The HOPACH clustering method (Figure 3) produced five distinctly separated clusters: 1) Y. pestis (KIM5 D27, India/P, and NYC), 2) Y. pseudotuberculosis, 3) Y. enterocolitica, 4) B. anthracis (Ames and Sterne), and 5) Control. This result is consistent with the findings using the correlation distance and the Euclidean distance with average linkage. In addition, HOPACH estimated the optimal number of clusters as five. That is, the Yersinia subcluster is best if it is divided into the three clusters specified by 1) through 5) above. Y. enterocolitica forms its own cluster, and so does Y. pseudotuberculosis. Y. pestis (KIM5 D27), Y. pestis (India/P), and Y. pestis (NYC) are grouped into one cluster. Further subdivisions lead to an overall clustering with inferior quality.
In addition to clustering the cytokine expression profiles across bacterial treatments, Figure 3 also groups the cytokines themselves and clusters the proteins based on their similarities across the pathogen exposures and reorders them accordingly. Interestingly, the three pro-inflammatory cytokines IL-1β, TNFα, and IL-6 clustered closely, and so did the three chemokines MCP-1, IP-10, and IL-8. Although these 6 cytokines do not cluster as a single group, they do cluster at a branch further away from the leaf node, which includes IL-10 and sCD95, to make a larger group of 8 proteins. Several of these proteins are involved in inflammatory conditions, such as IL-1beta, TNFα, IL-6,  and have been shown to be upregulated in cell culture and animal model specifically exposed to biothreat agents . Increased expression of IL-6 and TNFα clustered together in a study involving mouse splenic CD11b + cells following sub-lethal Y. enterocolitica infection . In addition, several of the cytokines in this cluster, namely TNF-alpha, IL1-beta, IL-10, and MCP-1 are expressed higher in exposed whole blood as compared to control in this study and in whole blood exposure to LPS from several other gram negative bacterial pathogens . In addition to expression differences, the absence of detected cytokine expression can also be helpful in discriminating pathogen exposure.
The multiplex detection of 30 cytokines in this study revealed the early phase cytokine expression profiles in human plasma following exposures to B. anthracis (Ames and Sterne), Y. pestis (KIM5 D27, NYC and India/P), Y. pseudotuberculosis, and Y. enterocolitica. The expression levels of 8 cytokines, IL-1α, IL-1β, IL-6, IL-8, IL-10, IP-10, MCP-1, and TNFα were significantly different from that of unexposed control (Figure 2). Although the focus of our work was to show that cytokine expression profiling can discriminate between different pathogen exposures in a human whole blood ex vivo model, these results also represent an initial attempt to characterize the full cytokine response to each individual pathogen. Our preliminary study using a single exposure protocol at a single time post-exposure will need to be supplemented with more thorough investigation in order to determine the usefulness of using cytokine levels for diagnosing pathogen exposure. However, the single time point chosen, 4 hours, is sufficient to detect proteomic changes and has been used in previous studies examining cytokine levels [25–27]. This time point represents a start towards a more complete temporal study, as has been done with gene expression patterns for two of the pathogens studied here [25, 27]. In addition, studies that provide expression patterns for a single cytokine using multiple time points will also be needed to make the results of this paper clinically useful, such as has been done by, Cooper and coworkers, who examined IL-12p40 and IL-12p70 levels following different growth conditions and exposure levels for a time course of Y. pestis exposed dendritic cells . The results of the current work shows a similar expression pattern trend to this previous work, in which, Y. pestis induces IL-12p40 and at a substantially higher level than IL-12p70.
Our results showed that the expression levels of 3 chemokines, IL-8, MCP-1 and IP-10, were induced by both Yersinia and B. anthracis exposures. No significant differences were found for these cytokines between Yersinia and B. anthracis exposures. IL-8, MCP-1 and IP-10 are chemokines that enable the migration of leukocytes from the blood to the site of inflammation. IL-8 is a key chemokine regulating neutrophil recruitment . The essential involvement of IL-8 in acute inflammation was demonstrated by neutralizing IL-8 with its antibody. When highly specific antibody against IL-8 was administered in acute inflammatory reactions induced by several stimuli including lipopolysaccharide, neutrophil infiltration was blocked . MCP-1 is known for its ability to act as potent chemoattractant and activator of monocytes/macrophages as well as NK cells but not neutrophils [31, 32] . IP-10 has no chemotactic activity for neutrophils but attracts monocytes, NK, and T cells to the site of infection and regulates T cell maturation [33, 34]. It was reported previously that elevated IL-8 and MCP-1 were secreted by human epithelial cells after Y. enterocolitica infection, but not IP-10 [35, 36]. Human dendritic cells, infected with B. anthracis spores, secreted high level of IL-8 at 7.5 hours . In our study, the fold increase of IL-8 was much greater than MCP-1 and IP-10 (Figure 2). For example, the induction of IL-8 by Ames strain of B. anthracis was 41 fold, while MCP-1 was 2 fold and IP-10 was 2.5 fold (Figure 2). This result may indicate that IL-8 is a dominant chemokine in early response (4 hours exposure in our study) and neutrophils are the major player in early inflammatory response.
Here we compared cytokines induced by B. anthracis and Yersinia exposures. Overall, Yersinia exposure induced higher levels of IL-1α, IL-1β, IL-6, IL-10 and TNFα than B. anthracis exposure, suggesting these cytokines could be used to develop an assay for discriminating Yersinia spp. from B. anthracis exposures. The vaccine strain (Sterne) of B. anthracis induced higher levels of IL-1β and TNFα than the virulent strain (Ames) (Figure 2), suggesting these cytokines can contribute to a biomarker panel to discriminate if a particular isolate of B. anthracis is virulent. There was also a difference in induction of IL-10 between Y. pestis and near neighbors (Figure 2), suggesting this cytokine is a candidate biomarker for discriminating the virulence of Yersinia species. These data regarding IL-10 expression following Yersinia spp. exposure are in agreement with published literature that shows Y. enterocolitica and Y. pestis can elicit statistically different levels of IL-10 expression . Differences in IL-10 induction may be due to differences in the lcrV protein among Yersinia spp.. The different cytokine profiles induced by B. anthracis and Yersinia here may be partially due to different surface antigens on the outermost part of these pathogens and the manner in which these bacteria were grown. Lipopolysaccharide (LPS), the main constituent of the outer membrane of Gram-negative bacteria, and peptidoglycan (PGN), the major cell wall component of Gram-positive bacteria, have been reported to elicit markedly different immune responses . However, virulence factors, such as B. anthracis lethal toxin and Yersinia virulence antigen, LcrV, may also play important roles in differential cytokine induction. This view is supported by numerous reports that B. anthracis toxin and virulence factors of Yersinia bacteria (Yops, invasin, LcrV) modulate host cytokine responses [40–51].
While the various clustering methods resulted in slightly different final hierarchies, all were consistent in separating the unexposed control from the samples exposed to B. anthracis or to the Y. pestis and near neighbors. Agreement on this level among the various clustering procedures lends more confidence to the overall results. On a more detailed level, the methods grouped slightly differently the samples exposed to the Y. pestis and near neighbors, which indicates that these samples cannot be unequivocally separated based on the current data and additional biomarkers or a larger sample set would be needed. The most advanced HOPACH method estimated the optimal number of clusters in the data as five, corresponding to the unexposed control, and the four species: B. anthracis, Y. pseudotuberculosis, Y. enterocolitica, and Y. pestis (avirulent and virulent) (Figure 3).
Information gained from the targeted protein array data for host response complements genomic [52–56], and other proteomic studies [57–60] of host-pathogen interactions. The success of the WEEM and computational method to distinguish pathogen exposure, based on host response in this initial study, is encouraging and suggests a number of possibilities for future studies to refine the findings. Comparative analysis, such as the current work, can potentially reveal the critical pathogenic mechanism(s) and host innate immune responses during infection as was previously shown for Y. pestis and Y. pseudotuberculosis. Opportunities include using statistical hypothesis tests based on analysis of variance to assess the significance of the observed differences among the host-pathogen cytokine concentration profiles, as well as performing follow-up studies to focus more on the Y. pestis and near neighbor cluster. In addition, the methods can be extended to investigate host responses to diverse pathogens in multiple host model systems to cross validate the significance of the biomarkers to distinguish pathogen exposures.