Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jul 17;9:1620.
doi: 10.3389/fimmu.2018.01620. eCollection 2018.

Footprints of Sepsis Framed Within Community Acquired Pneumonia in the Blood Transcriptome

Affiliations
Free PMC article

Footprints of Sepsis Framed Within Community Acquired Pneumonia in the Blood Transcriptome

Lydia Hopp et al. Front Immunol. .
Free PMC article

Abstract

We analyzed the blood transcriptome of sepsis framed within community-acquired pneumonia (CAP) and characterized its molecular and cellular heterogeneity in terms of functional modules of co-regulated genes with impact for the underlying pathophysiological mechanisms. Our results showed that CAP severity is associated with immune suppression owing to T-cell exhaustion and HLA and chemokine receptor deactivation, endotoxin tolerance, macrophage polarization, and metabolic conversion from oxidative phosphorylation to glycolysis. We also found footprints of host's response to viruses and bacteria, altered levels of mRNA from erythrocytes and platelets indicating coagulopathy that parallel severity of sepsis and survival. Finally, our data demonstrated chromatin re-modeling associated with extensive transcriptional deregulation of chromatin modifying enzymes, which suggests the extensive changes of DNA methylation with potential impact for marker selection and functional characterization. Based on the molecular footprints identified, we propose a novel stratification of CAP cases into six groups differing in the transcriptomic scores of CAP severity, interferon response, and erythrocyte mRNA expression with impact for prognosis. Our analysis increases the resolution of transcriptomic footprints of CAP and reveals opportunities for selecting sets of transcriptomic markers with impact for translation of omics research in terms of patient stratification schemes and sets of signature genes.

Keywords: blood disturbances; community-acquired pneumonia severity; epigenetics; immune suppression; infections; molecular subtypes; prognostic impact.

Figures

Figure 1
Figure 1
Self organizing map (SOM) portrayal of the blood transcriptome of community-acquired pneumonia (CAP). (A) The sample similarity (correlation) network and the mean expression portraits of the five sample groups, healthy control (C), discovery cohort groups 1 and 2 (D1 and D2), and validation cohort group 1 and 2 (V1 and V2) reveal two major dimensions of diversity (arrows). (B) The expression variance map indicates 12 major spots (A–L) of co-expressed genes. (C) Difference portraits between the sample groups reveal modes of differential expression. (D) The similarity score describes the similarity of each sample to its own cluster. Negative scores indicate that the sample is more similar to other clusters. The figure shows that D2 is an unspecific group for which most of the samples are more similar to C and D1 (see arrows) while C and V1 are specific groups. (E) The spot frequency distributions indicates that sample SOM portraits express more spots in direction from the left to the right, thus reflecting a larger heterogeneity of transcriptional programs activated in CAP compared with the controls.
Figure 2
Figure 2
Profiling of the expression modules: pairwise correlation maps of the sample self organizing maps and of the expression profiles of the spot modules were generated in two variants of sample sorting using either the group-classification (left part) or the inflammatory activity as estimated by the mean gene expression of spot L (right part). The barplots in the right part show the percentage of samples of each class expressing the respective spot. Healthy blood signatures enriched in the spots (p < 10−7, Fishers exact test) were taken from Ref. (20, 31) and other signatures from Ref. (32, 33).
Figure 3
Figure 3
Gene expression heatmaps of functional categories (A) gene ontology biological process, (B) lifestyle, aging, BMI, and blood pressure signatures of peripheral blood taken from Ref. (–38). (C) Selected genes with functions in immune response such as HLA type I and II, their regulators (BLM2, CIITA, RFX5), immune checkpoint inhibitors, and the inflammasome gene AIM2. All heatmaps reveal two major clusters of either decaying (green background) or increasing (apricot) expression with increasing community-acquired pneumonia (CAP) severity. Cytokines show similar profiles. Selected receptor–ligand pairs are connected by arrows and assigned to the respective immune cells according to Ref. (39). (D) Selected immunity genes were mapped onto the self organizing map portraits, where immune checkpoint inhibitors and MHC II-related genes accumulate in the region of spot A collecting genes that become deactivated in CAP. (E) GSZ-heatmaps “Immunome” signatures of immune cells and constituents, as provided by Ref. (40, 41). The part below shows gene maps and expression profiles of signature genes characterizing different stages of CD4 activation: upon activation the position of the respective genes shift from spot A to spot K (see gray arrows in the maps), where they show either decaying or increasing expression profiles.
Figure 4
Figure 4
Footprints of infections in the community-acquired pneumonia (CAP) transcriptome: (A) two-way hierarchical clustering of interferon-related, bacterial and viral response signatures reveals two clusters of IFN-high (IFN-H) and IFN-low (IFN-L) responsive cases, in addition to clusters of cases with up and down regulated expression according to CAP severity. The IFN-responsive genes accumulate strongly in spot D. (B) Gene sets from recent meta-analyses (46, 47) split into the same clusters as the gene sets in part A. The IFN-responsiveness associates with the signature for viral infections. The respective genes accumulate in different characteristic spot areas of the map. (C) The expression of gene signatures related to bacterial infections and endotoxin tolerance taken from Ref. (–51) increases with CAP severity. (D) Gene markers that differentiate between different types of infections taken from Ref. (46, 47, 52) are mapped into the expression self organizing map. They locate typically in/near spots D (viral markers), L (bacterial and ET markers), and A (sterile infection).
Figure 5
Figure 5
Analysis of the viral, erythrocyte, and platelet signatures as a function of the community-acquired pneumonia (CAP) severity scores. The scores are defined as GSZ-values of selected gene sets as indicated in the figure. The viral, erythrocyte, platelet, coagulation, and humoral response scores show different trends depending on the CAP severity score as indicated by the gray curves. The viral, erythrocyte, and also the coagulation scores positively correlate with each other as indicated by their joint accumulation within the same spot J and the respective scatter plots. The plot of the viral expression score as a function of the CAP severity score reveals a decaying baseline, which suggests exhaustion of interferon induction in more severe CAP cases. The gene set maps also show selected genes that regulate development of platelets and erythrocytes from their common precursors (58), as well as ribosome rescue in both types of cells (59). As a rationale that possibly explains co-expression of mRNA attributed to erythrocytes and platelets. Note also that the positions of the metagenes of highest gene density slightly differ between the platelet and erythrocyte signatures (see the gray circles), which reflects their similar but not identical profiles. The genes of both signatures do not overlap except for one gene (20).
Figure 6
Figure 6
Previous community-acquired pneumonia (CAP) and sepsis signatures were taken from Ref. (–11, 79) and mapped onto the data analyzed here. (A,B) Expression heatmaps (GSZ-score) of the gene signatures mostly show two types of signatures, which antagonistically switch between normal conditions and CAP. (C) Profiles and gene set maps (genes are indicated by dots in the maps) reveal further details about the disease course and the noisiness of gene expression. Regions of increased local densities of signature genes were indicated by red and green frames. (D) Maps of selected sets of marker genes accumulate in regions of the spots detected here. MARS genes were provided pair-wisely for up- and downregulation in each of the MARS endo-types (11). The pairs are connected by lines as a guide for the eye.
Figure 7
Figure 7
Expression signatures of genes referring to different chromatin states in lymphocytes are evaluated in the community-acquired pneumonia (CAP) blood transcriptome. Chromatin states as defined in Ref. (81) were taken from Ref. (28). (A) Two-way hierarchical clustering of the GSZ-heatmap reveals two major groups of chromatin states showing either high expression in predominantly control and group 2 samples (green) or in predominantly group 1 samples (red). Genes upregulated in the former cluster are commonly in transcriptional active states, whereas poised and repressed states on the average are at low expression level. This relation reverses in the latter cluster. (B) Boxplots of mean group expression of selected chromatin states show either decaying or increasing expression with increasing severity of CAP. Note the relatively large height of the boxes due to large variability of gene expression (GE) in most of the groups. (C) Bar plots of samples ranked with CAP severity underline the increasing and decreasing trends but also the noisy character of GE. Gene sets were taken from Ref. (31, 32, 82). (D) Gene maps indicate the accumulation of genes from transcribed and repressed chromatin states in T-cells in different areas of the map, as indicated by the dashed frames. The genes were either shown as dots, or as local percentages in areas determined by K-means clustering (12) (number of genes attributed to the state/all genes in the respective area). The population map visualizes the number of genes in each pixel. Chromatin states were defined as follows: active promoters (TssA), transcribed genes (Tx), active enhancer and enhancer-like states (Enh and EnhG), zinc finger proteins (ZNF_rpts), quiescent (Quies), heterochromatin (Het), poised promoters, and enhancers (TssBiv, EnhBiv), repressed polycomb states (ReprPC) (81).
Figure 8
Figure 8
(A) Expression heatmap of chromatin modifying enzymes in the five sample groups: the gene expression data of methyltransferases (MTs) and demethylases (DM’s) of DNA cytosines, histone lysine (H3K), and arginine (H3R) side chains split into two main clusters down- or upregulated in the control group compared with the community-acquired pneumonia groups. The plots in the right part show expression profiles for selected enzymes with single sample resolution. (B) Maps of genes encoding lysine MTs and demethylases of H3K4, H3K9, H3K26, and H3K36, and of their local percentage among all genes. The genes are accumulated in the region assigned to active chromatin states and are depleted in the region assigned to repressed and poised states as shown in Figure 7D. (C) Genes and gene signatures related to DNA methylation suggest activation of repressed and poised genes by DNA-demethylation.
Figure 9
Figure 9
Re-classification of community-acquired pneumonia samples into six novel groups and their prognostic impact. (A) The sankey diagram (R-program “riverplot”) illustrates the composition flow between the original and novel classes [low, medium and high severity, LS, MS, HS, respectively; blood disturbances (BDs); IFN high and low or middle severity, IFN LS and IFN HS, respectively], while the class-averaged mean portraits visualize the respective expression landscapes. The new subtypes differ with regard to severity levels (low, medium, high, spots A, H, and L) and the expression of spots J and D being assigned to BDs and interferon response, respectively (see also Table S4 in Supplementary Material). (B) The prognostic map links the expression level with the 28-day survival rate where the latter is calculated as the percentage of patients expressing the respective metagene and surviving 28-days after admission in intensive care unit. (C) Re-coloring of the correlation net according to new classification illustrates the asymmetrical distribution of the novel groups with respect to the discovery and verification cohorts (compare with the similarity net in Figure 1A). (D) The pie charts show the composition of the original groups with respect to the new ones. Note that about 65% of group 1 cases are collected in the novel HS group, while the remaining cases are mainly assigned to IFN HS. In contrast, about 50% of group 2 cases consist of LS and MS cases, while the remaining half of the cases distribute over the remaining groups.
Figure 10
Figure 10
Visualization of the transcriptome landscape on three levels (A), namely “personalized” expression portraits, a “feature” gene expression (GE) landscape and a sample diversity landscape and schematic overview of the major community-acquired pneumonia (CAP) footprints in the blood transcriptome (B). (A) The portraits show up- and downregulated modules of correlated genes in red and blue, respectively, where the vertical coordinate visualizes the expression and the horizontal one the self organizing maps-mosaic. The GE “feature” landscape summarizes these spots as maroon areas of variant GE (see also Figure 1B). For each spot A–L, we colored the cases showing this particular spot in the diversity network (see also Figure 1A). Spot activation associates with groups of samples in the network: increasing severity of CAP activates spots roughly along the arrow in the GE landscape, which in turn associates with cases located in direction of the arrows shown in the similarity nets.

Similar articles

See all similar articles

Cited by 6 articles

See all "Cited by" articles

References

    1. Morgan AJ, Glossop AJ. Severe community-acquired pneumonia. BJA Educ (2016) 16(5):167–72.10.1093/bjaed/mkv052 - DOI
    1. Pereira JM, Paiva JA, Rello J. Severe sepsis in community-acquired pneumonia — early recognition and treatment. Eur J Intern Med (2012) 23(5):412–9.10.1016/j.ejim.2012.04.016 - DOI - PubMed
    1. Seligman R, Ramos-Lima LF, do Amaral Oliveira V, Sanvicente C, Pacheco EF, Rosa KD. Biomarkers in community-acquired pneumonia: a state-of-the-art review. Clinics (2012) 67(11):1321–5.10.6061/clinics/2012(11)17 - DOI - PMC - PubMed
    1. Marti C, Garin N, Grosgurin O, Poncet A, Combescure C, Carballo S, et al. Prediction of severe community-acquired pneumonia: a systematic review and meta-analysis. Crit Care (2012) 16(4):R141.10.1186/cc11447 - DOI - PMC - PubMed
    1. Viasus D, Simonetti A, Garcia-Vidal C, Carratalà J. Prediction of prognosis by markers in community-acquired pneumonia. Expert Rev Anti Infect Ther (2013) 11(9):917–29.10.1586/14787210.2013.825442 - DOI - PubMed
Feedback