1. Introduction
Developing atherosclerotic plaque integrates multiple extra- and intracellular signals recognized by immune as well as vascular cells [
1,
2,
3]. Activation of multiple signaling pathways in all cell types present in the atheroma leads to cross-talk at the level of cell-cell interaction but also at that of transcription factor activation. For instance, endothelial and smooth muscle cells respond to damage and pathogen associated molecular patterns, which activate Toll-like receptors (e.g., TLR4) [
4]. The same ligands activate T
H1 lymphocytes to release pro-inflammatory interferon (IFN)γ, which also acts on vascular cells [
5]. Consequently, cell activation leads to release of inflammatory cytokines and chemokines and presentation of adhesion molecules, allowing leukocyte infiltration of the vessel wall and plaque progression [
6]. It has been shown in many studies that IFNγ plays a key role in atherogenesis [
7,
8]. IFNγ activates an inflammatory transcriptional program through the canonical janus-kinase-signal transducer and activator of transcription (JAK)-STAT signaling pathway. This involves early activation of STAT1 (Signal Transducer and Activator of Transcription-1) and, at a later stage, of IRFs (Interferon Regulatory Factors), to generate a prolonged interferon response [
9]. The immediate IFNγ response involves binding of STAT1 dimers to the Gamma Activated Sequence (GAS) element in the promoters of genes such as interferon-regulated factor 1 (IRF1), guanylate binding proteins (GBP) and intercellular adhesion molecule 1 (ICAM1). In contrast, the IRF-mediated response to IFNγ utilizes the Interferon Stimulated Response Element (ISRE) to induce expression of chemokine genes such as chemokine (C-C motif) ligand 5 CCL5 [
10] or C-X-C motif chemokine 10 (CXCL10) [
11].
Similar to IFNγ, TLR4 expression was detected in human and mouse atherosclerotic plaques. Moreover, circulating monocytes from acute coronary syndrome and coronary arteriosclerotic patients exhibit elevated TLR4 expression [
12]. Mice deficient in TLR4 show reduced atherosclerosis proving that Toll-like receptor-dependent signaling participates in disease development [
13]. Signaling of TLR4 involves two major pathways: 1. MyD88 (Myeloid Differentiation factor 88)-dependent NFκB (Nuclear Factor kappa B) activation and TRAM (TLR4 adaptor molecule)-dependent IRF3 activation. NFκB induces expression of many inflammatory genes, such as inducible nitric oxide synthase (iNOS), tumor necrosis factor alpha (TNFα) and interleukin 6 (IL6). IRF3, on the other hand, induces production of IFNβ, which in an auto- and paracrine manner activates STAT1 [
14]. Thus, IFNγ and TLR4 signaling pathways utilize common transcription factors, including STAT1 and IRFs.
Over the years, cross-talk has been shown to exist between IFNγ and lipopolysaccharide (LPS) in the different cell types building the atherosclerotic plaque, resulting in increased expression of inflammatory mediators [
15]. This cross-talk encompasses a complex, multi-layered mechanism relying on increased activation of STAT1 as well as on interactions of STAT1 with other transcription factors (
i.e., IRF1 and NFκB), resulting in increased expression of genes such as CXCL10 and ICAM1 [
11,
16,
17]. The interactions can be either direct at the protein level or indirect at the level of promoter binding sites. The latter requires specific regulatory modules containing STAT1 binding elements and NFκB or IRF binding sites in close proximity. Similar cross-talk phenomena have been proven to exist for other cytokine combinations (e.g., TNFα and IFNγ, and IL1β with IFNγ) and to particularly contribute to inflammatory gene expression. For instance, IFNγ and TNFα synergistically induced CXCL9 [
18], ICAM1 [
19] and iNOS [
20], which depended on GAS and NFκB elements in their promoters and involved interaction of bound STAT1 and NFκB factors with CREB-binding protein and enhanced recruitment of RNA polymerase II [
18]. Alternatively, the human IDO1 gene and murine Tap1 and Lmp2 genes were shown to possess combined ISRE and gamma-interferon-activated sites (GAS) elements in their promoters, both being necessary for maximum induction by IFNγ [
21,
22]. A third possible mechanism was revealed in the regulation of vascular cell adhesion molecule 1 (VCAM1) [
19] and CCL19 [
23] expression, where a combination of interferon-stimulated response element (ISRE) and NFκB elements appeared to be responsible for optimal transcription. In case of CXCL10, an even more complex mechanism seems to be involved, since synergistic induction by IFNγ and TNFα relied on ISRE and NFκB elements [
11], whereas IFNγ-TLR4 cross-talk was suggested to depend on STAT1 [
24].
Until now, the existence of these different cross-talk mechanisms has been revealed predominantly by
in vitro experiments conducted in the individual cell types involved in atherosclerotic plaque formation. To date there is limited information available on the role of IFNγ and TLR4 signaling cross-talk in the regulation of pathophysiological processes underlying atheroma development. Here, by applying an
in silico approach, we analyzed gene expression profiles in combination with gene ontology (GO) classification and promoter analysis of human coronary and carotid lesions (extracted from GEO: GSE40231 [
25] and GSE21545 [
26]) for potential evidence that STAT1-dependent inflammatory signal integration may be involved in plaque development. Indeed, our analysis highly suggests that STAT1-NFκB and STAT1-IRF regulatory modules are over-represented in promoters of inflammatory genes up-regulated in human coronary and carotid plaques and points to a possible involvement of IFNγ and TLR4 cross-talk.
Moreover, based on GO classification of these up-regulated genes, we detected high similarity in molecular processes and cellular interactions underlying plaque development in both vessel types, predicting overlap in pathophysiology. Finally, this comparative gene expression analysis revealed the presence of a common subset of inflammatory chemokine, cytokine and matrix remodeling genes, encoding for secreted proteins. These could serve as a basis of a non-invasive diagnostic assay for early detection and monitoring of the atherosclerotic process.
3. Discussion
IFNγ and TLR4-mediated signaling pathways, activated in different immunomodulatory and vascular cell types, have been implicated in plaque development and progression [
8,
13]. Both pathways utilize STAT1 [
15] to regulate expression of inflammatory and pro-atherosclerotic genes, such as chemokines (CXCL10) and adhesion molecules (ICAM1) [
38]. In our previous research we were able to show
in vitro that in endothelial cells (ECs) and vascular smooth muscle cells (VSMCs) cross-talk between IFNγ and LPS exists and
in vitro facilitates STAT1-dependent increase chemokine expression and monocyte to endothelial cells adhesion, a hallmark of early atherosclerosis [
17]. A similar STAT1-dependent mechanism was described by others in immune cells [
39]. Thus, IFNγ-TLR4 signaling cross-talk importantly controls behavior and interactions of all cells involved in atherosclerotic plaque formation, although this has not been studied in the context of atherosclerotic plaques.
Here we applied an
in silico approach on deposited in GEO gene expression profiles of coronary and carotid atherosclerotic plaques. Using GO classification, we first conducted functional analysis on 254 up-regulated genes selected from the coronary dataset. This disclosed a statistically significant over-representation of genes involved in cell adhesion, migration, response to external stimulus, and immune response (
Table 1). Likewise, functional analysis of 845 up-regulated genes from the carotid dataset revealed strong over-representation of GO terms such as: immune response, adhesion, migration,
etc. (
Table 3), which was similar to our findings in coronary plaques.
These results are in line with the current view of the different processes involved in plaque development and progression [
40,
41]. First, injured endothelium becomes pro-thrombotic and releases cytokines and chemokines, allowing for leukocyte adhesion and migration into the vessel wall. Then, lymphocytes differentiate into T cells (mostly of the T
H1 phenotype) and monocytes into macrophages. Subsequent production of inflammatory mediators by these cells activates proliferation and migration of smooth muscle cells facilitated by secretion of matrix-metalloproteinases (by macrophages and smooth muscle cells) [
1]. Together with a disturbed composition of plasma lipoproteins, a known risk factor for atherosclerosis [
42], all of these processes are reflected in the over-represented GO terms listed in
Table 1 and the
supplementary data file.
Interestingly, we have observed that 62 out of 254 of the up-regulated genes in coronary plaques possess potential STAT1 binding sites (GAS or ISRE) in their promoters (see Results,
Table 2 and
supplementary data) highly implicating STAT1 in the development of atherosclerotic plaques. Genes containing either GAS or ISRE elements included chemokines (CCL2, CCL5, CCL19, CXCL10, CXCL9), cytokines (VEGFC), adhesion molecules (integrins), and proteins involved in matrix remodeling (MMP19). In the carotid plaques 208 of 845 up-regulated genes demonstrated presence of putative STAT1 binding sites and involved chemokines (CCL2, CCL19, CCL5, CCRL2, CCL13), chemokine receptors (CX3CR1, CXCR6), cytokines, and cytokine receptors (IL18, IL2RG, IL2RB). These findings are in correlation with results from
in vitro studies, which showed that many of these genes are indeed STAT1 targets (e.g., CCL5 [
16], CXCL10 [
43]).
Since our aim was also to provide evidence for the potential role of IFNγ-TLR4 cross-talk in plaque development, we searched for the presence of transcriptional modules of STAT1 and NFκB or STAT1 and IRF binding sites (not more than 50 nucleotides apart). Indeed, we uncovered in coronary plaques that 116 genes contained a STAT1-NFκB module and 150 a STAT1-IRF module, pointing to a potential mechanism of STAT1 dependent co-regulation of gene expression in the cell types present in atherosclerotic plaques.
GO classification of up-regulated genes in coronary plaques with putative STAT1-IRF modules in their promoters reflect mechanisms underlying plaque formation: cell adhesion (NCAM1, VCAM1, THBS1) and migration (CCL19, CCL2, CCL4 and CCRL1), matrix remodeling and calcification (ADAMTS9, SPP1, MMP19), and inflammatory signaling (IL7R, IL13RA2, TLR4, TNFRSF10B, TNFRSF11B). In carotid plaques STAT1-IRF and STAT1-NFκB modules were present in a partially similar set of genes: adhesion molecules (ITGAM, PECAM1, VCAM1, ITGB2), chemokines (CCL2, CCL5, CCL18, CCL19, CCL7, CCL8), and matrix remodeling molecules (MMP1, MMP9, MMP12, SPP1).
Collectively, these results predict that key processes involved in plaque development are regulated by STAT1, either alone or in close co-operation with NFκB and IRFs providing a platform for cross-talk between different inflammatory stimuli. It also provides further proof for the crucial role of STAT1 in human atherosclerosis. This notion is further supported by an additional study from our group that was recently submitted elsewhere [Chmielewski S.
et al.], in which we uncovered a subset of 30 STAT1-dependent genes that are highly amplified by co-administration of IFNγ and LPS in VSMCs
in vitro [Chmielewski S.
et al.]. These genes included the chemokines CCL5, CXCL10, CCL8, CXCL9, CCRL2, which were also up-regulated in carotid or coronary human plaques, as shown here, and together reflect pro-atherogenic responses in human atherosclerosis. Based on our findings and other studies from the literature,
Figure 2 summarizes the complex role of STAT1 in signal integration between IFNγ and TLRs that we propose to take place in the forming atheroma.
Regulatory modules containing elements for STAT1 and IRF or NFκB have been shown to play key role in signal integration of various signaling pathways. For instance, maximal transcription of ICAM1 is possible only when both NFκB and STAT1 are bound to the promoter. The same mechanism was observed for CXCL10 in human smooth muscle cells [
44] and monocytic THP-1 cells [
45]. We observed a similar synergy between IFNγ and TLR4 signaling in human endothelial cells and murine smooth muscle cells, which was STAT1 dependent and led to amplified expression of CXCL10 and other genes [
17] [Chmielewski S.
et al.]. Furthermore, induction of iNOS expression by LPS is managed by NFκB binding, but presence of a GAS element in the promoter and co-treatment with IFNγ enables maximal expression [
46]. Co-operative action of binding sites for STAT1 and IRF1 has been shown to manage expression of indoleamine 2,3-dioxygenase 1 (IDO1) (also termed INDO) gene [
21], which is involved in sustaining chronic inflammation and is one of the hallmark genes of the M1 inflammatory macrophage phenotype [
47]. Involvement of regulatory modules in complex signal integration of IFNγ and TLR4 (also other TLRs), in addition to other pro-inflammatory signals, has long been appreciated by the scientific community and the evidence presented here suggests that it is a functional mechanism in the plaque development contributing to chronic production of inflammatory modulators.
Figure 2.
Signal integration between interferon gamma (IFNγ) and Toll-like receptors (TLRs) involves interactions of STAT1, interferon-regulated factor (IRF)1, IRF8 and nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB). Analysis of gene expression profiles of atherosclerotic plaques together with the results of microarray experiments conducted by our group on STAT1 -/- SMC [Chmielewski S. et al.] create this complex signal integration picture (see Discussion). STAT1 regulates genes by itself, but also interacting with IRF1, IRF8 or NFκB.
Figure 2.
Signal integration between interferon gamma (IFNγ) and Toll-like receptors (TLRs) involves interactions of STAT1, interferon-regulated factor (IRF)1, IRF8 and nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB). Analysis of gene expression profiles of atherosclerotic plaques together with the results of microarray experiments conducted by our group on STAT1 -/- SMC [Chmielewski S. et al.] create this complex signal integration picture (see Discussion). STAT1 regulates genes by itself, but also interacting with IRF1, IRF8 or NFκB.
It is known that the pathophysiology of coronary plaques resembles that of carotid plaques [
48]. Prevalence of stenosis in one artery type has been associated with prevalence of stenosis in the other. Both plaque types have been shown to respond in a similar way to shear stress and both display an inflammatory background [
49]. We were interested if the degree of similarity between coronary and carotid plaques could be recognized at the gene expression level. Indeed, overlap was observed in up-regulated genes from both datasets, together forming a 72-gene “plaque signature”. As expected, also distinct gene sets could be identified. This was in agreement with the fact that atherosclerosis is a multi-factorial disease, and it is highly likely that factors inducing and accelerating the disease will differ between various plaque types and even between different specimens [
50,
51,
52].
When we characterized in more detail the biological functions of these 72 “plaque signature” genes, we could distinguish many GO terms related to plaque development processes: cell adhesion (ALCAM [
53,
54], VCAM1 [
55]), T-cell migration (CCL5 [
56], CCR5 [
57], and CXCR4 [
58]), response to LPS (TLR4),
etc. (
Table 4 and
supplementary data file). In addition, promoter analysis of these genes revealed over-representation of STAT1 binding sites as well as STAT1 containing modules (
Table 2). The STAT1-NFκB module was present in 31 genes of this common group while the STAT1-IRF module could be detected in 45 genes. Some of these genes (CCL5, CCL19, CCL4, CXCL10, CXCL2, CXCL9 and MMP9) actually possessed binding sites for STAT1, NFκB as well as IRFs. The idea that STAT1-dependent cross-talk between IFNγ and TLR4 potentially exists in human plaques and the discovery of a “plaque gene signature” present the possibility for development of a novel non-invasive screening assay. Indeed, with respect to the cellular localization of the proteins encoded by the 72 common signature genes many of them are secreted and localize to the extracellular space or cell membrane making them ideal serum markers of atherosclerosis (
Figure 1 and
Table 5). In addition, the majority has a known connection to atherosclerosis, as determined by literature mining (
Figure 1).