If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address reprint requests to Claudia C. dos Santos, M.D., Interdepartmental Division of Critical Care, St. Michael's Hospital/University of Toronto, 30 Bond St., Room 4-011, Toronto, ON, M5B 1WB, Canada; or W. Conrad Liles, M.D., Ph.D., University of Washington, 1959 NE Pacific St., RR-511, Box 356420, Seattle, WA 98195-6420
Interdepartmental Division of Critical Care, The Keenan Research Centre of the Li Ka Shing Knowledge Institute of St. Michael's Hospital, University of Toronto, Toronto, Ontario, CanadaInterdepartmental Division of Critical Care, Institute of Medical Sciences, University of Toronto, Toronto, Ontario, Canada
Department of Medicine, Sandra A. Rotman Laboratory for Global Health Research, McLaughlin-Rotman Centre for Global Health, McLaughlin Center for Molecular Medicine, University of Toronto, Toronto, Ontario, CanadaDepartment of Pediatric Critical Care Medicine, Hospital for Sick Children, Toronto, Ontario, Canada
Interdepartmental Division of Critical Care, The Keenan Research Centre of the Li Ka Shing Knowledge Institute of St. Michael's Hospital, University of Toronto, Toronto, Ontario, Canada
Interdepartmental Division of Critical Care, The Keenan Research Centre of the Li Ka Shing Knowledge Institute of St. Michael's Hospital, University of Toronto, Toronto, Ontario, CanadaDepartment of Anesthesiology and Intensive Care, Lund University Hospital, Lund, Sweden
Address reprint requests to Claudia C. dos Santos, M.D., Interdepartmental Division of Critical Care, St. Michael's Hospital/University of Toronto, 30 Bond St., Room 4-011, Toronto, ON, M5B 1WB, Canada; or W. Conrad Liles, M.D., Ph.D., University of Washington, 1959 NE Pacific St., RR-511, Box 356420, Seattle, WA 98195-6420
Interdepartmental Division of Critical Care, Institute of Medical Sciences, University of Toronto, Toronto, Ontario, CanadaDepartment of Medicine, Sandra A. Rotman Laboratory for Global Health Research, McLaughlin-Rotman Centre for Global Health, McLaughlin Center for Molecular Medicine, University of Toronto, Toronto, Ontario, CanadaDepartment of Medicine, University of Washington, Seattle, Washington
Although bone marrow–derived mesenchymal stem cell (MSC) systemic administration reduces sepsis-associated inflammation, organ injury, and mortality in clinically relevant models of polymicrobial sepsis, the cellular and molecular mechanisms mediating beneficial effects are controversial. This study identifies the molecular mechanisms of MSC-conferred protection in sepsis by interrogating transcriptional responses of target organs to MSC therapy. Sepsis was induced in C57Bl/6J mice by cecal ligation and puncture, followed 6 hours later by an i.v. injection of either MSCs or saline. Total RNA from lungs, hearts, kidneys, livers, and spleens harvested 28 hours after cecal ligation and puncture was hybridized to mouse expression bead arrays. Common transcriptional responses were analyzed using a network knowledge-based approach. A total of 4751 genes were significantly changed between placebo- and MSC-treated mice (adjusted P ≤ 0.05). Transcriptional responses identified three common effects of MSC administration in all five organs examined: i) attenuation of sepsis-induced mitochondrial-related functional derangement, ii down-regulation of endotoxin/Toll-like receptor innate immune proinflammatory transcriptional responses, and iii) coordinated expression of transcriptional programs implicated in the preservation of endothelial/vascular integrity. Transcriptomic analysis indicates that the protective effect of MSC therapy in sepsis is not limited to a single mediator or pathway but involves a range of complementary activities affecting biological networks playing critical roles in the control of host cell metabolism and inflammatory response.
Severe sepsis remains the second leading cause of intensive care unit morbidity and mortality.
ACCP/SCCM Consensus Conference Committee, American College of Chest Physicians/Society of Critical Care Medicine Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis.
These derangements are complex processes that include activation of innate immunity, suppression of the adaptive immune responses, mitochondrial dysfunction, development of a hypermetabolic state with insulin resistance,
Although multiple potential therapeutic targets have emerged from studies of these derangements, the development of an effective therapeutic intervention has proved to be challenging.
Treatments for sepsis are largely limited to surgical removal or drainage of the infected site (ie, “source control”), antimicrobial agents, and supportive care. Recently, it has been shown that systemic administration of mesenchymal stem (stromal) cells (MSCs) reduces sepsis-associated inflammation and multiple organ injury in clinically relevant models of polymicrobial sepsis.
Bone marrow stromal cells attenuate sepsis via prostaglandin E(2)-dependent reprogramming of host macrophages to increase their interleukin-10 production.
Although these studies demonstrate the feasibility of exploiting cell-based technologies for the treatment of sepsis, the mechanism(s) underlying the beneficial effects of MSCs remains to be elucidated.
Although cell engraftment with differentiation, transdifferentiation, or cell fusion may contribute to the reconstitution of normal organ function, this phenomenon is now generally felt to be a rare occurrence of uncertain physiologic significance in models of sepsis-induced multiple organ dysfunction syndrome.
In contrast, several paracrine mediator–specific mechanisms have been proposed as effectors of MSC-conferred protection from injury, including anti-inflammatory mediators, such as IL-10, angiopoietin-1, keratinocyte growth factor,
Bone marrow stromal cells attenuate sepsis via prostaglandin E(2)-dependent reprogramming of host macrophages to increase their interleukin-10 production.
Recent work, however, points to the role of MSCs in “reprogramming” the inflammatory response network in sepsis. MSC administration in cecal ligation and puncture (CLP)–injured mice resulted in coordinated modulation of the host transcriptional response characterized by an overall down-regulation of inflammation and inflammation-related genes and a shift toward up-regulation of genes involved in effective antigen presentation, phagocytosis, and bacterial killing.
In contrast to traditional knowledge-based approaches to identify molecular mechanisms of MSC-conferred protection from sepsis-induced organ injury, whole genome approaches enable the transcriptional response of various septic organs to MSCs to be interrogated simultaneously independent of a priori knowledge. We previously investigated global transcriptional responses to injury and identified injury-specific expression profiles in comparable organ injury models.
The key biological concept is that an individual “disease phenotype” (organ injury) is composed of the sum of cell- and organ-specific, developmental stage, and metabolism-related changes in gene expression. Genome-wide gene regulatory networks govern this behavior. Theoretical studies of complex networks suggest that these can exhibit ordered (stable) dynamics, raising the possibility that molecular phenotypes of illness may represent high-dimensional attractor states that can be identified by whole genome analysis of expression patterns.
For the most part, genes that share transcription profiles are biologically related, suggesting that the information contained in expression profiles can help identify and inform regarding mechanisms of injury. Such otherwise “unsupervised” analysis strategies also offer an unprecedented opportunity to identify novel molecular targets for therapy.
In this study, we performed a network analysis of common transcriptional responses induced in major target organs affected in sepsis by systemic administration of bone marrow–derived MSCs with the goals of elucidating molecular mechanisms of disease and identifying novel potential targets for therapeutic intervention.
Materials and Methods
Animal experiments have been previously published.
All the studies were approved by the animal care committee at St. Michael's Hospital (Toronto, ON, Canada) in accordance with Canadian Council of Animal Care guidelines. All the studies used 8- to 14-week-old female C57Bl/6J mice (The Jackson Laboratory, Bar Harbor, ME).
MSC Culture
Murine MSCs (isolated from male C57Bl/6J mice, a gift from Dr. Darwin Prockop, Texas A&M Health Science Center, College Station, TX) were expanded in culture according to previously published literature.
Adult stem cells from bone marrow (MSCs) isolated from different strains of inbred mice vary in surface epitopes, rates of proliferation, and differentiation potential.
in which female C57Bl/6J mice (8 to 14 weeks old) were randomized to receive either sham surgery or CLP. Saline or MSCs (2.5 × 105 cells) were injected via the jugular vein 6 hours after surgery. Mice were sacrificed 28 hours after the initial sham or CLP procedure to collect tissue samples for analysis. Total RNA was isolated from whole hearts, kidneys, livers, lungs, and spleens as described elsewhere.
Female mice were anesthetized with 200 mg/kg ketamine (Ketalean, 100 mg/mL; Bimeda-MTC Animal Health Inc., Cambridge, ON, Canada) and 10 mg/kg xylazine hydrochloride (Rompun, 20 mg/mL; Bayer Inc., Toronto, ON, Canada) by i.p. injection. The mice were positioned in dorsal recumbency, and the chest/abdomen surfaces were shaved and prepared with 70% ethanol. A ventral midline incision (approximately 1 cm) was made to allow exteriorization of the cecum. The cecum was ligated 1 cm from the apex with 3–0 silk sutures and penetrated through and through using a 22-gauge needle. The abdominal incision was then closed in two layers with 4–0 nylon sutures. Sham surgery, in which the cecum was exteriorized and manipulated as described but not ligated or punctured, was performed in control animals. Immediately after surgery, the animals received fluid resuscitation with 50 mL/kg saline injected s.c.
MSC Administration
Saline or MSCs were injected 6 hours after the animals had undergone CLP to induce sepsis. Saline or cultured male MSCs (at 2.5 × 105 cells, 100 μL total volume) were slowly infused via a cannula inserted into the jugular vein 6 hours after sham operation or CLP. The mice were randomized to receive saline or MSCs. After infusion, the cannula was withdrawn, the vein was ligated, and the mice were returned to the vivarium and had free access to food and water. The mice were sacrificed 28 hours after the initial sham or CLP procedure to collect tissue samples for analysis. The heart, liver, spleen, lungs, and kidneys were snap frozen and stored at −80°C for microarray analysis.
Microarray Analysis
Total RNA from whole spleens, lungs, livers, hearts, and kidneys (collected 28 hours after CLP) from five animals per group (sham/saline, CLP/saline, and CLP/MSCs) was extracted using TRIzol reagent (Invitrogen, Burlington, ON, Canada) and purified using RNeasy (Qiagen Inc., Chatsworth, CA) per manufacturer specifications. RNA quality was ensured by spectrophotometric analysis (OD260/280) and gel visualization. Animals intended for microarray analysis were preselected on the basis of clear group assignment based on physiologic parameters of inflammation (P < 0.05). A total of 300 ng of mRNA was hybridized to the Illumina MouseWG-6 v2.0 expression bead array (Illumina Inc., San Diego, CA) per manufacturer specifications (n = 5 animals per group). The spleen samples were analyzed first, and part of that analysis has been previously published.
The raw image files from the spleen samples (n = 4 per group) previously profiled were merged with the data from the other four tissues, and batch analysis was performed using Partek Genomics Suite software version 6.5 (Partek Inc., St. Louis, MO) to remove any “batch” effects. Nonnormalized raw intensity values for all 72 profiled tissues were then uploaded to the R Project Bioconductor statistical tools package for normalization.
Microarray data have been deposited in a public functional genomics data repository [Gene Expression Omnibus, http://www.ncbi.nlm.nih.gov/geo (accession number GSE40180)].
Limma Analysis
Differential gene expression was determined by computing empirical Bayes-moderated t-statistics using the limma package (linear models for microarray data).
Briefly, limma starts by fitting a linear model for each spot/gene in the data, and then an empirical Bayes method is used to moderate the SEs for estimating the moderated t-statistics for each spot/gene, which shrinks the SEs toward a common value. This test is similar to a multivariate analysis of variance method for each spot/gene except that the residual SDs are moderated across genes to ensure more stable inference for each gene. The moderated SDs are a compromise between the individual genewise SDs and an overall pooled SD. Two main comparisons were considered: sham versus CLP + placebo and CLP + placebo versus CLP + MSCs. Limma considered the main effects of two factors: treatment and tissue. We adjusted the effect of tissue in the linear models when we explored the differences among treatment groups. Differentially expressed genes were defined as those having a false discovery rate (FDR)–corrected P < 0.05
(adjusted P value). For clustering, log2-transformed normalized expression data were gene centered (mean). Hierarchical clustering was performed using JMP software (http://www.jmp.com/software/jmp6; SAS Institute, Inc., Cary, NC) and Partek software using a correlation matrix and Ward's linkage.
Functional Enrichment and Network Analyses
Functional enrichment, pathway, and network analyses were performed using genes ranked by adjusted P values, fold change, and modified t-statistics from limma analysis (CLP versus MSC comparison). Gene set enrichment analysis (GSEA) was performed using GSEA version 3.0 software.
All 45,282 probes, ranked by limma-derived modified t-statistics, were imported into the GSEA software. The gene sets used are available from the Molecular Signatures Database C2 (curated gene sets) and include gene sets collected from sources such as online pathway databases, publications in PubMed, and the knowledge of domain experts. Chemical and genetic perturbations (CGPs) and canonical pathways (CPs) were analyzed separately. GSEA was used to detect coordinated expression in the three main treatment groups (sham, CLP, and MSC). GSEA was run according to default parameters: probes for the same gene were collapsed into a single gene symbol (identified by its HUGO gene symbol), permutation number = 1000, and permutation type = “gene sets.” By convention, an FDR of <25% was used as the cutoff value for significance.
Ingenuity Pathway Analysis
Functional enrichment analysis was performed by using Ingenuity Pathway Analysis (IPA) version 2.0 software (Ingenuity Systems Inc., Redwood City, CA; online registration required for access). To perform IPA analysis, we inputted all the limma-selected genes (adjusted P < 0.05) differentially regulated in CLP and MSCs in three columns: Illumina probe ID, t-value (fold change), and adjusted P value (FDR). By convention, genes that were up-regulated in animals exposed to CLP after MSC administration (that contribute to enrichment in gene sets up-regulated by MSC treatment) are shown in red and genes that were down-regulated (that contribute to enrichment in gene sets down-regulated by MSC treatment) are shown in green. By default, during IPA analysis, only molecules from the data set associated with the Ingenuity Knowledge Base repository (Ingenuity Systems Inc.) were considered. The significance of the association between the data set and the specific pathways of interest was determined in two ways: ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the Ingenuity Knowledge Base pathway, and Fisher's exact test was used to calculate a P value determining the probability that the association between the genes in the data set and the pathway of interest can be explained by chance alone.
Network Analysis
In biological pathways, many genes tend to be co-expressed; thus, it is natural to incorporate these correlations into a network-based framework. In this framework, pairwise correlations between genes are used to describe the “connectedness” of the network, and clusters of tightly correlated genes (modules) can define correlated pathways. To construct a co-expression network that characterizes the effects of MSC treatment on experimental sepsis, network analysis was performed using Cytoscape (http://www.cytoscape.org; online registration is required for access; last accessed July 21, 2012),
an open source bioinformatics software platform for visualizing molecular interaction networks and biological pathways and for integrating these networks with annotations, gene expression profiles, and other state data. A Cytoscape plugin that enables visualization and interpretation of GSEA results was used to focus the network analysis on limma-ranked/GSEA-identified pathways of interest. Only GSEA-significant CP-related gene sets were imported into Cytoscape. The parameters used for network analysis (similarity coefficients) were a Jaccard and overlap combined threshold of 0.2, P = 0.05, and a combined constant of 0.5. Cytoscape and IPA enrichment mapping were used for visualization of the GSEA results. Nodes represent enriched gene sets, which are grouped and annotated by their similarity according to related gene sets. Enrichment results are mapped as a network of gene sets (nodes). Node size is proportional to the total number of genes in each gene set. The proportion of shared genes between gene sets is represented by the thickness of the line between nodes. This network map was manually curated, removing general and uninformative subnetworks, resulting in a simplified network map.
Results
Effect of MSC Treatment on Global Gene Expression Profiles of Major Target Organs in Experimental Sepsis
A total of 45,282 genes from 72 different tissue samples (lung, liver, kidney, spleen, and heart) collected from four to nine different animals were profiled using Illumina bead arrays (see Supplemental Figure S1A at http://ajp.amjpathol.org). Primary component analysis (PCA) of normalized unfiltered genes revealed considerable differences in apparent gene expression in the five tissues profiled, with the first three primary components (PCs) explaining only 33.8% of the variability in the data (see Supplemental Figure S2A at http://ajp.amjpathol.org). A total of 11,187 probes were significantly different between sham and CLP (sham versus CLP + placebo; limma: 5675 up-regulated and 5512 down-regulated), and 4751 probes (corresponding to 3968 genes) were different between the placebo- and MSC-treated groups (CLP versus MSC; limma: 2305 up-regulated and 2446 down-regulated) (see Supplemental Figure S1B at http://ajp.amjpathol.org). Molecular classification of CLP treatment responses (placebo versus MSC) based on gene expression profiles were determined using PCA. The first three PCs explain 63.4% of the variability in the data (Figure 1A). A biplot of PC determinants revealed a distinct partition between mice exposed to different treatment strategies (Figure 1B). Hierarchical clustering for each individual tissue performed separately showed that MSC treatment returned transcriptional responses to a molecular profile similar to the sham profile for each specific tissue (Figure 1C). All subsequent analysis focused on the 4751 gene probes that were deemed to be differentially expressed by limma for the CLP versus MSC comparison (adjusted P ≤ 0.05).
Figure 1Molecular classification of MSC-mediated treatment of experimental sepsis based on gene expression profiles. A: PCA mapping (63.4%) by tissue. The figure shows the classification of all 72 tissue samples by PCA (in biplot representation). PCA of all genes deemed to be differentially expressed in mice with CLP-induced sepsis treated with placebo versus MSCs (N = 4751). The cutoff point for selection was set at adjusted P < 0.05. Each circle represents a sample (purple, lung; orange, heart; pink, spleen; blue, kidney; brown, liver), with all 72 samples presented on the PCA. The three-dimensional plot shows how individual tissues cluster independently. The percentage value in parenthesis indicates the proportion of the total variance described by each PC: PC1: x axis, PC2: y axis, and PC3: z axis. The first three PCs explain 63.4% of the variance in the data. Ellipsoids represent space 2 SD from the mean of the centroid for each treatment group. B: PCA mapping (63.4%) by treatment. The same PCA plot coded (colored) by treatment strategy (mouse 107 was excluded) reveals a distinct partition between mice exposed to different treatment strategies. Red, CLP treated with placebo; blue, CLP treated with MSCs; green, sham operated. The most discriminating genes have the highest absolute scores on the between-group analysis axes (ie, a large distance from the origin of the plot). The centroid for each treatment group is shown (N = 4751). C: Hierarchical clustering (Ward's method) of genes differentially expressed between CLP treated with either placebo or MSCs (N = 4751). Top panel: A dendogram for all 72 tissues profiled. The bars below the dendogram show the distribution of samples according to tissue and treatment strategies. Bottom panel: Heat map for hierarchical clustering results for each individual tissue performed separately showing that for each specific tissue, MSC treatment restored transcriptional responses to a molecular profile similar to the sham profile. By convention, up-regulated genes are shown in red and down-regulated genes are shown in blue.
Functional Map of the Global Transcriptional Response to MSC Treatment of Experimental Sepsis
To develop a functional map of the global transcriptional response to MSC treatment in septic mice, we used GSEA to test for functional enrichment of gene sets and used the FDR to assess which gene sets were more frequently altered by treatment with MSCs (default parameters: FDR <25%, nominal P < 0.05). To visualize enrichment gene sets, overlap scores were used to organize significant gene sets into a functional enrichment map (or network) using the Cytoscape software platform 2.6.
To better capture the network topology, and not overburden the network, we identified the “seed” gene set for network analysis as differentially enriched CPs at P = 0.05 and did not include curated gene sets for CGP data sets that are manually curated data sets from domain experts. Network analysis highlighted the critical effect of MSC treatment on bioenergy and mitochondria metabolism and on innate immune response (Figure 2). MSC administration resulted in down-regulation of gene sets for transcription and translation grouped as RNA and protein processing (see Supplemental Table S1 at http://ajp.amjpathol.org). Gene sets up-regulated by MSC treatment included a much more eclectic group of genes involved in a variety of related functions, such as cancer, cell-cell interaction, cell growth, and remodeling, and potentially unrelated genes, such as hemostasis and complement activation (Figure 2; see also Supplemental Table S2 at http://ajp.amjpathol.org). Results from IPA analysis are presented in Supplemental Table S3 (available at http://ajp.amjpathol.org) and were used to visualize the putative impact of changes in gene expression on specific CPs altered by MSC treatment of septic mice.
Figure 2Functional map of global changes in gene expression in response to MSC treatment of experimental sepsis. Enrichment results from GSEA were mapped as a network of gene sets (nodes) related by mutual overlap (edges), where red identifies up-regulated and blue down-regulated gene sets. The global network is presented in an organic layout that highlights the separation between functional gene set classes. Dashed circles indicate the class of gene set and solid circles the group of functionally related gene set classes: blue, innate immune response; pink, RNA and protein processing; red, mitochondria and bioenergy metabolism; green, cell-cell interaction and remodeling; purple, hemostasis and complement; and orange, growth factor and cancer. Node size is proportional to the total number of genes in each set, and edge thickness represents the number of overlapping genes between the sets. The FDR (q-value) is represented as the node color gradient.
MSC-Conferred Protection Against Sepsis-Induced Organ Injury Is Characterized by Transcriptional Reconstitution of Mitochondrial and Bioenergy-Related Pathways
Using the selection criteria described, 149 gene sets were shown to be negatively correlated (down-regulated; see Supplemental Table S1 at http://ajp.amjpathol.org) and 69 positively correlated (up-regulated; see Supplemental Table S2 at http://ajp.amjpathol.org) with MSC treatment (C2: CP collection online pathway databases, C2.cp.v3.0). Manually curated data sets from the biomedical literature (C2: CGP collection C2.cgp.v3.0) were analyzed separately. A total of 301 CGP gene sets were up-regulated and 80 down-regulated after MSC treatment. The top GSEA-identified gene sets encoded for mitochondrial and bioenergy-related genes, including oxidative phosphorylation (Mootha_VoxPhos and Mootha_FFA_oxydation), mammalian target of rapamycin (mTOR), peroxisome proliferator-activated receptor gamma coactivator 1-α (PGC-1alpha), and gluconeogenesis signaling (Table 1). Two of the top gene sets regulated by MSC treatment included estrogen receptor target genes (Stein_ESRRA_targets_up and Stein_ESRRA_targets). A heat map showing changes in specific genes in different tissues is shown for the top three enriched gene sets (FDR, 0.00; Figure 3A). MSC treatment also resulted in enrichment for CP gene sets associated with known pathologic mitochondrial dysfunction conditions (Parkinson's disease, Alzheimer's disease, and Huntington's disease) (Figure 3B). In addition, MSC treatment up-regulated gene sets involved in the detoxification and management of oxidative stress (Nrf2 pathway, glutathione, and cytochrome p450 metabolism) (see Supplemental Table S2 at http://ajp.amjpathol.org). In parallel, the putative impact of changes in gene expression on specific pathways was visualized using GSEA (Figure 3, A and B) and IPA (Figure 3C). These data indicate that reconstitution of transcriptional pathways involved in preserving bioenergy status and limiting oxidative stress play a fundamental role in mediating the beneficial effects of MSC therapy in polymicrobial sepsis (genes that are up-regulated after treatment with MSCs are colored red, and those that are down-regulated are colored green) (Figure 3C).
Table 1Top Curated Gene Sets Enriched (Up-Regulated) after MSC Treatment of Septic Mice
Figure 3MSC-conferred protection to sepsis-induced organ injury is characterized by transcriptional reconstitution of mitochondrial and bioenergy-related pathways. A: Heat map for the expression values for genes contained in the top three mitochondrial-related gene sets (n = 72 chips) significantly enriched (up-regulated) in GSEA by treatment with MSCs (Table 1). By convention, expression values are represented as a color spectrum going from red (up-regulated) to blue (down-regulated) depending on the correlation with each specific phenotype. Only top genes that contributed to the enrichment are included. FDRs for specific curated CGP pathways are shown on the left [Mootha Human Mitodb_6_2002 (FDR, 0.00%), Mootha Mitochondria (FDR, 0.00%), and Wong Mito Gene Module (FDR, 0.00%)]. Full gene sets can be downloaded from the Molecular Signatures Database. B: Expanded enrichment map (organic layout) showing the relationship between gene sets related to mitochondrial dysfunction. The gene symbols are described in Supplemental Table S2 (available at http://ajp.amjpathol.org). C: Differential expression genes involved in mitochondrial function. Genes contributing to mitochondria-related functional gene set enrichment were inputted into the IPA software. Changes in the transcriptional profile in MSC-treated mice are shown. The IPA library of CPs identified mitochondrial dysfunction as a top pathway altered in all 72 tissues after treatment with MSCs. Molecules from the data set (gene list presented in Supplemental Table S3 at http://ajp.amjpathol.org) that were associated using Ingenuity Knowledge Base were considered for the analysis. The significance of the association between the data set and the CP was determined in two ways: ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the CP (ratio, 0.52) (top), and Fisher's exact test was used to calculate a P value determining the probability that the association between the genes in the data set and the CP can be explained by chance alone (6.6 × 105) (bottom). By convention, up-regulated genes are shown in red and down-regulated genes in green. The legend for the IPA diagram is presented in Supplemental Figure S4 (available at http://ajp.amjpathol.org).
Effect of MSC Therapy on Inflammation and Immune Transcriptional Responses
Network analysis identified gene sets previously known to be down-regulated by MSC treatment in sepsis, such as those involved in innate immune response and inflammation (Figure 2): IL-6 and IL-10 responses (see Supplemental Tables S1 and S3 at http://ajp.amjpathol.org). Novel observations included down-regulation of genes involved in NOD (nucleotide-binding oligomerization domain) and Toll-like receptor (TLR) pathways that have important roles in innate immunity as sensors of microbial components derived from bacteria, viruses, and fungi (Figure 4A). Treatment with MSCs decreased expression of critical genes in immune response to lipopolysaccharide and other microbial products, such as CD14, IL-1β, MyD88, Tirap, Traf3 and Traf6, Tollip, and genes involved in NF-κB activity, such as IκB, p65, and Rel A (Figure 4B). Down-regulation of genes involved in the host response to viruses such as HIV and influenza were also significantly down-regulated. Of specific interest is the effect of MSC treatment on down-regulation of the IL-17 pathway (see Supplemental Figure S3 and Supplemental Table S3 at http://ajp.amjpathol.org).
Figure 4Effect of MSC treatment on innate immune gene transcription in experimental sepsis. A: Network analysis of MSC-conferred down-regulation of immune-regulated pathways: Left panel: Expanded enrichment map (circle layout) showing the relationship between gene sets down-regulated by MSC treatment involved in immune regulation. Enrichment results from GSEA were mapped as a network of gene sets (nodes) related by mutual overlap (edges), where blue denotes down-regulated gene sets. Node size is proportional to the total number of genes in each set, and edge thickness represents the number of overlapping genes between the sets. The FDR (q value) is represented as the node color gradient. Right panel: IPA analysis of MSC-dependent down-regulation of TLR pathway–related genes. The significance of the association between the data set and the CP was determined in two ways: ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the CP (ratio, 0.36), and the Fisher's exact test was used to calculate a P value determining the probability that the association between the genes in the data set and the CP can be explained by chance alone (2.39 × 10−4). By convention, up-regulated genes are shown in red and down-regulated genes in green. The legend for the IPA diagram is presented in Supplemental Figure S4 (available at http://ajp.amjpathol.org). The gene symbols are described in Supplemental Figure S4. B: Changes in TLR signaling pathways after treatment with MSCs. Left panel: IPA analysis shows down-regulation of critical genes involved in the response to lipopolysaccharide (LPS). Right panel: A GSEA enrichment plot for TLR signaling pathways (GSEA curated pathways; see Supplemental Table S1 at http://ajp.amjpathol.org). The top part of the plot shows progression of the running enrichment score (ES) and the maximum peak therein for gene sets enriched after MSC treatment. The middle part shows the genes in each gene set as “hits” (vertical black lines) against the ranked list of genes. The bottom part of the plot shows the value of the ranking metric as you move down the list of ranked genes. The ranking metric measures a gene's correlation with the phenotype. A positive value indicates correlation with the phenotype profile (placebo), and a negative value indicates no correlation or inverse correlation with the profile (MSC). The bottom part shows the corresponding heat map for the expression values for genes contained in the gene set (n = 72 chips) that are significantly differentially expressed between placebo and MSC treatment. By convention, expression values are represented as a color spectrum going from red (up-regulated) to blue (down-regulated) depending on the correlation with each specific phenotype. Only the top genes that contributed to the enrichment are included. Gene symbols are described in Supplemental Table S4 (available at http://ajp.amjpathol.org). Dendograms are shown highlighting how MSC treatment results in a transcriptional profile that is more similar to that of sham-treated mice than septic mice treated with placebo. In the far right, a tree graph shows fold change in gene expression after treatment with MSCs. Red represents CLP treated with placebo; blue, CLP treated with MSCs; and green, sham-operated mice.
Inflammation and innate immune response gene sets up-regulated after MSC treatment included pathways specifically involved in opsonization and clearance of bacteria, including activation of complement (FDR, 0.02), activation of coagulation (FDR, 0.02), activation of bioactive peptides (FDR, 0.04), endocytosis (FDR, 0.06), NFAT signaling (FDR, 0.017), and NOTCH signaling (FDR, 0.22). MSC treatment resulted in up-regulation of genes that differentiate M1 (proinflammatory) versus M2 (anti-inflammatory) macrophage subtypes (M1 COATES_MACROPHAGE_M1_VS_M2_UP; FDR, 0.013).
Together, the present data suggest that MSC-conferred protection to polymicrobial sepsis–induced organ injury is underscored by a profound programmatic change in the transcription of inflammation- and immune-related genes in all five of the major organs at risk for injury during severe sepsis.
MSC-Conferred Protection to Sepsis-Induced Organ Injury Is Characterized by Transcriptional Regulation of Gene Sets Involved in Control of Cellular Interaction and Remodeling
Network analysis highlighted the role of MSC treatment in up-regulating gene sets implicated in cell-cell signaling and cell-cell interaction. MSC treatment increased expression of genes involved in tightening gap junctions, calcium signaling, and focal adhesion (Figure 5A). MSC treatment was associated with changes in transcriptional regulation of genes previously implicated in cardiac function, such as those involved in muscle contraction, vascular smooth muscle contraction, and dilated and hypertrophic cardiomyopathy. Of specific interest, GSEA and IPA revealed up-regulation of the Tie-2 pathway, a critical pathway involved in the regulation of inflammation and vascular quiescence. IPA analysis showed that MSC administration was associated with up-regulation of angiopoietin-1, an important regulator of vascular function that is known to promote endothelial stabilization and quiescence, and its cognate receptor Tie-2 (Figure 5B). In keeping with an increase in expression of genes that may be involved in regulating membrane function and permeability, MSC treatment resulted in up-regulation of the CP for vasopressin-regulated water reabsorption.
Figure 5Effect of MSC treatment on cell behavior and interaction in experimental sepsis. A: Expanded enrichment map showing interphase between gene sets involved in remodeling, cell-cell interaction, innate immune response, and cell growth. Expanded enrichment map (organic layout) showing the relationship between gene sets related to immune regulation (blue nodes), complement and coagulopathy (red circles), remodeling (blue circles), and cell growth and migration (purple circles). Enrichment results from GSEA were mapped as a network of gene sets (nodes) related by mutual overlap (edges), where blue denotes down-regulated gene sets and red up-regulated gene sets. Node size is proportional to the total number of genes in each set, and edge thickness represents the number of overlapping genes between the sets. B: IPA analysis revealed up-regulation of the Tie-2 pathway, a critical pathway involved in the regulation of inflammation and vascular quiescence. Specifically, MSC treatment was associated with up-regulation of angiopoietin-1, which is known to stabilize the vascular endothelium via binding to its cognate receptor Tie-2. Other downstream transcriptional events are consistent with down-regulated NF-κB after MSC treatment of sepsis.
In this study, we present the first genome-wide analysis of functional transcriptional regulation in response to MSC treatment of sepsis in five major target organs (lungs, kidneys, liver, heart, and spleen). The present findings demonstrate that 25% of the probes interrogated were differentially expressed 28 hours after the induction of polymicrobial sepsis by CLP. This number may be large for a single cellular process, but it represents the cumulative response from five different tissues to complex sepsis-induced derangements that affect multiple biological processes. Of the differentially expressed probes, 42% were subsequently regulated by MSC administration in vivo, which has been demonstrated in this model to confer protection and reduce mortality from sepsis.
Given 30,000 as the estimated total number of genes in the mouse genome, the present data suggest that ≥13% of the murine genome is transcriptionally reprogrammed during MSC-induced protection from sepsis. Taken together, these data challenge the paradigm that a single, specific MSC-derived paracrine mediator is responsible for the global pleotropic effect of MSCs on transcriptional and network reprogramming in sepsis. A much more plausible explanation is that MSC-conferred protection from sepsis-related complications involves a range of complementary activities, resulting in mitigation of the innate and acquired immune and inflammatory responses while also affecting complex networks of host cell-cell, cell-matrix, metabolism, and bioenergy substrate utilization and functional pathways.
We hypothesized that the transcriptional effects of MSC administration could be made up of many small cumulative changes in gene expression. By combining GSEA expression profiles of gene sets (rather than genes) with cellular network information, we were able to document striking similarities in MSC-induced transcriptional changes in various critical pathways involved in sepsis, such as mitochondrial dysfunction, in major target organs of sepsis, including the liver, spleen, kidneys, heart, and (less pronounced in) lungs (Figure 3). The present study demonstrated that whereas CLP resulted in down-regulation of mitochondrial and bioenergy-related pathways, MSC treatment returned the transcriptional profile toward that seen in sham-operated animals. In this context, MSC administration in sepsis could possibly restore mitochondrial function to meet metabolic energy demands, thereby enabling cells to fulfill critical roles, including calcium homeostasis, maintenance of the cellular redox state, and cell signaling (Figure 5A).
Innate immune detection of danger signals and microbial motifs is achieved by distinct families of pattern-recognition molecules, including membrane-anchored TLRs,
Consistent with previous results, MSC treatment down-regulated critical innate immune response gene sets and pathways, such as TLR, NF-κB, IL-6, and IL-10 signaling (Figure 4).
The observation that MSC administration regulates critical pathogen recognition pathways in five different organs simultaneously reflects the ability of these cells to coordinate not only cell- or organ-specific responses but also global responses to sepsis (Figure 4A). Speculating on the role of MSCs, the present data suggest that a key effect might have been in reprogramming of regulatory immune cells and their pattern-recognition receptors. This concept is, in part, supported by novel observations, including down-regulation of genes involved in NOD-mediated signaling and IL-17 (TH17) functional responses.
The NOD proteins have important roles in innate immunity as intracellular sensors of pathogen-associated molecular patterns and damage-associated molecular patterns.
The biological importance of these molecules is underscored by the fact that mutations in the genes that encode these proteins have been associated with complex inflammatory disorders.
IL-17 is the founding member of a group of cytokines called the IL-17 family. Produced by a recently identified subtype of T-helper cells, IL-17 acts as a potent cytokine that augments the production of various chemokines and cytokines from a variety of cell types that attract monocytes and macrophages to sites of inflammation.
MSC administration also resulted in up-regulation of NFAT cell–related genes. These are transcription factors that are expressed in most immune cells and that play a pivotal role in the transcription of cytokine genes and other genes critical for the immune response.
The activity of NFAT proteins is tightly regulated by the calcium/calmodulin-dependent phosphatase calcineurin. Calcineurin controls the translocation of the Rel- family of proteins (a component of NF-κB). MSC treatment results in up-regulation of NFAT, calcium, and calcineurin gene sets (Figure 5) but down-regulation of RelA- and NF-κB–related pathways (Figure 4).
Administration of MSCs in the present study resulted in up-regulation of gene sets encoding for various genes involved in coagulation, complement regulation, and platelet activation. The implication of this finding needs to be investigated in functional biological studies. MSC treatment of septic mice also up-regulated neurite and axon outgrowth pathways previously unknown to play a role in sepsis-induced organ failure, such as those regulating semaphorin interactions, neural adhesion protein 1 (NCAM1), and NCAM signaling for neurite outgrowth (Figure 5A).
MSC treatment also caused down-regulation of neurotrophin and up-regulation of the receptor for neurotrophin (TrkA) signaling, suggesting downstream signaling from different neurotrophin pathways. Consistent with a profound change in transcriptional programming, MSC administration results in up-regulation of histone deacetylase–related pathways (Figure 5A). Sepsis induces epigenetic changes in macrophages, which results in transcriptional reprogramming, partially explaining the development of endotoxin tolerance
Of particular relevance for sepsis therapeutics is the effect of MSC administration on critical pathways known to be involved in the preservation of endothelial/vascular integrity.
Increased permeability of the endothelial monolayer is largely explained by gaps between cells (paracellular leak) or by transit through individual cells (transcellular leak).
In the present study, MSC treatment increased expression of the gap and adherens junction, as well as Tie-2–related signaling gene sets, strongly suggesting that one possible mechanism by which MSCs may be exerting their beneficial effects is by directly altering the expression of specific genes involved in preserving cell-cell interaction and endothelial quiescence.
In this study, we integrated expression and network data. The advantage of this approach is that the biologically relevant signals identified through these two approaches are more likely to be correct and reproducible.
The main limitation of inferring biological activity is the use per se of databases that are manually curated from the literature and, thus, dependent on data mining techniques and intrinsic biases of the literature. GSEA is unique in this respect because of its comprehensive inclusion of various database resources. Although not always accurate, such analyses have proved useful in aggregate.
Moreover, because we were unable to examine the structure and function of organs in animals that died, this study is biased toward survivors (although most of the mortality occurs at a later time point, ie, 48 hours). A time course experiment would enable critical understanding of molecular changes associated with increased mortality. Also, given our experimental design, it is unclear whether a single organ may be spearheading and coordinating MSC-dependent changes in gene transcription in distal organs. This would be a clinically relevant question since organ-specific MSC delivery may prove to be as effective as systemic MSC administration in polymicrobial sepsis.
An important point to highlight is the analysis approach undertaken in this study. The focus was on gene sets that were commonly and simultaneously regulated by MSC treatment in the five organs examined. We accounted for differences in tissue-specific responses by performing moderated t-statistics in limma. The initial approach was to capitalize on this global analysis strategy to identify common genes/pathways and expression networks altered in sepsis and in response to MSCs to inform regarding the broad biological effects of MSCs at the organism level. Currently, we are performing separate, detailed, organ-specific analyses to determine unique organ-specific transcriptional network responses mediated by MSC treatment of sepsis.
Conclusion
Transcriptomic analysis indicates that the protective effect of MSC treatment of sepsis is not limited to a single mediator or pathway but involves a coordinated expression of transcriptional programs that serve as the blueprint for MSC-conferred protection from sepsis at the cellular level in multiple organs. Networks of gene sets modulated by MSCs coordinate complementary activities, affecting host cell metabolic and inflammatory responses. The ability of MSCs to widely affect multiple common large-scale transcriptional profiles of critically important biological networks in a variety of major target organs underscores the rationale for and potential utility of developing cell-based technologies for the treatment of clinical sepsis.
Experimental design and analysis strategy. A: Schematic of the experimental design. Five different tissues were collected (lung, heart, liver, kidney, and spleen) from mice randomized to receive either sham surgery or polymicrobial sepsis (induced by CLP) treated with placebo (equal volume saline) or MSC administration (6 hours after induction of sepsis) 28 hours after sham surgery or CLP. Total RNA from 72 samples was hybridized to Illumina bead arrays (MouseWG-6 v2.0). B: Limma analysis. Changes in the expression of 42,282 probes were determined in 72 samples (>3 million data points) using limma for differential expression analysis of treatment by adjusting for the effects of tissue. A total of 11,187 genes were significantly changed between sham and CLP treatment (limma: 5675 up-regulated and 5512 down-regulated), and 4,751 genes were changed between the placebo- and MSC-treated groups (limma: 2305 up-regulated and 2446 down-regulated). A total of 4111 genes were significantly changed by CLP and MSC treatment.
Transcriptional profile of all 72 tissue samples. A: PCA of all genes profiled by Illumina MouseWG-6 v2.0. PCA was performed using a correlation matrix for normalized Eigenvalues are plotted. Each circle represents a sample, with all 72 samples (colored by tissue) presented on the PCA (purple = lung, orange = heart, pink = spleen, light blue = kidney, and brown = liver). The percentages in parenthesis indicate the proportion of the total variance described by each PC. PC1: x axis; PC2: y axis; and PC3: z axis. The first three PCs explain 33.8% of the variance in the data. Ellipsoids represent space 2 SD from the mean of the centroid for each treatment group (N = 42,282). B: PCA analysis of all 42,282 Illumina probes colored by treatment (red = CLP treated with placebo, blue = CLP treated with MSC, and green = sham operated). The three-dimensional plot shows how individual tissues cluster independently. For each individual tissue, samples cluster according to treatment strategy. C: Dissimilarity matrix of all Illumina probes for 72 tissues profiled (N = 42,282). Pairwise comparisons of sample interpoint distances are displayed as an intensity map. Similar samples are colored blue, and dissimilar samples are colored red. Samples are plotted in the same order along the top and the side: red = CLP treated with placebo, blue = CLP treated with MSCs, and green = sham operated. Arrows identify samples from the same animal (mouse 107) treated with CLP, which shows a divergent clustering behavior. D: PCA mapping of all genes selected by limma (N = 12,075) underscores the importance of individual tissue transcriptional phenotypes and the similarities between tissue expression profiles in response to different treatments. The first three PCs explain 72.9% of the variability in the data, and although tissues cluster separately, individual samples cluster according to treatment strategy.
IPA analysis of MSC-dependent up-regulation of IL-17 pathway–related genes: transcriptional profile in placebo-treated mice. The significance of the association between the data set and the CP was determined in two ways: ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the CP (ratio, 0.31), and Fisher's exact test to calculate a P value determining the probability that the association between the genes in the data set and the CP can be explained by chance alone (9.7E-4). By convention, up-regulated genes are shown in red and down-regulated genes in green. The legend for the IPA diagram is presented in Supplemental Figure S4.
Bone marrow stromal cells attenuate sepsis via prostaglandin E(2)-dependent reprogramming of host macrophages to increase their interleukin-10 production.
Adult stem cells from bone marrow (MSCs) isolated from different strains of inbred mice vary in surface epitopes, rates of proliferation, and differentiation potential.
Supported by the Canadian Institutes of Health Research (grants MOP-106545 to C.C.D.S. and MOP-74752 to D.J.S.), the Ontario Thoracic Society (OTS2010/2011/2012 to C.C.D.S.), the Physicians Services Incorporate (PSI 09-21 to C.C.D.S.), Northern Therapeutics, Natural Sciences and Engineering Research Council of Canada Doctoral Canada Graduate Scholarship, Ontario Graduate Scholarship (S.H.J.M.), the Weston Foundation (J.J.H.), the McLaughlin Center for Molecular Medicine, and a Canada Research Chair in Infectious Diseases and Inflammation (W.C.L.).
00606-2&id=gr1.jpg){kind=link}
00606-2&id=gr2.jpg){kind=link}
00606-2&id=gr3.jpg){kind=link}
00606-2&id=gr4.jpg){kind=link}
00606-2&id=gr5.jpg){kind=link}