ABSTRACT
It is unknown why only some individuals are susceptible to acute rheumatic fever (ARF). We investigated whether there are differences in the immune response, detectable by gene expression, between individuals who are susceptible to ARF and those who are not. Peripheral blood mononuclear cells (PBMCs) from 15 ARF-susceptible and 10 nonsusceptible (control) adults were stimulated with rheumatogenic (Rh+) group A streptococci (GAS) or nonrheumatogenic (Rh−) GAS. RNA from stimulated PBMCs from each subject was cohybridized with RNA from unstimulated PBMCs on oligonucleotide arrays to compare gene expression. Thirty-four genes were significantly differentially expressed between ARF-susceptible and control groups after stimulation with Rh+ GAS. A total of 982 genes were differentially expressed between Rh+ GAS- and Rh− GAS-stimulated samples from ARF-susceptible individuals. Thirteen genes were differentially expressed in the same direction (predominantly decreased) between the two study groups and between the two stimulation conditions, giving a strong indication of their involvement. Seven of these were immune response genes involved in cytotoxicity, chemotaxis, and apoptosis. There was variability in the degree of expression change between individuals. The high proportion of differentially expressed apoptotic and immune response genes supports the current model of autoimmune and cytokine dysregulation in ARF. This study also raises the possibility that a “failed” immune response, involving decreased expression of cytotoxic and apoptotic genes, contributes to the immunopathogenesis of ARF.
INTRODUCTION
Acute rheumatic fever (ARF) is a major cause of heart disease and premature death in developing countries and in indigenous populations in industrialized countries, including the Aboriginal population of Australia (1). ARF occurs as a result of pharyngeal infection with a rheumatogenic strain of Streptococcus pyogenes (group A streptococci, or GAS) in a genetically susceptible individual. This induces an abnormal autoimmune host response, thought to be mediated at least in part through molecular mimicry (2).
Epidemiological data show that only some individuals are susceptible to ARF (1, 3). The reason for this is unknown, but numerous studies have identified individual associations between genetic polymorphisms (including HLA and cytokine polymorphisms) and ARF/rheumatic heart disease (RHD) (4). However, the results are inconsistent between studies and populations, so the role of genetic susceptibility underlying the abnormal immune response remains poorly defined. It is likely that a number of genes are involved in the pathogenesis of ARF, and the disease therefore lends itself to a global genomic approach (5).
We hypothesized that there are differences in the immune response between individuals who are susceptible to ARF and those who are not that could be detected by investigating gene expression. Defining the immunological events that occur in a susceptible individual infected with rheumatogenic GAS that do not occur in a nonsusceptible individual is critical to understanding the pathogenesis of ARF.
To determine genes that might be involved in susceptibility to and pathogenesis of ARF, we compared global gene expression in vitro between ARF-susceptible and -nonsusceptible (control) individuals in response to stimulation with rheumatogenic (Rh+) GAS and nonrheumatogenic (Rh−) GAS strains.
MATERIALS AND METHODS
Participants.Adults were enrolled from the indigenous population in the Top End of the Northern Territory, Australia, and gave written informed consent. They were assigned to either the susceptible group on the basis of having rheumatic heart disease (RHD) or to the nonsusceptible control group on the basis that as an adult over the age of 30 years in a population universally exposed to GAS, failure to develop ARF or RHD equates to nonsusceptibility. Group assignment was further verified using B cell antigen D8/17 expression as a marker for susceptibility to ARF (6). There were 15 individuals in the RHD group and 10 in the control group, and using the criteria from the D8/17 study, all of individuals in the RHD group had positive D8/17 expression, and all of those in the control group were negative, confirming the clinically based assignment.
Bacterial preparation and peripheral blood mononuclear cell (PBMC) separation and stimulation.The GAS strains used for stimulation in this study were provided by Bart Currie of the Menzies School of Health Research in Darwin, Australia, and were all isolated from clinical samples from local indigenous patients. Three strains of Rh+ GAS were obtained from patients with ARF, and one strain of Rh− GAS was obtained from a patient without ARF. Three Rh+ GAS strains were used because of the difficulty in being certain that the particular strain caused ARF in a particular patient in this region of endemicity and to increase the likelihood of using a strain that was rheumatogenic in this population. The three strains were NS1713 (M type 53 [M53]), NS702 (M65), and NS52 (M nontypeable). Only one Rh− GAS strain was used as it is a strain associated with acute poststreptococcal glomerulonephritis and not ARF (NS819; M49). The Rh+ GAS and Rh− GAS strains were grown at 37°C and formalin killed. The three Rh+ strains were added in equal concentrations to make a final concentration of 108 CFU/ml. The Rh− strain was also diluted to 108 CFU/ml.
Each subject had 9 ml of blood collected into tubes containing endotoxin-free lithium heparin (Becton, Dickinson, Franklin Lakes, NJ, USA), and peripheral blood mononuclear cells (PBMCs) were separated by Ficoll-Hypaque gradient (Amersham Biosciences, Uppsala, Sweden). Aliquots of 1 × 106 PBMCs were simultaneously stimulated with 100 μl of Rh+ GAS or 100 μl of Rh− GAS. Samples were incubated at 37°C with 5% CO2 for 0, 3, and 24 h. The final concentration of the bacteria was 107 CFU/ml. The concentrations and time points that were optimal for detection of gene expression were determined in preliminary experiments. At each time point, after centrifugation, TRIzol (Invitrogen Life Technologies, Invitrogen Corporation, Carlsbad, CA, USA) was added to the sample before storage at −80°C.
RNA isolation, purification, and amplification.RNA from all samples was extracted using the chloroform-phenol method within 1 month after stimulation and before further storage at −80°C. Samples were then purified using an RNeasy 234 kit (Qiagen Pty., Ltd., Clifton Hill, Victoria, Australia) and amplified using a MessageAmp 234 II aRNA Kit (Ambion Inc., Austin, TX, USA) according to the manufacturers' protocols. All samples were analyzed postamplification using an Agilent 2100 Bioanalyzer (Agilent Technologies, Forest Hill, Victoria, Australia). Since the samples from one individual in the control group did not amplify, that individual was excluded.
Microarray hybridization.The study used 78 microarray slides printed with the Compugen human 19,000 oligonucleotide library (Compugen, South San Francisco, CA) with a selection of control probes at the Adelaide Microarray Facility (Adelaide, Australia). Amplified RNA was labeled by direct platinum-based labeling using a Universal Linkage System (ULS) kit (Kreatech Biotechnology, Amsterdam, The Netherlands) according to the manufacturer's protocol. RNA from each stimulation time point was competitively hybridized with RNA from 0 h from the same subject on two-color microarrays. For each pair of RNA samples to be hybridized to a slide, 2 μg of one sample was labeled with ULS-Cy3, and 2 μg of the other was labeled with ULS-Cy5. Slides were incubated overnight, washed, and scanned using Genepix Pro, version 6.0.
Microarray data normalization and analysis.Each scanned TIF image was quantified to obtain foreground and background intensity values for each spot. Genepix was configured to generate the custom morphological opening background estimator. Normalization and differential expression analysis were undertaken using the limma software package (7) for the R programming environment (http://www.r-project.org). Microarray data quality was checked using diagnostic image plots and MA plots (where M is log intensity ratio and A is average log intensity). Log ratios were print-tip loess normalized (8).
A linear model approach was used to analyze all the microarrays together, including a function for array weights, whereby slides were given weight in the analysis according to their quality and closeness to the linear model fit (9). In addition, a blocking algorithm with subject as the fixed effect was used to increase information from different stimulations from the same subject (10). Statistical significance was tested between the three time points for each stimulation using empirical Bayes moderated t tests, which borrow information between genes and give reliable inference even with small sample sizes (11). The P values were adjusted for multiple testing, across all genes and all comparisons, using the method of Benjamini and Hochberg (12) to control the expected false-discovery rate (q value) at less than 5%. For each gene and each comparison, the comparison was considered to be statistically significant if the q value was less than 5% and the fold change was greater than 50%. GoStat was used to investigate functional gene ontologies (13). For the investigation of gene networks, genes which demonstrated differential expression between the groups of 50% but which failed to meet the significance cutoff were used with Ingenuity web-based software (Ingenuity Systems, Inc., Redwood City, CA) (14).
This study received approval from the Human Research Ethics Committee at the Menzies School of Health Research (HREC/0026) and the Aboriginal Ethics Research Committee, and participants gave informed consent. The gene expression data are MIAME (minimum information about a microarray experiment) compliant.
RESULTS
Differences between ARF-susceptible and control (nonsusceptible) individuals.The primary question in this study was whether in vitro gene expression responses to Rh+ GAS differ between individuals who are susceptible to ARF and those who are nonsusceptible. For each individual in the ARF and control groups, RNA from in vitro stimulation of PBMCs with Rh+ GAS was competitively hybridized on the same two-color microarray against unstimulated RNA (0 h) from the same individual. This was done after both 3 h and 24 h of simulation with Rh+ GAS, resulting in two microarrays for each individual. Comparisons of gene expression were then made between the ARF and control groups (Fig. 1). The use of competitive hybridization with unstimulated RNA had the effect that gene expression at 0 h was subtracted from gene expression at 3 h to identify changes due to Rh+ GAS stimulation. The gene expression responses to Rh+ GAS were then compared between the ARF and control group individuals. Changes in gene expression in the control group were subtracted from changes in gene expression in the ARF group to identify differences between the two groups after 3 h of stimulation (analysis 1). The same was done for 24 h of stimulation with Rh+ GAS (analysis 2).
Experimental design. PBMCs from 15 subjects in the ARF group and from 10 subjects in the control group were stimulated with Rh+ GAS and Rh− GAS for 0, 3, and 24 h, and samples were amplified. Amplified samples from 3 and 24 h were hybridized with samples from 0 h to two-color microarray slides. Comparisons were made between the subject groups (analyses 1 and 2) and between the types of bacteria (analyses 3 and 4) (see Materials and Methods).
Significant gene expression differences were detected between the ARF and control groups at both time points (Fig. 2). After 3 h, the expression of seven genes was significantly different between the ARF group and the control group: five genes had increased expression, and two genes had decreased expression (Table 1). After 24 h, 27 genes were significantly differentially expressed between the ARF group and the control group: 7 genes were more highly expressed, and 20 genes had decreased expression, that is, had greater expression in the control group.
Volcano plots of differences in gene expression following in vitro stimulation with Rh+ GAS between the ARF and control groups for 3 h (analysis 1) and 24 h (analysis 2). Each point represents a gene. Genes with positive values are those that were more highly expressed in the ARF group, and genes with negative values are those more highly expressed in the control group. Log fold change, log2 change (M value); log odds, statistical significance (higher indicates more significant).
Details of genes significantly differentially expressed after 3 and 24 h in response to Rh+ GAS between the ARF group and the control group
Among the genes that responded differentially in response to in vitro stimulation with Rh+ GAS between the two groups, there were 12 that are known to be involved in the immune response (Table 1). These were predominantly less expressed in the ARF group after 24 h: gamma interferon (IFN-γ), interleukin-10 (IL-10), granzyme A (GZMA), granzyme B (GZMB), granulysin (GNLY), perforin 1 (PRF1), signaling lymphocytic activation molecule family member 1 (SLAMF1), inducible T cell costimulator (ICOS), Jun dimerization protein p21SNFT (SNFT), and B cell CLL/lymphoma 2 (BCL2). There was also one gene involved in the immune response that had greater expression in the ARF group at 24 h: TYRO protein tyrosine kinase binding protein (TYROBP). The only gene involved in the immune response that was different between the two groups at 3 h was the neutrophil elastase gene (ELA2), which was more highly expressed in the ARF group than in the control group.
The interaction between genes that differed in expression levels between the two groups was further investigated using network analysis of previously published gene interactions (14–16). Genes whose expression was different between the ARF and control groups were found to interact within several networks at both 3 and 24 h. Within a network there were interactions between the genes that were both more and less highly expressed. The network with the highest number of differentially expressed genes that are known to interact with each other is shown in Fig. 3. Every gene in this network was differentially expressed between the two subject groups in the study, with the majority having decreased expression in the ARF group. The genes in this network are involved in multiple signaling pathways in the immune response, including cytokine signaling and signaling to effect leukocyte extravasation. Specifically, there are 12 genes in this network involved in the activation of T lymphocytes, in particular, Th1 lymphocytes (IFN-γ, IL-2, and ACAM1), and 10 involved in T lymphocyte proliferation. There are four genes involved the activation of B lymphocytes (IFN-γ, IL-2, IL-10, and CD86), seven involved in B lymphocyte proliferation, and four relating to antibody responses (IFN-γ, CD83, ICOS, and IL-10). There are also several genes whose functions include the chemotaxis, recruitment, and infiltration of B and T lymphocytes and natural killer cells. Activation of monocytes involves eight genes in this network, with five genes affecting monocyte chemotaxis (IL-10, CXCL3, CCL3, CCL4, and CCL8).
Interactions between genes that were differentially expressed after 24 h using a threshold difference in expression of 50%. Genes that were increased in the ARF group compared to the control group are shown in red, and genes that were decreased are shown in green, with the intensity of color reflecting the magnitude of differential expression.
Differences in response to rheumatogenic and nonrheumatogenic bacteria.In addition to comparing gene expression in response to Rh+ GAS between the ARF and control groups, we also compared gene expression following in vitro stimulation with Rh+ GAS and Rh− GAS strains. The analysis comparing the responses to Rh+ GAS and Rh− GAS was analogous to the analysis in the first part of the study comparing ARF and control groups. For each individual in the ARF group, RNA from in vitro stimulation of PBMCs with Rh+ GAS and with R− GAS strains for 3 h and 24 h was hybridized against unstimulated RNA (0 h). Comparisons of gene expression levels were then made between responses to Rh+ GAS and Rh− GAS after stimulation for 3 h (analysis 3) and 24 h (analysis 4).
Significantly different gene expression responses were detected between responses to Rh+ GAS and Rh− GAS at both time points (Fig. 4). The scale of the differential response in this comparison was much larger than in the comparison between the ARF and control groups, indicating that there were more genes in this experiment for which the difference in expression levels was statistically significant. After 3 h, the expression of 343 genes was significantly different in response to Rh+ GAS compared to Rh− GAS: 142 genes were more highly expressed, and 201 genes were less expressed. After 24 h, 331 genes were significantly more highly expressed after stimulation with Rh+ GAS, and 308 genes were less expressed.
Volcano plots of differential gene expression in the ARF group following in vitro stimulation with Rh+ GAS and Rh− GAS for 3 h (analysis 3) and 24 h (analysis 4). Genes with positive values are those that were more highly expressed with Rh+ GAS stimulation, and genes with negative values are those more highly expressed with Rh− GAS stimulation. Log fold change, log2 change (M value); log odds, statistical significance.
Because of the large number of genes that were significantly differentially expressed, they were initially analyzed based on their functional gene ontologies. The gene ontology categories that were overrepresented compared to the genome as a whole in the genes with increased expression after stimulation with Rh+ GAS versus Rh− GAS relate to cellular processes including proliferation, differentiation, homeostasis, regulation of the cell cycle, and regulation of transcription. The gene ontology categories overrepresented in the genes with decreased expression (that is, those that had greater expression after stimulation with Rh− GAS) are related to cellular processes and the immune response. Immune response terms include response to external stimulus, cytokine biosynthetic process, chemotaxis, and inflammatory response. The overrepresentation of immune response functions in the genes with decreased expression indicates that more components of the immune response are decreased than are increased after stimulation with Rh+ GAS than after stimulation with Rh− GAS.
Further analysis focused on the individual genes that were significantly differentially expressed in the same direction in both the first (ARF versus control groups) and second (Rh+ GAS versus Rh− GAS) arms of the study. After 3 h of stimulation, there were no genes that were differentially expressed in the same direction in the two experiments. After 24 h of stimulation, there were 13 genes that were significantly differentially expressed in the same direction: 4 had increased expression and 9 had decreased expression (Table 2). Seven of these 13 genes (TYROBP, IFN-γ, GZMB, IL-10, SLAMF1, GNLY, and PRF1) were genes involved in the immune response (Fig. 5). Although the mean expression for each of these genes was significantly different between the groups and stimulations, there was individual variability in absolute expression changes, with no single gene being universally significantly differentially expressed between every different individual in the two groups (Fig. 5).
Genes that were significantly differentially expressed in the same direction between ARF subjects stimulated with Rh+ GAS and control subjects stimulated with Rh+ GAS and between ARF subjects stimulated with Rh+ GAS and ARF subjects stimulated with Rh− GAS
Immune genes with significantly decreased expression after 24 h between in vitro stimulation with Rh+ GAS in ARF subjects and both Rh+ GAS in control subjects and Rh− GAS in ARF subjects. Each point represents the log2 fold change in gene expression (M value) for one individual after stimulation; the horizontal bar represents the mean M value of the group. IFN-γ, gamma interferon; GZMB, granzyme B; GNLY, granulysin; IL-10, interleukin 10; PRF1, perforin 1; SLAMF1, signaling lymphocytic activation molecule family member 1. The one significantly increased gene (TYROBP) is not shown.
DISCUSSION
Our study is the first to describe differences in gene expression between individuals who are susceptible to ARF and those who are not. It is also the first study to describe differences in response to stimulation with rheumatogenic and nonrheumatogenic GAS. Notably, a common set of genes, the majority of which are involved in the immune response, had significantly different levels of expression in the same direction in both arms of the study. This both validates the methods of the study and adds weight to the result. These findings support the hypothesis that a difference in the immune response to rheumatogenic GAS underlies susceptibility to ARF and might explain why only some individuals develop ARF after infection with rheumatogenic GAS.
Gene expression differences.Gene expression that was significantly different in both arms of the study was predominantly decreased in the ARF group after in vitro stimulation with Rh+ GAS, corresponding to failure to respond in the same way as both control individuals and the ARF group after stimulation with Rh− GAS. These genes are involved in cytokine activity, chemokine signaling, leukocyte extravasation, cell-mediated cytotoxicity, and apoptosis.
The genes with decreased expression included IFN-γ and IL-10, which play a central role in cell-mediated immunity. The role of these cytokines in ARF/RHD has been explored in other studies. In the animal model of ARF (cardiac myosin-induced autoimmune myocarditis and valvulitis in the Lewis rat), protection is conferred by IL-10-producing lymphocytes (17). In our study, a decrease in expression or the number of such lymphocytes may account for the decrease in measurable IL-10 expression. In a different mouse model of autoimmune myocarditis, the condition is lethal in IFN-γ receptor-deficient mice (18). While this is different from the rat ARF model as there is no valvulitis, T cell mimicry is similarly believed to be involved (19), and it may be speculated that similar processes of inflammation/protection from autoimmunity are involved. In humans, cytokine dysregulation has been implicated in the damage in rheumatic valves, with a proinflammatory Th1 cytokine (tumor necrosis factor [TNF] and IFN-γ) response predominating in valvular cellular infiltrate (20, 21). Patients with ARF/RHD also have increased plasma T cell and macrophage-derived proinflammatory cytokines (22–24) and unregulated activated T cell responses (25, 26). It is not possible in this study to determine which cell types are responsible for the decrease in expression of IFN-γ as it is produced in NK cells and macrophages in addition to T cells. However, expression of cytokine genes involved in both the Th1 (IFN-γ) and Th2 (IL-10) responses was decreased in the ARF group in response to Rh+ GAS compared to the level in the control group and to the response to Rh− GAS, so the Th17 pathway was investigated. IL-17 expression was not significantly increased although this does not completely exclude involvement of the Th17 pathway as the proportion of cells is likely to be small and may not significantly impact the overall expression changes (27).
This study was not designed to determine whether the decreased expression of these cytokine genes in PBMCs from ARF-susceptible individuals was accompanied by changes in gene expression in the cellular infiltrate in the valves. Similarly decreased expression in the valves would suggest that the whole process is due to a failure of the immune system in these individuals to suppress autoimmune progression. Increased expression in the valves would support migration of cytokine-producing cells from the peripheries to the heart through chemotaxis, potentially initiated by a failure of suppression.
Several of the products of the differentially expressed genes in this study are either chemoattractants themselves (IFN-γ, IL-10, and GNLY) or induce or inhibit genes that are chemoattractants. This is significant in light of the proposed importance of chemotaxis and leukocyte extravasation in the maladapted immune response that characterizes ARF. Studies suggest that streptococcal antigen-activated T cells migrate from the periphery to the heart and extravasate through the endothelium to heart valve tissue (28–30). Changes in expression in genes with a role in chemotaxis may be associated with disrupted migratory processes in ARF.
The decrease in expression of genes encoding proteins within cytotoxic T cell granules—PRF1, GZMA, GZMB, and GNLY—might be explained by the decrease in cytotoxic T cells that has been found in both blood and valvular tissue from individuals with ARF and, to a lesser extent, with RHD (25, 26). Our group has previously shown that the composition of cell mixtures affects detected gene expression, and therefore a reduction in cytotoxic T cells may explain the decrease found in this study (27). This would also be consistent with the finding of lower expression of IFN-γ as decreased IFN-γ would result in decreased activation of cytotoxic T cells. A reduction in cytotoxicity could result in inappropriate persistence of cells with consequent overstimulation leading to an abnormally aggressive immune response and autoimmunity.
Several genes that mediate/induce apoptosis, including IFN-γ, GZMA, GZMB, GNLY, and PRF1, had decreased expression. Although it is not possible to determine in which cells the genes relating to apoptosis were less expressed, defective apoptosis, including lymphocyte apoptosis, is associated with autoimmunity (31, 32). The potential consequence of dysregulation of apoptosis is the inappropriate persistence of cells, leading to prolonged antigenic stimulation or excessive T cell activity, resulting in autoimmunity.
Individual differences.Although the expression levels of several genes were significantly different between the two groups, there was variability within each group. The range of values for each gene in each individual in response to Rh+ GAS stimulation indicates that, even within the ARF and control groups, individuals do not respond identically. Our hypothesis is that a combination of genes is expressed differently between individuals who are susceptible to ARF and those who are not and, further, that this combination may be different in different individuals. It is possible that the details of the pattern of immunological dysregulation following infection with rheumatogenic GAS are slightly different in different people but that the culmination is a final common pathway that underpins the development of ARF. This is consistent with the existence of many different genetic polymorphism associations with susceptibility to ARF (4).
Strengths and limitations.This is the first study to use global gene expression analysis to investigate susceptibility to ARF. Other strengths of our study include the accurate categorization of participant susceptibility to ARF in a homogenous population with universal exposure to rheumatogenic GAS. Our strategy of comparing both susceptible and nonsusceptible patients and both rheumatogenic and nonrheumatogenic strains enabled us to identify the genes most likely to be involved in susceptibility.
Potential limitations of this study include the possibility that a susceptible individual was misassigned to the nonsusceptible control group as a result of having not been exposed to rheumatogenic GAS. This is unlikely as adults in the Northern Territory are universally exposed to high levels of rheumatogenic GAS (1). In addition, the clinical assignment was confirmed by D8/17 expression being positive in all of the individuals in the ARF-susceptible group and negative in all of the controls. A further limitation is that while the GAS strains were selected from well-defined clinical cases, the association between those strains and concurrent ARF and acute poststreptococcal glomerulonephritis (APSGN) cannot be absolutely verified. Finally, it is not possible to distinguish between responses in the ARF-susceptible group that reflect susceptibility from those that might result from changes attributable to a previous episode of ARF. However, without samples taken prior to the development of ARF, this cannot be explored.
Conclusion.The fact that a high proportion of the differentially expressed genes have roles in apoptosis and in the immune response supports the current model of autoimmune and cytokine dysregulation in ARF. This study also raises the possibility that a failed response to GAS, involving decreased expression of cytotoxic and apoptotic genes and other regulatory/suppressive genes, potentially allows the autoimmune process, contributing to the immunopathogenesis of ARF. This questions the likely effectiveness of suppressive treatment alone. In addition, the variability between individuals suggests that a common target for effective immunomodulatory treatment is likely to be late in the pathogenic pathway.
ACKNOWLEDGMENTS
This work was supported by a grant from the National Heart Foundation of Australia and the Australian National Health and Medical Research Council (NHMRC grant number 216716). P.A.B. was the recipient of a European Society of Pediatric Infectious Diseases Fellowship Award and Melbourne International Research Scholarships from The University of Melbourne.
We acknowledge the laboratory assistance of Rebecca Towers, the field assistance of local nursing staff, and the generosity of the indigenous individuals enrolled in the study.
FOOTNOTES
- Received 10 September 2013.
- Returned for modification 21 October 2013.
- Accepted 24 November 2013.
- Accepted manuscript posted online 2 December 2013.
- Copyright © 2014, American Society for Microbiology. All Rights Reserved.
