Introduction

Endotoxin is a major component of the outer membrane of gram-negative bacteria, and the inflammation caused by the potent activation of the innate immune system by this moiety may be a complicating factor in many serious diseases including trauma, burns, invasive surgery and organ-specific illnesses (cystic fibrosis, inflammatory bowel disease, liver disease, kidney dialysis complications, asthma and autoimmune diseases). Bacterial endotoxins are associated with systemic inflammation is a major defining characteristic of gram-negative sepsis. Septic shock affects approximately 750,000 cases annually in the USA. Due to its complexity therapeutic strategies have been elusive (1). Although systemic inflammation is but one component of the sepsis syndrome, it can be studied in the absence of complex pathophysiology and co-morbidities of sepsis by administration of a small dose of bacterial endotoxin to otherwise healthy human volunteers. The ensuing response encompasses many of the characteristics observed in critically ill patients albeit with lower intensity and a self-limited time course of 12-24 hours. The responses in the human endotoxin model include core temperature, cardiac, vasomotor, hematologic, metabolic, hormonal, acute phase reactant, and cytokine components that have been well described. Innate immune cell activation leads to production and release of proinflammatory cytokines, which are proximal mediators of the systemic inflammatory response. Although the bulk of this proinflammatory mediator release likely originates in cells of the reticuloendothelial system, the leukocytes present in peripheral blood are also activated, and importantly, are available for sampling with minimal invasiveness. One of the challenging tasks in the post-genomic era is to extract new biological insight from available high throughput genomic studies. Given that microarrays measure simultaneously thousands of genes, a critical step is the extraction and classification of genes that are responsive to particular systemic perturbations. We present a novel approach for the global analysis of temporal expression profiles driven by the premise that informative subsets of genes are able to reconstruct the inherent dynamics of the system.

Materials and Methods

Human endotoxin model and data collection (2) Human subjects were injected intravenously with endotoxin (CC-RE, lot 2) at a dose of 2-ng/kg body weight. Blood samples were collected before endotoxin infusion (0hr) and 2, 4, 6, 9 and 24 hours after infusion. Cellular RNA was isolated from the leukocyte pellets and a total of 44,924 probe sets on the Hu133A and Hu133B arrays were analyzed thus generating the expression measurements of thousands of genes that are activated/or repressed in response to endotoxin (3).

Identification of informative expression motifs We recently proposed a novel algorithm for the identification and classification of relevant gene (4). The key motivation for this method is the realization that in the presence of noise and uncertainties looking for exact correlations or distance metrics between gene pairs may not necessarily yield the most informative interpretation. Robust, coherent and dominating qualitative features and similarities are more informative proxies of the expression experiment. The raw data is transformed into sequences symbols and these are further analyzed for consistencies. We first perform a fine-grained clustering resulting in a large number of coherent responses and then evaluate the ability of a subset of these individual clusters to reproduce the overall dynamic response of the host organism. The advantage is that we are able to perform data selection with clustering quality in mind and parse the contribution of each cluster to the overall dynamics of the system exploring the notion that genes which are part of large set of highly correlated set of genes are more likely to be significant based on the assumption that an organism responds to outside challenges to homeostasis through the utilization of a set of genes which are highly controlled in both their expression levels and temporal evolution. In order to determine the biological coherence of the selections we characterize the selected genes in terms of the ability to describe the inflammatory response through enrichment analysis (5) and gene ontology annotations (6, 7).

Results and Discussion

We identified 7,441 informative probe sets distributed across 7 informative clusters. We further reduce the number of informative probes to 5,498 by accounting for consistent magnitude changes. These probe sets account for 4055 genes. The aforementioned clusters are further classified in three categories:

Class I Three clusters (16% of probes) exhibit early up-regulation peaking at around 2hr post-exposure followed by return to base. Class I is enriched in genes involved in cytokine secretion, leukocyte adhesion, lipid transport, defense response to bacteria, cytoskeleton organization, fever, MAPKKK cascade as well as catecholamine metabolism. Previous clinical studies (8) have demonstrated that endotoxin induces a rapid physiological response in humans as well as early circulation of proinflammatory cytokines at around 2 h after endotoxin infusion. We identified genes involved in the acute phase response related to the development of the central nervous system (neurotransmitter catabolism). Their primary role is the release of hormones which include catabolism, gluconeogenesis, thyroxin (9). Specific inflammatory biomarkers included lipocalin –2 (LCN2), TIMP3 (metallopeptidase inhibitor), as well as genes involved in cholesterol biosynthesis, positive regulation of IkB/NFkB pathway, Rho protein signal transduction, protein catabolism and modification as well as in complement activation. We also identified members of TNF superfamily such as TNFRSF12 known to induce the activation of IL-8, as proinflammatory interleukin.

Class II Includes 5% of probes exhibiting delayed up-regulation with peak response in the time window 4-6 hr post-exposure and eventual return to base. Class II is enriched in genes associated with oxidative stress response, fatty acid metabolism, induction of apoptosis by extracellular signals, positive regulation of interleukins, macrophage activation as well as to ubiquitin cycle – all of them are strongly related to inflammatory pathways. An increased expression was also observed in the study (10). Biomarkers included membrane-associated proteins such as IL2RG (severe combined immunodeficiency) and IL10RB were found to dominate. Class II is also enriched in CCAAT/enhancer binding protein gene families.

Class III Three clusters (79%) of the probes exhibit clear down-regulation until about 4hr post-exposure and return to base line. Class III is enriched in genes associated with ATP – dependent proteolysis, ATP biosynthesis, mitochondrial transport, electron transport, ribosome biogenesis and assembly, proteasomal ubiquitin – dependent protein catabolism, translational elongation, cyclic- nucleotide mediate signaling as well as Ras signaling cascade and antigen processing via MHC class II and we did identify both proteasomes (PSM genes) and ribosomal genes (RPL, RPS). Inflammatory biomarkers included UQCR and COPS as well as suppressors of cytokine signaling (SOC) were identified to dominate. Class III is enriched in genes associated with antigen processing (via MHC complex II), as well as genes that affect cell-cell adhesion (integrin alpha and beta chain), cell cycle regulation and motility. Moreover, biomarkers including gene tumor necrosis factor p53 (TP53) as well as genes that are associated with b – amyloid metabolism were identified in this Class.

Conclusions and Future Research

Our preliminary results demonstrated the ability of the method to construct an integrated picture of the inflammatory response due to exposure to endotoxins. The informative genes capture the most critical features of the host response. We are currently assembling and quantifying putative regulatory interactions driving the response using NCA (11). We recently demonstrated how to further evaluate the possibility of constructing alternative regulatory sub-structures in order to evaluate the relative importance of specific regulatory interactions (12) leading to a more complete picture of how to modulate the inflammatory response.

References

1. G. S. Martin, D. M. Mannino, S. Eaton, M. Moss, N Engl J Med 348, 1546-54 (Apr 17, 2003).

2. The data used was generated by the Inflammation and Host Response to Injury Large Scale Collaborative Projected funded by the USPHS (U54 GM621119). Parts of these data have been used in previously published reports (PNAS 2005;102:4801-4806; Nature 2005; 437:1032-1037, however, the analyses in this abstract represent an approach that has not been previously published.

3. S. E. Calvano et al., Nature (Aug 31, 2005).

4. E. Yang, F. Berthiaume, M. L. Yarmush, I. P. Androulakis, paper presented at the 16th European Symposium on Computer Aided Process Engineering and 9th Int. Symp. Process Systems Engineering Garmisch-Partenkirchen / Germany 2006.

5. N. Bluthgen et al., Genome Inform Ser Workshop Genome Inform 16, 106-15 (2005).

6. Nucleic Acids Res 34, D322-6 (Jan 1, 2006).

7. D. J. Allocco, I. S. Kohane, A. J. Butte, BMC Bioinformatics 5, 18 (Feb 25, 2004).

8. K. J. Van Zee et al., J Immunol 154, 1499-507 (Feb 1, 1995).

9. P. Barany, Nephrol Dial Transplant 16, 224-7 (Feb, 2001).

10. S. E. Calvano et al., Nature 437, 1032-7 (Oct 13, 2005).

11. J. C. Liao et al., Proc Natl Acad Sci U S A 100, 15522-7 (Dec 23, 2003).

12. E. Yang, T. Maguire, M. L. Yarmush, I. P. Androulakis, Computers and Chemical Engineering submitted for publication (2006).

Extended Abstract Status: Not Uploaded