Myofibroblasts revert to an inactive phenotype during regression of liver fibrosis

Regression of Liver Fibrosis Is Accompanied by Loss of Myofibroblasts.

Our study was designed to determine the fate of aHSCs/myofibroblasts (α-SMA⁺ColI⁺ cells) during regression of hepatic fibrosis. For this purpose, reporter Col-GFP mice, expressing collagen-α1(I) promoter/enhancer-driven GFP, were subjected to CCl₄-induced liver injury for 2 mo. After cessation of the toxic agent, mice recuperated for 1 or 4 mo, and regression of liver fibrosis was evaluated by measuring collagen deposition and myofibroblast number (Fig. 1 A and B). CCl₄-treated mice developed severe fibrosis with activated myofibroblasts (Fig. 1 A and B), which decreased markedly after 1 and 4 mo of recovery. After 1 mo recovery, hydroxyproline levels and expression of the fibrogenic genes collagen-α1(I) and α-SMA were significantly decreased, compared with CCl₄-treated mice (7.8 ± 1.2% Col-GFP and 8 ± 1.5% α-SMA; Fig. 1B), confirming that CCl₄-activated myofibroblasts disappear during recovery from liver fibrosis. Thus, because Col-GFP mice undergo regression of liver fibrosis after 1 mo of recovery, it is appropriate to study the fate of aHSCs/myofibroblasts for this time period.

Hepatic Stellate Cells Are the Major Source of CCl₄-Activated Myofibroblasts.

The contribution of aHSCs to liver myofibroblasts in CCl₄-treated Col-GFP mice was determined using flow cytometry of the isolated nonparenchymal liver cell fraction, which contains aHSCs/myofibroblasts, inflammatory cells, and endothelial cells (8). Myofibroblasts were identified by Col-GFP expression, and HSCs were identified by vitamin A expression (1, 4, 8) (detected at 405 nm as an autofluorescent signal quenched by a violet laser) (Fig. 1C; SI Appendix, Fig. S1). GFP⁺ cells (92 ± 3%) coexpressed vitamin A, demonstrating that aHSCs represent the major population of fibrogenic myofibroblasts in CCl₄-injured liver, as predicted by previous qualitative studies (9). Therefore, aHSCs can be genetically labeled on the basis of specific up-regulation of type I collagen expression (SI Appendix, Fig. S2) in CCl₄-induced liver fibrosis because other cellular sources do not make a significant contribution to the myofibroblast population.

Some aHSCs Apoptose During Regression of Liver Fibrosis.

We hypothesize that the disappearance of aHSCs/myofibroblasts during regression of liver fibrosis may result from cell death by senescence (3) and apoptosis (2), inactivation, or both (Fig. 1D). Apoptosis of HSCs during regression of liver fibrosis is well documented (2). In agreement, we detected apoptotic aHSCs/myofibroblasts (2.6 ± 0.7%) by colocalization of cleavable caspase-3⁺ and GFP⁺ cells in the livers of Col-GFP mice 7 d after CCl₄ cessation, when apoptosis of hepatic cells was highest (SI Appendix, Fig. S3). Overall, early (7 d) recovery from liver fibrosis is accompanied by apoptosis of some aHSCs/myofibroblasts.

Genetically Labeled aHSCs/Myofibroblasts Persist in the Liver After 1 mo of Recovery from CCl₄.

To determine if some liver myofibroblasts survive the regression of fibrosis, Col-α2(I)^Cre-YFP mice [collagen-α2(I)^Cre × Rosa26^{flox-Stop-flox-YFP} mice; SI Appendix, Fig. S2] were treated with CCl₄ (2 mo), allowed to recover (1 mo), and then analyzed for the persistence of genetically labeled YFP⁺ cells (Fig. 2A). HSCs were identified by expression of GFAP and Desmin, and aHSCs/myofibroblasts were detected by expression of α-SMA. HSCs (98 ± 2%) were activated (e.g., upregulated YFP) in response to CCl₄ treatment, and YFP expression was detected in 94 ± 4% of myofibroblasts (α-SMA⁺). Although myofibroblasts had completely disappeared in livers after 1 mo of recovery, YFP⁺ cells surprisingly persisted. In particular, expression of YFP was detected in 38 ± 8% of Desmin⁺ and 41 ± 5% of GFAP⁺ cells, consistent with being HSCs that had been previously activated (Fig. 2A).

The immunohistochemistry (Fig. 2A) and flow cytometry (Fig. 2B) of gradient purified HSCs from Col-α2(I)^YFP mice identified three HSC phenotypes: (i) qHSCs (vitamin A⁺ YFP⁻ α-SMA⁻), (ii) aHSCs (vitamin A⁺ YFP⁺ α-SMA⁺), and (iii) inactivated HSCs [(iHSCs), vitamin A⁺ YFP⁺ α-SMA⁻]. After recovery from fibrosis, 56 ± 4% of HSCs coexpressed YFP⁺ and vitamin A⁺, indicating that these iHSCs had a history of type I expression but reverted to an inactivated phenotype (Fig. 2B).

Collagen-α2(I) and -α1(I) form a triple helix to produce collagen type I and are coexpressed in aHSCs/myofibroblasts (10). To provide independent confirmation of the above findings, we used Col-α1(1)^Cre-YFP mice, generated by crossing collagen-α1(I)^Cre mice (SI Appendix, Fig. S4A) with Rosa26^{flox-Stop-flox-YFP} mice. As expected, CCl₄ treatment of Col-α1(1)^Cre-YFP mice produced aHSCs (Desmin⁺YFP⁺α-SMA⁺ cells; SI Appendix, Fig. S4 B and C). Although α-SMA⁺ myofibroblasts were no longer detected in livers after 1 mo recovery, 37 ± 9% of Desmin⁺ HSCs still expressed YFP. In fact, genetically labeled YFP⁺ HSCs persisted after 4 mo recovery (SI Appendix, Fig. S4D). Similarly, flow cytometry demonstrated that 38 ± 7% of YFP⁺ vitamin A⁺ HSCs expressed YFP after 1 mo recovery, compared with 83 ± 6% of YFP⁺ vitamin A⁺ aHSCs in fibrotic liver (SI Appendix, Fig. S4E). In the recovered liver, these iHSCs resided in the peri-sinusoidal space of Disse and exhibited a stellate shape (SI Appendix, Fig. S4F).

HSCs Transiently Express Collagen Type I During Development.

Detection of YFP⁺ qHSCs in Col-α2(1)^Cre-YFP and Col-α1(1)^Cre-YFP in adult livers before injury (Fig. 2B; SI Appendix, Figs. S4E and S5) may reflect transient collagen gene expression activating Cre during development. To prove this hypothesis, expression of collagen-α1(I) in real time was examined in livers of Col-GFP mice during embryogenesis (SI Appendix, Fig. S5). Indeed, transient expression of collagen-α1(I)-GFP was detectable in HSCs, identified by vitamin A, Desmin, and GFAP expression, between embryonic day 16.5 (E16.5) and postnatal day 14 (P14) (SI Appendix, Fig. S6A). At P14, 46 ± 8% of HSCs up-regulated collagen-α1(I)-GFP in real time but lacked α-SMA expression (SI Appendix, Fig. S6B). These GFP⁺ HSCs did not exhibit characteristics of myofibroblasts (SI Appendix, Fig. S6 C and D), but were more similar to qHSCs than to aHSCs (SI Appendix, Fig. S6 E and F).

The fate of embryonic collagen⁺ HSCs was examined in adult Col-α2(1)^Cre-YFP mice (8 wk old). Consistent with our findings, YFP⁺ qHSCs with a history of collagen expression and YFP⁻ qHSCs had identical gene expression profiles characteristic of a quiescent phenotype (SI Appendix, Fig. S6E).

Tamoxifen-Induced Genetic Labeling of aHSCs/Myofibroblasts in Adult Mice Confirmed Their Persistence in the Liver After 1 mo of Recovery from CCl₄.

Tamoxifen-inducible Col-α2(1)^ER-Cre-GFP mice were generated by crossing Col-α2(1)^ER-Cre mice × Rosa26^{flox-mTRed-Stop-flox-mGFP} mice (SI Appendix, Fig. S2). Genetic labeling of HSCs was achieved in adult CCl₄-treated Col-α2(1)^Er-Cre-GFP mice by daily tamoxifen administration during the last week of CCl₄ treatment (SI Appendix, Fig. S7A). Genetically labeled aHSCs were visualized by loss of mTRed expression and gain of GFP expression upon Cre-loxP recombination. Desmin⁺ HSCs (35 ± 6%) expressed GFP after CCl₄, and 14 ± 4% of HSCs were still GFP⁺ after 1 mo recovery (Fig. 2C), confirming that CCl₄-activated HSCs (and their progeny) persist in the liver after regression of fibrosis. Consistently, GFP⁺ iHSCs expressed Desmin, but not α-SMA (SI Appendix, Fig. S7B). Thus, three independent transgenic mice demonstrated that aHSCs/myofibroblasts revert to an inactive phenotype during regression of fibrosis.

Livers Recovering from Fibrosis Have Fewer HSCs.

To quantify the number of HSCs during fibrosis and its regression, we generated GFAP^Cre-GFP mice (GFAP^Cre mice × Rosa26^{flox-Stop-mTRed-flox-mGFP} mice; SI Appendix, Fig. S8A). In uninjured mice, qHSCs were distributed throughout the hepatic acinus and represented 10.6 ± 0.8% of total liver cells. CCl₄ induced HSC activation, proliferation (14.3 ± 1.5% of total liver cells), and accumulation of aHSCs in the pericentral area. One month after recovery, the number of HSCs was reduced (5.6 ± 1.8% of total liver cells), and the distribution of HSCs was again similar to qHSCs. On the basis of immunostaining for GFAP after recovery from fibrosis in Col-α2(I)^Cre-YFP and Col-α1(I)^Cre-YFP mice (Fig. 2B; SI Appendix, Fig. S4E), iHSCs constitute 2% of total liver cells in the recovered liver (SI Appendix, Fig. S8B).

Genetically Labeled aHSCs/Myofibroblasts Persist in the Liver After 7 wk of Recovery from Alcohol-Induced Liver Fibrosis.

We next determined if survival of aHSCs/myofibroblasts occurs during regression of alcohol-induced liver fibrosis. Liver fibrosis (and steatosis) was induced in Col-α2(I)^Cre-YFP mice (collagen-α2(I)^Cre × Rosa26^{flox-Stop-flox-YFP} mice) by intragastric alcohol feeding for 2 mo (SI Appendix, Fig. S9 A and B). Liver fibrosis (and steatosis) regressed in these mice 7 wk after withdrawal from ethanol feeding. Flow cytometry demonstrated that genetic labeling (YFP⁺) was achieved in 64 ± 5% of myofibroblasts and persisted in 36 ± 4% of vitamin A⁺ YFP⁺ HSCs upon recovery from fibrosis (SI Appendix, Fig. S9C). Our findings were confirmed by immunohistochemistry (SI Appendix, Fig. S9 D and E). YFP expression persisted in 38 ± 7% of Desmin⁺ HSCs/myofibroblasts following regression of liver fibrosis after withdrawal from ethanol despite the disappearance of myofibroblasts (α-SMA expressed in 1.4 ± 1% of YFP⁺ HSCs/myofibroblasts; SI Appendix, Fig. S9D). Thus, two models of regression of liver fibrosis demonstrate survival of iHSCs.

iHSCs Demonstrate an Increased Response to Repeated Fibrogenic Stimuli.

Purified iHSCs had a phenotype similar to that of qHSCs (Desmin⁺, GFAP⁺, Synemin⁺, α-SMA⁻; SI Appendix, Figs. S9E and S10). However, expression of myofibroblast-specific genes [Col-α1(I), α-SMA, TIMP-1] was induced more strongly in cultured TGF-β1–treated iHSCs than in qHSCs (Fig. 3A). In concordance, Col-GFP mice subjected to two rounds of CCl₄ injury separated by a 6-mo interval to allow complete recovery (2 × CCl₄) developed more severe fibrosis than littermates treated with one round of CCl₄ (1 × CCl₄; Fig. 3B). Thus, our in culture and in vivo data indicate that iHSCs with a history of activation are more effectively activated than qHSCs.

Adoptively Transferred HSCs (1-mo Recovery), but Not qHSCs, Contribute to Liver Fibrosis in Mice.

To test this hypothesis, HSCs were isolated from Col-GFP⁺/β-actin-RFP⁺ mice that were uninjured and after 7 d and 1 mo recovery from CCl₄-induced fibrosis and adoptively transferred into livers of the newborn Rag2^−/−γc^−/− mice (11) (Fig. 3C). One month later, these Rag2^−/−γc^−/− mice were subjected to CCl₄ injury, and fibrotic livers were analyzed for the presence of GFP⁺RFP⁺ HSCs. Highest engraftment (70–78%) was achieved in mice transplanted with HSCs after 7 d or 1 mo recovery (versus 50% for qHSCs; SI Appendix, Fig. S11A). Unlike qHSCs, which were mostly scattered under the capsule or in liver parenchyma and constituted only 0.5 ± 0.2% of total HSCs, HSCs from the recovering livers were incorporated into the myofibroblast population in recipient mice and contributed 19 ± 2.3% and 13 ± 2.0% of total HSCs, respectively (Fig. 3C). Moreover, despite poor engraftment, comparable results were observed in CCl₄-treated wild-type mice adoptively transferred with qHSCs or HSCs (2 wk recovery) from Col-α1(I)^Cre-YFP mice (SI Appendix, Fig. S11B). Taken together, iHSCs are primed to differentiate into myofibroblasts more rapidly in response to recurrent stimuli.

Inactivated HSCs Gradually Down-Regulate Collagen-α1(I).

To further characterize iHSCs, Col-α1(1)^Cre-YFP mice were crossed with Col-GFP mice, and genetically labeled HSCs (YFP⁺) were analyzed for expression of collagen-α1(I) in real time (GFP⁺; SI Appendix, Fig. S12). Following CCl₄ treatment (2 mo, SI Appendix, Fig. S12A), all YFP⁺ HSCs expressed GFP. After 1 mo recovery from fibrosis, YFP⁺ HSCs had decreased GFP expression. Similar results were obtained by flow cytometry (SI Appendix, Fig. S12 B and C), which allowed simultaneous detection of vitamin A, YFP, and GFP expression (12) in isolated HSCs. As expected, qHSCs lacked GFP expression and HSCs expressed GFP in response to CCl₄ (87 ± 5%; SI Appendix, Fig. S12B). Following a 2-wk recovery from CCl₄, decreased GFP expression was observed in 75 ± 3% of HSCs, of which 92 ± 4% still expressed YFP. The mean fluorescent intensity (mfi) of GFP expression was strongly reduced in YFP⁺ HSCs at this time (∼4 × 10³ mfi, compared with aHSCs of ∼6 × 10⁴ mfi; SI Appendix, Fig. S12B). GFP expression (∼1 × 10³ mfi) decreased further in 42 ± 4% of HSCs after 1 mo recovery and correlated with the number of YFP⁺ iHSCs (55 ± 3%). Thus, inactivation of HSCs occurs gradually and steadily during recovery from CCl₄-induced fibrosis. Interestingly, 45% of HSCs after 1 mo recovery had no history of collagen expression (YFP⁻) and represent new qHSCs (SI Appendix, Fig. S12B).

iHSCs Acquire a New Phenotype Distinct from qHSCs.

To assess changes in global gene expression, inactivated YFP⁺ HSCs (iHSCs, 1 mo recovery) were evaluated by the whole-mouse genome microarray and compared with qHSCs, aHSCs, and HSCs after 7 d recovery (Fig. 4A). We confirmed that YFP⁺ iHSCs down-regulated fibrogenic genes (Col-1α1, Col-1α2, Col-1α1, α-SMA, TGFβRI, and TIMP1) during recovery from fibrosis, but failed to obtain a quiescent phenotype [up-regulated peroxisome proliferator-activated receptor γ (PPARγ) and Bambi, but not the other quiescence-associated genes adipose differentiation relate protein (Adfp), Adipor1, or GFAP] (5) (Fig. 4B). Unsupervised clustering of gene expression profiles revealed that YFP⁺ iHSCs (1 mo) exhibit an intermediate profile between that of qHSCs and YFP⁺ HSCs (7 d recovery), but share more similarity to qHSCs than to aHSCs (Fig. 4 C and D). Similar results were obtained using correlation coefficient analysis comparing expression profiles to qHSCs (Fig. 4C) and unsupervised clustering of gene-specific expression profiles (SI Appendix, Fig. S13 A and C).

Activation of Heat-Shock Proteins 1a/b May Promote Survival of iHSCs at Day 7 of Recovery from Liver Fibrosis.

To understand how YFP⁺ iHSCs escape apoptosis, we examined the signaling pathways in YFP⁺ HSCs after 7 d recovery (SI Appendix, Fig. S13 B and E). In particular, expression of the anti-apoptotic heat-shock proteins 1a/b (Hspa1a/b) genes was strongly but transiently induced in these HSCs (Fig. 5A; SI Appendix, Fig. S14A) to the levels comparable to qHSCs, but was dramatically down-regulated in aHSCs and HSCs after 1 mo recovery (Fig. 5A; SI Appendix, Fig. S14A).

We examined if Hspa1a/b would impact survival of cultured HSCs. For this purpose, HSCs were isolated from CCl₄-treated Hspa1a/b^−/− and wild-type mice (SI Appendix, Fig. S14B) and cultured 5 d on plastic. Hspa1a/b^−/− HSCs had a rounded shape and exhibited growth retardation (cell-number ratio knock out:wt: 1:1.7; SI Appendix, Fig. S14C). Moreover, Hspa1a/b^−/− HSCs were more susceptible to glyotoxin- (13) and TNF-α–induced apoptosis (14) (Fig. 5B; SI Appendix, Fig. S14C). Therefore, up-regulation of Hspa1a/b genes may promote survival of iHSCs during recovery from fibrosis.

Resolution of CCl₄-Induced Fibrosis Is Expedited in Hspa1a/b^−/− Mice.

We hypothesized that the loss of survival signals in Hspa1a/b^−/− HSCs would result in increased clearance of aHSCs after recovery from CCl₄-induced fibrosis. To test this, Hspa1a/b^−/− and wild-type mice were subjected to CCl₄-induced liver injury. As expected, Hspa1a/b^−/− mice developed more severe fibrosis (probably due to increased hepatocyte death) (15) than did the wild-type littermates (Fig. 5C). However, after stopping CCl₄ treatment, regression of liver fibrosis was strongly accelerated in Hspa1a/b^−/− mice compared with wild-type mice (decreased 49 vs. 20%, respectively, by Sirius red staining). Hspa1a/b^−/− livers also had a greater loss of α-SMA⁺Desmin⁺ aHSCs compared with wild-type mice (decreased 68 vs. 40% of Desmin⁺ positive area, respectively; Fig. 5C). Thus, Hspa1a/b is required so that iHSCs persist in the recovering liver.

Mice.

Expression of collagen type I in real time was studied using reporter Col-GFP mice (24). Cell fate mapping of aHSCs was studied using collagen-α2(I)^Cre (25) and collagen-α1(I)^Cre (SI Appendix) and tamoxifen-inducible collagen-α2(I)^ER-Cre crossed to Rosa26^{flox-Stop-flox-YFP} mice (or Rosa26^{flox-mTRed-Stop-flox-mGFP} mice) (Jackson Labs). GFAP^Cre mice were used to determine the total number of HSCs.

Liver Fibrosis.

Liver fibrosis was induced in mice by intragastric gavage with CCl₄ (at 16 × 1:4 dilution in 100 μL of corn oil) over 2 mo (8) or by intragastric ethanol feeding combined with Western diet (for 2 mo) (SI Appendix) (26). Reversal of liver fibrosis was studied 1 mo after CCl₄ cessation and 7 wk after withdrawal from alcohol feeding. Recurrent injury in Col-GFP mice was induced for 1 mo with CCl₄ (8 × 1:4). Liver injury in Rag^−/−γc^−/− and Hspa1a/b^−/− mice was gradually induced with CCl₄ (4 × 1:16; 2 × 1:8; 2 × 1:4) for 1 mo. Collagen content was measured by hydroxyproline and Sirius Red staining.

Isolation of Nonparenchymal Cell Fraction and Primary HSCs.

Livers are perfused and digested using pronase/collagenase and the gradient centrifugation method, as previously described (8). Freshly isolated HSCs were analyzed by flow cytometry or cultured in DMEM (Gibco-BRL) + 10% FCS and 2 mM l-glutamine + antibiotics.

Flow Cytometry.

Phenotyping of the hepatic nonparenchymal cells was performed on FACSCanto II RUO (BD). Activated myofibroblasts were visualized by GFP expression (488 nm) using argon laser, and vitamin A⁺ expression (405 nm) was detected by violet laser. Cell sorting was performed on a MoFlo (Beckman Coulter) for GFP (488 nm, using Lyt-200S laser, iCYP Visionary Bioscience Inc.) and vitamin A (350 nm, UV laser, JDSU-Excyte).

Whole-Mouse Genome Gene Expression Microarray.

The gene expression profile of HSCs was studied using Whole Mouse Genome Microarray (Agilent). Vitamin A⁺YFP⁺ and vitamin A⁺YFP⁻ HSCs were purified by cell sorting from collagen-α2(I)^Cre-YFP mice with no injury, from mice after CCl₄ treatment (2 mo), and from mice after 7 d or 1 mo recovery from CCl₄. mRNA was purified using RNAeasy columns (Qiagen), and 160 ng of purified RNA per sample was labeled using the LRILAK PLUS and hybridized to a Whole Mouse Genome Microarray 4 × 44,000 60-mer slide (Agilent). Slides were analyzed using the Gene Spring Software (Agilent) as described in SI Appendix. Quantitative RT-PCR and apoptosis of aHSCs are described in SI Appendix.