Fibroblasts in liver cancer: functions and therapeutic translation

Abstract

Accumulation of fibroblasts in the premalignant or malignant liver is a characteristic feature of liver cancer, but has not been therapeutically leveraged despite evidence for pathophysiologically relevant roles in tumour growth. Hepatocellular carcinoma is a largely non-desmoplastic tumour, in which fibroblasts accumulate predominantly in the pre-neoplastic fibrotic liver and regulate the risk for hepatocellular carcinoma development through a balance of tumour-suppressive and tumour-promoting mediators. By contrast, cholangiocarcinoma is desmoplastic, with cancer-associated fibroblasts contributing to tumour growth. Accordingly, restoring the balance from tumour-promoting to tumour-suppressive fibroblasts and mediators might represent a strategy for hepatocellular carcinoma prevention, whereas in cholangiocarcinoma, fibroblasts and their mediators could be leveraged for tumour treatment. Importantly, fibroblast mediators regulating hepatocellular carcinoma development might exert opposite effects on cholangiocarcinoma growth. This Review translates the improved understanding of tumour-specific, location-specific, and stage-specific roles of fibroblasts and their mediators in liver cancer into novel and rational therapeutic concepts.

Introduction

Liver cancer is the sixth most common cancer and fourth most common cause of cancer-related death worldwide.¹ In contrast to the decreasing incidence and mortality of most solid tumours, the incidence of liver cancer has tripled in the USA between 1975 and 2005,² and liver cancer-related mortality has increased by 43% between 2000 and 2016.¹ Approximately 80% of primary liver cancers are hepatocellular carcinomas and 15% are cholangiocarcinomas.¹ 80–90% of hepatocellular carcinomas arise in patients with underlying liver fibrosis or cirrhosis. With increasing global rates of obesity and non-alcoholic fatty liver disease (NAFLD), NAFLD is emerging as a leading cause of hepatocellular carcinoma for the coming decades, whereas hepatitis B virus (HBV)-associated tumours are expected to decrease with efficient vaccination and hepatitis C virus (HCV)-associated tumours are expected to decrease with efficient virus eradication regimens.¹ Like hepatocellular carcinoma, cholangiocarcinoma is often associated with chronic liver disease. Anatomically distinct forms of cholangiocarcinoma (ie, intrahepatic cholangiocarcinoma, perihilar cholangiocarcinoma, and distal cholangiocarcinoma) can be associated with different underlying diseases, such as liver fluke infections (all forms of cholangiocarcinoma), primary sclerosing cholangitis, liver cirrhosis, and chronic HBV and HCV infections (primarily intrahepatic cholangiocarcinoma), but also occur frequently without an identifiable underlying cause.^3,4 Similar to hepatocellular carcinoma, incidence and mortality rates of cholangiocarcinoma have increased in most countries, with increased rates of intrahepatic cholangiocarcinoma but stable or decreasing rates of hilar and distal cholangiocarcinoma.^3,4

Despite improvements in medical therapy for hepatocellular carcinoma, 5-year survival rates of 18% worldwide make liver cancer one of the most lethal solid tumours.¹ However, there are large stage-specific differences. Early-stage tumours amenable to liver transplantation have high 5-year and 10-year survival rates and low tumour recurrence,¹ whereas advanced stages display a median survival of less than 2 years even with the best systemic treatments. There is a plethora of therapies, ranging from transplantation and resection, to local therapies such as radiofrequency ablation and transarterial chemoembolisation, to systemic therapies with tyrosine-kinase inhibitors and immune checkpoint inhibitors. The most effective first-line systemic treatment for advanced hepatocellular carcinoma is the combination of PD-L1-blocking antibody atezolizumab and anti-VEGF antibody bevacizumab, which when compared with sorafenib increase both 12-month survival (67·2% vs 54·6%) and median survival (19·2 vs 13·4 months).^5,6 Additional first-line systemic therapies include the recently approved combination of immune checkpoint inhibitors tremelimumab and durvalumab,⁷ as well as monotherapy with tyrosine-kinase inhibitors lenvantinib or sorafenib.¹ Tyrosine-kinase inhibitors such as sorafenib decrease liver fibroblast proliferation, activation, and survival and liver fibrosis via inhibition of PDGFRB, ERK, and AKT,⁸ which might contribute to antitumour action in addition to direct effects on tumour cells. Immune checkpoint inhibitor-based first-line therapies are more efficient than tyrosine-kinase inhibitors, but do not exert known direct effects on fibroblasts. Although the IMbrave050 trial (NCT04102098) showed improved recurrence-free survival for atezolizumab and bevacizumab as adjuvant therapy compared with active surveillance, there is no approved therapy for primary hepatocellular carcinoma prevention besides the treatment of the underlying liver disease. In summary, currently approved therapies for hepatocellular carcinoma do not include specific fibroblast-targeted therapies for hepatocellular carcinoma treatment or prevention.

Cholangiocarcinoma has a dismal prognosis with overall 5-year survival rates of 7–20%.⁹ Treatment options for cholangiocarcinoma include surgical resection and liver transplantation for early-stage disease, with transplantation being less established than for hepatocellular carcinoma. In addition, patients can be treated with locoregional therapies (intra-arterial or external beam radiation therapies), chemotherapy, and immune checkpoint inhibitors either alone or in combination, as well as targeted therapies for tumours with specific mutations.^3,4 The combination of cisplatin and gemcitabine is the best-established systemic therapy³ and has also been approved in combination with the immune checkpoint inhibitor durvalumab based on studies done in the past few years.¹⁰ The immune checkpoint inhibitors pembrolizumab and nivolumab are approved for patients with cholangiocarcinoma and microsatellite instability-high tumours. The US Food and Drug Administration (FDA) has approved the IDH1 inhibitor ivosidenib for patients with cholangiocarcinoma and IDH1 mutations; pemigatinib, infigratinib, or futibatinib for patients with cholangiocarcinoma and FGFR2 fusions; and dabrafenib in combination with trametinib for patients with BRAF^V600E mutations.⁴ Similar to hepatocellular carcinoma, the approved therapies for cholangiocarcinoma are not directed at fibroblasts or their mediators, thereby, leaving fibroblast-targeting as an untapped therapeutic opportunity. This is in part also due to the absence of efficient drugs for liver fibrosis.¹¹

Fibroblasts and liver cancer: a wound that does not heal

To understand the roles of fibroblasts in cancer development, progression, and therapy, it is important to recognise their diverse functions in non-malignant settings. These mesenchymal cells were first described by Rudolf Virchow as spindle-shaped cells of the connective tissue and later termed fibroblasts by Ernst Ziegler¹² for their ability to produce large amounts of extracellular matrix. Besides wound healing and fibrosis, fibroblasts regulate normal tissue and organ homoeostasis, which also includes interactions with immune cells in organs and in lymphoid structures.^12,13 By embedding neighbouring cells in a physiological matrix environment and producing growth factors, fibroblasts contribute to proper epithelial proliferation, differentiation, tissue homoeostasis, mechanical stability, and organ function.¹² Most liver fibroblasts are derived from hepatic stellate cells (HSCs).^14,15 Besides secreting extracellular matrix, HSCs exert cytoprotective and growth-promoting effects on hepatocytes via the secretion of growth factors such as HGF.¹⁶ In the liver, HGF is stored bound to extracellular matrix, allowing rapid release in critical conditions such as liver regeneration. Therefore, HSC-secreted extracellular matrix and growth factors exert collaborative effects on liver epithelial health in normal and pathophysiological conditions.

Fibroblasts activate, proliferate, and migrate in response to acute and chronic tissue injury. Multiple signalling molecules and convergent pathways regulate the initiation and perpetuation of fibroblast activation.¹⁵ TGF-β is a powerful mediator of HSC activation, whereas PDGF triggers HSC proliferation.¹⁵ Molecular mechanisms of HSC and cancer-associated fibroblast (CAF) activation and migration have been reviewed elsewhere.^15,17–19 The progressive accumulation of fibroblasts and liver fibrosis cause loss of hepatic function and architecture in chronic liver diseases, including NAFLD, chronic HBV and HCV infections, and alcohol-associated liver disease, causing significant morbidity and mortality.²⁰ Extracellular matrix produced by activated HSCs and CAFs undergoes active remodelling and degradation, mediated largely by macrophages and matrix metalloproteinases, but also additional cell types and molecular mediators.²¹ The balance between extracellular matrix production and degradation not only controls properties such tissue stiffness, elasticity, and viscosity in injured and tumour tissue, and thereby cell migration and immune cell infiltration, but also regulates cellular signalling processes (eg, through the degradation of extracellular matrix components that activate specific receptors, through mechanosensitive signalling pathways, or via the release of growth factors bound to extracellular matrix). Extracellular matrix receptor such as CD44, integrins, and discoidin domain receptors, and extracellular matrix ligands (such as cleaved forms of collagen) have been shown to participate in this remodelling during tumourigenesis.^16,22,23 Although there is a strong focus on type I collagen, a plethora of additional extracellular matrix components in the fibrotic liver form a complex network whose dynamic changes and contributions to disease progression are incompletely understood.^24,25

Cancers have been likened to wounds that do not heal.^26,27 In the liver, wound healing processes can contribute to carcinogenesis or even be actively hijacked by tumour cells. At the same time, fibroblasts can react to encapsulate and possibly shield some cancers in ways similar to a foreign body reaction. The view that cancer represents a wound that does not heal can be applied to both hepatocellular carcinoma and cholangiocarcinoma, but there are conceptual differences to the role and timing. The accumulation of activated HSCs and HSC-secreted extracellular matrix exerts its main effects within the premalignant environment of hepatocellular carcinoma, thereby contributing to a stiff and tumour-prone environment with increased ability of damaged hepatocytes to transform, survive, and expand to macroscopic tumours (figure 1).¹⁶ Epidemiological studies,²⁸ showing that more than 80% of human hepatocellular carcinomas develop in fibrotic or cirrhotic livers, further support this concept. In cholangiocarcinoma, CAFs accumulate within the tumour microenvironment, in which chronic wounding processes with increased secretion of extracellular matrix and growth factors such as HGF²⁹ support malignant growth (figure 1).

Figure 1: Spatial and temporal differences in fibroblast accumulation and interactions in hepatocellular carcinoma and cholangiocarcinoma.

In hepatocellular carcinoma, fibroblasts accumulate in the premalignant environment of the liver, where they interact with hepatocytes but also immune and endothelial cells to regulate the development of hepatocellular carcinoma. In hepatocellular carcinoma tumour lesions, fibroblasts are less abundant than in the surrounding liver but might also exert biological functions. In intrahepatic cholangiocarcinoma, fibroblasts accumulate in the tumour microenvironment, where they interact with tumour cells but also immune and endothelial cells to promote intrahepatic cholangiocarcinoma progression. Both hepatocellular carcinoma and cholangiocarcinoma are often surrounded by cancer-associated fibroblasts and a fibrotic capsule.

Tumour-promoting and tumour-suppressive roles of fibroblasts in hepatocellular carcinoma

Fibroblast origin and subtypes in hepatocarcinogenesis

HSCs are considered the major source of myofibroblasts in fibrotic livers, the setting in which most hepatocellular carcinomas develop, as shown by genetic tracing and single-cell RNA sequencing in different aetiologies, including NASH.^14,30 Portal fibroblasts,³¹ mesothelial cells,³² vascular smooth muscle cells,³³ and bone marrow-derived fibrocytes³⁴ represent additional but quantitively less relevant sources of liver myofibroblasts. Besides different ontogeny, fibroblasts in the premalignant liver also exist in diverse functional states. Single-cell RNA sequencing has revealed HSC heterogeneity with two functionally antagonistic subpopulations. A less activated HSC subpopulation is inflammatory and enriched in cytokine and growth factors, whereas highly activated myofibroblastic HSCs express high amounts of extracellular matrix.^16,35 These two subpopulations have a dynamic transcriptome suggesting their potential to switch from one state to another depending on the surrounding milieu.¹⁶ The presence of inactivated HSCs, which have similarity to quiescent HSCs in a NASH regression mouse model,³⁶ further support the dynamic state of fibroblasts in chronic disease. With NASH expected to become the leading cause of hepatocellular carcinoma in the near future, understanding HSC biology in NASH-related hepatocarcinogenesis is becoming highly relevant. Bulk and single-cell RNA sequencing analysis of multiple mouse models of hepatic fibrosis, including toxic, cholestatic, and NASH, show similar gene expression patterns or follow the same activation trajectory.^16,37–39 However, further confirmation, such as in patients, is required and it is possible that although many pathways are conserved between aetiologies, that there are also some aetiology specific features in HSC activation. Consistent with similar activation pathways, HSCs appear to exert similar roles through similar pathways in NASH-associated hepatocellular carcinoma as in hepatocellular carcinomas associated with other causes.¹⁶ By contrast, the origin and diversity of CAFs within hepatocellular carcinoma tumour lesions are not as well characterised. Most CAFs within hepatocellular carcinoma lesions are derived from Lrat-positive HSCs and carry an HSC signature,¹⁶ but their origin, diversity, and transcriptomic profiles require further investigations. Likewise, comparison of CAF activation between different patients and hepatocellular carcinoma subtypes is required to reveal if this is similar to HSC activation with a conserved core programme.

Fibroblast localisation and cell–cell interactions in hepatocellular carcinoma

One characteristic feature of hepatocellular carcinoma is the lower degree of fibrosis and fibroblast accumulation in tumours than in the surrounding non-malignant fibrotic liver tissue in mouse models and most patients.¹⁶ This anatomical distribution suggests a role for HSCs in the premalignant fibrotic niche, in which hepatocellular carcinoma develops. Accordingly, signatures for activated HSCs or an imbalance between myofibroblastic HSCs and cytokine and growth factor-expressing HSCs in non-tumour liver correlate with hepatocellular carcinoma risk and survival.^16,40 Within hepatocellular carcinoma lesions, CAFs expressing alpha-smooth muscle actin (α-SMA) are localised in the fibrous septum, tumour capsule, and hepatic blood sinusoids,⁴¹ and CAF density is correlated with poor survival.^42,43 Therefore, activated HSCs in the premalignant environment as well as CAFs in the tumour microenvironment might exert tumour-promoting roles in hepatocellular carcinoma. Ligand–receptor analyses have revealed HSCs as a hub of cell–cell communication in the fibrotic liver, interacting strongly with endothelial, immune, and epithelial cells.^16,44,45 This dynamic crosstalk can be divided into HSC-afferent signals that regulate HSC activation (eg, via interactions with macrophages, endothelial cells, and epithelial cells) and HSC-efferent signals, through which HSCs regulate other cell types, including hepatocytes. In the premalignant liver, HSCs are the cell type that interacts most intensely with hepatocytes, regulating hepatocyte death and proliferation and thereby affecting hepatocellular carcinoma development.^16,18,46

Fibroblasts and their mediators in hepatocellular carcinoma: balancing tumour suppression and promotion

Although fibroblasts are considered tumour promoting in most tumours, there is growing evidence that they also exert an antitumour effect. Via a process termed neighbour suppression, normal fibroblasts can inhibit the growth of transformed cells in vitro and in vivo.^47–49 In the liver, soluble factors such as HGF, secreted by cytokine and growth factor-expressing HSCs, protect from injury and hepatocellular carcinoma development.^16,50 This protective role of HSC-secreted HGF occurs in the healthy liver and early disease stages (figure 2). By contrast, the HGF–MET axis within tumours might promote tumour cell survival and malignant growth and is therefore being evaluated for hepatocellular carcinoma and cholangiocarcinoma therapy.^16,51 As neighbour suppression can also be mediated in both contact-dependent and contact-independent manner, protective effects of HSCs via HGF-independent and contact-dependent signals needs further investigation. The paucity of HSCs, especially quiescent HSCs, within hepatocellular carcinoma lesions indicates a loss of protective neighbour suppression. Due to a shift from tumour-suppressive towards tumour-promoting HSC subpopulations in chronically injured livers,¹⁶ the overall role of HSCs in hepatocellular carcinoma is tumour promoting. Extracellular matrix from activated HSCs increases tumour-promoting liver stiffness (figure 2).¹⁶ Accordingly, liver stiffness above 12 kPa predisposes patients to hepatocellular carcinoma development.²⁴ Moreover, persistence of increased liver stiffness seems to be responsible for continuing risk of hepatocellular carcinoma development in patients with successful HCV eradication.⁵² Mechanistically, type I collagen from activated HSCs contributes to tumour initiation and growth through multiple pathways. In vitro, type I collagen and increased stiffness promotes epithelial–mesenchymal transition, proliferation, migration, survival, and stemness of hepatocellular carcinoma tumour cells.²⁴ In vivo, type I collagen promotes stiffness and hepatocellular carcinoma development in multiple mouse models, including NASH-associated hepatocellular carcinoma, but was only observed in models with strong fibrosis and stiffness,¹⁶ suggesting the existence of a threshold similar to the 12 kPa threshold that predisposes patients to development of hepatocellular carcinoma. Collagen-induced tumour-promoting stiffness activates TAZ and triggers hepatocyte proliferation within the premalignant environment.¹⁶ In the tumour microenvironment, type I collagen promotes hepatocellular carcinoma proliferation in receptor-dependent manner via DDR1¹⁶ or ITG-β1.⁵³ Of note, a non-cleavable form of type I collagen reduces hepatocellular carcinoma development despite increased fibrosis,⁵⁴ suggesting collagen cleavage is an additional regulator of cancer growth. Accordingly, a signature of collagen-cleaving matrix metalloproteinases was associated with worse survival in patients with hepatocellular carcinoma.¹⁶ Other extracellular matrix components such as HSC-secreted hyaluronic acid and its receptor CD44 on hepatocytes also promote hepatocellular carcinoma development.^16,22 Additional extracellular matrix molecules produced by HSCs such as tenascin-C, laminins, fibronectin, and periostin are upregulated in fibrosis and hepatocellular carcinoma, and have been correlated with tumour growth.²⁴ However, further in-vivo studies are needed to better define their roles in hepatocellular carcinoma and the underlying mechanisms.

Figure 2: Dual role of fibroblasts and their mediators in chronic liver injury and hepatocarcinogenesis.

(A) In early disease stages, cytokine and growth factor-expressing HSCs predominate. Cytokine and growth factor-expressing HSCs-produced HGF predominantly protects from hepatocyte death, which leads to reduced inflammation and fibrosis. Together, this protects the liver from hepatocellular carcinoma development. (B) In advanced disease stages, myofibroblastic HSCs dominate. Via production of type I collagen, they increase stiffness, TAZ signalling in hepatocytes, and promote hepatocyte proliferation, which increases the risk for hepatocellular carcinoma development. In addition, HSCs and HSCs secreted mediators might also contribute to immune cell exclusion and inhibit antitumour immunity and increase angiogenesis (via VEGF secretion), which both also increase hepatocellular carcinoma risk in the liver. HSC=hepatic stellate cells.

As most patients develop hepatocellular carcinoma after 20–40 years of liver injury and fibrogenesis, fibroblasts might undergo replicative senescence and display a senescence-associated phenotype. However, the role of HSC senescence and the senescence-associated phenotype remains controversial and has been shown to restrict,⁵⁵ promote,^56,57 or not affect¹⁶ hepatocellular carcinoma development in mice. Therapeutic approaches using senolytic drugs such as dasatinib plus quercetin, the BCL2 inhibitor navitoclax,⁵⁷ or urokinase-type plasminogen activator receptor-specific (uPAR) CAR T-cell therapy⁵⁸ inhibited Nras^G12V-driven hepatocellular carcinoma progression in mice but were not specific to HSCs.

With the success of immune checkpoint inhibitor-based therapies in hepatocellular carcinoma treatment, understanding the contribution of activated HSCs and CAFs to an immunosuppressive liver environment and the potential role of HSCs-targeted or CAF-targeted therapies for the restoration of cancer immunity become increasingly relevant. Co-transplantation of HSCs and hepatocellular carcinoma cells in subcutaneous or orthotopic mouse models reduced T-cell infiltration.^59,60 Mechanistically, HSCs can directly regulate T-cell activation or apoptosis via CD54 or PD-L1,⁶⁰ or indirectly via DIgR2 on dendritic cells.^59,60 Moreover, activated HSCs can promote Treg expansion,^59,60 natural killer cell dysfunction,⁶¹ and the accumulation of myeloid-derived suppressor cells⁶⁰ and tumour-associated macrophages.⁶² Accumulation of cross-linked collagen might create a barrier that impedes immune cell infiltration in the highly fibrotic liver and tumours.^16,63 Furthermore, immune cells express both inhibitory and stimulatory collagen receptors through which HSCs-secreted collagen could affect antitumour immunity.⁶⁴

Additional pathways through which HSCs might affect hepatocellular carcinoma development include hepatocyte and tumour cell metabolism, growth factors, and angiogenesis.^24,28 A study published in 2022 showed that extracellular vesicles containing HK1 from activated HSCs promote glucose uptake and glycolysis in tumour cells and thereby hepatocellular carcinoma growth.⁶⁵ HSC-derived epiregulin might promote hepatocellular carcinoma growth.⁶⁶ Activated HSCs express VEGF-A and angiopoietins and can thereby promote angiogenesis.⁵⁹ Moreover, HSC-derived CAFs release GDF15 in an autophagy-dependent manner, which stimulates hepatocellular carcinoma tumour growth.⁶⁷ In addition, CLCF1 from CAFs was shown to induce CXCL6 and TGF-β secretion by hepatocellular carcinoma cells, which in turn contributes to hepatocellular carcinoma stemness and the recruitment of tumour-associated neutrophils and tumour growth.⁶⁸

Functions of fibroblasts in cholangiocarcinoma

Consistent with the desmoplastic nature of cholangiocarcinoma, CAFs constitute one of the most abundant cell types in the cholangiocarcinoma tumour microenvironment. Although their overall role is thought to be tumour growth-promoting in cholangiocarcinoma, potential tumour-suppressive functions as described for hepatocellular carcinoma¹⁶ and pancreatic ductal adenocarcinoma^69,70 need to be further investigated.

Cancer-associated fibroblast subtypes in cholangiocarcinoma

HSCs seem to be the main source of CAFs in mice and patients, whereas mesothelial signature CAFs are a minority,²⁹ as shown by Cre-mediated tracing strategies and single-cell RNA sequencing. There is a high degree of CAF heterogeneity with distinct origins and functional states. Single-cell RNA sequencing classified CAFs in cholangiocarcinoma into myofibroblastic CAFs enriched in extracellular matrix transcripts, and inflammatory or vascular CAFs, both enriched in cytokines and growth factors transcripts.^29,71,72 The presence of antigen-presenting CAFs, first described in pancreatic ductal adenocarcinoma, remains to be confirmed in cholangiocarcinoma. Further subdivision has been done by some studies increasing the resolution of the analysis, with the identification of matrix CAF, epithelial-to-mesenchymal CAFs, and lipofibroblast CAFs.⁷¹

Fibroblast localisation and cell–cell interactions in cholangiocarcinoma

CAFs are in close contact with tumour cells within the tumour as well as at the tumour–liver interface in both murine and human cholangiocarcinoma, but the spatial organisation of the CAF subtypes and their proximity to tumour cells and other cells in the tumour microenvironment has not been determined at higher resolution. To date, cell–cell interactions and the involved ligand-receptor pairs have been predicted in cholangiocarcinoma from single-cell RNA sequencing data without spatial context using algorithms such as CellPhoneDB. Based on these studies, CAFs are the most interactive cell type in cholangiocarcinoma, engaging in ligand–receptor-based crosstalk with tumour cells, endothelial cells, immune cells, and other CAFs.²⁹ Through these interactions, CAFs might regulate multiple aspects of cholangiocarcinoma biology including neoangiogenesis, chemoresistance, stiffness, immunosuppression, and metastasis.⁹ It is not known whether fibroblasts in the non-malignant liver participate in early stages of cholangiocarcinoma carcinogenesis or influence cholangiocarcinoma in other relevant ways (eg, via interactions with immune cells or through long-distance signals).

Roles of fibroblasts and their mediators in intrahepatic cholangiocarcinoma

The overall tumour-promoting role of CAFs in intrahepatic cholangiocarcinoma (figure 3) has been shown by pharmacological and genetic depletion in mice^29,73 and is further supported by the clinical correlation of CAFs expressing α-SMA with poor prognosis in patients.⁹ CAFs crosstalk with other stromal cells in the tumour microenvironment via growth factors and cytokines, largely secreted by inflammatory CAF; via extracellular matrix, largely secreted by myofibroblastic CAF; and extracellular vesicles. Conversely, tumour cells and other cell types in the tumour microenvironment secrete mediators, in particular growth factors and cytokines, that might act on CAFs and their HSCs precursors, promoting activation, proliferation, and migration (figure 3).

Figure 3: Key fibroblast mediators involved in cholangiocarcinoma growth.

(A) Cholangiocarcinoma contains diverse CAF subpopulations including vascular CAF, inflammatory CAF, and myofibroblastic CAF. (B) Vascular CAF promote tumour growth via IL-6 secretion. (C) Inflammatory CAF promote cholangiocarcinoma growth through direct interactions with tumour cells via HGF secretion, which activates tumour cell-expressed MET. (D) Myofibroblastic CAF promote tumour growth via hyaluronic acid secretion. Myofibroblastic CAF-secreted type I collagen does not appear to promote cholangiocarcinoma growth but might have parallel tumour-suppressive functions (eg, physical tumour restraint) and tumour-promoting functions (eg, stiffness and mechanosensitive pathway activation or immune physical and functional exclusion). CAF=cancer-associated fibroblast.

Among the best-characterised growth factors and cytokines secreted by CAF are HGF, VEGF, HB-EGF, IL-6, and CXCL12. Single-cell RNA sequencing-based ligand receptor analysis and subsequent studies in mice with CAF-selective knockout of Hgf and epithelial knockout of Met have revealed the HGF–MET axis as a tumour-promoting pathway that directly links CAF to tumour cells (figure 3). CAF secrete VEGF-A and VEGF-C in response to PDGF-D, and HB-EGF in response to TGF-β.^74,75 CAF-secreted VEGF has been suggested to promote, via increased recruitment and permeability of lymphatic endothelial cells, lymphatic vascularisation, and lymph node metastasis.⁷⁴ CAF-secreted HB-EGF might promote tumour cell migration and invasion via EGFR activation.²⁸ CAF-derived CXCL12 promotes the migration and survival of cholangiocarcinoma lines via CXCR4.²⁸ Vascular CAFs secrete IL-6, which induces epigenetic alterations and the malignant potential in intrahepatic cholangiocarcinoma cells via upregulation of EZH2.⁷¹ However, further in-vivo confirmation of the roles of CAF-derived VEGF, HB-EGF, IL-6, and CXCL12 is required. In addition, CAFs might affect antitumour immunity. Although genetic depletion of CAFs in mice did not significantly alter immune cell composition, it has been suggested that the release of IL-6, IL-8, MIF, CCL2, IL-13, IL-34, osteoactivin, and TSP-1 from CAFs inhibits antitumour immunity via M2 macrophage polarisation, MDSC, and Treg recruitment, thereby promoting cholangiocarcinoma progression.⁷⁶

Type I collagen represents a major component of cholangiocarcinoma extracellular matrix and contributes to liver stiffness.¹⁶ However, COL1A1 deletion from CAFs, although strongly reducing stiffness, does not affect cholangiocarcinoma growth²⁹ and even increases tumour growth of metastatic pancreatic ductal adenocarcinoma in the liver in mice.³⁷ This finding is surprising as stiffness is widely believed to be a powerful promoter of tumour growth.⁷⁷ Moreover, deletion of collagen receptor DDR1 did not consistently reduce cholangiocarcinoma growth. By contrast, the deletion of Has2, one of the enzymes responsible for hyaluronic acid production, strongly reduced cholangiocarcinoma growth.²⁹ Analysis using CellPhoneDB suggested CD44 on tumour cells as potential receptor for hyaluronic acid and HAS2 from CAFs in murine and human cholangiocarcinoma.²⁹ However, the biological effects of hyaluronic acid might also be mediated by interactions with CD44 on other cells or with other hyaluronic acid receptors (figure 3).

Towards rational development of fibroblast-targeted therapies

Therapeutic concepts for targeting fibroblasts in liver cancer—one size does not fit all

Preclinical studies in mice support the concept of targeting fibroblasts in hepatocellular carcinoma and cholangiocarcinoma.^16,29,37 Translating these animal studies into clinical trials is an opportunity to target an additional cell compartment that could be potentially synergistic with therapies targeting tumour cells, immune cells, endothelial cells, or their mediators. However, because no efficient antifibrotic treatments exist for the liver, developing such synergistic concepts and integrated fibroblast-targeted therapies into clinical practice will be a greater challenge than for tumour-targeting and immune-targeting approaches, for which multiple drugs and treatment regimens had been established in other tumour entities. Moreover, with the advent of fibroblast-selective genetic knockout and depletion as well as single-cell RNA sequencing studies, the simplistic view of CAFs solely promoting liver tumour growth has turned into a more complex scenario, in which fibroblasts can mediate tumour-promoting and tumour-suppressing effects. The relevance of understanding the underlying biology is highlighted by the often opposing functions of mediators such as HGF and type I collagen in hepatocellular carcinoma and cholangiocarcinoma. With a high degree of functional heterogeneity, therapeutic concepts might need to focus on specific tumour-promoting mediators or fibroblast subpopulations. Inhibiting the most potent tumour-promoting mediators in a tumour-specific manner seems to be the most straightforward approach. Therapeutically changing the fibroblast landscape towards decreased tumour-promoting (ie, myofibroblastic HSCs) and increased tumour-suppressive (ie, cytokine and growth factor-expressing HSCs) fibroblasts might be a more powerful therapeutic approach, especially for hepatocellular carcinoma. This could be achieved either by targeting key regulators of this transition or by therapeutic depletion of tumour-promoting fibroblasts. Importantly, studies in hepatocellular carcinoma and cholangiocarcinoma have shown the power of combination therapies,^3,4,6,7 and it will be important to understand the additive and synergistic effects of fibroblast-targeted therapies with already approved therapies. Also, fibroblast-targeted therapeutic concepts need to clearly distinguish between therapy and prevention. Differences in the underlying biology suggest that distinct therapeutic concepts might be needed for fibroblast-targeted therapies of hepatocellular carcinoma and cholangiocarcinoma.

The capability to identify CAFs in human liver cancers and assess their response to therapies is emerging with the use of PET imaging. FAP-α is overexpressed on CAFs in most human malignancies and ⁶⁸Ga-labelled FAP inhibitors have been developed; FAP-based PET imaging is being explored as an imaging modality for liver cancers.⁷⁸ We anticipate this imaging modality will permit exploration of CAF-targeted therapies in hepatocellular carcinoma and cholangiocarcinoma and might allow stratification of patients with more and less fibrotic cancers.

Concepts for targeting fibroblasts and their mediators for hepatocellular carcinoma prevention and therapy

Because most of fibroblast effects are in non-tumour areas or early-stage lesions, targeting fibroblasts and their mediators seems to be most suited for primary or secondary hepatocellular carcinoma prevention (figure 4, table). Among therapeutically targetable mediators, HGF appears promising, based on its ability to protect from liver injury and hepatocellular carcinoma development.^16,50 With the short half-life of HGF in the range of minutes, HGF mimetics with longer half-life, which have shown effects in the liver,⁷⁹ could be a more applicable therapeutic strategy. Because treatment for hepatocellular carcinoma prevention might require long time frames, studying the effects of HGF mimetics in high-risk patient populations for primary or secondary hepatocellular carcinoma prevention seems most straightforward. Moreover, the hepatoprotective effects of HGF mimetics in the context of tumour prevention might depend on the presence of liver injury and patients with increased alanine aminotransferase (ALT) and active disease might be more suitable than with burnt-out cirrhosis and normal ALT concentrations. Importantly, transgenic mice that globally overexpress and have very high concentrations of HGF have a two-times increase in liver size, and increased liver tumour development.^80,81 Therefore, the dose of HGF as well as the underlying disease might determine the protumour or antitumour effects of HGF. Hence, proof-of-concept preclinical studies in animal models of cirrhosis-associated hepatocellular carcinoma will be important to further advance the concept of HGF-mediated tumour prevention. These studies also need to determine dosing of HGF mimetics and test effects in different models and at different time points of chronic liver disease, and carefully exclude tumour-promoting effects in any setting.

Figure 4: Targeting fibroblasts and their mediators for hepatocellular carcinoma as single therapy or in combination with targeting of other cellular compartments.

(A) Hepatocellular carcinoma prevention strategies include HGF mimetics, decreasing hepatocellular carcinoma risk through decreased hepatocyte death, and a subsequent reduction of inflammation and fibrosis; inhibition of type I collagen production (eg, via small interfering RNA) to decrease liver stiffness, hepatocyte TAZ, and hepatocyte proliferation; or killing of activated HSC (eg, via HSC-targeted liposome HSP47 small interfering RNA), which in addition to decreased stiffness, TAZ, and proliferation might also decrease angiogenesis or improve immunosurveillance. Inhibition or killing of senescent HSCs or fibroblasts (via senolytics or CAR T cells) might achieve similar effects as killing activated HSC and reduce fibroblast and HSCs senescence-associated phenotype. (B) Inhibition of type I collagen production or killing of activated HSCs or cancer-associated fibroblast might synergise with immune checkpoint inhibitors and VEGF inhibition for hepatocellular carcinoma therapy by increasing hepatic or tumour immune cell recruitment and activation, and possibly angiogenesis. Furthermore, DDR1 inhibitors might decrease tumour cell proliferation. HSC=hepatic stellate cells.

Table:

Drugs targeting fibroblasts and their mediators for hepatocellular carcinoma and cholangiocarcinoma therapy or prevention

	Drug	Mechanism of action	Application	Clinical feasibility
Hepatic stellate cells	ND-L02-s0201 (BMS-986263)	Small interfering RNA against HSP47; killing hepatic stellate cells via collagen misfolding	Hepatocellular carcinoma prevention; possibly hepatocellular carcinoma therapy	In phase 2 clinical trial for patients with compensated non-alcoholic steatohepatitis-associated cirrhosis (NCT04267393)
Senescent CAFs and HSCs	Senolytics such as navitoclax and dasatinib plus quercetin combination or CAR T cells (eg, targeting uPAR)	Navitoclax (BCL2 inhibitor) and dasatinib (SRC kinase inhibitor) with quercetin (BCL-xL inhibitor) induce apoptosis in senescent cells, including fibroblasts and CAF; removal of senescent cells (eg, via uPAR CAR T cells recognising senescent cells with upregulated uPAR); decreases SASP-mediated promotion of hepatocellular carcinoma development	Cholangiocarcinoma therapy; hepatocellular carcinoma prevention and possibly therapy	Navitoclax and dasatinib are approved by the US FDA; dasatinib plus quercetin combination in phase 1–2 clinical trials for non-alcoholic fatty liver disease-associated fibrosis (NCT05506488)
TGF-βR1	Galunisertib (LY2157299)	Galunisertib targets TGF-β signalling and can reduce HSC-CAF activation but also promotes antitumour immunity through direct effects on T cells; however, systemic promotion of T-cell activation can also lead to severe side-effects	Hepatocellular carcinoma prevention and possibly therapy; cholangiocarcinoma therapy	Multiple completed phase 1 and 2 trials as monotherapy or in combination with immunotherapy, sorafenib, ramucirumab, or radiotherapy (NCT02906397, NCT02423343, NCT02240433, NCT02178358, NCT01246986)
HGF	HGF mimetics (eg, ANG-3777)	HGF agonism with prolonged half-life; protects hepatocytes from cell death	Hepatocellular carcinoma prevention	ANG-3777 in phase 2 clinical trials for cardiac surgery-associated acute kidney injury (NCT02771509) and for acute lung injury in COVID-19 (NCT04459676)
MET	MET inhibitors	Inhibiting the HGF–MET axis reduces cholangiocarcinoma growth	Cholangiocarcinoma therapy	Drugs approved by the US FDA for various non-liver tumours: capmatinib and tepotinib
IL-6	Anti-IL-6	IL-6 neutralising antibody blocks crosstalk between vascular CAF and tumour cells	Intrahepatic cholangiocarcinoma therapy	Approved by the US FDA: tocilizumab (for rheumatoid arthritis and other forms of arthritis)
Type I collagen	COL1A1 small interfering RNA (no drug in clinical studies)	Reduced production of type I collagen reduces liver stiffness and tumour-promoting mechanosensitive signals	Hepatocellular carcinoma prevention and therapy	No studies applying collagen small interfering RNA or synthesis inhibition on ClinicalTrials.gov
Hyaluronic acid	Inhibition of hyaluronic acid production or hyaluronic acid receptors	Reduces hyaluronic acid-mediated tumour-promoting signals	Hepatocellular carcinoma prevention	Hymecromone currently studied in patients with COVID-19 (NCT05386420); planned phase 2 trial for primary sclerosing cholangitis (NCT05295680)
DDR1	DDR1 inhibitors	Blocking DDR1 kinase activity or ligand binding to reduce hepatocellular carcinoma proliferation	Hepatocellular carcinoma therapy	Dasatinib (approved by the US FDA) has activity against DDR1; no specific DDR1 inhibitor in clinical trials

CAF=cancer-associated fibroblast. CAR=chimeric antigen receptor. FDA=Food and Drug Administration. SASP=senescence-associated secretory phenotype.

As genetic depletion of HSCs or fibroblasts expressing α-SMA reduced hepatocellular carcinoma development in mice,¹⁶ pharmacological depletion of HSCs (eg, via small interfering RNA against the collagen chaperone HSP47, which kills activated HSCs via collagen misfolding) might represent a potential and clinically applicable hepatocellular carcinoma prevention strategy (figure 4, table). Another therapeutic approach is the depletion of senescent HSCs and fibroblasts via the senolytic drug combination dasatinib and quercetin⁵⁷ or via uPAR-specific CAR T-cell therapy,⁵⁸ which inhibited hepatocellular carcinoma progression. As these approaches do not specifically target HSCs, it is conceivable that some of the effects of senolytic therapies are via the epithelial macrophage or tumour compartment (figure 4, table). It will be important to determine in animal models if protective cytokine and growth factor-expressing HSCs and HGF are affected by depletion of activated or senescent HSCs, because these therapies should ideally only affect tumour-promoting fibroblasts and spare those that are protective.

Reducing liver stiffness via inhibition of collagen synthesis or crosslinking, or via increased collagen degradation appears a second promising strategy for tumour prevention (figure 4, table). Liver-targeted, HSC-targeted, or fibroblast-targeted delivery of small interfering RNA against type I collagen would be a feasible approach. This could potentially be combined with collagen degradation therapies as collagen turnover might be slow in patients with liver cirrhosis and a high amount of crosslinked collagen. Stimulating collagen degradation through the infusion of macrophages has been shown to be feasible in patients with cirrhosis.⁸² Such therapies are likely to work best in advanced stages and in patients with high liver stiffness. Further preclinical studies in animal models of cirrhosis-associated hepatocellular carcinoma are needed for proof of concept. Hyaluronic acid is another extracellular matrix candidate that could be targeted for tumour prevention (figure 4, table). As targeting hyaluronic acid by recombinant human hyaluronidase PEGPH20 was detrimental in pancreatic ductal adenocarcinoma,⁸³ probably mediated by biologically active hyaluronic acid degradation products, other therapeutic concepts such as inhibiting the expression or activity of HAS2, the main enzyme in the liver that produces hyaluronic acid,^16,84 or blocking hyaluronic acid receptors such as CD44 need to be developed.

Targeting fibroblasts or their mediators for hepatocellular carcinoma therapy is likely to be less effective than for hepatocellular carcinoma prevention due to the predominant accumulation of fibroblasts in non-tumour tissue and absent effects of HSC depletion on established tumours in mice. However, it is conceivable that patients, who’s tumor tissue contains large amounts of extracellular matrix and fibroblasts, might be good candidates for the approach. Moreover, targeting extracellular matrix might synergise with immune checkpoint inhibitor-based therapy (figure 4, table), enabling easier access for immune cells. These opportunities should be tested in preclinical animal models with high accumulation of collagen in tumours, focusing primarily on silencing or therapeutically inhibiting type I collagen or its receptor DDR1 (figure 4, table). There is a wide range of DDR1 inhibitors, including existing drugs such as dasatinib.⁸⁵

Concepts for targeting fibroblasts and their mediators for cholangiocarcinoma therapy

Based on the strong accumulation of fibroblasts and extracellular matrix in cholangiocarcinoma and functional studies in mice, fibroblasts and their mediators might be suitable targets for the therapy of established cholangiocarcinoma (figure 5, table). Without known protective roles of fibroblasts in cholangiocarcinoma, direct targeting or depletion of fibroblasts appears straightforward. Options include HSC-targeted small interfering RNA against the collagen chaperone HSP47, an approach which is in clinical trials for liver cirrhosis (NCT04267393), as well as the BH3 mimetic navitoclax, which induces apoptosis in CAFs in cholangiocarcinoma (table).⁷³ In view of decreased tumour burden but increased mortality after targeting fibroblasts in pancreatic ductal adenocarcinoma,^69,70 further preclinical studies are needed to assess the effects of fibroblast-targeting survival in multiple models of cholangiocarcinoma. Inhibiting CAF activation is a second therapeutic option. TGF-β inhibition seems particularly attractive as it might not only reduce fibrosis, but also in parallel increase immune cell activation. However, due to strong side-effects of systemic TGF-β inhibition, local inhibition therapy might need to be developed. HGF and hyaluronic acid appear to be among the most potent fibroblast mediators that promote cholangiocarcinoma growth (figure 5, table). MET inhibitors have shown effects in preclinical cholangiocarcinoma models.⁸⁶ Given the detrimental effects of hyaluronic acid targeting with PEGPH20 in pancreatic ductal adenocarcinoma, hyaluronic acid should be targeted by reducing the activity or expression of HAS2 via drugs or small interfering RNA, or by inhibiting hyaluronic acid receptors. For all above-described approaches, further preclinical studies are needed and should also test their combination with cisplatin plus gemcitabine, the current standard of care. Likewise, the effect of combinations with immune checkpoint inhibitors should be determined as reducing fibrosis might improve infiltration and activity of immune cells.

Figure 5: Targeting fibroblasts and their mediators for cholangiocarcinoma therapy.

(A) Killing of activated hepatic stellate cells (eg, with navitoclax) might synergise with immune checkpoint inhibitors by increasing immune cell recruitment and activation or with chemotherapy (cisplatin plus gemcitabine) by increasing tumour cell sensitivity to cell death and reducing tumour cell proliferation. (B) Inhibition of HGF or MET via specific inhibitors might synergise with chemotherapy by increasing tumour cell sensitivity to cell death and reducing tumour cell proliferation. It is also conceivable that inhibition of hyaluronic acid production (eg, via inhibition or silencing of HAS2) or its receptors or neutralisation of IL-6 (eg, via tocilizumab) might synergise with chemotherapy via synergistic effects on tumour cell proliferation or through increased sensitivity of tumour cells to cell death. CAF=cancer-associated fibroblast.

Summary and future directions

Fibroblast-targeted and fibroblast mediator-targeted therapy is a therapeutic opportunity that has not yet been leveraged for the treatment of liver cancer. Targeting fibroblasts in addition to cancer, immune, and endothelial cells might further enhance synergies in combination treatments for cholangiocarcinoma and hepatocellular carcinoma, or provide more opportunities for hepatocellular carcinoma prevention strategies. However, although there is increasing evidence for a tumour-promoting role of fibroblasts in the development and progression of liver cancer, additional and carefully done preclinical studies are needed to avoid repeating mistakes made in targeting fibroblasts in patients with pancreatic ductal adenocarcinoma, in whom inhibition of the profibrotic pathway or treatment with hyaluronidase were detrimental in clinical studies.^83,87 More detailed studies will also be needed to understand the relationship between fibroblasts, extracellular matrix, and the immune system to design combination therapies in which immune checkpoint inhibitors and fibroblast-directed therapies could be synergistic. Finally, better understanding of regulators that control the transition between different functional states such as myofibroblastic HSCs and cytokine and growth factor-expressing HSCs will be important to reprogramme the fibroblast landscape in the premalignant or malignant liver with the goal of activating protective pathways and reducing those that are tumour promoting. Although TGF-β is a key mediator in this transition, knowledge about additional and more targetable mediators might open up novel therapeutic avenues for tumour prevention or treatment.

Search strategy and selection criteria.

We based our Review on existing research. To do this, we searched PubMed and ClinicalTrials.gov for literature published between 2005 and 2023 using the terms “fibroblasts”, “HSCs”, “cancer-associated fibroblasts”, “hepatocellular carcinoma” and “cholangiocarcinoma” in various combinations.