INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Community-acquired pneumonia (CAP) is a major health problem worldwide. In the United States, CAP is the leading cause of death resulting from infection and the sixth most common cause of death overall (1, 2). Clinically CAP exhibits an enormous variety in the severity of presentation, from fulminant septic shock at one end of the spectrum to almost asymptomatic disease at the other. Increasing antibiotic resistance, particularly in Streptococcus pneumoniae, has focused attention on the need to develop nonantibiotic strategies for both the prevention and therapy of CAP. A better understanding of what determines individual immune responses to CAP is therefore crucial.

Several factors could play a role in the variability of the presentation of CAP. First, different pathogens, as well as variable virulence in different strains of a pathogen, or combinations of pathogens are likely to be important. Second, underlying chronic illnesses would be expected to have an important impact. However, apparently similar patients with identical pathogens can have vastly different presentations and outcomes (3). Finally, an individual's ability to respond to infection is variable and a predisposition to death from infection is clearly inheritable (4). Therefore, a tendency to particular presentations of CAP may also be genetically determined.

The proinflammatory cytokine tumor necrosis factor-alpha (TNF) is an essential component of the host immune response to infection (5), and is responsible for the release of other inflammatory mediators. TNF also plays a major role in the clinical manifestations of septic shock (6), and serum levels inversely correlate with survival from severe sepsis (7, 8).

Guanine to adenine transitions at the +250 site within the lymphotoxin-alpha (LT), also known as TNF-beta(TNF), gene (9) and the -308 site in the TNF promoter region (10, 11) have both been associated with variability in TNF secretion after endotoxin and other stimuli, with increased levels associated with the A allele at both sites. The presence of the A allele at the LT+250 polymorphic site is associated with a significantly higher mortality from septic shock (12), with AA homozygotes at the greatest risk. Similar findings have been reported with the TNF-308 polymorphism (13), although this finding has not been universal (14).

We hypothesized that the LT+250 and TNF-308 gene polymorphisms may have an impact on the clinical variability in the presentation of CAP. Specifically, we examined whether individuals with CAP who are high secretors of TNF based on their genotype would have a higher mortality, and a greater risk of developing septic shock or respiratory failure, the main organ failures associated with CAP. Because patients with Type II respiratory failure are likely to have chronic hypoxia, we restricted our primary hypothesis testing to Type I respiratory failure.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

Our study was a prospective cohort study of patients admitted to the Memphis Methodist Healthcare System with CAP between November 1, 1998 and January 31, 2000. Patients were enrolled once they met the inclusion criteria as outlined subsequently. Informed consent was obtained from all patients, and this study was approved by the institutional review board of Methodist Le Bonheur Healthcare.

Inclusion Criteria

For the purpose of this study CAP was defined as: an acute illness (< 14 d of symptoms), the presence of a new chest radiographic infiltrate as confirmed by either a radiologist or a pulmonary/critical care physician, and clinical features suggestive of acute pneumonia. The clinical features required were one of Group A (fever [> 37.8° C], hypothermia [< 36.0° C], cough, sputum production); or two of Group B (dyspnea, pleuritic pain, physical findings of lung consolidation, and leukocyte count of > 12 × 10⁹/L or < 4.5 × 10⁹/L (15).

Exclusion Criteria

Exclusion criteria included: (1) patients with severe immunodeficiency as defined by the Centers for Disease Control criteria (16) for patients with acquired immune deficiency syndrome; (2) patients receiving chemotherapy in the past 60 d; (3) patients receiving treatment with corticosteroids equivalent to prednisolone > 20 mg/d for more than 14 d; (4) patients receiving immunosuppression after organ transplantation; (5) patients on cyclosporine, cyclophosphamide or azathioprine; (6) patients from nursing homes who were nonambulatory; (7) patients hospitalized within the past 30 d.

Data Collection

All patients were assessed by a pulmonary physician (GWW or RGW) within 24 h of presentation. The majority of patients were seen in the emergency department at the time of admission. Pneumonia severity index (PSI) scores using the clinical data available at the time of presentation were calculated as described by Fine and colleagues (17). Acute physiologic and chronic health evaluation (APACHE) II scores (18) were calculated, using the worst physiologic values during the first 24 h after presentation. Results of microbiologic and other laboratory tests as ordered by the treating physician were recorded. A history of chronic obstructive pulmonary disease (COPD), cardiac failure, renal insufficiency, diabetes, or cerebrovascular disease was recorded. Alcohol consumption was also noted.

Definitions

Septic shock was defined using American College of Chest Physicians/ Society of Critical Care Medicine (ACCP-SCCM) criteria (19). To meet the criteria for septic shock, a documented systolic blood pressure of < 90 mm Hg for at least 30 min in the absence of any other causes of shock, and at least 4 h of inotropic support after adequate fluid replacement were required. Respiratory failure was defined as an oxygen saturation of < 90% on room air. If the corresponding arterial PCO₂ was < 45 mm Hg, the patient was classified as having Type I respiratory failure. A corresponding arterial PCO₂  45 mm Hg defined Type II respiratory failure.

Septic shock or respiratory failure had to occur within 48 h of presentation to hospital for the patient to be classified as having one of these two endpoints. Patients were classified in a blinded fashion, i.e., without any knowledge of genotype information. Delayed antibiotic therapy was defined as any subject receiving his or her first dose of antibiotics more than 8 h after presentation to hospital.

Blood Collection and Processing

Whole blood for genotypic analysis was collected, transferred into 1.5 ml cryotubes, and stored at 70° F until processed. DNA was extracted from the whole blood samples using the Genomic DNA Purification Kit (Promega, Madison, WS). The genotypic analysis was also performed in a blinded fashion, that is, the analysis was performed without knowledge of any clinical data including endpoints such as mortality, septic shock, and respiratory failure.

Genotypic Analyses

The LT+250 polymorphism contains an NcoI restriction site when the G allele is present. We amplified a 782 base pair (bp) fragment in a polymerase chain reaction (PCR) mixture containing 20 ng of DNA, 20 pmol each of the primers TNF+250-1 (5'-CCGTGCTTCGT GCTTTGGACTA-3') and TNF+250-2 (5'-AGAGGGGTGGAT GCTTGGGTTC-3') (12), 1 unit of Taq polymerase, 1× reaction buffer (Promega), 500 µM each of deoxyadenosine triphosphate, deoxycytidine triphosphate, deoxyguanosine triphosphate, deoxythymidine triphosphate, and 2.5 nM of MgC1₂. Reaction conditions were as follows: 35 cycles of denaturation at 94° C for 30 s, annealing at 69° C for 30 s, and extension at 74° C for 42 s. The amplified DNA was incubated with NcoI and the fragments were analyzed by electrophoresis in a 1% agarose gel and visualized by ethidium bromide staining. Interpretation was as follows: a single band of 782 bp identified individuals homozygous for an adenine at the LT+250 locus; two bands at 586 and 196 bp identified individuals homozygous for a guanine at the LT+250 locus; three bands at 782, 586, and 196 bp identified individuals heterozygous at the LT+250 locus.

The region containing the TNF-308 locus was amplified using the primers TNF-308-1 (5'-AGGCAATAGGTTTTGAGGGCCAT-3') and TNF-308-2 (5'-ACACTCCCCATCCTCCCTGCT-3') (20). The TNF-308-1 primer contains 4 bp of the NcoI recognition sequence, including a mismatched cytosine as shown by the C in the TNF-308-1 primer sequence. This mismatched cytosine allows for creation of an NcoI restriction site (CCATGG) when the G allele is present at position 308. A 116 bp PCR product was generated using the following reaction conditions: 35 cycles of denaturation at 95° C for 30 s, annealing at 64° C for 15 s, and extension at 74° C for 15 s. The amplified DNA was incubated with NcoI and the treated fragments were analyzed by electrophoresis in an 8% polyacrylamide gel and visualized by ethidium bromide staining. Interpretation was as follows: a single band at 116 bp identified individuals homozygous for an adenine at the TNF-308 locus; two bands at 96 and 20 bp identified individuals homozygous for a guanine at the TNF-308 locus; three bands at 116, 96, and 20 bp identified individual heterozygous at the TNF-308 locus.

Statistical Evaluation

All statistical calculations, including multivariate analysis, were performed using the statistical package JMP version 3.2.2 (SAS Institute Inc., Cary, NC). Unless otherwise stated, results are expressed as mean ± SD. Relative risks (RR) are reported as RR (95% confidence intervals). The statistical significance of differences in continuous variables was calculated using Student's t test (after confirming they were normally distributed), and for categorical variables with Fisher exact test. The significance of trends was assessed using chi-square analysis, except for the haplotype analysis where the Kruskal-Wallis test of association with one ordered category was used. All reported p values are two-tailed with a value of < 0.05 considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographics

A total of 300 subjects consented to participate in the study. Twenty subjects were subsequently determined to have a diagnosis other than CAP and excluded from analysis. In 17 of the 20 subjects excluded, the infiltrate seen on admission was determined to be chronic through subsequent review of old chest radiographs. The three other subjects excluded were also subsequently determined to have a diagnosis other than pneumonia (malignancy, two; pulmonary embolus, one).

The mean age of the 280 study patients was 57.9 y (range 18 to 98). There were 146 (52.1%) female subjects and 134 (47.9%) male subjects, with 158 (56.4%) African American, 121 (43.2%) white, and one (0.3%) Asian subject. The distribution of subjects by PSI grade was I36 (12.9%), II78 (27.9%), III59 (21.1%), IV72 (25.7%), and V35 (12.5%).

A pathogen was identified from blood cultures in 30 patients (10.7%) and from sputum cultures in an additional 12 patients (4.2%), giving an etiologic diagnosis in 42 patients (15.0%). The most common pathogens isolated were Streptococcus pneumoniae (24 patients), Pseudomonas aeruginosa (4), Streptococcus viridans (4all bacteremic), and Hemophilus influenzae (3). Four patients had more than one pathogen identified. An additional 50 patients (17.9%) had negative blood and sputum cultures but the sputum Gram stain was consistent with infection due to S. pneumoniae.

No patient had a pathogen identified not covered by the empiric antibiotic regime. Thirty-seven subjects (13.2%) had antibiotics before presentation to hospital.

As has been noted previously (9, 11), we found a significant linkage disequilibrium between the two polymorphisms (p < 0.001). The distribution of the combinations of genotypes is shown in Figure 1. From this figure, it is apparent that although all A homozygotes at the TNF+250 locus were G homozygotes at the TNF-308 locus, the A allele at TNF+250 was not always associated with the G allele at TNF-308.

View larger version (16K): [in this window] [in a new window]   Figure 1.   Linkage disequilibrium between LT+250 and TNF--308.

Mortality

There were 25 deaths (8.9%). Table 1 shows a comparison of age, mortality, APACHE II scores, and PSI scores by genotype at each locus. There were no significant differences between genotypes at either locus. The trend to decreasing APACHE II scores between LT+250 AA and LT+250 GG genotypes did not reach statistical significance (p = 0.12). There were no significant differences between genotypes with respect to the proportion of patients who received antibiotics before presentation. Table 2 shows a comparison of known clinical risk factors for adverse outcome from CAP by genotype. There were no significant differences between genotypes for any risk factor.

Septic Shock

All 31 subjects who met the criteria for septic shock required inotropic support for greater than 24 h, except two subjects who died within the first 24 h after admission. Figure 2 shows the proportion of patients within each LT+250 genotype who developed septic shock. AA homozygotes were significantly more likely to develop septic shock than non-AA homozygotes (p = 0.01), with RR of 2.48 (1.28 to 4.78). This attributable risk of the LT+250 AA genotype to the development of septic shock in our population was therefore 30.7%.

View larger version (11K): [in this window] [in a new window]   Figure 2.   Proportion of patients developing septic shock within each LT+250 genotype.

The proportion of subjects developing septic shock within each TNF-308 genotype was AA-0.16, GA-0.06, and GG-0.12 (p = not significant [NS]). Carriage of an AA genotype at either the LT+250 locus or the TNF-308 locus was associated with a significantly increased risk of septic shock [18.0% versus 6.8%, p = 0.006, RR 2.51 (1.30 to 4.87)].

In nominal logistic regression analysis incorporating age, sex, the presence of chronic pulmonary, cardiac, renal, hepatic, or neurologic disease, alcohol consumption (both Alcohol Use Disorders Identification Test (AUDIT)-C total score and subjects consuming > 80 g of alcohol/day), prior antibiotic exposure, delayed antibiotic therapy, LT+250 and TNF-308 genotype, the only factors that remained significant predictors of septic shock were LT+250 genotype (p = 0.03) and increasing age (p = 0.048). The age-adjusted odds ratio for septic shock in carriers of the LT+250 AA genotype was 3.64 (1.28 to 10.66).

The data were then reanalyzed by LT+250:TNF-308 haplotype carriage to determine whether this made the association between the LT+250 A allele and septic shock more specific. Because haplotypes cannot be unequivocally assigned to heterozygotes at both loci, they are scored as a 0.5 probability for each haplotype. Carriage of the 250A:308G haplotype was associated with a significantly greater risk of septic shock (p = 0.014 by Kruskal-Wallis test of association with one ordered category, Table 3). No association between the risk of septic shock and carriage of the 250A:308A or 250G:308A haplotypes was found. However, the trend to a decreased risk of septic shock with the 250G:308G haplotype was significant (p = 0.011, Table 2).

Fifteen patients with septic shock died (48.4%). The mortality rate from septic shock within each genotype was LT+250: AA50.0%, GA50.0%, GG40.0%; TNF-308 AA 66.6%, GA25.0%, GG50.0%. There were no statistically significant differences between mortality rates for individual genotypes with either polymorphism or by LT+250: TNF-308 haplotype.

There were insufficient subjects with a definitive microbiologic diagnosis to analyze the polymorphism data by specific pathogens. However, pooling all patients with proven or suspected pneumococcal disease (n = 74), the same trends were seen as were observed in the CAP cohort as a whole. There were trends to greater bacteremia in subjects with LT+250 AA (p = 0.07) and TNF-308 (p = 0.08) genotypes that did not reach statistical significance.

Respiratory Failure

A total of 103 subjects (36.8%) met the criteria for respiratory failure, 80 with Type I and 23 with Type II. With respect to the arbitrary cutoff between Type I and Type II, all subjects classified as Type II respiratory failure had at least one PCO₂ > 48 mm Hg.

The proportion of subjects developing Type 1 respiratory failure within each genotype was LT+250: AA0.28, GA 0.25, GG0.37; TNF-308 AA0.32, GA0.30, GG0.28. For Type II respiratory failure, the proportion was LT+250: AA0.07, GA0.11, GG0.03; TNF-308 AA0.11, GA 0.14, GG0.06. There were no significant differences between genotypes at either locus for Type I or Type II respiratory failure.

The risk of Type I respiratory failure by carriage of LT+ 250:TNF-308 haplotypes is shown in Table 4. No statistically significant associations between the risk of Type I respiratory failure and any LT+250:TNF-308 haplotype was demonstrated.

Given the suggestion of different genotype associations between septic shock and respiratory failure, we analyzed the subgroup of patients with respiratory failure in the absence of septic shock. In each genotype, the proportion in this subgroup was LT+250 AA0.15, GA0.20, GG0.31 (p = 0.03 for trend); TNF-308 AA0.16, GA0.28, GG0.19 (p = NS).

Nominal logistic regression was performed incorporating the same variables assessed in the septic shock model to assess the interaction between clinical factors and cytokine genotype on the development of both Type I and Type II respiratory failure. Increasing age (p = 0.02) and prior antibiotic exposure (p = 0.003) were independent predictors of Type I respiratory failure, whereas COPD (p < 0.0001) and TNF-308 GG genotype (p = 0.03) were independent predictors of Type II respiratory failure.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have found a significant association between the LT+ 250 genotype and the risk of septic shock in patients with CAP. Whereas our finding that the LT+250 AA genotype is associated with the greatest risk of septic shock was expected, the finding that Type I respiratory failure is not associated with this genotype has significant implications for the definition of the systemic inflammatory response syndrome (SIRS). In addition, analysis of LT+250:TNF-308 haplotypes strongly suggests that the LT+250 locus is not the causative polymorphism. Finally, our data shed some light on the relative importance of the LT+250 and TNF-308 loci in the setting of CAP.

Given the previous studies showing an association between the AA genotype at the LT+250 locus and severe sepsis (12, 21), finding the same relationship with CAP was not surprising. As the LT+250 AA genotype has been shown to be associated with greater TNF secretion after a variety of stimuli both in vitro (9, 10) and in vivo (12, 21), this association was consistent with our current understanding of the pathogenesis of septic shock (6, 22).

A number of possible explanations exist for our observation of no association between TNF-308 genotype and septic shock. The TNF-308 locus may not influence the risk of septic shock, as has been previously suggested (14). A more likely explanation is the linkage disequilibrium between the two loci, with the G (TNF low secretor) allele at TNF-308 locus almost always associated with the A (TNF high secretor) allele at the LT+250 locus. Supporting this possibility is the haplotype analysis showing the lowest risk of septic shock is in patients with two 250G:308G haplotypes. The haplotype analysis suggests that the TNF-308 locus does exert an influence that is largely masked by the much greater prevalence of the A allele at the LT+250 locus in our population. This finding emphasizes the important role that differences in local allelic frequencies may play in evaluating the significance of gene polymorphisms on disease susceptibility and outcome. Although another possible explanation is the low frequency of the TNF-308 A allele in our CAP cohort (0.18), our frequency is not dissimilar to that reported (0.21) by Mira and coworkers where a positive association in a sepsis group was found (13). Another confounding factor may be that the relative influence of each locus is dependent on the pathogen, and the proportion of gram-negative pathogens in our cohort would be expected to be much lower than that of Mira and coworkers.

Why do patients with a TNF low secretor genotype still develop septic shock? The stimulus required to develop septic shock may be determined genetically, but will be modified by other factors, such as pathogen virulence and comorbid illnesses. A large number of other cytokine gene polymorphisms may also be important, not only within the TNF locus (13) but also in other cytokine genes, including interleukin (IL)-1 (23, 24), IL-1 receptor antagonist (24, 25), IL-6 (26), and IL-10 (23). Sorting out the relative contributions and interaction between all of these polymorphisms, and ones yet to be described, will require an extensive genetic and clinical database.

Our findings have important implications for understanding the inflammatory response to severe infections. In an effort to identify patients at high risk for subsequent septic shock, the SIRS criteria were developed (19). Hypoxemia was included as an organ dysfunction and is one of the two most common organ dysfunctions in most studies of SIRS (22). Hypoxemia (and other organ dysfunctions in SIRS) was assumed to occur by the same mechanism and reflect the same proinflammatory state as septic shock. Our unexpected finding that respiratory failure is not associated with a TNF hypersecretor genotype, with a trend to greater respiratory failure with a TNF hyposecretor genotype (LT+250 GG) is not consistent with this understanding of SIRS. The lack of association between hypoxemia and a TNF hypersecretor genotype may explain some of the disappointing Phase III anti-TNF sepsis trials (27, 28) after encouraging initial studies. If this genotype association is confirmed in SIRS due to infections other than pneumonia, use of hypoxemia in the absence of septic shock as inclusion criteria for anti-inflammatory sepsis trials may need reevaluation. In our study, respiratory failure with or without septic shock had different genotype associations, which may affect the TNF response.

One potential explanation for an increased risk of hypoxemia in patients with a low TNF secretor genotype is that they may have an attenuated immune response to pneumonia. This relative reduction in immune response may in turn lead to an increased bacterial burden. Immunostimulation, such as use of granulocyte colony stimulating factor, of this subgroup of patients may be of potential benefit but more studies will be required.

The increased risk of septic shock with the 250A:308G haplotype but not the 250A:308A haplotype is difficult to interpret. The most likely explanation is that LT+250 is not directly causative, but is a marker for the "real" polymorphism that is located within the 250A:308G haplotype. This has important implications not only for understanding the molecular basis for susceptibility to septic shock, but also for other studies showing association between disease states and LT+250 genotype. A second, less likely possibility is that both polymorphisms are causative, but an A allele at one locus in some way interferes with the mechanism leading to increased TNF- production (such as conformation change that affects the binding of transcription activating factors) with the A allele at the other locus.

There are limitations to our study and the interpretation of our data. First, we have shown association, not causation. The TNF and LT genes are in the Class III region of the major histocompatibility complex (MHC), so it is possible that the polymorphisms are not directly related to the variable immune response but are linked to MHC genes that are the real determinants. One such possibility is that the TNF-308 A allele has been reported to be in linkage disequilibrium with human leukocyte-associated antigen-DR3 (HLA-DR3) (11, 13). Although HLA-DR3 positive individuals are reported to have higher TNF secretion than HLA-DR3 negative individuals (29), this appears to only be the case when it is associated with the TNF-308 A allele (11). HLA-DR3, unlike TNF-308, had no association with clinical outcomes in the study by Mira and coworkers (13). However, while this association between HLA-DR3 and high TNF production appears to be due to its linkage disequilibrium with the TNF-308 A allele, other sites within the MHC may be the actual causative gene.

Second, clearly nongenetic factors such as the length of time to initial therapy and adequacy of therapy also play important roles in the ultimate development of the clinical presentation, including the development of septic shock. We did not control for the length of time from onset of symptoms to presentation. However, our data support the hypothesis that genetic polymorphisms play a role in the variable presentation of CAP.

Differences in the virulence of pathogenes will also clearly have an impact on outcome of CAP. At present, we are unable to analyze the influence of genotype on CAP for individual pathogens owing to the lower number of cases with a definite microbiologic diagnosis. The problem of the low sensitivity of traditional culture techniques in CAP is well known, and our diagnostic rate of 15% is consistent with culture yield in other CAP studies (30). However, the pattern in patients with proven or suspected pneumococcal CAP parallels that of the entire cohort, a finding which is not surprising given that the pneumococcus is the most common cause of CAP. The genotype associations we have demonstrated may be pathogen-specific, even stronger in some while unrelated in others. The trend to increased bacteremia with LT+250 AA genotype or TNF-308 GG genotype is interesting and does not fit with the hypothesis that LT+250 AA is associated with an enhanced or excessive immune response. However, further studies are required to assess whether this is a real association, because statistical significance was not reached and the association was only found on post-test analysis.

Although several studies have shown correlation of both LT+250 (9, 10, 12, 21) and TNF-308 genotype (31) with serum TNF levels, we have not measured them in our cohort. Any meaningful interpretation of serum TNF levels taken at varying time points after the onset of illness and commencement of antibiotics is difficult. Even in the setting of acute septic shock, an increased serum TNF is not detectable in all patients (27, 28). Whether serum TNF levels correlate well with tissue concentrations is also not clear (32, 33). Genotype associations that correlate with an overall phenotypic response may be more useful than correlation with single point cytokine measurements.

In summary, we have shown that septic shock is associated with the TNF high secretor genotype (AA) at the LT+250 locus whereas Type I respiratory failure in the absence of septic shock is associated with the low secretor (GG) genotype. The TNF-308 polymorphism appears to have similar associations, but a lesser influence in our cohort because of the comparative rarity of the TNF-308A allele and linkage disequilibrium with the LT+250 polymorphism. The finding that these two presentations of CAP have opposite associations with respect to TNF secretion raises significant concern regarding the validity of the SIRS definition as an inclusion criteria for anti-inflammatory sepsis trials. As genotyping can be performed in less than 1 h (34), determination of cytokine genotypes, both prospectively and retrospectively, is likely to be important for future trials of immunomodulatory therapy in both sepsis and CAP.