Transforming growth factor β (TGFβ) cross-talk with the unfolded protein response is critical for hepatic stellate cell activation

Transforming growth factor β (TGFβ) potently activates hepatic stellate cells (HSCs), which promotes production and secretion of extracellular matrix (ECM) proteins and hepatic fibrogenesis. Increased ECM synthesis and secretion in response to TGFβ is associated with endoplasmic reticulum (ER) stress and the unfolded protein response (UPR). TGFβ and UPR signaling pathways are tightly intertwined during HSC activation, but the regulatory mechanism that connects these two pathways is poorly understood. Here, we found that TGFβ treatment of immortalized HSCs (i.e. LX-2 cells) induces phosphorylation of the UPR sensor inositol-requiring enzyme 1α (IRE1α) in a SMAD2/3-procollagen I–dependent manner. We further show that IRE1α mediates HSC activation downstream of TGFβ and that its role depends on activation of a signaling cascade involving apoptosis signaling kinase 1 (ASK1) and c-Jun N-terminal kinase (JNK). ASK1–JNK signaling promoted phosphorylation of the UPR-associated transcription factor CCAAT/enhancer binding protein β (C/EBPβ), which is crucial for TGFβ- or IRE1α-mediated LX-2 activation. Pharmacological inhibition of C/EBPβ expression with the antiviral drug adefovir dipivoxil attenuated TGFβ-mediated activation of LX-2 or primary rat HSCs in vitro and hepatic fibrogenesis in vivo. Finally, we identified a critical relationship between C/EBPβ and the transcriptional regulator p300 during HSC activation. p300 knockdown disrupted TGFβ- or UPR-induced HSC activation, and pharmacological inhibition of the C/EBPβ–p300 complex decreased TGFβ-induced HSC activation. These results indicate that TGFβ-induced IRE1α signaling is critical for HSC activation through a C/EBPβ-p300–dependent mechanism and suggest C/EBPβ as a druggable target for managing fibrosis.

Hepatic stellate cell (HSC) 4 activation increases production of extracellular matrix (ECM) proteins, leading to fibrogenesis (1,2). ECM proteins undergo folding and processing within the endoplasmic reticulum (ER) prior to trafficking through the secretory pathway. Canonically, increased production of ECM proteins leads to ER stress and activation of the unfolded protein response (UPR), which facilitates their folding and trafficking out of the cell (3,4). Based on this, several studies show that the UPR is critical for HSC activation. Indeed, stimuli that promote HSC activation, such as TGF␤, are associated with ER stress and UPR induction (5)(6)(7). Furthermore, chemical induction of the UPR in HSCs in vitro increased expression of ECM proteins in HSCs, whereas pharmacological inhibition of the UPR attenuated HSC activation (8 -11). Despite these observations, the mechanisms by which the UPR regulates HSC activation, and in turn how HSC activation modulates UPR signaling, are unclear.
One of the major ER stress sensors that initiates the UPR is inositol-requiring enzyme 1␣ (IRE1␣). IRE1␣ is critical for UPR signaling to accommodate increased protein folding and trafficking (12)(13)(14). Upon sensing increased levels of unfolded proteins in the ER, IRE1␣ dimerizes and undergoes autotransphosphorylation. This kinase activity is necessary for activation of its endonuclease domain as well as mediating downstream signaling cascades involving apoptosis-signaling kinase 1 (ASK1) (15)(16)(17)(18)(19). Both the kinase and endonuclease domains of IRE1␣ are associated with HSC activation (5,8,11), but the mechanisms that regulate and propagate IRE1␣ signaling during HSC activation and fibrogenesis are not completely understood.
We sought to better understand the mechanisms of IRE1␣ activation and signaling during fibrogenesis. First, we established that TGF␤ induces IRE1␣ activation downstream of SMAD2/3 signaling and procollagen I expression. Mutational analysis revealed that IRE1␣ signaling through its kinase domain, but not its endonuclease domain, is sufficient to pro-mote HSC activation. IRE1␣ signaling through ASK1 and JNK is critical for TGF␤ up-regulation and phosphorylation of the UPR-responsive transcription factor CCAAT/enhancer binding protein ␤ (C/EBP␤), which in turn is critical for HSC activation. Pharmacological antagonism of C/EBP␤ through adefovir dipivoxil disrupted HSC activation in vitro and fibrogenesis in vivo. Finally, we found that the transcriptional regulator p300 is critical for both TGF␤and UPR-mediated HSC activation and may act through a mechanism involving C/EBP␤. Together, our work shows that TGF␤ induction of the UPR leads to a feed-forward signaling pathway that acts through C/EBP␤-p300 to promote fibrogenesis and that C/EBP␤ can be therapeutically targeted to limit fibrosis in vivo.

TGF␤ induction of IRE1␣ signaling is critical for HSC activation
To better understand the relationship between UPR signaling and TGF␤, we first established a model to characterize and understand IRE1␣ signaling in the presence of TGF␤. Immortalized HSCs (LX-2 cells) treated with TGF␤ (5 ng/ml) showed increased phosphorylation of IRE1␣, indicative of its activation (Fig. 1A). TGF␤ canonically signals through a pathway involving the transcription factors SMAD2/3, which up-regulates transcription of several genes involved in HSC activation and fibrogenesis, including procollagen 1␣1 and 1␣2 and fibronectin (20). To test whether TGF␤ induced IRE1␣ phosphorylation Figure 1. TGF␤ induction of IRE1␣ is critical for HSC activation. A, LX-2 cells were stably infected with an shRNA targeting procollagen 1␣1 (shCollagen 1␣1) or an NT control. Cells were treated with 5 ng/ml TGF␤ for 0, 1, 2, 4, or 24 h, followed by assessment of phosphorylation and protein levels of IRE1␣, fibronectin (FN), ␣SMA, and collagen I. HSC70 served as a loading control. Quantification is shown adjacent. B, LX-2 cells were transfected with an siRNA targeting SMAD2 (siSMAD2) or a nontargeting siRNA (siControl). 24 h post-transfection, cells were serum-starved and treated with TGF␤ for 0, 1, 2, 4, or 24 h. Cell lysates were harvested, and levels of phosphorylated and total IRE1␣, SMAD2/3, and HSC70 (loading control) were assessed by immunoblotting. Quantification is shown adjacent. C, primary HSCs were isolated from IRE1␣ fl/fl mice and infected with adenovirus expressing Cre recombinase (AdCre) or LacZ as a control. 48 h postinfection, cells were treated with TGF␤ or vehicle. Cell lysates were harvested and analyzed by immunoblotting for FN, collagen I, ␣SMA, and HSC70 (loading control). Quantification is shown adjacent. D and E, LX-2 cells were pretreated with the IRE1␣ inhibitor 48C (15 M) for 1 h, followed by TGF␤ or vehicle for 24 h. Cell lysates or mRNA were harvested and analyzed by either immunoblotting (D) for FN, collagen I, and HSC70 (loading control) or qPCR (E) to assess fibronectin, procollagen 1␣1 and 1␣2, ␣SMA, or PPAR␥ expression. Quantification for D is shown below the blots. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
downstream of SMAD2/3, we disrupted either SMAD2 or procollagen I␣1 expression using RNAi (siSMAD2 or shCollagen 1␣1). IRE1␣ phosphorylation in response to TGF␤ was attenuated in shCollagen 1␣1 or siSMAD2 cells compared with nontargeting (NT) controls ( Fig. 1 (A and B), quantification adjacent). Furthermore, loss of procollagen I␣1 reduced TGF␤ up-regulation of fibronectin, as well as ␣SMA, a marker of activated HSCs. These observations led us to examine whether IRE1␣ signaling was critical for TGF␤-induced HSC activation. HSCs were isolated from mice harboring loxP sites within the gene encoding IRE1␣ and were treated with an adenovirus expressing Cre recombinase (AdCre) to disrupt IRE1␣ expression or with LacZ as a control (21,22). Following TGF␤ treatment, protein levels of fibronectin, collagen I, and ␣SMA were reduced in cells with Cre-mediated IRE1␣ knockdown (Fig. 1C, quantification adjacent). Similarly, treatment of LX-2 cells with the IRE1␣ inhibitor 48C (15 M) effectively blocked TGF␤ induction of collagen I and fibronectin protein (Fig. 1D, quantification below). 48C also reduced gene transcription of fibronectin, procollagen 1␣1 and 1␣2, and ␣SMA (Fig. 1E). Increased mRNA expression of peroxisome proliferator-activated receptor ␥ (PPAR␥), which is associated with quiescent HSCs, was also observed. Thus, inhibition of IRE1␣ signaling limits TGF␤ activation of HSCs.
We next asked whether IRE1␣ signaling was sufficient to mediate profibrotic gene expression. A doxycycline-inducible IRE1␣ construct was stably expressed in LX-2 cells (doxIRE1␣), and upon doxycycline treatment (5 g/ml), increased protein levels of IRE1␣, as well as fibronectin and collagen I, were observed ( Fig. 2A, quantification adjacent). To elucidate the mechanisms downstream of IRE1␣ signaling responsible for this effect, we utilized doxycycline-inducible IRE1␣ mutants that specifically inactivate either the IRE1␣ kinase domain (K599A) or the endonuclease domain (K907A) (16). Overexpression of the K907A mutant recapitulated the effects of the WT IRE1␣ construct; however, overexpression of IRE1␣ K599A failed to up-regulate fibronectin or collagen I ( Fig. 2A). This implicates the IRE1␣ kinase domain in HSC activation independent of IRE1␣ endonuclease activity. A major downstream target of the IRE1␣ kinase domain is ASK1, which in turn phosphorylates and activates signaling cascades through Jun N-terminal kinase (JNK) or p38 MAPK. To explore whether . Cells were treated with doxycycline (5 g/ml) or vehicle for 24 h, cells were lysed, and protein levels of IRE1␣, fibronectin, and collagen I levels were assessed. HSC70 served as a loading control. Quantification is shown adjacent. B, doxIRE1␣ cells were treated with GS-444217 (2 M), SP600125 (10 M), or SB203580 (0.5 M) for 1 h to inhibit ASK1, JNK, or p38, respectively, followed by doxycycline treatment for 24 h. Cell lysates were harvested, and IRE1␣, fibronectin, collagen I, phosphorylated and total JNK, and phosphorylated and total p38 were assessed by immunoblotting. HSC70 served as a loading control. Quantification is shown adjacent. C, doxIRE1␣ cells were pretreated with U0126 (5 M) to inhibit ERK1/2 phosphorylation, followed by doxycycline treatment for 24 h. Cells were lysed and assessed by immunoblotting for IRE1␣, fibronectin, collagen I, phosphorylated and total ERK1/2, and HSC70 as a loading control. Quantification is shown adjacent. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01). n Ն 3 biological replicates for each experiment. Error bars, S.E.

TGF␤ induces expression and phosphorylation of C/EBP␤ in an IRE1␣-dependent mechanism
The role of IRE1␣ signaling in regulating expression of TGF␤ target genes led us to explore the mechanisms responsible for this effect. The ENCODE database showed binding sites for the UPR-associated transcription factor C/EBP␤ along several gene promoters, including procollagen 1␣1 and fibronectin. C/EBP␤ is a b-ZIP domain-containing protein transcribed from a single-exon gene as three different isoforms: liver-activating protein 1 and 2 (LAP1 and LAP2), which are transcriptional co-activators, and liver inhibitory protein (LIP), which acts as a co-repressor (25,26). To establish a role for C/EBP␤ in HSC activation, C/EBP␤ protein levels were examined in response to TGF␤ and found to be up-regulated; however, this up-regulation was disrupted by loss of collagen I (Fig. 3, A and B, quantification below). We also observed translocation of C/EBP␤ into the nucleus after TGF␤ treatment (4 h) by immunofluorescence (Fig. 3C). We next assessed whether TGF␤-induced C/EBP␤ expression required IRE1␣ signaling. Pretreatment with 48C attenuated TGF␤-induction of C/EBP␤ protein expression (Fig. 4A, quantification adjacent). Phosphorylation of C/EBP␤ at site Thr-235 (corresponding to Thr-223 on LAP2 and Thr-37 on LIP) is known to mediate its activation; thus, we also assessed whether C/EBP␤ phosphorylation was modulated by a TGF␤/IRE1␣ mechanism (27). Indeed, C/EBP␤ phosphorylation increased with TGF␤ treatment but was attenuated in the presence of 48C. Due to the increase in both total expression and phosphorylation of C/EBP␤ in response to TGF␤, the ratio of phosphorylated/total C/EBP␤ was unchanged. We further assessed C/EBP␤ expression and phosphorylation in the doxIRE1␣ cells. Increased expression of C/EBP␤ (total LAP1/2 and LIP) was observed with overexpression of WT, K599A, or K907A IRE1␣; however, C/EBP␤ phosphorylation (p-LAP1/2 or p-LIP) failed to be induced with the K599A mutant, which suggests dependence on an IRE1␣-dependent kinase cascade ( Fig. 4B, quantification adjacent). Similarly, both the ASK1 and JNK inhibitors blocked C/EBP␤ phosphorylation downstream of IRE1␣ expression (Fig. 4C, quantification adjacent). Interestingly, whereas doxycycline treatment did not alter the ratio of phosphorylated/total C/EBP␤ in doxIRE1␣ WT or K907A cells, phosphorylated/total C/EBP␤ was significantly reduced following doxycycline treatment in doxIRE1␣ K599A cells or with inhibition of ASK1 or JNK. Thus, ASK1/JNK signaling promotes C/EBP␤ activity downstream of IRE1␣. Together, these data suggest that IRE1␣ signaling promotes TGF␤-mediated HSC activation through up-regulation of both C/EBP␤ protein levels and phosphorylation.

Loss of C/EBP␤ disrupts TGF␤-induced HSC activation
To determine whether C/EBP␤ is critical for HSC activation, LX-2 cells were infected with a lentivirus expressing an shRNA against C/EBP␤ (sh-C/EBP␤), or an NT shRNA. Two clones were selected that exhibited reduced expression of the activating and inhibitory isoforms of C/EBP␤, LAP1/LAP2 (69 and 68% reduction) and LIP (82 and 78% reduction) (Fig. 5A, quantification below). Reduced C/EBP␤ expression attenuated the effects of TGF␤ on fibronectin and collagen I protein levels . TGF␤ promotes C/EBP␤ expression. A, LX-2 cells were treated with 5 ng/ml TGF␤ for 0, 1, 2, 4, or 24 h, harvested, and assessed by immunoblotting for expression of C/EBP␤ isoforms (LAP1/2 and LIP). HSC70 served as a loading control. Quantification is shown below. B, shCollagen 1␣1 or NT cells were treated with TGF␤ for 0, 1, 2, 4, or 24 h, and cell lysates were analyzed by immunoblotting for collagen I or C/EBP␤ isoforms. HSC70 served as a loading control. Quantification is shown below. C, LX-2 cells were treated with 5 ng/ml TGF␤ for 4 h, fixed, permeabilized, and stained for C/EBP␤ (green) and DAPI (blue). Representative images are shown. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01). n Ն 3 biological replicates. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
compared with controls in both clones ( Fig. 5 (B and C), quantification below). Loss of C/EBP␤ expression also attenuated fibronectin, procollagen 1␣1, procollagen 1␣2, and ␣SMA mRNA expression in response to TGF␤, indicative of a crucial role for C/EBP␤ in TGF␤-mediated gene transcription (Fig. 5D, clone 1). To determine whether C/EBP␤ acted downstream of IRE1␣, doxIRE1␣ cells were stably infected with the shRNA targeting C/EBP␤ and treated with doxycycline. Despite induction of IRE1␣, the associated up-regulation of fibronectin, collagen I, and ␣SMA was lost in doxIRE1␣ sh-C/EBP␤ cells (Fig.  5E, quantification adjacent). Together, these data highlight a crucial role for C/EBP␤ in mediating mRNA and protein expression of ECM proteins and ␣SMA downstream of TGF␤ and IRE1␣.

Adefovir dipivoxil inhibits C/EBP␤ expression and blocks HSC activation in vitro and in vivo
The crucial role of C/EBP␤ in HSC activation led us to ask whether C/EBP␤ could serve as an antifibrotic target. Recently, the hepatitis B antiviral drug adefovir dipivoxil was shown to antagonize C/EBP␤ expression in vitro (28). Indeed, adefovir dipivoxil treatment (10 M) of LX-2 cells attenuated TGF␤

. C/EBP␤ expression and phosphorylation is increased by TGF␤ in an IRE1␣-dependent manner.
A, LX-2 cells were pretreated for 1 h with 15 M 48C, followed by TGF␤ (5 ng/ml) treatment for 0, 1, 2, 4, or 24 h. Cell lysates were harvested, and protein expression or phosphorylation of C/EBP␤ isoforms on Thr-235 (corresponding to Thr-223 on LAP2 and Thr-37 on LIP) was examined by immunoblotting. HSC70 served as a loading control. Quantification is shown in adjacent graphs. B, doxIRE␣ WT, K599A, or K907A cells were treated with doxycycline (5 g/ml) or vehicle for 24 h. Cells were lysed, and C/EBP␤ phosphorylation, total C/EBP␤, and IRE1␣ expression were analyzed by immunoblotting. HSC70 served as a loading control. Quantification is shown in adjacent graphs. C, doxIRE1␣ cells were treated with GS-444217 (2 M) or SP600125 (10 M) for 1 h to inhibit ASK1 or JNK, respectively, followed by doxycycline treatment for 24 h. Cells were lysed, and phosphorylation of C/EBP␤ isoforms and total C/EBP␤ expression, as well as IRE1␣ expression, was analyzed. HSC70 served as a loading control. Quantification is shown in adjacent graphs. Statistics for A and C were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). For B, paired t tests were performed. n Ն 3 biological replicates for each experiment. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
induction of C/EBP␤ in response to TGF␤ and blocked the effects of TGF␤ on fibronectin and procollagen 1␣1 at the mRNA and protein level (Fig. 6 (A and B), quantification adjacent) despite having no effect on TGF␤ induction of SMAD2/3 phosphorylation (Fig. 6C, quantification below). The conserved SMAD2/3 phosphorylation indicates that adefovir dipivoxil is acting downstream of SMAD2/3. We also observed a significant increase in PPAR␥ mRNA expression. Adefovir dipivoxil similarly reduced collagen I and fibronectin protein levels in isolated rat primary HSCs in response to TGF␤ or stiffnessinduced activation (Fig. 6 (D and E), quantification below).
We next tested adefovir dipivoxil as an antifibrotic agent in vivo, injecting 10-week-old, sex-matched C57Bl/6J mice for 6 weeks with CCl 4 or vehicle (2 days/week) and 10 mg/kg adefovir dipivoxil (or vehicle) on the remaining 5 days/week. Livers were then harvested and assessed for fibrosis. Mice receiving adefovir dipivoxil in conjunction with CCl 4 exhibited significantly less collagen deposition as assessed by sirius red staining (Fig.  7A, quantified below). Hydroxyproline analysis showed a significant increase in collagen content in CCl 4 -treated mice compared with controls, but no significant difference between Adefovir-CCl4 and Adefovir-olive oil groups (Fig. 7B). Furthermore, adefovir dipivoxil reduced CCl 4 -induced protein expression of fibronectin and ␣SMA (Fig. 7C, quantification adjacent) and gene expression of procollagen 1␣1, TIMP1, ␣SMA, and PDGFR␣ (Fig. 7D). Finally, immunofluorescence was performed on liver sections and showed reduced ␣SMA, desmin (a marker of HSCs), and collagen I in adefovir-CCl 4 -treated mice Figure 5. C/EBP␤ is critical for HSC activation in response to TGF␤ or IRE1␣ signaling. A, LX-2 cells were infected with a lentivirus expressing an shRNA against C/EBP␤ (sh-C/EBP␤) or an NT control. Two clonal cell lines were engineered and were assessed by immunoblotting for expression of C/EBP␤ isoforms (LAP1/2 and LIP). HSC70 served as a loading control. Quantification is shown below. B and C, sh-C/EBP␤ cells (clones 1 and 2) or NT cells were treated with TGF␤ (5 ng/ml) for 24 h. Cell lysates were harvested and assessed by immunoblotting for collagen I, FN, C/EBP␤, and HSC70 (loading control). Quantification is shown below. D, sh-C/EBP␤ or NT cells were treated with TGF␤ (5 ng/ml), after which mRNA was harvested and qPCR was performed to assess fibronectin, procollagen 1␣1 and 1␣2, or ␣SMA expression. E, DoxIRE1␣ cells were infected with lentivirus expressing shRNA targeting C/EBP␤ (sh-C/EBP␤) or an NT control, and clonal cell populations were selected. Cells were treated with doxycycline for 24 h, after which cell lysates were harvested and assessed by immunoblotting for IRE1␣, fibronectin, collagen I, ␣SMA, C/EBP␤, and HSC70 (loading control). Quantification is shown adjacent. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). n Ն 3 biological replicates for each experiment. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
compared with CCl 4 alone (Fig. 7E). Assessment of C/EBP␤ protein levels in whole-liver tissue revealed that LIP increased in response to CCl 4 but was significantly decreased with adefovir dipivoxil treatment (Fig. 7F, quantification below). Alternatively, LAP1/2 protein levels were unaffected. We also analyzed IRE1␣ phosphorylation in whole liver and observed increased phosphorylation of IRE1␣ with CCl 4 treatment, whereas treatment with adefovir dipivoxil blocked this effect (Fig. 7F). Finally, analysis of the data revealed no differential effect of CCl 4 or adefovir dipivoxil between male and female mice (data not shown). Thus, adefovir dipivoxil reduces HSC activation in vitro and fibrogenesis in vivo.

C/EBP␤ mediates HSC activation through a mechanism involving p300
With a link between C/EBP␤ and HSC activation identified, we questioned the mechanism of C/EBP␤ regulation of profibrotic genes. C/EBP␤ canonically acts as a homo-or heterodimer to mediate transcription. Analysis of the ENCODE database revealed C/EBP␤-binding sites in close proximity to sites bound by the transcriptional regulator p300 along the promoters of TGF␤-responsive genes, including procollagen 1␣1 and fibronectin. p300 has been implicated in HSC activation and is a known C/EBP␤-binding partner (29,30). In vivo analysis of p300 protein levels showed increased p300 in mice treated with CCl 4 , but this effect was blocked in mice receiving both CCl 4 and adefovir dipivoxil (Fig. 8A, quantification below). This prompted us to study p300 in both TGF␤-and UPR-induced HSC activation. LX-2 cells were infected with a lentivirus expressing an shRNA that targets p300 (sh-p300). sh-p300 or NT cells were treated with TGF␤ or tunicamycin (Tm; 1 g/ml), a potent inducer of ER stress, for 24 h or vehicle alone. TGF␤ or Tm treatment increased protein levels of fibronectin, collagen I, and ␣SMA, but p300 knockdown attenuated this effect (Fig. 8, B and C). Additionally, p300 knockdown in doxIRE1␣ cells (doxIRE1␣-sh-p300) attenuated IRE1␣-mediated up-regulation of fibronectin, collagen I, and ␣SMA protein levels (Fig. 8D, quantification below).
The similar effects of p300 and C/EBP␤ knockdown on TGF␤-and IRE1␣-mediated HSC activation suggested that these proteins may act together. Previous studies showed that p300 binding to C/EBP␤ led to transcriptional activation in a mechanism that involved C/EBP␤ phosphorylation at Thr-235. Thus, we asked whether p300 is responsible for the effects of C/EBP␤ during TGF␤-mediated HSC activation (29). To this Figure 6. Adefovir dipivoxil limits HSC activation through decreasing C/EBP␤ expression. A and B, LX-2 cells were pretreated with 10 M adefovir dipivoxil (ADV) for 1 h, followed by TGF␤ (5 ng/ml) for 24 h. Cell lysates or mRNA were harvested and assessed by immunoblotting (A) for FN, collagen I, C/EBP␤ isoforms LAP1/2 and LIP, and HSC70 (loading control) or qPCR (B) for fibronectin, procollagen 1␣1 and 1␣2, or PPAR␥ expression. Quantification for A is shown in adjacent graphs. C, cells were treated with TGF␤ in the presence of adefovir dipivoxil or vehicle for 0, 1, 2, 4, or 24 h. Cell lysates were harvested and assessed for phosphorylated SMAD3 or total SMAD2/3. HSC70 served as a loading control. D, primary HSCs were harvested from rats and pretreated with adefovir dipivoxil for 1 h, followed by either vehicle or TGF␤ for 24 h. Lysates were harvested and analyzed by immunoblotting for fibronectin or collagen I. HSC70 served as a loading control. Quantification is shown in the graphs below. E, primary HSCs were harvested from rats, followed by adefovir dipivoxil treatment on days 2, 4, and 6 post-isolation. Cell lysates were harvested on day 7 and analyzed by immunoblotting for fibronectin, collagen I expression, or HSC70 (loading control). Quantification is shown in adjacent graphs. Statistics were performed using ANOVA followed by Tukey post hoc analysis for A-D and paired t test for E (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). n Ն 3 biological replicates for each experiment. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
end, we utilized helenalin acetate, a sesquiterpene lactone reported to disrupt the interaction between C/EBP␤ and p300 (31). Co-treatment of LX-2 cells with TGF␤ and helenalin acetate (1 M) disrupted TGF␤ induction of fibronectin, procollagen 1␣1, procollagen 1␣2, and ␣SMA mRNA levels, as well as increased PPAR␥ mRNA (Fig. 9A). Helenalin acetate also attenuated TGF␤ induction of fibronectin and collagen I protein levels, despite no effect on TGF␤-mediated SMAD2/3 phosphorylation (Fig. 9 (B and C), quantification below). Helenalin acetate similarly reduced TGF␤ induction of collagen I and fibronectin in primary HSCs isolated from rats (Fig. 9D, quantification adjacent). Together, these data show that p300 is critical for mediating HSC activation downstream of TGF␤-IRE1␣ signaling and may act through a mechanism involving C/EBP␤.

Discussion
UPR signaling plays an important role during HSC activation, although the mechanisms that govern the relationship between HSC activation, UPR signaling, and fibrogenesis are not fully understood. Here, we make four key findings that provide mechanistic insight into how IRE1␣ drives fibrogenesis in response to TGF␤ (Fig 10). We show that TGF␤ induces the UPR and IRE1␣ signaling through a SMAD2/3-collagen I-dependent mechanism. Next, we identify C/EBP␤ as a crucial mediator of HSC activation that is regulated by IRE1␣ in an Figure 7. Adefovir dipivoxil-treated mice display reduced fibrogenesis. Age-and sex-matched C57Bl/6 mice (n ϭ 7-8 mice/group) were treated with CCl 4 or olive oil twice a week for 6 weeks, in conjunction with injections of 10 mg/kg adefovir dipivoxil or 0.05 M citric acid (vehicle) 5 times a week during the same time period. Whole liver was harvested and analyzed for fibrosis. A, liver sections underwent sirius red staining, and quantification was performed using ImageJ (shown below). B, hydroxyproline analysis for collagen content. C, whole liver lysates were harvested and immunoblotted for fibronectin, ␣SMA, and HSC70 (loading control). Quantification is shown adjacent. D, qPCR was performed on mRNA harvested from whole liver to analyze expression of procollagen 1␣1, TIMP1, ␣SMA, and PDGFR␣. E, immunofluorescence was performed on frozen liver sections to assess ␣SMA, desmin (a marker of HSCs), and collagen I (green). DAPI was used as a nuclear stain (blue). Representative images are shown. Staining was quantified using ImageJ and is shown in the adjacent graphs. F, whole-liver lysates were harvested and immunoblotted for phospho-IRE1␣, total IRE1␣, C/EBP␤, and HSC70 (loading control). Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
ASK1/JNK-dependent mechanism. Furthermore, C/EBP␤ can be pharmacologically targeted by adefovir dipivoxil to block HSC activation in vitro and fibrogenesis in vivo. Finally, we show that p300 is a crucial mediator of HSC activation downstream of the UPR, which may act together with C/EBP␤ in this role. Together, these data map out a UPR-dependent feed-forward mechanism that drives fibrogenesis through C/EBP␤-p300.
UPR signaling in activated HSCs is thought to accommodate increased protein translation, trafficking, and secretion. Chemical induction of the UPR can also drive fibrogenesis independently of canonical activation signals. The relationship between canonical HSC activation mechanisms and the UPR, however, is still unclear. We show here that IRE1␣ signaling through its kinase domain is crucial for HSC activation. IRE1␣ kinase activity leads to a signaling cascade involving ASK1 and JNK, which are critical for phosphorylation of C/EBP␤. Phosphorylation of the Thr-235/Thr-223/Thr-37 on C/EBP␤ is associated with increased transcriptional activity and is a known target of the MAPK family of kinases, consistent with our data (27). This role of IRE1␣ is interesting for a few reasons. First, IRE1␣ signaling has been implicated in HSC fibrogenic gene expression through both its endonuclease domain and through p38 signaling; however, we find that neither of these mechanisms are critical for HSC activation in our inducible IRE1␣ system or necessary for C/EBP␤ phosphorylation. p38 activation was previously associated with SMAD3 phosphorylation and collagen I expression in response to chemically induced ER stress (11), whereas activation of the transcription factor XBP1 downstream of IRE1␣ endonuclease activity is associated with ER dilation and up-regulation of genes involved in protein trafficking and secretion (5,6,8). Furthermore, we show that overexpression of either the K907A or K599A IRE1␣ mutant still led to up-regulation of total C/EBP␤ expression, implicating a kinase-and endonuclease-independent mechanism of C/EBP␤ regulation. Thus, IRE1␣ acts through several mechanisms to contribute to HSC activation. The role of ASK1-JNK signaling in this process is also interesting, as a separate ASK1 inhibitor, selonsertib, is associated with reduced collagen content and liver stiffness when used in a clinical trial to treat patients with non-alcoholic Figure 8. p300 is critical for TGF␤-or UPR-mediated HSC activation and may involve a C/EBP␤-dependent mechanism. A, whole-liver lysates harvested from mice treated with olive oil, CCl 4 , adefovir dipivoxil, or CCl 4 ϩ adefovir dipivoxil and assessed by immunoblotting for protein levels of p300. HSC70 served as a loading control. B and C, LX-2 cells were infected with a lentivirus encoding an shRNA targeting p300 (sh-p300) or NT control and were selected to yield a clonal cell population. NT or sh-p300 cells were treated with 5 ng/ml TGF␤ (B) or 1 g/ml Tm (C) for 24 h, and cell lysates were harvested and analyzed by immunoblotting for p300, fibronectin, collagen I, ␣SMA, and HSC70 (loading control). Quantification is adjacent. D, doxIRE1␣ cells were stably infected with a lentivirus encoding an shRNA targeting p300 (doxIRE1␣ sh-p300) or NT, treated with doxycycline, and assessed by immunoblotting for IRE1␣, p300, fibronectin, collagen I, ␣SMA, and HSC70 (loading control). Quantification is adjacent. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). n Ն 3 biological replicates for each experiment. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
steatohepatitis (32,33). Our work suggests that this reduction may be due to inhibition of C/EBP␤ phosphorylation, subsequently limiting HSC activation. Finally, the loss of procollagen and fibronectin expression in the absence of IRE1␣ signaling provides insight into HSC activation mechanisms. We propose here that initial SMAD2/3-dependent up-regulation of ECM proteins leads to UPR signaling and that the UPR is critical for driving and maintaining the fibrogenic phenotype, thus placing procollagen I as both a cause and a consequence of UPR signaling during HSC activation. Alternatively, UPR induction prior to increased protein load has been observed in immune cells (34). Further assessment of UPR induction in response to different activating stimuli may provide insight into the role of the UPR in HSCs.
Our data indicating that C/EBP␤ is critical for HSC activation in a UPR-dependent manner provide some context for the regulation of C/EBP␤ and ECM production in HSCs. Previous studies show that C/EBP␤ can positively or negatively regulate procollagen 1␣1 expression, and these effects are dependent on the stimulus, isoform, and coregulators (35)(36)(37)(38). C/EBP␤ transcription is unique, in that three isoforms are transcribed from the same gene. The activating isoforms of C/EBP␤ (LAP1 and LAP2) bind DNA as well as interact with several transcriptional co-activators, whereas the inhibitory LIP isoform serves as a dominant negative. Thus, the ratio of LAP/LIP greatly influences transcriptional programming. This is evident during the UPR, where LAP1/LAP2 are typically upregulated during early adaptive UPR, and LIP is up-regulated later and associated with proapoptotic UPR signaling (39). LIP overexpression also inhibits collagen I expression (38). Thus, targeting C/EBP␤ or a specific isoform, such as the full-length LAP1, which can bind to p300, could reduce the activated HSC population in vivo (31).
UPR-mediated phosphorylation of C/EBP␤ may influence recruitment of transcriptional coactivators during HSC activation. Mutation of mouse C/EBP␤ at Thr-188 to alanine (corresponding to Thr-235 in human C/EBP␤) blocks C/EBP␤ induction of the c-fos promoter as well as p300 recruitment to the same site (40,41). A similar mechanism may be responsible during HSC activation, as loss of Thr-235 phosphorylation is associated with attenuated ECM production in response to IRE1␣ overexpression, and helenalin acetate also limits the effects of TGF␤. The mechanisms and the transcriptional complexes involved in TGF␤-and IRE1␣-mediated gene transcription through C/EBP␤ merit further study, as both C/EBP␤ and p300 can interact with several transcription factors, including SMADs. Of further interest is the requirement for p300 in UPRinduced HSC activation. There is limited evidence of p300 Figure 9. Helenalin acetate limits TGF␤-induced HSC activation. LX-2 cells were pretreated with 1 M helenalin acetate (HA) for 1 h, followed by TGF␤ treatment for 24 h. mRNA (A) or cell lysates (B and C) were harvested and analyzed by qPCR for fibronectin, procollagen 1␣1 and 1␣2, ␣SMA, or PPAR␥ or by immunoblotting for fibronectin, collagen I, p-SMAD3, total SMAD2/3, and HSC70 (loading control). Quantification for B and C is shown below the blots. D, primary rat HSCs (rHSCs) were pretreated with helenalin acetate for 1 h, followed by TGF␤ for 48 h. Cells were lysed and assessed for protein expression of fibronectin, collagen I, and ␣SMA. HSC70 served as a loading control. Quantification is shown in adjacent graphs. Statistics were performed using ANOVA followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). n Ն 3 biological replicates for each experiment. Error bars, S.E.

C/EBP␤ mediates HSC activation downstream of the UPR
involvement downstream of the UPR, aside from acetylation of the transcription factor XBP1 (34). p300 may complex with additional UPR-responsive transcription factors to regulate transcriptional activity.
Whereas we have focused here on UPR signaling through IRE1␣, signaling through the other two UPR sensors, PERK and ATF6␣, also likely plays a role in HSC activation (4). In vivo studies show a positive relationship between PERK signaling and HSC activation, and subsequent studies showed that PERK increases SMAD2 expression through a mechanism involving phosphorylation of heterogeneous nuclear riboprotein A1 and inhibiting the degradation of SMAD2 mRNA (9). Additionally, C/EBP␤ is canonically downstream of the PERK pathway and thus may provide mechanisms for further regulation of C/EBP␤ in HSCs (39). Less is known regarding ATF6␣ in HSCs. Inhibition of ATF6␣ reduced HSC activation in response to chemically induced ER stress, but no direct mechanisms for ATF6␣ signaling during HSC activation have been identified (11). Further studies are needed to fully understand the integration of IRE1␣, PERK, and ATF6␣ signaling during HSC activation and how crosstalk between activation signals (such as TGF␤) and UPR signaling influences HSC activation and survival.
Adefovir dipivoxil was identified as a pharmacological inhibitor of C/EBP␤ during a screen for compounds that preferentially targeted the LIP isoform of C/EBP␤. In the drug screen, adefovir dipivoxil reduced the ratio of LIP/LAP but also impacted expression of the activating isoforms, prompting us to test whether adefovir dipivoxil could limit HSC activa-tion through targeting C/EBP␤. Adefovir dipivoxil effectively reduced protein levels of all C/EBP␤ isoforms in LX-2 cells and attenuated TGF␤-stimulated HSC activation, as well as fibrogenesis in vivo. Interestingly, LIP, but not LAP1/2, was significantly reduced in whole liver in vivo. These data suggest a crucial and targetable role for C/EBP␤, and the ratio of LIP to LAP1/2 in fibrosis, further supported by previous findings where mesenchymal C/EBP␤ deletion attenuated pulmonary fibrosis (27). Based on the known role of C/EBP␤ in both proliferation and apoptosis, we anticipate that adefovir dipivoxil may also limit HSC proliferation or promote HSC apoptosis, both of which are favorable for fibrosis regression and resolution (42)(43)(44)(45). Finally, whereas we anticipate that C/EBP␤ is one of the major targets of adefovir dipivoxil in HSCs, off-target effects may contribute to the antifibrotic nature of the drug. Another acyclic nucleotide analog used in hepatitis B patients, tenofovir, was associated with reduced sirius red staining in a thioacetamide model of liver injury, but its effects on C/EBP␤ or UPR signaling were not assessed (46).
In conclusion, we show that TGF␤ induces UPR signaling in HSCs, which acts in a feed-forward mechanism through IRE1␣ to promote fibrotic gene expression through C/EBP␤-p300.

Primary HSC isolation and infection
HSCs were harvested from IRE1␣ fl/fl mice (21,47,48) or adult female Lewis rats (Charles River Laboratories) (49) as described previously. HSCs isolated from IRE1␣ fl/fl mice were infected with adenovirus encoding for Ad-Cre-eGFP to induce gene recombination or with an Ad-LacZ control, as described previously (50). All procedures were reviewed and approved by the Mayo Clinic Institutional Animal Care and Use Committee.

In vivo treatment and analysis
Eight-week-old C57Bl/6J mice were purchased from Envigo. Following a 1-week acclimation period, age-and sex-matched mice were separated into four treatment groups: olive oil ϩ vehicle (n ϭ 7), CCl 4 ϩ vehicle (n ϭ 8), olive oil ϩ adefovir dipivoxil (n ϭ 8), and CCl 4 ϩ adefovir dipivoxil (n ϭ 8). CCl 4 (0.5 l/mg, 0.08 ml/mouse) was administered by intraperitoneal injection twice a week for 6 weeks. Each group consisted of four males and four females, which were housed together according to their treatment group. Mice were housed in transparent polycarbonate cages subjected to 12-h light/dark cycles under a temperature of 21°C and a relative humidity of 50%. A standard chow diet and freshwater were provided ad libitum. Adefovir dipivoxil (Sigma-Aldrich, SML0240) was dissolved in sterile in citric acid (0.05 M, pH 2.0) and administered at a concentration of 10 mg/kg/day 5 days a week for 6 weeks. Mice were observed daily for distress and tolerated the experimental protocol without adverse effects, aside from one mouse unexpectedly reaching end point criteria (olive oil ϩ vehicle group). After 6 weeks, mice were sacrificed, and livers were harvested for analysis of fibrosis. Assessment of collagen content by sirius red staining or hydroxyproline was performed as described previously (7). For analysis of gene expression, mRNA was harvested using an RNeasy kit (Qiagen), followed by reverse transcription into cDNA and assessment by qPCR. For protein assessment, livers were homogenized in lysis buffer as described previously and analyzed by Western blotting (7). For immunofluorescence, liver samples were snap-frozen in OCT embedding compound, and 7-m frozen sections were obtained using a Leica cryostat. Sections were fixed with 4% paraformaldehyde, followed by incubation with primary antibodies against collagen I, desmin, and ␣SMA. Anti-goat IgG and antirabbit IgG conjugated with Alexa fluorochromes were used as secondary antibodies (Molecular Probes, Inc., Eugene, OR). DAPI was used at 1:2000 for counterstaining. Microscopy was performed by Axio Observer (Carl Zeiss, Thornwood, NY), in which appropriate light and filter combinations were selected according to excitation and emission spectrum features of the Alexa fluorochromes. All procedures were reviewed and approved by the Mayo Clinic Institutional Animal Care and Use Committee.

Immunoblotting
Following the indicated treatment, cells were lysed in a modified radioimmune precipitation buffer and spun down to remove cell debris (7). Lysates were denatured using 6ϫ sample buffer (Boston Bioproducts BP-111R), resolved using SDS-PAGE, transferred to nitrocellulose membrane, and blocked in 5% BSA. Membranes were incubated in the indicated primary antibody, followed by the appropriate secondary antibody, and developed. All experiments were performed in triplicate biological replicates at a minimum, with quantification and statistics performed using ImageJ and PRISM, respectively. HSC70 or GAPDH were used as loading control for all immunoblots.

qPCR analysis
Following treatment, cells were harvested using an RNeasy kit (Qiagen). Equal amounts of mRNA were reversetranscribed, and qPCR was performed. Human GAPDH or mouse 18S primers were used for normalization in experiments using human cells or mouse liver lysates, respectively. Primer sequences used for qPCR analysis are listed in Table 1.

Statistical analysis
For all experiments with three or more groups, statistics were performed using analysis of variance (ANOVA) followed by Tukey post hoc analysis (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001). All statistical analysis comparing two groups utilized paired t tests. n Ն 3 biological replicates for each experiment.