Fibrosis is a pathology associated with many liver diseases, particularly chronic conditions, that can develop into cirrhosis, liver failure and death if left untreated (5, 37). Liver fibrogenesis often occurs when HSCs become activated following liver damage or disease and respond with the secretion of ECM proteins (38). HSCs can activate in response to a range of stimuli, with one of the most common being TGF-β1 (16, 17). TGF-β1 is a potent cytokine expressed throughout mammalian tissues, and is involved in a wide variety of key cellular processes (17). Despite the physiological importance of TGF-β1 and its potent HSC activating ability, the specific responses TGF-β1 induces in HSCs have yet to be fully characterised due to the complexity and far reaching nature of TGF-β1 signalling (1).
Several studies have explored HSC activation at the transcriptomic level using various methods and cell lines (39–44). The first such study utilised microarray analysis to investigate the effects of culture-induced activation (where HSCs activate over time on tissue culture plastic) on gene expression in LI90 cells, another immortalised human HSC line, when cultured on Matrigel (43). This study identified 3,350 differentially expressed genes and led to the identification of myocardin (MYCOD) as an activator of HSCs (43). Similarly, a second microarray study identified transcriptomic changes following culture-induced activation in primary rat HSCs (41). In their study, over 2,000 differentially expressed genes were identified with a fold change ≥ 2 and the Wnt5a signalling pathway was identified as a facilitator of rat HSC activation (41). A third study used RNA sequencing of primary human foetal HSCs exposed to TGF-β1 to identify differentially expressed long non-coding RNAs (lncRNAs) (40). This study found that TGF-β1 influences the expression of 381 lncRNAs in human foetal HSCs (40). Another RNA sequencing study was conducted on primary rat HSCs following culture-induced activation (42). A total of 553 genes were identified as being differentially expressed (42). Yet another RNA sequencing study investigated the differences in gene expression between quiescent and culture-activated primary human HSCs, with valproic acid used to maintain quiescence (39). Overall, the differential expression of 5,449 genes were detected and three genes which regulate the expression of connective tissue growth factor (CTGF), fibroblast growth factor 2 (FGF2) and netrin 4 (NTN4), each associated with HSC activation and liver fibrosis, were identified (39). The most recent study examining transcriptional regulation in HSCs used microarray analysis to investigate the differences between quiescent and culture-activated LX-2 cells, with MDI solution used to maintain quiescence (44). This study identified 3,424 differentially expressed genes with a fold change ≥ 2 (44). This same group had previously reported the RNA sequencing of primary HSCs taken from patients suffering from NAFLD (45). Comparison of the gene expression profiles of culture-activated LX-2 cells and primary NAFLD-associated HSCs revealed that 1,138 genes are differentially expressed in common (44).
As noted in the most recent study of Gerhard et al., the aim of all of these previous studies can be put simply as characterising the changes in gene expression that occur in HSCs during activation, and yet the findings show a large amount of variation in both the identity and number of differentially expressed genes (44). It is clear that the methods used to provoke HSC quiescence or activation, and detect gene expression, as well as the specific cell lines assayed, have a strong impact on the final results (44). To our knowledge, the current study is the first attempt to use RNA sequencing to investigate the changes in gene expression induced by TGF-β1 treatment in LX-2 cells.
A summary of the effects of TGF-β1 on the genes and signalling pathways discussed below can be found in Table 3.
Table 3
Summary of the deregulating effects of transforming growth factor-β1 (TGF-β1) on genes and pathways in LX-2 cells.
Gene | Effect of TGF-β1 | Phenotype | Pathway | Effect of TGF-β1 | Phenotype |
CIITA | Downregulated | Not characterised | Actin nucleation by ARP-WASP complex | Upregulated | Activated |
COL17α1 | Downregulated | Not characterised | Apelin signalling | Downregulated | Activated |
COL4α6 | Downregulated | Not characterised | EIF2 signalling | Upregulated | Activated |
EGR2 | Upregulated | Activated | ERK5 signalling | Upregulated | Activated |
FAP | Upregulated | Activated | Ethanol degradation | Downregulated | Activated |
FN1 | Upregulated | Activated | PI3K/AKT signalling | Upregulated | Activated |
FOXS1 | Upregulated | Not characterised | PPAR signalling | Downregulated | Quiescent |
HES1 | Upregulated | Activated | PPARα/RXRα activation | Downregulated | Quiescent |
NOX4 | Upregulated | Activated | PTEN signalling | Downregulated | Quiescent |
PI16 | Upregulated | Not characterised | STAT3 signalling | Upregulated | Activated |
PRG4 | Upregulated | Not characterised | tRNA charging | Upregulated | Activated |
PSG1 | Downregulated | Not characterised | Unfolded protein response | Upregulated | No effect |
SEMA3A | Downregulated | Not characterised | | | |
SERPINB2 | Downregulated | Not characterised | | | |
TGFβI | Upregulated | Not characterised | | | |
VCAM1 | Downregulated | Not characterised | | | |
VIP | Upregulated | Not characterised | | | |
The deregulating effects of TGF-β1 on the genes and signalling pathways discussed above, and the result of this deregulation on the activation status of hepatic stellate cells (HSCs). Genes whose role within HSCs is unknown are listed as "not characterised". |
Genes deregulated by TGF-β1 in LX-2 cells
Several genes described in Table 1 have known roles in promoting HSC activation and liver fibrosis downstream of TGF-β1, including EGR2, FAP, FN1, HES1 and NOX4 (46–50). While the function of these genes in relation to liver fibrosis is known, their highly upregulated status in this context may indicate that they are particularly significant mediators of early HSC activation or TGF-β1 signalling, and therefore worthy of more attention as potential markers for activating HSCs or fibrogenesis.
Other genes were identified in Table 1 that do not have clearly reported roles in HSCs. These genes have instead been associated with either the activity of fibroblasts or fibrogenesis in other tissues, including FOXS1, TGFβI, PI16, VIP and PRG4. FOXS1 promotes the activation of primary human skin fibroblasts (51), while TGFβI has been shown to interact with ECM proteins, including collagen type 1 (COL1), to inhibit the cell-ECM adhesion of skin and scleral fibroblasts (52, 53). The upregulation of these genes seen here may indicate that FOXS1 also promotes HSC activation downstream of TGF-β1, while TGFβI is likely involved in facilitating the migration of early activating HSCs from the space of Dissé. The overexpression of PI16 has been shown to reduce the proliferation of, and expression of COL1 in, murine cardiac fibroblasts (54). Similarly, the reduced expression of VIP correlates with progressive cardiac fibrosis in murine models, which can be reversed by VIP overexpression (55). PRG4 is associated with protection functions in the connective tissues and reduced fibroblast activation in the synovial tissue (56, 57). Assuming these genes carry out similar functions in HSCs, their upregulation by TGF-β1 is indicative of negative regulation of HSC activation, likely as a means of controlling fibrosis progression.
Several downregulated genes whose function likely influences HSC activity were identified in Table 2, including CIITA, SERPINB2 and PSG1. The upregulation of CIITA results in the increased expression of major histocompatibility complex class II (MHCII) genes (58), which have been shown to reduce HSC collagen expression and fibrotic potential during schistosomiasis infection (59). It can therefore be assumed that the downregulated CIITA expression seen here would increase HSC collagen expression and contribution to fibrosis (59). A deficiency of SERPINB2 in the livers of murine models of the helminth Schistosoma japonicum infection results in a reduction in the deposition of collagen within the egg-induced granuloma (60). Given the role of HSCs within the granuloma, it is highly likely that SERPINB2 deficiency reduces HSC activity to bring about this effect and, if so, would implicate SERPINB2 as a promoter of HSC activity. PSG1 has been shown to stimulate the release of active TGF-β1 protein in vitro (61), and therefore its reduced expression in this context would inhibit TGF-β1 signalling and subsequent HSC activation.
Several other genes associated with fibroblast activation or fibrosis in other tissues were also identified in Table 2, including SEMA3A and VCAM1. Increased expression of SEMA3A has been shown to activate corneal fibroblasts downstream of TGF-β1 and contribute to corneal fibrosis (62). VCAM1 upregulation is typically observed in cirrhotic liver tissues (63) and in pulmonary fibrosis (64), and the depletion of VCAM has been shown to reduce the activation of pulmonary fibroblasts (64).
The downregulated status of many of these genes was unexpected, as it suggests a reduction in HSC activity that conflicts with the activating influence of TGF-β1. The reduced expression of PSG1 in this context is likely a mechanism for controlling the levels of active TGF-β1 protein in order to regulate TGF-β1 signalling. For the other genes, it is possible that they too are involved in negative regulation of TGF-β1 signalling in HSCs. Alternatively, they may carry out different functions in HSCs than in other cells and tissues, and so are not relevant to HSC-induced fibrosis, or that their fibrotic responses require cues from particular pathogens or from other signalling mediators besides TGF-β1. The short TGF-β1 exposure time used in these experiments should also be considered; it could be the case that some of these genes are only involved in chronic fibrosis or cirrhosis, rather than the initial events surrounding fibrogenesis, and so a longer duration of experiment may be needed to observe the previously reported expression pattern of these genes. Indeed, the literature often describes gene expression associated with fibrosis from the prospective of established models of fibrosis or patients suffering chronic disease, rather than short term in vitro studies.
The expression of COL17A1 and COL4A6 were also downregulated, despite COL4 having been shown previously to be upregulated in HSCs following TGF-β1 exposure (65). One previous study has shown that COL17 and COL4 interact together in skin and oral keratinocytes to assist cell-ECM adhesion (66). COL4 has been identified as an ECM component in the space of Dissé, the storage site of quiescent HSCs (67), while COL17 is a transmembrane collagen that interacts with both extra- and intracellular structural components to facilitate cell linkage to the epithelium (68). Given that activating HSCs must migrate from the space of Dissé towards the provoking stimuli, it is possible that the expression of these collagens, perhaps working in tandem with TGFβI, might be initially downregulated in order to allow the cell to disengage from the anchoring ECM in the space of Dissé, and thus allow migration. Type 4 collagens are known to be involved in maintaining normal liver architecture and are typically degraded in fibrosis to allow for the deposition of fibrillar collagens (69).
Signalling pathways upregulated by TGF-β1 in LX-2 cells
The most strongly upregulated signalling pathway in Fig. 3 was that of transfer (t)-RNA charging, a pathway involved with protein translation. Increased tRNA charging activity is synonymous with the increased level of protein synthesis that occurs in HSCs during, and following, activation (10). Similarly, eukaryotic translation initiation factor 2 (EIF2) signalling is important in the initiation of protein synthesis in eukaryotic cells, and its upregulation in this context is likely concurrent with the increase in translational activity displayed by aHSCs (10, 70). However, one study has reported that a component of the S. mansoni EIF2 signalling pathway, the subunit EIF2α, can interact with the TGF-β receptors TGFβRI and TGFβRII to inhibit TGF-β signalling (71). The nature of the enhanced EIF2 signalling in aHSCs following TGF-β1 exposure could therefore also double as a negative regulator of TGF-β1 responses.
Several pathways in Fig. 3, including ERK5, PI3K/AKT and STAT3 signalling, represent signalling pathways downstream of TGF-β1 that involve transcriptional regulators capable of driving HSC activation (72–74). TGF-β1 carries out physiological functions by inducing cellular gene expression, and the SMAD family of transcriptional regulators are generally responsible for transducing signals from TGF-β ligands to the cell nucleus (17). However, SMAD-independent TGF-β signalling is also common, including these three aforementioned pathways (1, 17). The absence of SMAD signalling in Fig. 3 is interesting; in HSCs, the level of phosphorylated SMAD2 protein reaches a peak within 1.5 hours of TGF-β1 exposure, and returns to basal levels within the following six hours (75). The absence of such signalling from the data could suggest that, while highly active immediately following TGF-β1 exposure, by the 24-hour timepoint SMAD signalling gives way to alternative SMAD-independent pathways, perhaps to prevent excessive HSC activation whilst simultaneously inducing the appropriate gene expression profiles associated with activation.
The assembly of organised ACTA2 filaments is a strong marker of myofibroblasts (10) (see Fig. 1). These filaments carry out several functions in aHSCs, including supporting the expanding cell cytoplasm, facilitating cell motility and acting as a method of attaching to, and signalling between, the ECM and other cells (10, 76). Therefore, it is unsurprising that the activity of the actin nucleation by ARP-WASP complex was upregulated by TGF-β1 exposure.
The unfolded protein response (UPR) pathway is activated following cellular stress induced by the build-up of improperly folded proteins within the endoplasmic reticulum (ER) (77). Activation of this pathway increases the expression of chaperone proteins to assist protein folding, reduces protein translation and degrades improperly folded proteins (77). If the ER stress cannot be lifted, the UPR will move to trigger cellular apoptosis (77). One recent study has shown that the sudden increase in protein production that accompanies HSC activation can trigger UPR activation, potentially as a compensatory mechanism (78). This pathway does not significantly affect overall HSC activation, but from the results presented here and those of others (78), it could be considered an early marker of HSC activation.
Signalling pathways downregulated by TGF-β1 in LX-2 cells
The most strongly downregulated signalling pathway in Fig. 4 was that of PPARα/RXRα activation, and PPAR signalling was also found to be downregulated. Quiescent HSCs take up and store vitamin A (retinol) within lipid droplets (10, 79) following its metabolism into lipid-soluble derivatives (80). HSCs regulate the expression of genes involved in fatty acid uptake and metabolism via peroxisome proliferator-activated receptors (PPARs) and the retinoid X receptor (RXR), which heterodimerise together to act as a transcription factor for these genes (80). Upon activation, HSCs lose the ability to store vitamin A and, as such, display reduced retinol-related signalling (81). Studies have shown that the expression of both PPAR-γ, a relative of PPAR-α, and RXR are reduced in aHSCs (81, 82). Furthermore, agonism of PPAR-γ signalling in aHSCs has been shown to suppress the expression of ACTA2 and collagen type 1α1 (COL1A1), and to facilitate aHSC reversion back into a quiescent state (82). As such, the downregulated activity of the PPARα/RXRα activation and PPAR signalling pathways following TGF-β1 exposure was expected.
Apelin is an endogenous ligand of the G protein-coupled APJ receptor (83). Apelin signalling is associated with a diverse range of tissue-specific functions (84). In the liver, apelin signalling is strongly associated with fibrosis (84). Several studies have highlighted how components of the apelin signalling pathway induce the expression of pro-fibrotic genes in LX-2 cells, including COL1, ACTA2 and platelet-derived growth factor receptor-β (PDGFRβ) (83, 85). Furthermore, the inhibition of apelin signalling has been shown to reduce the intensity and burden of liver fibrosis in murine models (86). Paradoxically however, other studies have linked apelin signalling with the inhibition of TGF-β1 responses; one study has shown that apelin inhibits the TGF-β1-induced activation of SMAD proteins and subsequent upregulation of ACTA2, COL1 and FN1 expression in epithelial cells (87), while another described how apelin inhibits the TGF-β1-induced upregulation of ACTA2 and COL1A1 expression in cardiac fibroblasts (88). These findings highlight the tissue-specific nature of apelin signalling and could indicate an interesting situation in HSCs whereby apelin increases fibrotic gene expression whilst simultaneously inhibiting TGF-β1 signalling.
Phosphatase and tensin homolog (PTEN) is a tumour suppressor protein that regulates cell cycle progression (89). Several studies have shown that PTEN signalling inhibits HSC activation; one study demonstrated that the downregulation of miR-181b, an inhibitor of PTEN expression, results in the suppression of HSC activation as determined by reduced ACTA2 expression and collagen deposition (90). Another study showed that PTEN-deficient mice develop progressive liver fibrosis characterised by the increased expression of ACTA2, COL1 and tissue inhibitor of matrix metalloproteinase (TIMP)-1 (91). HSCs isolated from these PTEN-deficient mice displayed higher levels of activation on average compared to HSCs in wild type mice (91). Similarly, one final study has described how the overexpression of PTEN in rat HSCs prevents the morphological changes associated with activation, and reduces the expression of ACTA2 and COL1A1 (92). Taken together, PTEN signalling is a strong negative regulator of HSC activation.
Ethanol and its metabolites have been shown to promote HSC activation through several mechanisms (93). Firstly, ethanol interferes with retinol-related metabolism and signalling in HSCs by decreasing the uptake of vitamin A and degrading vitamin A in the liver into inactive metabolites (93, 94). As mentioned above, retinol-related signalling promotes HSC quiescence, and so reduced levels of vitamin A uptake will encourage HSC activation (81, 82). Acetaldehyde, a product of ethanol metabolism, has been shown to induce the expression of both latent and active TGF-β1, as well as the receptor TGFβRII, in HSCs, which induces further HSC activation via auto- and paracrine TGF-β1 signalling (93, 95). Finally, both ethanol and acetaldehyde upregulate the expression of pro-fibrotic genes including COL1A1, COL1A2, matrix metalloproteinase (MMP)-2 and FN1 in HSCs (93, 96–98). Given the strong activating influence of ethanol and acetaldehyde in HSCs, it is unusual that TGF-β1 exposure would downregulate the activity of several ethanol degradation pathways. HSCs express enzymes involved in ethanol degradation, however, it is possible that activated HSCs may inhibit the expression of these enzymes in an attempt to regulate ethanol-induced activation and fibrosis as a protective mechanism (99).