in the United States.^{1 - 2} Sepsis-induced organ failure leading to death appears to be due to the activation of a mediator cascade initiated by microbial components.^{1 - 2} Although the pathophysiology of systemic inflammation and organ dysfunction is complex, tumor necrosis factor alpha (TNF-α) has emerged as a proximal and central cytokine of septic shock.^{2 - 3} Its administration reproduces essentially all the deleterious effects of endotoxin and bacteria, including hypotension, activation of the coagulation cascade, and organ dysfunction.^{2 - 4} Increased TNF-α plasma levels have been reported after endotoxin challenge in healthy volunteers⁵ and in septic shock patients with both gram-positive and gram-negative bacteremia.^{6 - 7} Finally, inhibition of TNF-α, by administration of anti–TNF-α antibodies or soluble TNF-α receptors, protects animals from the lethal effects caused by lipopolysaccharide, gram-negative, and gram-positive bacteria challenge.⁸ Despite this strong evidence for a causal relationship between TNF-α and development of septic shock, the three, phase-3, anti-TNF clinical studies did not show any improvement in survival rates.^{9 - 11}

Human monocytes, the main source of TNF-α synthesis, have large and stable differences in TNF-α production levels.^{12 - 13} This diversity in TNF-α synthesis was initially correlated with class II HLA genotypes.^{12 - 13} However, the location of the TNF-α gene within the major histocompatibility complex raised the possibility that genetic alterations in the TNF-α locus may be involved directly in high TNF-α production. Several polymorphisms have been identified inside the TNF-α promoter.¹⁴ Among these TNF-α variants, a polymorphism that directly affects TNF-α expression was located at nucleotide position −308.¹⁵ This polymorphism results in 2 allelic forms, 1 in which a guanine defines the common allele TNF1 and 1 in which an adenosine defines the uncommon alleleTNF2. The TNF2 polymorphism has been reported to be strongly linked to HLA-DR3, raising the possibility that the association of HLA-DR3 with high TNF-α production may be related directly to the TNF2 allele .¹⁵ Moreover, the TNF2 allele has been found to correlate with enhanced spontaneous and stimulated TNF-α production both in vitro^{16 - 17} and in vivo.^{18 - 19} Furthermore, the TNF2 polymorphism has been associated with morbidity and mortality of severe forms of cerebral malaria,²⁰ fulminans purpura,²¹ and mucocutaneous leishmaniasis.²²

To our knowledge, no study has evaluated the association between septic shock and the TNF2 allele. Strüber et al²³ failed to find a relationship between sepsis and the TNF2 allele. However, their finding does not exclude a potential association between the TNF2 allele and septic shock, which represents the most severe form of gram-positive and gram-negative infections.^{2 ,4 ,8} Indeed, the reported associations between the TNF2 polymorphism and infectious diseases have been observed only in the most severe infections.^{20 - 22}

Therefore, we evaluated in a multicenter study the frequency of the TNF2 polymorphism among patients with septic shock compared with those in a control group.

This multicenter study was conducted in 7 academic adult ICUs in France between March 1996 and June 1997. The protocol was approved by the institutional review board of Cochin Hospital, Paris, France. Informed consents were obtained from control subjects and patients or their relatives. Obtaining informed consent created a delay between the beginning of septic shock and study entry of a mean (SD) 3.9 (3.4) days, so that only 24 (27%) of the 89 patients could be included during the first 36 hours of the onset of septic shock.

The control group comprised 87 healthy unrelated white blood donors to the Cochin Hospital blood bank. Sixty-five were men and 22 were women; the mean (SD) age was 37.4 (11.1) years.

The septic shock group, defined by the criteria of the consensus conference,²⁴ included 89 ICU patients. To be eligible for enrollment, the patients had to be white and to have had 6 of the following inclusion criteria of septic shock within a 12-hour period: (1) clinical evidence of infection; (2) hyperthermia (temperature >38°C) or hypothermia (temperature ≤35.6°C); (3) tachycardia (>90 beats/min); (4) tachypnea (>20 breaths/min) or need for mechanical ventilation; (5) use of vasopressor to maintain systolic blood pressure higher than 90 mm Hg or hypotension defined as systolic blood pressure less than 90 mm Hg for more than 30 minutes or a decrease in systolic blood pressure of more than 40 mm Hg from previously established values for more than 30 minutes (hypotension had to be present at enrollment and refractory to an intravascular volume challenge of at least 500 mL); and (6) evidence of inadequate organ function or perfusion within 12 hours of enrollment, as manifested by at least 1 of the following syndromes (previously described²⁴ ): acute deterioration of patients' mental status; arterial hypoxemia (PAO₂/FIO₂ <280); plasma lactate concentrations above the normal range or metabolic acidosis; oliguria; and disseminated intravascular coagulation.

The exclusion criteria were the following: (1) older than 80 years; (2) cardiac failure (class III or IV); (3) liver insufficiency (Child C); (4) bone marrow aplasia (white blood cell count <0.50 × 10⁹ /L); (5) immunosuppression (positive human immunodeficiency virus serologic result, current immunosuppressive therapy including corticosteroids [equivalent prednisone >0.5 mg/kg per day], cancer).

Patients were followed up throughout their stays in the ICU. Age, sex, primary site of infection, infection-related organisms, and severity indexes including Simplified Acute Physiologic Score (SAPS II)²⁵ and Organ System Failure score (OSF)²⁶ were collected at each patient's entry into the ICU. The plasma of these patients was tested at the time of study entry for TNF-α concentrations by immunoassay (Immunotech, Paris, France). The lower limit of detection was 5 pg/mL. Values lower than 20 pg/mL were considered normal.

All the samples and information used in the study were coded and patient confidentiality was preserved according to the guidelines for studies of human subjects.

For the detection of polymorphisms in the −501/+11 region of the TNF-α gene promoter, we used denaturing gradient gel electrophoresis (DGGE). Among the various screening methods, this technique is emerging for its sensitivity and accuracy.²⁷ Denaturing gradient gel electrophoresis is based on 2 principles: (1) partial denaturation of double-stranded DNA, melting into single strands, retards its mobility on gel electrophoresis, and (2) thermal DNA denaturation is sequence dependent. This mutation scanning procedure could separate DNA fragments differing by as little as 1 base. To improve its sensitivity, we used psoralen-modified oligonucleotide primers and photo-induced cross-linking.²⁸ Melting analysis of the 500 base pair (bp) of the TNF-α promoter has been performed by computer algorithm McMelt 1.0 (computer program provided by L. S. Lerman). Three genomic DNA sequences have been defined: fragment A from −501 to −323; fragment B from −371 to −205; fragment C from −246 to +11 (Figure 1).

We obtained DNA by phenol-chloroform extraction from 10 mL of EDTA–anticoagulated blood.²⁹ Each fragment was initially amplified by polymerase chain reaction (PCR) using specific primers. After clamping of the PCR products by exposure to UV light, DGGE was performed. Polymerase chain reaction and DGGE conditions, including primer sequences, are reported in Table 1. Interpretations of different profiles were completed by 2 independent persons who had no access to the clinical data.

DNA that showed migration abnormalities in the DGGE experiment were subjected to automatic DNA sequencing of all the 512 bp on 377 PE/ABI automatic sequencer (Perkin-Elmer Cetus, Norwalk, Calif) (Table 1).

HLA antigens were typed in all controls and septic patients except for 2 of the latter patients due to technical problems. Antigens (HLA-A, HLA-B, and HLA-C) were typed using standard microlymphocytotoxicity assays. Generic HLA-DR genotyping was performed by PCR followed by hybridization of the product with allele-specific oligonucleotide probes (Innogenetics, Gent, Belgium).

Descriptive results of continuous variables were expressed as mean (SD). Plasma TNF-α levels were reported as median and range. Variables were tested for their association with mortality using χ² test for categorial data (sex ratio, medical or surgical patients, primary site of infection, microorganisms, polymorphisms) and Mann-Whitney U test for numerical data (age, OSF, SAPS II, TNF-α concentration). A multiple logistic regression model including age and the SAPS II–derived probability of dying was used to determine the respective role of the TNF2 polymorphism for ICU mortality. Analysis was completed on a personal computer using STATA software (STATA Corp, College Station, Tex) and a 2-tailed P<.05 was considered significant.

Eighty-nine white patients meeting the criteria of septic shock²⁶ were enrolled. The characteristics of patients with septic shock and the primary sites of infection are listed in Table 2. The mean (SD) age was 58 (15) years, and 68% were men. Mean (SD) SAPS II score was 55 (19), and the ICU mortality rate of 54% was in agreement with the predicted risk of death based on the SAPS II score.²⁵ All patients required mechanical ventilation at the time of study entry and the mean OSF score was 2.9 (1). Values of SAPS II and OSF underlined the severity of the selected patients. At study entry, 54 (61%) of 89 patient samples had a plasma TNF-α concentration above 20 pg/mL. The median plasma TNF-α concentration was 33 pg/mL.

When comparing the clinical characteristics of the survivors with those of nonsurvivors, only age and SAPS II values were significantly different (Table 2). The TNF-α levels were comparable between the 2 subgroups as already reported.¹¹ Analysis of TNF-α concentrations as a function of the shock duration before drawing the blood sample did not reveal any difference between either survivors and nonsurvivors (data not shown).

The complete 500 bp promoter region upstream of the start codon of TNF-α gene was analyzed by DGGE. Such an extensive study of this area of the genome had never been performed previously. Table 3 shows the different genotypes and their frequencies among the 2 populations. We detected the 4 previously described polymorphisms at position −238, −244, −308, and −376.³⁰ Moreover, we found 2 new DNA variants at location −419 and −49 (Figure 1 and Table 3), confirming the high sensitivity of DGGE. Except for the −419 variant, which was a G→C inversion, all the polymorphisms are G→A transitions. Only the TNFA-A and the TNF2 alleles (G→A transitions at position −238 and −308, respectively) were present either as heterozygous or as homozygous (Table 3).

Interestingly, this analysis revealed different associations of polymorphisms (Table 3). Hence, polymorphisms at −419, −244, and −49 appeared to be always associated with the TNF2 allele. Likewise, TNF2 and TNFA-A polymorphisms were present simultaneously in 6 patients (1 control and 5 septic shock patients). The G→A transition at −376 was constantly associated with the TNFA-A allele. To our knowledge, most of these allele associations have never been reported before and their functional importance in terms of TNF-α production remain to be determined.

The frequencies of the TNFA-A and TNF2 polymorphisms in the French control population (Table 3) were close to the published average frequencies and similar to the previously studied Danish,³¹ Spanish,³² and North American³³ groups. Moreover, we found similar results in a group of 65 unrelated white French patients with ankylosing spondylitis (M. Djouadi, MD, unpublished data, 1997). To confirm the validity of our control group, we performed HLA analysis on this population and did not find any significant difference with the reference HLA frequencies of the French population (data not shown).³⁴

The polymorphism frequencies of the controls and the septic shock patients were significantly different (Pearson χ², P=.008), which was essentially a result of the TNF2 polymorphism. Indeed, this TNF variant was found in 35 (39%) of the 89 septic shock patients and only in 16 (18%) of the 87 controls (Pearson χ², P=.002).

Within the septic shock group, patients who did not survive had significantly more uncommon alleles inside the TNF-α promoter (Pearson χ², P=.01; Table 4). The TNF2 allele was mainly responsible for this difference. Interestingly, every patient who was TNF2 homozygous or had an association of the TNF2 allele with another variant had a fatal outcome. Moreover, the 5 patients carrying the G→α transition at −376 died. However, the low number of patients involved did not allow us to perform a statistical analysis for these genotypes.

To study the specific effect of the TNF2 polymorphism on mortality, we compared subgroups of septic shock patients according to the presence (TNF2 group) or the absence (TNF1 group) of the G→A transition at −308 (Table 5). No difference existed between the 2 subgroups for age, SAPS II, OSF, sex ratio, primary sites of infection, and microorganisms (Table 5 and data not shown). Plasma TNF-α concentrations were elevated in 24 (68%) of the 35 patients in the TNF2 group and in 28 (53%) of the 54 patients TNF1 group, respectively, but their median values were not significantly different (48 vs 29 pg/mL for TNF2 and TNF1, respectively; P=.31). However, mortality of patients carrying the TNF2 allele was significantly higher than it was among patients carrying only the TNF1 allele (71.4% vs 42.6%, respectively; P=.008). To confirm that the TNF2 allele might be an independent predictor of severity during septic shock, we used a multiple logistic regression model to determine the predictive factors of mortality (Table 6 ). This analysis showed that the TNF2 polymorphism was strongly associated with an increased relative risk of death. For an identical rank of SAPS II value, patients with the TNF2 allele had a 3.75-fold increased risk of death.

High TNF-α production had been initially associated with the presence of the HLA DR3 allele.^{12 - 13} More recently, the TNF2 allele has been reported to be linked with the HLA-A1/B8/DR3 haplotype, which is in strong disequilibrium in Europeans.^{15 ,34} To evaluate the relative importance of the TNF2 allele and of the HLA genotypes in the outcome of septic shock, we analyzed HLA classes I and II. There were no significant differences in individual allele frequencies at any studied locus among the septic shock group, the control group, and a French reference population³⁴ or between the survival and nonsurvival groups (Table 7 and data not shown). In contrast, the TNF2 polymorphism was associated with HLA DR3, HLA B8 genotypes, and the HLA haplotype A1, B8/DR3 (Table 7). Thus, despite the presence of a linkage disequilibrium between TNF2 and HLA A1/B8/DR3, this haplotype did not appear to be associated with severity of septic shock.

This study demonstrates for the first time, to our knowledge, that the rare TNF2 allele of the TNF-α promoter is associated with susceptibility to and outcome of septic shock. The TNF2 polymorphism frequency was significantly higher in the studied septic shock group than in the control population (Table 2) and appeared to be an independent risk factor for death due to septic shock (Table 6).

A new important topic of medical research is the localization of genes implicated in the susceptibility to common diseases. This has been greatly facilitated by genomic sequencing and the discovery of polymorphisms. Reports of the association of polymorphisms with common diseases, such as this study, have quickly increased in number during the last 5 years.^{18 ,20 - 22 ,35 - 37} However, to prevent spurious conclusions from these studies, 3 questions have to be answered.³⁸

First, are the studied populations homogeneous? To avoid an artifact in population admixture, we selected only white patients in France and determined their HLA genotypes as an internal control of homogeneity. No significant difference existed between the control and septic shock groups nor between the control group and a French reference population³⁴ in terms of HLA class I or class II. Moreover, the TNF2 allele frequency in the control group was close to those previously reported^{31 - 33} and similar to the TNF2 allele frequency in another group of French patients with ankylosing spondylitis, confirming a recent report.³⁹

Second, does the product of the studied gene play an important role in the pathogenesis of the common disease? The central role for TNF-α in the occurrence and severity of septic shock has been clearly demonstrated by experimental studies.^{2 - 8}

Third, does the polymorphism of the gene under study cause a relevant alteration in the level or function of the gene product? In vitro, the TNF2 polymorphism has been shown to increase TNF-α synthesis.^{16 - 17 ,19} Both transcription studies and lipopolysaccharide stimulation of whole blood cell culture have shown than TNF-α production was significantly higher among TNF2 allele carriers than among persons homozygous for TNF1. Recently, the effect of the TNF2 allele was confirmed in vivo by Warzocha et al,¹⁸ who reported that this polymorphism was associated with higher plasma levels of TNF-α, and by Conway et al,⁴⁰ who found a correlation between this polymorphism and the higher levels of TNF-α detectable in tear fluid of patients with scarring trachoma. Hence, it seems conceivable that the TNF2 allele, which increases TNF-α production, might contribute to the occurrence and outcome of septic shock. Similarly, we cannot conclude that the association of the 2 polymorphisms located at positions −238 and −376 were associated with an unfavorable prognosis of the syndrome because the functional relevance of such variants on cytokine production is still unknown.

This study found an association between the TNF2 allele and mortality but failed to demonstrate a significant correlation between plasma TNF-α concentrations and TNF2 polymorphism carriage. This apparent discrepancy could be explained by the delay between the beginning of septic shock and the timing of the blood sample (mean, 3.9 days). Animal and volunteer studies had previously shown that the peak of plasma TNF-α production occurs during the first hour after endotoxin challenge.^{3 - 5} Moreover, Dhainaut et al,⁴¹ using serial blood samples for determination of TNF-α concentrations before and after anti-TNF antibody infusion, found in the placebo group that TNF-α concentration peaked during the first day of septic shock at 58 pg/mL, and progressively decreased to 27 pg/mL 3 days later. Based on these reports, it is likely that we missed the intravascular secretion of this cytokine.

However, the significance of the plasma TNF-α concentration in clinical studies is unknown. There is an evident gap between well-defined experimental conditions and septic shock in humans. First, the percentage of patients with an elevated plasma TNF-α concentration during the first 12 hours of septic shock varied from 86% in the International Sepsis Trial Study Group (INTERSEPT) study¹⁰ to 65% in the North America Sepsis Trial Study Group (NORASEPT) II study.¹¹ Second, it is clear that the detected plasma TNF-α does not represent the concentration of cytokine locally produced in the site of infection.^{42 - 44} Third, this detectable level does not take into account the membrane-bound form of TNF-α, which is up-regulated on many cells including endothelial cells and leukocytes of patients with multiple-organ failure.⁴⁵ Furthermore, the nature of microorganisms and primary sites of infection may influence the kinetic and the extent of plasma TNF levels, making data difficult to compare.^{46 - 47}

Our study confirmed that the TNF2 allele is strongly linked with the HLA haplotype A1/B8/DR3 (Table 7). However, no correlation existed between this haplotype or any of its components and septic shock or mortality. The HLA DR3 genotype was initially associated with high TNF-α production.^{12 - 13} However, Bouma et al⁴⁸ showed in vitro that HLA DR3 monocytes did not have higher capacity to synthesize TNF-α, except if they carried both the DR3 and TNF2 alleles.⁴⁸ Then, it seems probable that the described association between DR3 genotype and high TNF production^{12 - 13} was due to the linkage disequilibrium between the TNF2 alleles and HLA DR3.

The pathophysiology of septic shock is a complex and multifactorial process, involving an imbalance between proinflammatory and anti-inflammatory cytokine release.^{1 - 4} In this study, 39% of septic shock patients carried the TNF2 polymorphism, which may create such an imbalance of the immune response. Other factors such as surgery, trauma, or underlying diseases may modulate the cytokine response. Moreover, other genetic polymorphisms may be involved in the occurrence and/or the outcome of this syndrome. Recently, polymorphisms of inflammatory cytokines, such as interleukins 1 and 10, were described, but their significance in septic shock is unknown.⁴⁹

In contrast to the effect of antiendotoxin therapy, TNF monoclonal antibodies have been effective in gram-positive and in gram-negative models of septic shock. Despite encouraging preclinical studies, all multicenter trials studying the effect of such therapy on mortality have been negative.^{4 ,9 - 11} Various factors have been proposed to explain this difference between clinical studies and experimental models.⁵⁰ Among them, the inclusion of heterogeneous clinical trial populations and a lack of consensus on risk stratification appeared to be predominant. This study offers new opportunities for studying intervention with anti-TNF therapies. Determining a patient's TNF2 genotype before starting the treatment may permit the selection of a homogeneous group of high-risk patients who could benefit from treatment with anti-TNF. Such a possibility deserves further study, since an effective therapy for patients with septic shock syndrome would have important clinical and economic consequences.