Modular Transcriptional Networks of the Host Pulmonary Response during Early and Late Pneumococcal Pneumonia
- 8 Downloads
Streptococcus pneumoniae (Spneu) remains the most lethal bacterial pathogen and the dominant agent of community-acquired pneumonia. Treatment has perennially focused on the use of antibiotics, albeit scrutinized due to the occurrence of antibiotic-resistant Spneu strains. Immunomodulatory strategies have emerged as potential treatment options. Although promising, immunomodulation can lead to improper tissue functions either at steady state or upon infectious challenge. This argues for the availability of tools to enable a detailed assessment of whole pulmonary functions during the course of infection, not only those functions biased to the defense response. Thus, through the use of an unbiased tissue microarray and bioinformatics approach, we aimed to construct a comprehensive map of whole-lung transcriptional activity and cellular pathways during the course of pneumococcal pneumonia. We performed genome-wide transcriptional analysis of whole lungs before and 6 and 48 h after Spneu infection in mice. The 4,000 most variable transcripts across all samples were used to assemble a gene coexpression network comprising 13 intercorrelating modules (clusters of genes). Fifty-four percent of this whole-lung transcriptional network was altered 6 and 48 h after Spneu infection. Canonical signaling pathway analysis uncovered known pathways imparting protection, including IL17A/IL17F signaling and previously undetected mechanisms that included lipid metabolism. Through in silico prediction of cell types, pathways were observed to enrich for distinct cell types such as a novel stromal cell lipid metabolism pathway. These cellular mechanisms were furthermore anchored at functional hub genes of cellular fate, differentiation, growth and transcription. Collectively, we provide a benchmark unsupervised map of whole-lung transcriptional relationships and cellular activity during early and late pneumococcal pneumonia.
Streptococcus pneumoniae (Spneu) remains the most lethal bacterial pathogen and the dominant agent of community-acquired pneumonia (1,2). Considerable research has been dedicated to defining risk factors and designing risk-scoring systems associated with mortality that could aid in important site-of-care decisions regarding the treatment patients with pneumonia (3). Treatment has relied heavily on the immediate initiation of empirical antibiotic therapy, which in recent years has received substantial scrutiny owing to the emergence of antibiotic-resistant Spneu strains (3). The development of new antibiotics has provided promising results (4); however, the ever-evolving nature of commensal bacteria coupled with their ability to suppress the normal host defense mechanisms by virtue of their virulome (5,6) dictate studies that provide more insight into the immune response to pneumococci that enter the airways. Modulating the immune responses to infection, either by suppression or activation, has emerged as a useful therapeutic approach in the prevention and treatment of bacterial infections (7). Despite such advantages, improper activation and/or suppression of innate immune reactions can result in detrimental inflammatory responses and tissue damage (8). Thus, it is important to assemble a detailed pattern of pangenomic responses and related cellular biological pathways representative of the whole lung and the host pulmonary immune response to pneumococcal infection.
Employing a mouse model of acute pneumococcal pneumonia, we analyzed the genome-wide transcriptional responses in the whole lung at the early (6 h) and late (48 h) stages of the host response. By combining linear modeling to assess differential gene expression and the concepts of scale-free network biology (9), we constructed a gene coexpression network based on the most variable genes across the early and late phases of the host pulmonary transcriptional response. This network was organized into transcriptional modules of highly significant cellular signaling pathways anchored at regulatory “driver” genes, which are understood to be fundamental components of biological response systems, oftentimes directly targeted by bacterial pathogens to favorably modulate host immunity (10). These findings represent a benchmark characterization of the modular properties underlying the whole lung and the host pulmonary response to early and late acute pneumococcal infection.
Materials and Methods
Mice and Husbandry
C57BL/6 female mice (9–12 wks old) were purchased from Charles River (Maastricht, the Netherlands). All mice were maintained in pathogen-free conditions in the animal facility of the Academic Medical Center (Amsterdam, the Netherlands). Mice were supplied with the same rodent chow and water ad libitum and maintained on a 12-h light-dark cycle in a temperature- and humidity-controlled environment. The Animal Care and Use Committee of the University of Amsterdam approved all experiments.
Induction of Pneumonia and Bacterial Burden Assessment
The S. pneumoniae serotype 2 D39 strain was grown for 5 h to the midlogarithmic phase at 37°C using Todd-Hewitt broth (Difco, Detroit, MI) with yeast extract (0.5%), harvested by centrifugation at 2,683g for 15 min, and washed twice in isotonic saline. A suspension of 2 × 107 colony-forming units (CFUs) in 50 µL isotonic saline was used as the inoculum. Mice were lightly anaesthetized by inhalation of 2.0%–2.5% isoflurane (Upjohn, Ede, the Netherlands) mixed with O2 (1–2 L/min) and intranasally challenged as previously described (11,12). Mice were humanely euthanized at 6 or 48 h postinfection (n = 4 at each time point). Lung tissue was harvested and processed for the determination of bacterial outgrowth as described (11,12). CFUs in lung homogenates were determined from serial dilutions plated on blood agar plates incubated at 37°C for 16 h before colonies were counted. The lungs from 3 uninfected mice were used as controls.
Lungs were fixed in formalin and embedded in paraffin. Sections (5 µm) were stained with hematoxylin and eosin (HE). Granulocyte staining was performed by anti-Ly6G hybdridization and quantified by means of ImageJ software (Bethesda, MD, USA; https://doi.org/rsb.info.nih.gov/ij/) as previously described (12,13).
RNA Preparation and Genome-Wide Transcriptional Profiling
Total RNA was isolated from lung homogenates using the Nucleospin RNA II kit (Machery-Nagel [BIOKÉ, Leiden, the Netherlands]). Yield and purity (260 nm:280 nm) were determined by use of the Nanodrop ND-1000 (Thermo Scientific [Thermo Fisher Scientific, Waltham, MA, USA]). The integrity (RNA integrity number [RIN] >7.0) of the resuspended total RNA was determined by using the RNA Nano Chip Kit on the Bioanalyzer 2100 and the 2100 Expert software (Agilent, Amstelveen, the Netherlands). The Illumina TotalPrep-96 RNA amplification kit (Illumina, Eindhoven, the Netherlands) was used to generate biotinlabeled (biotin-16-UTP) amplified cRNA starting from 200 ng total RNA. A total of 750 ng biotinylated cRNA was hybridized onto the Illumina MouseRef-8v2 Expression BeadChip. The samples were scanned using the Illumina iScan array scanner. Preparation of cRNA, chip hybridization, washing, staining and scanning were carried out at ServiceXS (Leiden, the Netherlands). The raw scan data were read using the BeadArray package (version 1.12.1) (14) available through Bioconductor (15) using the R statistical environment (version 2.13.2; R Foundation for Statistical Computing, Vienna, Austria). Estimated background was subtracted from the foreground for each bead. For replicate beads, outliers greater than 3 median absolute deviations from the median were removed and the average signal was calculated for the remaining intensities. For each probe a detection score was calculated by comparing its average signal with the summarized values for the negative control probes. Resulting data were neqc normalized (16). Quality control was performed both on the bead level and on bead summary data. The arrayQualityMetrics package v2.6.0 (17) was used for further quality assessment of the normalized bead summary data. Probes were reannotated using the package illuminaMousev2.db (18) from Bioconductor. All nonnormalized and normalized data are available at the gene expression omnibus of NCBI (GEO) with accession number GSE42464.
Differential Gene Expression Analysis and Modular Network Construction
Differential probe intensities on microarrays were identified using the R package limma (version 3.8.3) that implements linear models for microarray data (19). Empirical Bayesian methods are used to generate p values obtained from moderated t statistics or f statistics. These p values are then adjusted for multiple comparisons with Benjamini-Hochberg (BH) method to control the false discovery rate (20). The weighted gene coexpression network construction algorithm was used to build the lung gene coexpression network as described previously (21,22). Briefly, a pair-wise Spearman correlation matrix of the top 4,000 most variable unique genes (ranked by coefficient of variation) was transformed into an adjacency matrix by using a “soft” power function of 7 ensuring scale-free topology (21). The adjacency matrix was further transformed into a topological overlap matrix to enable the identification of modules (clusters) of highly correlating genes by implementing a dynamic tree cut algorithm (21). Each module represents a cluster of coregulated genes with a distinct expression pattern from other identified modules. To define module driver genes we made use of the module eigengene concept, defined as the first principal component of the module expression matrix and the module membership measure, k (21,23). To provide robustness to our findings we performed weighted gene coexpression network analysis (WGCNA) on an independent publicly available lung transcriptome dataset generated from female mice (9–12 wks old) uninfected (n = 5) and 6 h after S. pneumoniae serotype 2 strain D39 intransal infection (n = 5) (GSE49533) (24).
Pathway analysis of the gene coexpression modules in connection with differential gene expression between experimental groups was performed using a comparison-based analysis in IPA (Ingenuity® Systems, Qiagen, Valencia, CA, USA; https://doi.org/www.ingenuity.com). Specifically, biological function overrepresentation and canonical pathway enrichment were performed using default settings. Module genes presenting differential expression with multiple-comparison corrected p values <0.05 in the univariate analysis between uninfected controls and Spneu at 6 h and between uninfected controls and Spneu at 48 h were eligible. Coexpression network visualizations were mapped by using Cytoscape version 2.8.2 (25). In silico cell-specific coexpression predictions were performed by testing for overrepresentation in the co-expression atlas of the Immunological Genome Project (Immgen.org) (26) available in the ToppGene bioinformatics suite (27). This analysis was performed with default parameters. All probabilities were adjusted for multiple comparisons by the BH method. BioMart (Ensembl, Hinxton, UK) was used to identify network genes that coincided within the pneumonia susceptibility quantitative trait locus (QTL) support interval (28).
Wilcoxon rank sum tests were performed to assess the HE pathology and anti-Ly6G+ histology scores. Logistic regression implemented in the globaltest R package (29) was used to assess the statistical significance of each module per infection time point. BH-adjusted p values <0.05 demarcated significant associations.
Differential Host Gene Expression during Pneumococcal Pneumonia
Modular Network Analysis Reveals Host Cellular Biological Pathways Shaping the Pulmonary Defense Response
Temporal Patterns and Regulatory Drivers of the Host Modular Transcriptional Network during Pneumococcal Pneumonia
Lung transcriptional network genes physically located within the chromosome 7 Spir1 quantitative trait locus support interval.
In this study a mouse model of early and late acute pneumococcal pneumonia was employed to construct a scale-free modular network of host pulmonary transcriptional responses to infection. This network encompassed modules of highly interconnected transcripts anchored at potentially crucial module driver genes. Modules were determined to be enriched for genes involved in distinct biological pathways, which included IL17A/IL17F signaling, NF-κB signaling, interferon signaling, lipid metabolism and the complement system pathway. These modules were predicted to constitute gene expression patterns attuned to shared and distinct cell-types including myeloid cells, such as granulocytes and macrophages, lymphoid cells, which included α-β and γ-delta T cells, as well as stromal cells. The early immune response phase was predominantly associated with an increase in granulocyte (IL17A/IL17F and NF-κB signaling) and stromal cell-associated (lipid metabolism) pathways, whereas the late phase of the host immune response was substantially impacted by macrophage (interferon signaling) and lymphoid cell-associated (complement system) pathways. Regulatory driver genes of the host pulmonary defense response included the cell fate, differentiation and proliferation genes, Wnt3a and Csf1, the transcriptional factors Irf7 and Fam111a, a glucose-regulated protein binding factor, Tmem132a, a phospholipid scramblase, Plscr1, and an interferon-induced transmembrane protein, Iftm5. These findings suggest a previously undisclosed, yet potentially integral role of these network driver genes in the host lung response to Spneu infection.
Components of complex systems, such as the cellular and tissue responses to infectious pathogens, are not randomly connected. Signal transduction pathways that are activated upon pathogen recognition by the innate immune system follow an intricately coordinated path converging to specific nodes or hubs; therefore, they can be considered as scale free in character (9). This is exemplified by the toll-like receptor (TLR)-initiated signal transduction pathway that primarily culminates in the activation of the master transcription factor NF-κB. The biological importance of such networks has been highlighted by evidence of bacterial virulence factors directly targeting network hub components to suppress the hosts’ immune response (10). In the same fashion therapeutic agents, either in isolation or in combination with antibiotics as adjunctive therapies, have been proposed to treat bacterial infections; for example, the innate defense-regulator peptide-1 (IDR-1) for protection against gram-positive and gram-negative pathogens, including methicillin-resistant Staphylococcus aureus (MRSA) (34). The nonstochastic nature of the hosts’ defense mechanisms against bacterial infection (as well as other infectious pathogens) permits the emergence of immunomodulatory strategies. Although bacterial pathogens have evolved such immunomodulating mechanisms by selection pressures imposed by the hostile host environment, our knowledge of such host response topologies in the primary site of infection is lacking in comparison. To the aim of garnering a systems-based viewpoint of the host lung before and after 6- and 48-h pneumococcal infection, we utilized a weighted gene coexpression approach (21,22) to construct a scale-free network representative of the whole lung (Figure 3). This approach allowed us to define 13 transcriptional modules, of which 7 modules were significantly associated with distinct biological signaling pathways. Five of these functionally annotated modules, including the well-established protective host pathways IL17A/IL17F signaling (35), NF-κB signaling (36) and the complement system (37), were also, not surprisingly, significantly altered due to pneumococcal infection (Figure 4). Six network modules were not significantly associated with known biological pathways (Figure 3), which suggests biological interactions of these module genes have yet to be experimentally revealed. Moreover, two of these nonannotated modules (mod1 and mod9; Figures 5D, G) were significantly altered due to Spneu infection and predicted to be driven by Tmem132a (mod1) and Ifitm5 (mod9). These nonannotated modules (mod1 and mod9) and a stromal cell lipid metabolism module (Figure 5F) were uniquely significant to the early (6-h) host pulmonary response (Figure 4). Of note, the putative driver gene for the stromal cell lipid metabolism module, Plscr1, encoding phospholipid scramblase 1, was recently unmasked as a mediator of the type-I interferon-induced protection against staphylococcal α-toxin in the human epithelial A549 cell line (38).
Host responses to infectious or any other deleterious agent are characterized by an intricate pattern of cell mobilization, differentiation, sacrifice and, as important, cell-to-cell communication. Whole-tissue transcriptomics during infection, as employed in this study, certainly presents a challenge in interpretation because differential gene expression is often the result of cell influx into the tissue rather than a bona fide transcriptional activation or repression. However, tissue-wide transcriptional profiling coupled with current knowledge-based analytical tools can provide a unique opportunity to delineate biological functions and canonical signaling pathways pertaining to predicted cell-type-specific phenotypes in vivo. This approach, which circumvents artifacts that may occur as a consequence of cell sorting, has enabled us to map the transcriptional modules to various shared and distinct cell-types (Figure 3B). Although these results showed predominantly nonspecific mapping to cell-types, for example, the IL17A/IL17F signaling module was enriched for gene markers of granulocyte, dendritic cell, macrophage, monocyte and B-cell types (Figure 3B). Such nonspecific mapping to cell subsets may be due to transcriptional network modules harboring submodules that might in turn map to specific cell-types, or certain biological pathways, for example NF-κB signaling (Figures 3A, B), are shared across cell subsets (39).
Host defense against pathogenic microbes (and other parasites) constitutes two primary components of conceptually different mechanisms (40). The first is defined as the ability of the host to limit microbial burden, termed resistance. The second is defined as the ability of the host to limit disease severity induced by a given microbial burden, which is termed tolerance (40). Notably, resistance has been defined as the inverse of pathogen burden (40,41). Through our systems-based approach we showed an inverse of lung bacterial burden (Figure 1E) and transcriptional modules enriched for interferon-signaling genes and, not surprisingly, complement system genes during the late phase of the host pulmonary response (Figure 4). Moreover, we predict macrophages (interferon signaling; Figures 3B and 5B) and γ-delta T cells (complement system pathway; Figures 3B and 5E) as key mediators. Monocytes, macrophages, dendritic cells and their precursors are understood to form networks of phagocytic cells, altogether called the mononuclear phagocyte system, which play important roles in development, inflammation and antipathogen defense (42,43). γ-Delta T cells have been proposed as key mediators in the immune response by virtue of their interactions with antigen-presenting cells (44). Our results provide further indication for an intricate network of macrophage and γ-delta T-cell cross-talk, primarily via interferon signaling and complement system pathways, in the resistance phase of the host defense response, perhaps also driving the differentiation of progenitor/precursor cells in the mononuclear phagocytic niche. Indeed, recent and past research highlighted the importance of complement factors in promoting the differentiation of myeloid-derived suppressor cells (45) and modulating T-cell responses (46,47). Thus, our findings that show one module overrepresented for both the complement system and (γ-delta) T-cell markers (Figure 3) lends weight to the concept of nonconventional pleiotropic effects of complement system signaling.
Our study has limitations. The lung transcriptional network during early and late pneumococcal pneumonia was constructed using a single mouse strain (C57Bl/6). Although we provide initial validation of the transcriptional modules in a BALB/c mouse model, further exploration of the lung transcriptome during pneumococcal pneumonia in other mouse strains is needed. Moreover, our model relied on only one S. pneumoniae serotype strain. Evaluation of the lung transcriptional response to early and late pneumococcal pneumonia using other virulent S. pneumoniae strains is certainly warranted. Lastly, the mice used for this study were relatively young (9–12 wks old), considering that the majority of pneumococcal pneumonia cases are in elderly humans.
Here, we report an unbiased investigation of the host pulmonary transcriptional alterations that occur as a consequence of acute pneumococcal infection coupled with immunohistochemical analysis. By means of systems-based genomics and bioinformatics approaches we constructed a dynamic network of modular transcriptional programs that shape the early and late phases of the host defense response. Early host responses to pneumococcal pneumonia were characterized by heightened expression of granulocyte-associated IL17A/IL17F and NF-κB signaling, established as protective mechanisms during Spneu infection, as well as a novel stromal cell lipid metabolism pathway. The late host responses were characterized by a γ-delta T-cell complement system and macrophage interferon signaling.
The authors declare they have no competing interests as defined by Molecular Medicine, or other interests that might be perceived to influence the results and discussion reported in this paper.
We thank Joost Daalhuisen and Regina de Beer for expert technical assistance and the animal caretakers of the AMC animal research core facility.
- 20.Benjamini Y, Hochberg Y. (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Series B Stat. Methodol. 57:289–300.Google Scholar
- 35.Tsai HC, Velichko S, Hung LY, Wu R. (2013) IL-17A and Th17 cells in lung inflammation: an update on the role of Th17 cell differentiation and IL-17R signaling in host defense against infection. Clin. Dev. Immunol. 2013:267971.Google Scholar
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, and provide a link to the Creative Commons license. You do not have permission under this license to share adapted material derived from this article or parts of it.
The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
To view a copy of this license, visit (https://doi.org/creativecommons.org/licenses/by-nc-nd/4.0/)