Searching for Candidate Genes in Acute Lung Injury: SNPS, CHIPS and PBEF

Abstract

Acute lung injury (ALI) is a devastating illness, occurring in the setting of sepsis, with genetic variations contributing to ALI susceptibility and severity. We utilized the “candidate gene approach” with extensive expression profiling in animal and human ALI models to identify novel candidate genes. We noted significant expression of pre-B-cell colony enhancing factor (PBEF), a gene not previously associated with lung pathophysiology. This finding was validated by molecular, biochemical and immunohistochemical approaches with increased levels of PBEF also detected in human BAL and serum. DNA sequencing identified two single nucleotide polymorphisms (SNPs) in the PBEF promoter (T-1001G, C-1543T), which were genotyped in a Caucasian cohort of sepsis-associated ALI patients. Carriers of the GC haplotype exhibited a 5.7-fold relative ALI risk compared to controls associated with increased PBEF promoter activity. These studies demonstrate the successful application of genomic technologies in the identification of novel candidate genes in complex lung disease.

INTRODUCTION

The study of the molecular basis of complex lung disorders, until recently, was the study of individual genes and limited to a gene-by-gene approach. The nineteenth century German mathematician, David Hilbert, aptly stated, “significant advances require the development of sharper tools for exploration” (1). With the completion of the Human Genome Project, the availability of high-throughput biology and parallel developments in computational analysis have heralded the era of molecular medicine and revolutionized the concept of translational biomedical research, especially in complex disorders. Acute lung injury (ALI) is a complex and devastating respiratory illness, often occurring in the setting of sepsis, with an annual mortality rate of 30–50% (2). Although the genetic basis of ALI has not been fully established, increasing evidence derived from association-based studies suggests that genetic variations contribute to ALI susceptibility and severity (3–12). Significant difficulty exists, however, in defining the exact nature of ALI genetic factors including large phenotypic variance, incomplete gene penetrance, complex gene-environment interactions and a strong potential for locus heterogeneity. Moreover, ALI arises in a critically ill population with diverse precipitating factors. The sporadic nature of ALI with the lack of affected families, precludes a conventional genomic approach such as linkage mapping (or “positional cloning”).

In the current report, we utilized a “candidate gene approach” with extensive gene expression profiling studies in animal and human models of ALI (rat, murine, canine, human) to identify potential ALI candidates. These studies identified a novel ALI gene, pre-B-cell colony enhancing factor (PBEF), suggesting that the candidate gene approach is a robust strategy to provide novel insights into the genetic basis of ALI, and the identification of potentially novel therapeutic targets.

MATERIALS AND METHODS

Animal Models of ALI

All animal models were institutionally approved. Two canine models were utilized: unilateral saline lavage-induced lung injury (13) and intrabronchially-delivered endotoxin (LPS) (14). In the first model, the injured left and uninjured right lungs were independently mechanically ventilated for six hours (8 ml/kg, 0 PEEP, and 10 ml/kg and PEEP 5 cm H₂O, respectively). The second canine model utilized high-tidal volume mechanical ventilation (6 hours, 17 ml/kg), as recently reported (14). Control animals received endobronchial saline with identical ventilation strategies. Lung tissues were processed for microarray analysis and BAL/serum were collected for protein analyses. Two murine ALI models were utilized one with intratracheal LPS and a second model with two hours of 17 ml/kg mechanical ventilation as we recently described (14,15). Control groups were spontaneously ventilated. Lung tissue and BAL were collected for microarray and protein analyses.

Human ALI

Human protocols were approved by the Johns Hopkins University Institutional Review Boards. Human BAL (n = 3 each) and serum samples (n = 8 each) were obtained from ALI patients and healthy controls.

Gene Expression Profiling and Validation

The affymetrix GeneChip Microarray System was utilized as we described previously (16). Semi-quantitative RT-PCR, western blot and real-time PCR were utilized to validate PBEF expression in animal lung tissues and human BAL, respectively.

Localization of PBEF Expression in Canine Lung

To evaluate the spatial localization of PBEF expression, we performed triple immunohistochemical staining in canine lung tissue (17) using an anti-canine PBEF polyclonal antibody (18), and antisera raised against Factor VIII (to visualize vascular endothelium). In addition, 4’6 diamidino-2-phenylinodole (DAPI) was utilized to visualize cell nuclei.

Western Blotting Analysis of PBEF Protein

The total protein content in each sample was quantified using the BCA Protein Assay Kit (Pierce, Rockford, IL). PBEF proteins were assessed by Western blotting with densitometric quantification.

Genotyping of PBEF Gene Promoter SNPs

Leukocyte DNA from subjects with sepsis-associated ALI, sepsis alone and healthy controls was obtained (Johns Hopkins University, Medical College of Wisconsin) according to consensus diagnostic criteria (19,20) with recording of APACHE II scores. SNP discovery of the human PBEF gene was performed in 36 subjects (12 per group) by direct DNA sequencing. Genotyping of the PBEF SNPs (T-1001G and T-1543C) in Caucasian subjects was performed using a restriction site polymorphism assay and an Assays-by-Design^SM Service-SNP Genotyping method (Applied Biosystems, CA), respectively.

Transient Transfection Assays

A 272 bp fragment (either TT or GG at—1001 position) or a 134 bp fragment (either TT or CC at —1543 position) of the PBEF gene promoter was sub-cloned into pGL3-basic vector (Promega, WI) and transiently transfected into human pulmonary microvascular endothelial cells. After transfection (6 hours), cell lysates were retrieved for luciferase activity determination.

Statistical Analyses

Statistical analyses were performed using SigmaStat (v 3.1) and/or Stata (v 8.0).

RESULTS

Regional Microarray Analysis and Increased PBEF Expression in Animal and Human Models of ALI

Initial microarray studies in a canine model of human ALI utilized samples derived from corresponding regions of the injured and control lungs of each animal. Results from each regional injury sample were normalized to the within-animal control, a unique feature of this unilateral injury model. Expression profiling revealed extensive expression of genes with differential regional gene expression changes along specific ontologies (Figure 1A). A selection of genes with apex vs. base expression changes in opposite directions or greater than two-fold differences in the same direction are presented in Figure 1B. These genes have been grouped into gene ontologies consisting of coagulation, protein synthesis/degradation, and inflammatory/immune response categories.

Fig. 1.

Regional heterogeneity in canine gene expression and PBEF gene expression in canine acute lung injury. Panel A. Gene expression was assessed in both apex and base regions of a canine model of ALI and its corresponding control lung. Each column represents an experimental condition of corresponding sample location. Red color indicates up-regulation and green color indicates down-regulation of gene expression relative to corresponding region of control lung in same animal, with color intensity corresponding to the fold-change amplitude (fold-change scale shown on the top). Panel B. Regional differences of gene expression (fold change, injured/control) in lung tissues (n = 6) were evaluated for genes with significant changes in expression in at least one region (apex or base) and a difference in regional expression 50% or higher. These were linked to gene ontology terms using MAPPFinder. Fold changes of genes from identified gene ontologies represented by horizontal bars ±SEM. PBEF expression was the highest of all the genes examined in these ontologies.

Examination of some of the highly differentially regulated genes between apex and base regions (Figure 1) reveals several genes commonly associated with ALI, such as VEGF, TLR, PAI-1, and TGFβ, as well as some novel genes not previously described in this context. In particular, the highest level of gene expression was pre-B cell colony enhancing factor (PBEF), a novel finding as only 10 publications existed in the PubMed literature concerning this gene and no PBEF -related publication existed which involved the role of PBEF in lung pathophysiology. This gene encodes for a pro-inflammatory cytokine, originally described for its role in the maturation of B-cell precursors with gene expression up-regulated in amniotic membranes from patients undergoing premature labor, especially with amniotic infections (21–23). As this was the first demonstration of PBEF expression in lung tissues, we confirmed these results with in lung tissues with RT-PCR in several ALI models (Figure 2). Furthermore, western blot results also confirmed the increased PBEF protein expression in muring ALI lung tissues and in canine ALI BAL fluid and serum (Figure 2). Finally, PBEF protein levels in human ALI were significantly increased in BAL relative to healthy controls (Figure 2). These results support PBEF as a potential biomarker in ALI and further validate the microarray-based enhanced PBEF expression in animal and human ALI.

Fig. 2.

Analysis of PBEF gene expression and protein levels in canine BAL/serum, and murine BAL in ALI. The left panel demonstrates that in each ALI model (canine or murine) as well as in BAL cells RNA from humans with ALI, that PBEF mRNA levels were validated by semi-quantitative duplex RT-PCR. Statistical comparative analysis of integrated density value (mean ± SEM) between control and ALI groups was performed using the paired t-test and unpaired t-test. The right panel reflects Western blot results that confirm PBEF protein content in lung tissue, BAL or serum. For BAL protein determinations, total protein (10 μg) of each sample was separated by 12% SDS-PAGE and immunodetected by western blot using the anti-canine PBEF antibody. Statistical comparative analysis of integrated density value (mean ± SEM) between control and ALI groups was performed using the unpaired t-test. PBEF protein levels in human ALI BAL and serum were significantly increased relative to healthy controls. These results support PBEF as a potential biomarker in ALI and further validate the microarray-based enhanced PBEF expression in animal and human ALI.

Spatial Localization of PBEF Expression in the Acutely Injured Canine Lung

To evaluate spatial localization of PBEF expression, we performed triple immunohistochemical staining in canine lung tissue samples from the unilateral lavage injury model and co-localized PBEF expression in vascular endothelial cells, neutrophils and Type II alveolar epithelial cells. Antibodies to Factor VIII (an endothelial marker), neutrophils, and ProSPC (a Type II alveolar epithelial cell marker) were obtained commercially. Figure 3 depicts strong canine PBEF expression in the vascular endothelium (Figure 3, Panel C). PBEF was also detected within infiltrating leukocytes in the surfactant-depleted, injured canine lung (data not shown, see ref 24) and Type II alveolar epithelial cells (data not shown, see 24) whereas PBEF immunoreactivity was minimally detectable in the uninjured but ventilated control canine lung (Figure 3, Panel B). Pre-immune serum did not show any significant staining of PBEF in an injured/ventilated canine lung (Figure 3, Panel A). These results indicate that lung vascular endothelial cells, type II alveolar epithelial cells and infiltrating neutrophils express PBEF in the injured lung.

Fig. 3.

Spatial localization of PBEF expression in canine lung. Panel A represents background staining with only secondary antibodies of injured/ventilated canine lungs. Panel B demonstrates mild sporadic staining of PBEF in uninjured/ventilated canine lung detected by a rabbit anti-canine PBEF affinity purified polyclonal antibody followed by a goat anti-rabbit IgG conjugated with Alexa Fluor 488 (green color). Panel C demonstrates significant staining of PBEF (green color) in injured/ventilated canine lung with positively stained endothelial cells identified by a rabbit anti-human von Willebrand Factor followed by a goat anti-rabbit IgG conjugated with Texas Red (C, red color). Blue color indicates nuclei staining by DAPI. Pre-immune serum did not show any significant staining of PBEF in an injured/ventilated canine lung (data not shown). Photos were at 400 × magnifications.

SNP Discovery and T-1001G and T-1543C PBEF Allelic Associations in Sepsis-Associated ALI

As our findings strongly implicated PBEF as a novel candidate gene in ALI, we next examined whether common variants in the human PBEF gene might be associated with susceptibility to sepsis-associated ALI. We identified, via direct DNA sequencing (24), 11 PBEF single nucleotide polymorphisms (SNPs) (24) with 2 SNPs, T-1001G and T-1543C transversions, in the human PBEF gene immediate promoter (-1 to -3,000 bp) having the highest degree of representation in 12 ALI subjects. The PBEF (T-1001G) SNP was genotyped in a case-control population of African-American and Caucasian subjects with sepsis-associated ALI (n = 87), sepsis alone (n = 94) and healthy controls (n = 84) with additional relevant characteristics of the study population available elsewhere (24). Both the T-1001G and T-1543C SNPs were in Hardy-Weinberg equilibrium (p = 0.50) with the overall frequency of the minor allele (24). Figure 4 depicts the minor allele frequency of both SNPs observed in subjects with ALI and the healthy control group with significantly greater G variant of T-1001G observed in the ALI (P = 0.004). The frequency of the G variant among sepsis subjects (25%) was also higher than healthy controls but this was not statistically significant (p = 0.09) (24). The second SNP, T-1543C, was also in Hardy-Weinberg equilibrium (p = 0.46) with the overall frequency of the minor C allele 25% (24). The C-allele frequency observed in subjects with ALI (18.9%) was significantly lower than the frequency observed in the healthy Caucasian control group (30.9%) (P = 0.008). The frequency of the C variant among sepsis subjects (23%) was also lower than healthy controls but this was not statistically significant (p = 0.155). In a univariate analysis, carriers of the G allele had a 2.75-fold increased risk of ALI compared to controls (p = 0.002). Multiple logistic regression analysis using relevant clinical risk factors revealed that, after controlling for age and sex as well as other co-morbidity factors (cancer, immunosuppression, etc.), the G mutant allele remained an independent risk factor for ALI susceptibility (OR = 2.16, 95% CI = 1.01–4.62) but not for sepsis without ALI (24). A borderline association was observed between the PBEF (T-1543C) genotype and ALI (p = 0.064), but not observed with either sepsis subjects or healthy controls (p = 0.265).

Fig. 4.

Minor allele frequencies of PBEF promoter SNPs (T-1001G and T-1543C) in a cohort of sepsis-associated ALI patients. Using a cohort of African American and Caucasian patients with ALI derived from sepsis and controls, we examined whether common variants in the human PBEF gene might be associated with susceptibility to sepsis-associated ALI. The minor allelic frequencies of the two PBEF promoter single nucleotide polymorphisms (SNPs) (24), T-1001G and T-1543C transversions, were assessed. The G variant of T-1001G appears to be associated with ALI susceptibility in blacks and whites (P = 0.004). The T variant T-1543C is associated with protection from ALI in whites (P = 0.008) but not African Americans.

Haplotype weighted analysis of T-1001G and T-1543C SNPs revealed four haplotypes GT, GC, TT, TC (Table 1). The frequency of the GT haplotype was >two-fold higher in ALI or sepsis group and represented a susceptible haplotype. In contrast, the TC haplotype is >two-fold lower in ALI and represents a protective haplotype. Since there were limited subject numbers, the haplotype TT was not subjected to further analysis. Univariant logistic regression analysis revealed that carriers of the haplotype GC from -1001G and -1543C alleles had a 7.71-fold higher risk of ALI (95% CI, 3.01–19.75, p <0.001) and 4.84-fold higher risk of sepsis (95% CI, 1.97–11.90, p = 0.001) (Table 1) while carriers of the haplotype TC from -1001T and -1543C alleles had a 20 percent-fold lower risk of ALI but the difference does not reach significance (data not shown).

TABLE 1.

Haplotype weighted analysis of PBEF T-1001G and T-1543C SNPs in patients with sepsis-associated ALI, sepsis and healthy controls

Haplotype*
Case	GT	GC	TT	TC	Total
Control	14.28	93.72	2.72	47.28	158
ALI	34.635	67.365	2.365	21.635	126
Sepsis	27.635	79.365	3.365	27.635	138
Total	76.55	240.45	8.45	96.55	422
χ2 test: p<0.01
Logistic regression analysis for the haplotype GT [Odds ratio (95% confidence interval)]
ALI	7.71 (3.01–19.75), P <0.001
Sepsis	4.84 (1.97–11.90), P = 0.001

All subjects are Caucasian. “Sepsis” denotes severe sepsis or septic shock.

^*T-1001G was genotyped by a RSP method while T-1543C was genotyped by an Assays-by-Design^SM Service-SNP Genotyping method (Applied Biosystems, CA). The sample size differences were due to the failure of the latter assays in some samples.

^*In each haplotype, first allele is from T-1001G and second allele from T-1543C.

Effect of PBEF Promoter SNPs on PBEF Expression and Luciferase Activity in Cytokine-Stimulated Endothelium

To examine whether the promoter SNPs (T-1001G and T-1543C) directly alters gene transcription, we generated a transient luciferase PBEF reporter construct transfecting TT-1001- or GG-1001-pGL3 basic vector as well as TT-1543- or GG-1543-pGL3 basic vector firefly luciferase reporter constructs (cotransfected with luciferase controls) into HMVEC-L for four hours. Cell lysates were harvested for luciferase activity assay. Renila luciferase activity was assayed as a base for the transfection efficiency normalization with activity expressed as relative light unit. Significant differences in luciferase activities between TT-containing- and GG-containing pGL3 basic vector constructs were not observed, however, the CC genotype in the PBEF promoter T-1543C decreased luciferase reporter gene expression (1.80 decrease, p <0.01) relative to controls.

DISCUSSION

The capacity for high throughput technologies (like sequencing) has been greatly advantaged by the Human Genome project, which heralded additional revolutionary technologic breakthroughs with rapid, high-throughput gene expression profiling and genotyping. Access to the complete genome sequences of prokaryotes, eukaryotic model organisms, and the mouse, rat and canine genome sequences has sparked efforts to identify specific gene expression patterns via large-scale microarray analysis that will ultimately help diagnose, prognosticate, guide therapy, or otherwise contribute to our overall understanding of human disease. Candidate gene identification in a complex lung disorder such as ALI poses a serious challenge due to the heterogeneity in inciting stimuli and the lack of available linkage studies. We applied emerging functional genomic technologies, specifically DNA microarray profiling and genotyping, to the study of the ALI pathogenesis in hope of providing mechanistic insights and identifying novel biomarkers and therapeutic targets. Gene expression profiling in lung tissue from animal models of ALI identified PBEF as a highly upregulated gene in ALI, results reinforced and validated by several complementary approaches (molecular cloning of canine PBEF, RT-PCR, immunohistochemical analysis). Furthermore, PBEF protein levels were significantly increased in BAL, serum and lung tissues from canine, murine and human ALI models, suggesting its potential as biomarker.

The published literature on PBEF is quite sparse (21–23), with our studies providing the first observation that PBEF is significantly upregulated in the lung as well as in models of lung injury (24). PBEF was first isolated from an activated peripheral blood lymphocyte cDNA library and found to be involved in B-cell precursor maturation (22). Subsequently, dysregulated PBEF gene expression was described in human fetal membranes of severe chorioamnionitis (23), with increased expression in an amniotic epithelial cell line following challenge with inflammatory cytokines (LPS, IL-1β, TNFα, IL-6) (23) as well as during IFNγ-induced maturation of pre-B cells and a B lymphoma cell line. Recombinant PBEF protein significantly increased expression of IL-6 and IL-8 in amniotic epithelium (21). Despite the findings of PBEF expression in non-lung tissues, the molecular physiological and pathophysiological relevance of PBEF to lung pathophysiology is completely unknown. The robust expression of PBEF in murine and canine models of ALI in our study suggests that PBEF may be an inflammatory signal transducer in the pathogenesis of ALI. Immunohistochemical colocalization studies revealed increased PBEF expression in lung endothelium, type II alveolar epithelial cells and infiltrating neutrophils as well as upregulation of PBEF expression in inflammatory cytokine-stimulated human pulmonary microvascular endothelial cells in vitro (24). These results strongly support a potentially important role for PBEF in the inflammatory lung processes observed in ALI. The immunohistochemical co-localization of increased PBEF expression in infiltrating neutrophils and lung endothelium is suggestive of a novel role for PBEF as a signal transducer during lung inflammation. This notion is supported by a recent report (25) that PBEF expression is significantly increased in circulating peripheral blood neutrophils derived from patients with sepsis, including data that convincingly demonstrated PBEF to inhibit neutrophil apoptosis. As the rate of clearance of apoptotic neutrophils is associated with resolution of neutrophilic lung inflammation, prolonging neutrophil survival via PBEF inhibition of apoptosis may sustain neutrophilic inflammation and contribute to the pathogenesis of ALI and other neutrophil-mediated disorders.

In addition to inflammatory cytokines, another clinically relevant stimulus for PBEF expression is increased mechanical stress, a major contributing factor to both ALI mortality as well as ventilator-associated ALI (26,27). The PBEF promoter contains two NF-κB binding elements that may potentially participate in conferring mechanical stretch responsiveness. Besides demonstrating the survival benefit of a lung-protective ventilatory strategy, the landmark ARDS Net study also highlighted a marked reduction in the number of neutrophils and the concentration of pro-inflammatory cytokines released into the airspaces of the injured lung (26). Studies are ongoing to establish the potential contribution of PBEF to both the pathogenesis and resolution of ALI. At a minimum, however, increased PBEF protein expression, either in BAL fluid or serum, has promise as a novel and useful biomarker to assist in the clinical diagnosis of inflammatory lung disease.

Given that our candidate gene approach identified PBEF as a viable candidate gene and potential biomarker in ALI, we investigated PBEF genetic variants, obtaining association data on the human PBEF promoter T-1001G and T-1543C SNPs. Haplotype analysis of these two SNPs revealed four haplotypes GT, GC, TT, TC. Carriers of the haplotype GT from -1001G and -1543T alleles had a 7.71-fold higher risk of ALI (p <0.001) and 4.84-fold higher risk of sepsis (p <0.001). Carriers of the haplotype TC from -1001G and -1543C alleles had a 20 percent lower risk of ALI, although the difference does not reach significance (data not shown) with this modest sample size. Multiple logistic regression analysis revealed that after controlling for 12 other risk factors (24), the G mutant allele remains an independent risk factor for ALI susceptibility. The G allele from T-1001G SNP and the C allele from T-1543C were not independently associated with sepsis or mortality among sepsis or ALI patients, which may reflect, in part, a limited sample size preventing the detection of a difference between sepsis and ALI groups. Because ALI patients in this study were complicated with sepsis, further analysis of DNA from patients with ALI from causes other than sepsis may be necessary to distinguish whether the haplotype GT is a risk factor or whether the haplotype TC a protective factor for ALI, and whether these effects pertain to ALI or rather severe sepsis, which frequently leads to ALI.

Preliminary studies addressing the functionality of the T-1001G variant using the luciferase reporter gene assay did not demonstrate a significant role for this variant in gene transcription regulation, however, the C-allele in the T-1543C SNP, in the PBEF promoter region, was associated with nearly 2-fold decrease in the reporter gene expression. The frequency of the C-allele was significantly lower in patients with acute lung injury (ALI) than that in normal controls (p = 0.008). This result is consistent with our observations from animal models of ALI, human ALI patients and in vitro cell culture experiments, and suggests that higher expression of PBEF is implicated in the pathogenesis of ALI. These results further suggest that genetically determined increased PBEF expression contributes to susceptibility to ALI.

SUMMARY

The availability of high-throughput biology and parallel developments in computational analysis have ushered in the era of molecular medicine, and revolutionized the concept of translational biomedical research. Significant challenges to the exploration of the genetic basis of complex lung disorders exist for diseases such as ALI. However, our study underscores the powerful potential of utilizing genomic approaches to deciphering the genetic basis of complex lung disorders. Utilizing a candidate gene approach and a series of diverse cellular, animal, and human studies, we identified PBEF as a potential novel biomarker and candidate gene in sepsis- and mechanical stress-induced inflammatory lung disease such as ALI. While further studies are clearly required to both define the pathophysiological role of altered PBEF expression in ALI and to more clearly link PBEF variants to ALI susceptibility, our results strongly support that PBEF may be a potential novel biomarker in ALI. Further analysis of select candidate genes by additional SNP discovery and mid- or high-level throughput genotyping will undoubtedly provide important insights into the genetic basis for ALI susceptibility and severity. The era of molecular medicine, in its truest sense, represents the capacity of genomics to bring clinicians, clinician-scientists and basic biomedical biologists together for a common goal in a way not previously imagined. Critical care physicians of the future will be armed with high-throughput technologies and phenotyping protocols which will customize care of the ICU patient, improve the survival of patients with critical illness and herald a new era in critical care medicine.

Fig. 5.

Effect of minor allele SNPs on PBEF promoter activity in cytokine-stimulated endothelium. Measurements of T-1543T- or C-1543C-pGL-3 basic vector firefly luciferase activity were performed in transiently transfected HMVEC-L. T-1543T- or C-1543C-pGL-3 basic vector firefly luciferase reporter constructs were cotransfected with luciferase controls into HMVEC-L for 4 hours. Cell lysates were harvested for luciferase activity assay. Renila luciferase activity was assayed as a base for the transfection efficiency normalization. The activity was expressed as relative light unit. Whereas the GG genotype T-1001G promoter failed to alter PBEF luciferase activity, the CC genotype in the PBEF promoter T-1543C decreased luciferase reporter gene expression (1.80 decrease, n = 6, p <0.01) relative to controls (n = 6). *, P <0.01. The C-1543T SNP alters PBEF expression, potential mechanism of protection from ALI in Caucasians.

DISCUSSION

Sacher, Cincinnati: Very nice presentation, thank you. Have you looked at perhaps a cleaner model, transfusion related acute lung injury with the passive transfer of alloantibodies for the most part?

Garcia, Baltimore: That's a great question. There are certainly other models of acute lung injury that would be very preferable to the kind of genes we're talking about here. But because of the wide heterogeneity we've kind of focused on just sepsis in order to kind of narrow down and increase our power for the genetic analysis. I will say that we're very interested and currently exploring commonalities particularly in the African American population that have acute lung injury. Commonalities between the chest syndrome that sickle cell patients endure and the acute lung injury that we've seen in our population, but I think that's a really great question.

Mackowiak, Baltimore: We heard from Dr. McKusick there are somewhere around 30,000 genes, and through various mechanisms, each produces a variety of different proteins. It occurs to me that as you are looking at genes, if you examine enough of them in relation to a particular disease, you're going to find associations. How do you deal with this problem in your statistical analysis to insure that a relationship related to the number of variables being considered simultaneously?

Garcia: I think that's a great question Phil. Certainly the folks that are genetically epidemiologists are grappling with that very same problem. And I think that really kind of doubles back to what we talked a little bit about at the beginning which is the need for extraordinary pheno-typing. I think that you find that you'll lose power as you go across these broad gene approaches, if you start grouping large groups of patients together. And that's one of the reasons, in answer to the previous question, why we kind of focus on sepsis induced acute lung injury. I think you will find that for a lot of genes the post translational modification of genes are going to be critically important. Several of the SNPs here are coding SNPs where there's very strong changes in amino acids and this is also a gene with a lot of splice variants. So I think you're right. But I think there are tools which exist for sorting those things out. And I think that really is a challenge of the genetic epidemiology field at the present time.