Cross-talk between Notch and Hedgehog regulates hepatic stellate cell fate in mice


  • Potential conflict of interest: Nothing to report.


Liver repair involves phenotypic changes in hepatic stellate cells (HSCs) and reactivation of morphogenic signaling pathways that modulate epithelial-to-mesenchymal/mesenchymal-to-epithelial transitions, such as Notch and Hedgehog (Hh). Hh stimulates HSCs to become myofibroblasts (MFs). Recent lineage tracing studies in adult mice with injured livers showed that some MFs became multipotent progenitors to regenerate hepatocytes, cholangiocytes, and HSCs. We studied primary HSC cultures and two different animal models of fibrosis to evaluate the hypothesis that activating the Notch pathway in HSCs stimulates them to become (and remain) MFs through a mechanism that involves an epithelial-to-mesenchymal–like transition and requires cross-talk with the canonical Hh pathway. We found that when cultured HSCs transitioned into MFs, they activated Hh signaling, underwent an epithelial-to-mesenchymal–like transition, and increased Notch signaling. Blocking Notch signaling in MFs/HSCs suppressed Hh activity and caused a mesenchymal-to-epithelial–like transition. Inhibiting the Hh pathway suppressed Notch signaling and also induced a mesenchymal-to-epithelial–like transition. Manipulating Hh and Notch signaling in a mouse multipotent progenitor cell line evoked similar responses. In mice, liver injury increased Notch activity in MFs and Hh-responsive MF progeny (i.e., HSCs and ductular cells). Conditionally disrupting Hh signaling in MFs of bile-duct–ligated mice inhibited Notch signaling and blocked accumulation of both MF and ductular cells. Conclusions: The Notch and Hedgehog pathways interact to control the fate of key cell types involved in adult liver repair by modulating epithelial-to-mesenchymal–like/mesenchymal-to-epithelial–like transitions. (Hepatology 2013;58:1801–1813)






bile duct ligation


N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester


dimethyl sulfoxide


double transgenic


fluorescence-activated cell sorting


glial fibrillary acidic protein




hairy and enhancer of split


hairy/enhancer-of-split related with YRPW


high-fat diet




hepatocyte nuclear factor


hepatic stellate cells








messenger RNAs


octamer-binding transcription factor 4


peroxisome proliferator-activated receptor gamma






quantitative reverse-transcriptase polymerase chain reaction


Sonic Hedgehog




alpha smooth muscle actin


SRY (sex determining region Y)-box 9


transforming growth factor beta




wild type.

The outcome of liver injury is dictated by the efficiency of repair responses that replace damaged liver tissue with healthy hepatic parenchyma. Defective repair of chronic liver injury can result in cirrhosis, a scarring condition characterized by dramatic changes in the cellular composition of the liver. Outgrowth of progenitors and myofibroblasts (MFs) is particularly prominent during scarring.[1] Because these cell types are critical for successful regeneration of damaged livers,[1, 2] their accumulation in cirrhotic liver suggests that scarring may occur because regenerative mechanisms become stalled prematurely. Therefore, to restore healthy wound healing, it is necessary to characterize and prioritize the key signals that regulate the fate of cells that are required for liver repair.

Reconstruction of damaged adult liver utilizes several highly conserved signaling pathways that orchestrate organogenesis during fetal development, including Wnt, Hedgehog (Hh), and Notch.[3] During embryogenesis, these pathways interact to modulate survival, proliferation, and differentiation of their target cells so that developing organs become appropriately populated with all of the cell types necessary for tissue-specific functions. For example, cross-talk between Hh and Notch controls the fate of embryonic stem cells,[4] zebrafish neural progenitors,[5] and Drosophila eye precursors.[6] In cancer biology, the importance of cell-autonomous cross-talk between Hh and Notch is also emerging. Overexpression of both the Notch- and Hh-signaling pathways occurs in a subpopulation of chemotherapy-resistant cancer stem cells, and targeting Notch and Hh depleted this population.[7] However, whether similar cross-talk occurs when damaged adult livers are regenerated, which cell types are involved, and whether or not such signaling becomes deregulated during defective repair, is not well understood. Also uncertain is if and how these newly uncovered pathways in the damaged adult liver fit into the classical paradigms for cirrhosis pathogenesis, and whether they are more or less important for that process than well-established regulators of adult liver growth, such as transforming growth factor beta (TGF-β), which is generally credited for driving defective liver repair in adults.[1]

Therfore, the aims of this study were to investigate if and how Notch signaling regulates damage-related outgrowth of liver MFs. We focused on MF derived from HSCs because adult HSCs are TGF-β-responsive cells that are also influenced by developmental morphogenic pathways, such as Wnt and Hh, which reactivate during adult liver repair. Adult HSCs require Hh signaling to become and remain MFs.[8] Recent lineage tracing studies in adult mice with injured livers demonstrated that some MFs became multipotent progenitors that regenerated hepatocytes, cholangiocytes, and HSCs. In parallel experiments, Cre recombinase-mediated knockdown of canonical Hh signaling in cells expressing the MF gene, alpha smooth muscle actin (α-SMA), both blocked MF accumulation and inhibited outgrowth of ductular cells during cholestatic liver injury.[9] Both autocrine and paracrine signaling regulated by the Hh pathway might be involved. For example, Sonic hedgehog ligand is known to promote the transcription of Jagged-1,[10] and MF-derived Jagged-1 is thought to work in a paracrine fashion to promote ductular differentiation of Notch-responsive liver progenitors.[2] Previous work suggested that HSCs themselves may also be capable of Notch signaling.[11] Most recently, Chen et al. reported that N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT), a γ-secretase inhibitor that blocks Notch signaling, decreased expression of various MF genes in a rat HSC line (HSC-T6).[12] They also found that DAPT inhibited CCl4-related fibrosis in rats and showed that this was accompanied by reduced hepatic expression of TGF-β, Snail, and various mesenchymal genes, but up-regulation of E-cadherin, suggesting that blocking Notch promoted mesenchymal-to-epithelial transitions.[13] However, an earlier study of cultured HSCs correlated induction of Notch-1 and Hes1 with suppression of α-SMA expression and proliferation, and showed that knocking down expression of Notch-1 enhanced HSC growth.[14]

Indeed, the effects of Notch on MF differentiation and growth are complex and appear to vary according to the type of MF precursor. Notch signaling inhibits myofibroblastic differentiation of myoblast precursors and some types of fibroblasts.[15, 16] In contrast, it enhances MF differentiation of lung MF precursors,[17] airway epithelial cells,[18] and dermal fibroblasts.[19] Activating Notch also promotes epithelial-to-mesenchymal transition in kidney cells,[20] stimulates expansion of cardiac progenitors at the expense of MFs,[21] and promotes an epithelial-to-mesenchymal transition process that enhances the stem-like properties of cancer stem cells.[22]

Notch signaling is critical for biliary morphogenesis during development.[23-25] As mentioned earlier, the fate of adult liver progenitors is also directed by Notch: Increasing Notch signaling promotes differentiation along the biliary lineage, whereas suppressing the Notch pathway shifts progenitors toward an hepatocytic fate.[2] Deregulated Notch signaling has been implicated in the pathogenesis of hepatocellular carcinoma and cholangiocarcinoma.[26, 27] Despite growing evidence for Notch pathway involvement in liver cancer and fibrosis, it is unclear how Notch interfaces with other key signaling pathways that have been implicated in those disorders, or how Notch signaling in one type of liver cell (e.g., MFs) might influence the accumulation of other types of liver cells (e.g., epithelial progenitors) that are required for adult liver repair.

In this study, we evaluated the hypothesis that Notch pathway activation in HSCs stimulates them to become (and remain) MFs through a mechanism that involves an epithelial-to-mesenchymal–like transition requiring cross-talk with canonical (i.e., TGF-β-independent) Hedgehog signaling.

Materials and Methods

Full methods are available in the Supporting Information.


Male C57BL/6 mice and Smotm2Amc/J (Smoothened [Smo]/flox) mice were obtained from The Jackson Laboratory (Bar Harbor, ME).[28] Smo/flox mice were crossed with α-SMA-Cre-ERT2 transgenic mice[29] to generate double-transgenic (DTG) mice in which treatment with tamoxifen induces conditional deletion of Smo in α-SMA-positive cells.[9] Mice (8-12 weeks old) were subjected to bile duct ligation (BDL) or sham surgery for 14 days. Other 8-10-week-old wild-type (WT) mice were fed with a high-fat diet (HFD) and given intraperitoneal injection of either vehicle (olive oil) or CCl4 (1 μL/g body weight, prediluted 1:3 in olive oil) twice per week for 2 weeks and sacrificed 72 hours after last CCl4 injection.[30] Animal experiments fulfilled National Institutes of Health (Bethesda, MD) and Duke University Institutional Animal Care and Use Committee (Durham, NC) requirements for humane animal care.


Formalin-fixed, paraffin-embedded livers were prepared for immunohistochemistry (IHC).[9] Protocols and antibodies used are listed in the Supporting Information.

Molecular Techniques

Quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) and immunoblottings were performed as previously described.[31]

Cell Isolation

Primary HSCs were isolated from C57BL/6 mice using standard approaches. Purity of the preparations was rigorously analyzed as previously described.[9]

Pharmacological Manipulation of Notch and Hh Signaling

Day 4 primary HSC cultures were treated with the γ-secretase inhibitor, DAPT (10 µM; Sigma-Aldrich, St Louis, MO), or the Smoothened agonist, GDC-0449 (1 µM; Selleck Chemicals, Houston, TX), for 3 days. Controls were treated with dimethyl sulfoxide (DMSO). 603B cells were treated the same way for 2 days.

Statistical Analysis

Results are expressed as mean ± standard error of the mean. Analyses were performed using the Student t test. P < 0.05 was considered significant.


Activation of Notch Signaling in Desmin-Expressing Cells During Hepatic Injury

We found up-regulation of messenger RNAs (mRNAs) for Notch-2, Jagged-1, and several Notch-target genes (Hes1, Hey1, Hey2, and HeyL) in a mouse BDL model (Fig. 1A), consistent with previous reports that adult liver injury activates Notch signaling.[2, 23] In addition to ductal cells (known Notch targets),[23] stromal cells expressed Notch-2, Jagged-1, and Hey2 post-BDL (Fig. 1B and Supporting Fig. 1A). Some of these stromal cells costained with the HSC marker, Desmin, suggesting that activated Notch signaling occurs in MFs/HSCs during liver injury. Quantitative IHC indicated that approximately 60% of the Desmin(+) cells coexpressed Notch-2 and/or Jagged-1 and 30% coexpressed Hey2. These findings were confirmed with fluorescence-activated cell sorting (FACS) analysis of HSCs isolated from BDL mice, which showed increased Notch-2, Jagged-1, and Hey2, compared to HSCs harvested from sham controls (Fig. 1C and Supporting Fig. 1B).

Figure 1.

Liver injuries increase Notch signaling in Desmin-expressing stromal cells. (A) qRT-PCR analysis of total liver mRNA from WT mice 14 days after sham or BDL surgery for expression of Notch-pathway genes. *P < 0.05 versus sham; n = 4. (B) Double IHC for Notch-2, Jagged-1, or Hey2 (brown) with Desmin (green) in BDL mouse livers demonstrates colocalization (inset) of these markers. Percentages of double-positive cells among Desmin(+) cells were also quantified in 10 randomly selected fields. *P < 0.001 versus sham; n = 3 mice/group. (C) FACS analysis of HSCs isolated from WT mice 14 days after sham or BDL surgery for expression of α-SMA, Notch-2, Jagged-1, or Hey2. Desmin was used as a marker for HSCs. (D) qRT-PCR analysis of total mRNA from livers of WT mice treated for 14 days with HFD/CCl4. *P < 0.05; **P < 0.01 versus HFD controls; n = 3. (E) Double IHC for Notch-2, Jagged-1, or Hey2 (brown) and Desmin (green) in HFD/CCl4 mouse livers demonstrates colocalization (inset) of these markers. Magnification, ×40. Percentages of double-positive cells among Desmin(+) cells were also quantified in 10 randomly selected fields. *P < 0.001 versus HF Ctrl; n = 3 mice/group.

We also examined mice treated with HFD ± CCl4 for 2 weeks to provoke liver sinusoidal fibrosis. Compared to HFD-fed controls, mice treated with HFD/CCl4 demonstrated increased mRNA expression of Notch-2, Jagged-1, Hes1, Hey1, and Hey2, as well a ductular marker, keratin (Krt)19 (Fig. 1D). As noted in BDL mice with portal-based fibrosis (Fig. 1B,C), quantitative IHC also demonstrated increased Notch-2, Jagged-1, and Hey2 expression in Desmin-positive cells of mice with CCl4-induced sinusoidal fibrosis (Fig. 1E and Supporting Fig. 1C).

Up-Regulation of Notch Signaling During HSC Activation In Vitro

Although it is established that cholangiocytes and their precursors are capable of Notch signaling,[24, 25, 27] it is uncertain whether primary HSCs and/or their progeny (e.g., MFs/HSCs) respond to Notch. Because IHC and FACS revealed Notch signaling components in Desmin-expressing cells that accumulate in fibrotic livers (Fig. 1B,C,E), we evaluated the expression of Notch-pathway genes in primary mouse HSCs (both freshly isolated HSCs and 7-day, culture-activated MFs/HSCs; Fig. 2A,B). Results in HSCs were compared to those in a mouse ductular cell line (603B), which served as a positive control for Notch signaling (Fig. 3). FACS showed that 603B cells express the cholangiocyte marker, Krt19, progenitor markers (SRY [sex determining region Y]-box 9 [Sox9], FN14, and CD24), and Notch pathway components (Notch-2 and Jagged-1) at very high levels, confirming that such cells are immature ductular-type cells with Notch-signaling capability (Fig. 3A). FACS similarly revealed that HSCs express proteins that regulate Notch signaling, including the Notch ligand, Jagged-1, Notch-1, and Notch-2 receptors, and Numb, a Notch-signaling repressor (Fig. 2A and Supporting Fig. 2A). qRT-PCR analysis readily demonstrated mRNA for these factors (Fig. 2B), whereas expression of another Notch ligand (Jagged-2) and other Notch receptors (Notch-3 and Notch-4) was detected at much lower levels (Supporting Fig. 2B).

Figure 2.

Notch signaling is activated during transdifferentiation of primary HSCs. (A) FACS analysis of quiescent (freshly isolated, day 0) and myofibroblastic (culture day 7) HSCs. Desmin and α-SMA were used as markers for quiescent or myofibroblastic HSCs, respectively. (B) qRT-PCR analysis of Notch inhibitor (Numb), receptors (Notch-1 and Notch-2), ligand (Jagged-1), and target genes (Hes1, Hey1, Hey2, and c-Myc) in quiescent and myofibroblastic HSCs. Results were compared to gene expression in ductular progenitor cells (603B). *P < 0.05; **P < 0.01; ***P < 0.001; n = 3.

Figure 3.

Notch-responsive liver progenitors (603B) coexpress ductular, hepatocytic, HSC, and mesenchymal markers. (A) FACS analysis confirmed that 603B cells are mouse ductular progenitors with active Notch signaling. Gray lines indicate isotype controls. (B) FACS analysis of 603B cells demonstrated expression of other ductular markers (Krt7 and HNF-6), but also hepatocytic markers (HNF-4α, AFP, and ALB), Hh-signaling factors/target genes (Ptc, Gli1, and Gli2), mesenchymal markers (Vimentin and α-SMA), and HSC-associated markers (Desmin and GFAP). (C) Comparison of gene expression in 603B cells with primary mouse hepatocytes (mHep) and freshly isolated or culture-activated primary mouse HSCs (d0 mHSC and d7 mHSC, respectively) by qRT-PCR analysis; n = 3/group. #Nondetectable signal.

Compared to freshly isolated (day 0) HSCs, which were relatively enriched with cells expressing Notch-1 and Numb proteins, MFs/HSCs demonstrated much lower expression of Notch-1 and Numb, but much higher expression of Jagged-1 and Notch-2 (Fig. 2A and Supporting Fig. 2A), consistent with a previous report showing decreased Notch-1 expression during rat HSC culture activation.[11] Thus, expression of proteins regulating Notch signaling changed substantially during MF transdifferentiation. To determine whether pathway activity also changed as quiescent (Q)-HSCs transitioned into MFs/HSCs, qRT-PCR analysis was performed to assess the expression of various Notch target genes (Hes1, Hey1, Hey2, and c-Myc; Fig. 2B). Hey2 and c-Myc mRNA expression increased significantly during HSC activation. This induction of Notch target genes occurred in conjunction with up-regulation of Jagged-1 and Notch-2 mRNAs and coincided with down-regulation of mRNAs for Notch-1 and Numb. The results suggest that HSCs activate Notch signaling as they become MFs. This possibility is supported by evidence that several Notch target gene (Hes1, Hey1, and Hey2) mRNA levels in HSCs are generally equal to or higher than their levels in ductular-type cells with acknowledged Notch-signaling capability (Fig. 2B).

Phenotypic and Genotypic Similarities in Notch-Responsive Liver Cells

Notch regulates the fate of bipotent liver epithelial progenitors,[2, 25] and lineage-tracing evidence in adult mice indicates that bipotent liver epithelial progenitors and HSCs derive from a common multipotent progenitor that is controlled by the Hh pathway.[9, 32] Thus, it is conceivable that Notch interacts with Hh to direct the differentiation of adult progenitors during liver injury. We began to examine this issue by further characterizing 603B cells by FACS (Fig. 3A,B) and using qRT-PCR to compare gene expression in 603B cells, mature liver cells (primary mouse hepatocytes), and freshly isolated or culture-activated primary HSCs (Fig. 3C).

FACS showed that although 97%-99% of 603B cells express well-accepted markers of ductular progenitors (Krt19, Krt7, and Sox9), only approximately one third express the biliary-associated transcription factor, HNF6. Hepatocyte nuclear factor (HNF)−4α, a hepatocyte-associated transcription factor, is evident in ∼50%, suggesting that 603B cells are capable of differentiating along both biliary and hepatocytic lineages. Consistent with that concept, virtually all of the cells (97%-99%) express established markers of hepatoblasts (a.k.a. oval cells), such as CD24, FN14, and albumin (ALB). More than 80% of 603B cells also express a putative HSC marker, glial fibrillary acidic protein (GFAP), suggesting that 603B cells may be multipotent (i.e., capable of differentiating into hepatocytes, cholangioctyes, and HSCs). Indeed, approximately one third of 603B cells express Desmin and approximately 25% are α-SMA positive. Coexpression of ductular, hepatocytic, and HSC markers occurs in Hh-responsive multipotent liver progenitors that are undergoing epithelial-mesenchymal transitions.[9] Ninety-nine percent of 603B cells coexpress Krt7 (epithelial marker), vimentin (mesenchymal marker), and one or more Hh target genes (Patched [Ptc], glioblastoma [Gli]1, and Gli2), exhibiting the phenotype of multipotent liver progenitors that are in the midst of epithelial-mesenchymal transitions (Fig. 3A,B).

qRT-PCR analysis provided additional evidence that 603B cells are transitioning multipotent liver progenitors. Compared to freshly isolated primary hepatocytes from healthy adult mice, 603B cells express significantly higher mRNA levels of Hh target genes (Ptc and Gli2), cholangiocyte-associated genes (e.g., Krt19 and HNF-6), and HSC-associated genes (e.g., Desmin and GFAP), but significantly lower mRNA levels of HNF-4α, a transcription factor that is strongly expressed by mature hepatocytes. As reported for transitional multipotent progenitors,[9] gene expression in 603B cells is more similar to HSCs than hepatocytes. For example, primary HSCs and 603B cells express comparable mRNA levels of Krt7, HNF-6, alpha-fetoprotein (AFP), Ptc, and Gli2. However, mRNA levels of Desmin and GFAP are significantly lower in 603B cells than freshly isolated HSCs, and this discrepancy is magnified when HSCs undergo culture activation to become MFs (Fig. 3C). Nevertheless, the aggregate data demonstrate genotypic and phenotypic similarities in Notch-responsive liver cells, and indicate that such cells are Hh responsive and inherently plastic (i.e., capable of undergoing epithelial-mesenchymal transitions).

DAPT Inhibits Notch Signaling in Both Progenitors and HSCs In Vitro

To investigate the functional significance of Notch signaling in HSCs, the Notch pathway was suppressed by treating cultured primary MFs/HSCs with a γ-secretase inhibitor (DAPT). Results in HSCs were compared to those in multipotent progenitor cells (603B), which served as a positive control for Notch signaling. As expected, studies in 603B cells showed that DAPT treatment significantly reduced expression of Jagged-1, Notch-2, and Notch target genes (Hes1, Hey1, and Hey2; Fig. 4). Inhibiting Notch signaling in 603B cells suppressed the expression of cholangiocyte-associated genes (Krt7, Krt19, HNF-1β, and HNF-6) and permitted induction of hepatocyte lineage markers (AFP, HNF-1α, and HNF-4α), consistent with previous reports that activation of Notch signaling drives liver progenitors toward the biliary lineage, whereas its suppression promotes differentiation along the hepatocytic lineage.[2, 24, 25] Blocking Notch signaling in 603B enhanced expression of GFAP, a Q-HSC marker, but reduced α-SMA, an MF/HSC marker, and TGF-β, a profibrogenic cytokine that promotes ductular differentiation of liver progenitors in developing embryos.[33] Blocking Notch also down-regulated key Hedgehog target genes (Gli1 and Ptc) in 603B cells. The aggregate findings suggest that Notch signaling interfaces with fibrogenic signals that are transduced by TGF-β and the Hh pathway in multipotent liver progenitor cells. This is particularly intriguing because both TGF-β and Hh signaling promote epithelial-to-mesenchymal transitions in developing embryos,[34] and Hh has been proven to stimulate epithelial-to-mesenchymal–like transitions in both adult HSCs and progenitor cells.[8, 35]

Figure 4.

Inhibiting Notch signaling suppresses Hedgehog signaling and promotes a mesenchymal-to-epithelial–like transition and hepatocytic differentiation in ductular-type progenitor cells. qRT-PCR analysis of 603B cells treated with DAPT (a γ-secretase inhibitor) for 48 hours for changes in (A) Notch-pathway genes, (B) epithelial/quiescence genes, and (C) MF/Hh genes. *P < 0.05 versus DMSO control; n = 3.

Having confirmed that DAPT performed as anticipated in Notch-responsive liver progenitor cells, we evaluated its actions in HSCs. For these studies, primary murine HSCs were cultured for 4 days to induce MF transdifferentiation and then treated with DAPT for an additional 3 days. As in 603B cells (Fig. 4), MFs/HSCs showed DAPT-inhibited expression of Notch-2, Jagged-1, and several Notch target gene (Hey1, Hey2, and HeyL) mRNAs (Fig. 5A). IHC confirmed that mRNA suppression was accompanied by decreased protein expression (Fig. 5E). Blocking Notch signaling in MFs/HSCs also repressed typical MF-associated genes (α-SMA, collagen, and TGF-β) and Hh target genes that are known to be expressed by MFs/HSCs (Gli2, Ptc, and Sonic Hedgehog [Shh]; Fig. 5B). In contrast, mRNA levels of various epithelial genes (bone morphogenic protein-7, desmoplakin, E-cadherin, AFP, HNF-4α, and Krt19) and Q-HSC markers (peroxisome proliferator-activated receptor gamma [PPAR-γ] and GFAP) were up-regulated (Fig. 5C). Immunocytochemistry confirmed the DAPT-induced reversion of MFs/HSCs to a more quiescent phenotype, showing decreased staining for α-SMA and Ki67 (proliferation marker) and increased Oil Red O staining, indicative of neutral lipid accumulation (Fig. 5F). Interestingly, when Notch signaling was inhibited and MFs/HSCs reverted to a more quiescent phenotype, mRNA expression of delta-like 1 homolog, a Notch-related gene that marks liver progenitors,[36] and mRNAs encoding other progenitor cell markers (e.g., Nanog, octamer-binding transcription factor 4 [Oct4], and FN14) were down-regulated (Fig. 5D). Thus, Notch signaling is activated during culture-induced primary MF/HSC transdifferentiation, and this permits the cells to acquire a more mesenchymal phenotype with progenitor-like features. This process parallels activation-associated induction of Hh signaling and might be regulated by cross-talk between the Notch and Hh pathways, because HSCs require Hh signaling to become MFs.[8, 31]

Figure 5.

Notch inhibition suppresses Hedgehog signaling and promotes a mesenchymal-to-epithelial–like transition in primary HSCs. qRT-PCR analysis of primary MFs/HSCs treated with DAPT for 3 days for changes in (A) Notch genes, (B) MF/Hh target genes, (C) epithelial/quiescence genes, and (D) progenitor genes. *P < 0.05 versus DMSO control; n = 3. (E) DAPT-treated MFs/HSCs were stained for cleaved Notch-2, Jagged-1, and Hey2 protein. Scale bar: 150 μM. (F) Effect of DAPT on HSC expression of α-SMA, proliferation (Ki67), and lipid content (Oil Red O) was examined.

Inhibiting Hedgehog Signaling Blocks Notch Signaling In Vitro

To further examine possible cross-talk between Notch and Hh signaling, the two Notch-responsive cell types (603B and primary MFs/HSCs) were treated with an Hh-signaling antagonist (GDC-0449). GDC-0449 directly interacts with and inhibits the Hh coreceptor, Smoothened.[37] Earlier work has proven that GDC-0449 recapitulates the effect of Smoothened gene knockdown in MFs/HSCs, with both approaches inhibiting canonical Hh signaling, thereby blocking the nuclear localization and transcriptional activation of Gli DNA-binding proteins.[31] In both cell types, antagonizing Smoothened caused suppression of Notch-2, Jagged-1, and Notch target genes (Fig. 6A,B), demonstrating that canonical Hh-pathway activity promotes the expression of Notch-signaling pathway genes. Given that DAPT, a γ-secretase inhibitor that specifically blocks Notch signaling, suppressed expression of Shh ligand, Gli2 (Hh-regulated transcription factor), and Ptc (a direct transcriptional target of Gli) (Fig. 5), the Notch pathway seems to stimulate Hh-pathway activity. Hence, the results identify a previously unsuspected Hh-Notch-positive feedback loop that regulates cell-fate decisions in immature ductular-type cells and MFs/HSCs. In certain types of adult liver injury, these two cell types accumulate and intermingle within fibrotic septae that extend outward from portal tracts to cause bridging fibrosis, an antecedent to cirrhosis.[38] This suggests that Notch-Hh interactions might regulate cirrhosis pathogenesis by controlling the fate of two key cell types that are involved in liver repair.

Figure 6.

Blocking Hedgehog signaling in myofibroblastic liver cells inhibits Notch signaling. (A) qRT-PCR analysis of 603B cells treated with an Hh inhibitor (GDC-0449) or DMSO for 48 hours for changes in Hh target genes (Ptc and Gli1), Notch genes (Notch-2, Jagged-1, Hey1, and Hey2), and epithelial genes (AFP and HNF-4α). *P < 0.05 versus DMSO control; n = 3. (B) qRT-PCR analysis of primary MFs/HSCs treated with GDC-0449 for 3 days. *P < 0.05; **P < 0.01 versus DMSO control. (C-E) α-SMA/Cre-ERT2–Smo-flox (DTG) mice were subjected to BDL and treated with vehicle (VEH, olive oil; n = 3) or TMX (n = 4) every other day from days 4 to 10 post-BDL. (C) qRT-PCR analysis of total liver mRNA; *P < 0.05. (D) Representative IHC and quantification for Notch-2 and Hey2. Scale bar: 100 µm. *P < 0.05; **P < 0.01. (E) Double staining of Notch-2 or Hey2 (brown) with Desmin (green) in liver sections described in the legend to Fig. 5D. Percentages of Notch-2/Desmin or Hey2/Desmin double-positive cells among Desmin(+) cells were also quantified. At least 10 fields were counted per mouse. *P < 0.05; n = 3.

Blocking Hh Signaling in MF Inhibits Notch Signaling In Vivo

To verify that Hh signaling regulates Notch signaling in vivo, as observed in vitro, and to evaluate the functional implications of this interaction for liver repair, we used a genetic approach to conditionally delete Smoothened in MFs/HSCs. DTG mice were created by crossing Smoflox/flox mice with α-SMA/Cre-ERT2 mice. Treating such DTG mice with tamoxifen (TMX) induced selective deletion of the floxed Smo gene, but only in α-SMA-expressing cells,[31] providing a useful tool for examining the effects of Hh signaling in MFs/HSCs and their progeny.[9] DTG mice underwent BDL to provoke liver injury and compensatory repair responses. Four days later, treatment with either vehicle or TMX was initiated and given every other day through day 10; mice were sacrificed on day 14 post-BDL for liver tissue analysis. In an earlier study, we showed that this approach knocked down expression of Smo in the liver, reduced the hepatic content of α-SMA(+) cells by >85%, and significantly decreased collagen gene expression, hepatic hydroxyproline content, and Sirius Red staining, as well as accumulation of Krt19(+) ductular cells.[9] In this study, we confirmed that TMX reduced both Smo and α-SMA expression (Fig. 6C), and showed that decreasing Hh-responsive MFs dramatically decreased numbers of Notch-2(+) and Hey2(+) cells, both along liver sinusoids (colocalized with Desmin(+) cells) and in residual ductular structures (Fig. 6D). qRT-PCR analysis of whole-liver RNA demonstrated that loss of Notch-2-expressing cells in TMX-treated DTG mice was accompanied by significantly reduced whole-liver expression of Notch target genes, compared to vehicle-treated controls (Fig. 6C). Immunoblotting analysis of whole-liver lysates confirmed that suppression of Notch signaling was accompanied by the expected loss of proteins that mark ductular-type cells and their progenitors (e.g., Krt19 and HNF-6), with concomitant induction of the hepatocyte-enriched transcription factor, HNF-4α (Supporting Fig. 3C). Interestingly, however, we were unable to detect differences in expression of Jagged-1 mRNA (Fig. 6C) or protein (Supporting Fig. 3A) in our BDL mice, despite significant reductions in α-SMA-expressing cells at the time point we examined. IHC demonstrated colocalization of Jagged-1 in Desmin(+) stromal cells that persisted after Smo deletion, suggesting that unlike culture-activated MFs/HSCs (Fig. 5A), in vivo–activated HSCs maintain Jagged-1 expression for at least a while after they revert from a myofibroblastic state to a more quiescent HSC phenotype. To determine whether or not Jagged-1 is able to activate Notch signaling after Smo knockdown, we tested responses to recombinant Jagged-1 ligand in primary HSCs from Smoflox/flox mice after HSCs were culture activated to MFs and treated with Cre-recombinase adenoviral vectors to delete Smo. Results were compared to Smoflox/flox HSCs treated with control adenoviral vectors (adenovirus encoding green fluorescent protein). Jagged-1 significantly increased expression of Notch 2 and Notch target genes in control HSCs, but had no effect in Smo-depleted HSCs (Supporting Fig. 3B). Thus, the aggregate in vivo and in vitro data suggest that the Hh pathway modulates Notch signaling downstream of Jagged-1 in liver cells, at least in part, by promoting expression of Notch-2. Abrogating canonical Hh signaling prevents Jagged-1 from inducing Notch-2 and is sufficient to cause liver cells to become relatively resistant to Jagged-1, thereby inhibiting Jagged-Notch signaling and blocking induction of Notch target genes. This blocked the outgrowth of both myofibroblastic and ductular cells and reduced fibrosis during cholestatic liver injury (present data and previous work[9]). Given that blocking Notch inhibited Hh in cultured MFs (Fig. 5B), and inhibiting Notch signaling also decreased liver fibrosis in rats treated with CCl4,[13] it seems likely that the Hh and Notch pathways interact to control HSC fate in vivo, as they do in vitro. Future experiments that conditionally disrupt Notch signaling in MFs are needed to resolve that issue.


This study demonstrates, for the first time, that primary HSCs use the Notch-signaling pathway to regulate their transdifferentiation. We found that as HSCs become MFs in culture, they up-regulate their expression of the Notch ligand, Jagged-1, as well as the Notch-2 receptor, while down-regulating their expression of Notch-1 receptor and Numb, a Notch-signaling inhibitor. Our findings in primary mouse HSCs differ somewhat from those that were reported on recently in a T-antigen-transformed rat HSC line, which was shown to express mainly Notch-3.12 However, as was noted in that immortalized rat HSC line, we also found that primary MFs/HSCs reverted to a less myofibroblastic phenotype when treated with DAPT, a specific Notch-signaling inhibitor. Moreover, we showed that inhibiting Notch permitted the primary MFs/HSCs to reacquire markers of Q-HSC (e.g., GFAP and PPAR-γ), reaccumulate lipid, become less proliferative, and express several genes that typify epithelial cells (e.g., E-cadherin and Desmoplakin). Evidence that blocking Notch signaling permits a mesenchymal-to-epithelial–like transition in primary MFs/HSCs is novel, but consistent with the known ability of Notch to promote epithelial-to-mesenchymal transitions.[39] Indeed, we observed that DAPT also decreased Notch signaling and mesenchymal gene expression in an immature ductular cell line (603B) with multipotent liver epithelial progenitor features. During this process, we observed that 603B exhibited not only the expected down-regulation of ductular progenitor markers (e.g., HNF-1β, HNF-6, and Krt19) and reciprocal up-regulation of hepatocytic progenitor markers (e.g., HNF-4α and AFP), but also showed increased expression of the Q-HSC gene, GFAP.

Evidence that a Notch-regulated progenitor for hepatocytes and cholangiocytes can also differentiate into Notch-sensitive cells that express markers of HSCs is consistent with an earlier lineage tracing study in adult mice, which suggested a common lineage for such bipotent liver epithelial progenitors and HSCs,[32] as well as a more recent lineage tracing study, which proved that α-SMA- and GFAP-expressing cells give rise to hepatocytes and ductular cells during adult liver injury.[9] MFs derived from HSCs express several markers of multipotent progenitors, including Oct4.40 Other adult epithelial tissues are known to harbor subpopulations of differentiated (nonstem) cells that are capable of dedifferentiating into stem-like cells41; passage of such nonstem cells through epithelial-to-mesenchymal transitions has been closely connected to their entrance into the stem cell state.[42] These findings have prompted speculation that stem cell compartments in adult tissues might be replenished by contextual signals within the microenvironment that reactivate pluripotency factors, such as Oct4, in subpopulations of mature cells with intrinsic phenotypic plasticity.[41]

During liver injury, the hepatic microenvironment changes dramatically, and factors that are not expressed in healthy adult livers, such as Jagged and Hh ligands, accumulate. Many of the cell types required for liver repair are Hh responsive, including HSCs and bipotent liver progenitors. Activating Hh signaling in such cells globally affects their fate, provoking epithelial-to-mesenchymal–like transitions, stimulating proliferation, and enhancing survival.[43] Here, we demonstrate, for the first time, that Hh interacts with Notch to orchestrate these cell-fate changes in primary HSCs. We showed that blocking Notch signaling with DAPT inhibited expression of Hh target genes, such as Ptc, whereas GDC-0449, a direct antagonist of Smoothened, reduced expression of Notch-2, Hes1, Hey2, and HeyL. MFs/HSCs require cross-talk between the Notch and Hh pathways to retain their myofibroblastic phenotype, because blocking either pathway suppressed expression of typical MF markers (e.g., α-SMA and collagen) while inducing reexpression of quiescent markers (e.g., PPAR-γ and GFAP). Parallel studies in 603B cells confirm that similar Hh-Notch interactions regulate cell-fate decisions in multipotent liver progenitors. In addition, cross-talk with other key repair-related signaling pathways is likely to be involved because we found that DAPT suppressed expression of TGF-β mRNA in both MFs/HSCs and the progenitor cell line, and GDC-0449 has been reported to inhibit TGF-β expression in MFs/HSCs.[44] TGF-β interacts with its receptors to initiate signals that activate Gli-family factors independently of Smoothened,[45] suggesting that Notch-Hh cross-talk might promote activation of other signaling pathways that reenforce their actions on downstream targets.

Therefore, to clarify the ultimate biological relevance of Hh-Notch interactions in adult liver repair, we used a Cre-recombinase-driven approach to target α-SMA-expressing cells and deleted Smoothened to abrogate canonical (i.e., TGFβ-independent) Hh signaling in mice with ongoing cholestatic liver injury induced by BDL. We found that knocking down Hh signaling in MFs significantly inhibited Notch signaling, decreasing whole-liver expression of various Notch target genes by 40%-60%. This inhibited accumulation of cells that express ductular markers, such as Krt19 and HNF-6 (P < 0.05 and 0.005 versus respective vehicle-treated controls). As expected by data generated here and in our earlier work,[9, 31] blocking Hh signaling in MFs significantly decreased accumulation of collagen-producing cells and decreased liver fibrosis post-BDL. However, contrary to our prediction, depletion of MF did not appreciably reduce hepatic expression of Jagged-1. IHC localized Jagged-1 to Desmin(+) stromal cells that persisted after Smo depletion, suggesting that MFs/HSCs that revert to quiescence when Hh signaling is abrogated in vivo retain Jagged-1. However, Hh-deficient cells are relatively resistant to Jagged-Notch signaling, because treating Smo-depleted cells with recombinant Jagged-1 failed to evoke induction of Notch-2 or increase expression of Notch-regulated genes. Given present and published evidence for the inherent plasticity of HSCs and HSC-derived MFs,[40] additional research will be necessary to determine whether the outcomes observed after Smo knockdown in MFs of BDL mice reflect disruption of Hh-Notch interactions that control epithelial-to-mesenchymal–like/mesenchymal-to-epithelial–like transitions in these wound-healing cells. In any case, the new evidence that Hh signaling influences Notch-pathway activity in the injured adult mouse livers complements data that demonstrate mutually reenforcing cross-talk between these two signaling pathways in cultured adult liver cells. Stated another way, both in vitro and in vivo, activating the Hh pathway stimulates Notch signaling, and the latter further enhances profibrogenic Hh signaling. The newly identified positive feedback loop provides a previously unsuspected mechanism that helps to explain why a recent study found that treating rats with a Notch inhibitor reduced CCl4-induced liver fibrosis.[13]

In summary, our latest discoveries complement work by other groups and, together, extend growing evidence that adult liver repair is controlled by reactivated morphogenic signaling pathways that orchestrate organogenesis during development, such as Notch and Hedgehog. These pathways clearly act in concert during adult organ repair and likely coordinate during development as well. In the adult liver, these mechanisms appear to involve modulation of fundamental fate decisions in subpopulations of adult liver cells that retain high levels of inherent plasticity. Although additional research is needed to clarify the nuances of this insight, it has already identified a myriad of novel diagnostic and therapeutic targets that might be exploited to improve outcomes of adult liver injury.