Trends in Molecular Medicine

Review
Immunomodulatory TGF-b Signaling in
Hepatocellular Carcinoma
Jian Chen,1,* Julian A. Gingold,2 and Xiaoping Su3

Hepatocellular carcinoma (HCC) is an inflammation-induced and chemotherapy-resistant cancer. Highlights

Dysregulated signaling in the transforming growth factor beta (TGF-b) pathway plays a central Highly activated TGF-b in hepato-
role in inflammation, fibrogenesis, and immunomodulation in the HCC microenvironment. This cellular carcinoma (HCC) is signifi-
review dissects the genetic landscape of the TGF-b superfamily genes in HCC and discusses cantly associated with immune cell
the essential effects of this pathway on the tumor immune microenvironment. We highlight exhaustion, while inactivated TGF-
the TGF-b signature as a potential biomarker for identifying individualized immunotherapeutic b is correlated with impaired DNA
approaches in HCC. An improved understanding of the detailed mechanisms of liver cancer repair.
immunogenicity and the specific role of TGF-b in mediating immunotherapy resistance in HCC
The combination of TGF-b in-
will provide important insights into HCC immune escape and promote the development of hibitors and immune-checkpoint
biomarker-derived combination immunotherapies for HCC. inhibitors successfully induced
complete responses in mouse can-
Therapeutic Challenges for HCC cer models, highlighting the key
immunosuppressive role of TGF-b
HCC is the solid tumor type with the fastest-rising incidence in the USA and is the second-leading signaling in regulating HCC
cause of cancer-related death worldwide [1,2]. Attributable HCC risk factors are highly heteroge- immunotherapy.
neous across patients and regions but include chronic hepatitis C virus (HCV) or hepatitis B virus
(HBV), alcoholic cirrhosis, obesity/diabetes, and nonalcoholic fatty liver disease (NAFLD)/nonalco- A novel immune-based classifica-
holic steatohepatitis (NASH). While HCV can now be cleared in almost all patients using antiviral ther- tion identified three HCC tumor
apies, patients infected with HBV typically remain infected for life [3]. Additionally, the dramatic rise in classes in the tumor microenviron-
ment: the immune class, the inter-
obesity, and with it diabetes, NAFLD/NASH, and chronic liver inflammation, is expected to be the pri-
mediate immune class, and the
mary driver of increasing HCC incidence globally [4].
immune exclusion class. Treatment
of tumors in each class is likely to
Regardless of the cause, HCC remains a highly lethal malignancy, particularly for the 40% of affected require specific immunothera-
patients diagnosed with advanced-stage disease. Sorafenib is a multitarget tyrosine kinase inhibitor peutic approaches.
and was the only systemic therapy with FDA approval for the treatment of advanced HCC prior to
2016 [5]. Recently, three new multikinase inhibitors, lenvatinib [6], regorafenib [7], and cabozantinib Deep single-cell RNA sequencing
[8], and an antibody-based VEFGR2 antagonist, ramucirumab, have received FDA approval for of tumor-infiltrating lymphocytes
identified molecular characteristics
advanced HCC [9], although the median overall survival for patients treated with them remains under
in the microenvironment that may
15 months. Several immune-checkpoint inhibitors (see Glossary) have recently shown promising re-
guide future therapeutic strategies.
sults in select patients with advanced HCC. Two monoclonal antibodies targeting the programmed
cell death 1 (PD-1) pathway, nivolumab [10] and pembrolizumab [11], were recently approved by the
FDA for treating patients with advanced-stage HCC previously treated with sorafenib and early-stage
results have been reported with the anti-CTLA-4 therapy astremelimumab [12]. However, tumor
response rates for these immune-checkpoint inhibitors are reported in the 14–17% range. Thus, there
remains a large unmet medical need for novel and effective (immuno)therapeutic strategies against
HCC [13].

It is thought that the complexity of HCC treatment and its high therapeutic resistance may arise from 1Department of Gastroenterology,

the multifactorial process that leads to its development, often involving the interaction of two or more Hepatology, & Nutrition, The University of
risk factors including cirrhosis, NASH, viral infection, and impaired hepatic function. The lack of un- Texas MD Anderson Cancer Center,
Houston, TX, USA
derstanding of the heterogeneous mechanisms of HCC tumorigenesis and progression, the complex 2Women’s Health Institute, Cleveland
interactions of liver tumors with their immune microenvironment, and the lack of validated diagnostic Clinic Foundation, Cleveland, OH, USA
and prognostic biomarkers have markedly hindered the development of clinically validated agents 3Departments of Bioinformatics and

against HCC. In this review, we provide an overview of the immune landscape in HCC with a particular Computational Biology, The University of
focus on signaling from TGF-b, a master immune regulator. We review the interaction of the TGF-b Texas MD Anderson Cancer Center,
Houston, TX, USA
pathway with the HCC tumor immune microenvironment and discuss the TGF-b signature as a poten-
*Correspondence:
tial biomarker for identifying the ‘exhausted’ immune signature in HCC. Finally, we discuss jianchen@mdanderson.org

1010 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 https://doi.org/10.1016/j.molmed.2019.06.007
ª 2019 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Trends in Molecular Medicine

opportunities and challenges for combination therapy for HCC targeting both TGF-b and immune
checkpoints. Glossary
Antigen-presenting cells (APCs):
cells that present antigens to
Immunosuppressive Role of TGF-b in HCC T cells. The antigens presented on
their MHC proteins interact with
The TGF-b superfamily of ligands includes TGF-bs, bone morphogenetic proteins (BMPs), inhibins,
TCRs. Professional APCs include
and nodal and growth and differentiation factors (activins), the receptors of which activate SMAD- macrophages, B cells, and DCs.
dependent and SMAD-independent pathways to regulate transcriptional/translational cascades CD4+ T cells: also known as Th
(Box 1). The TGF-b pathway plays a context-dependent role in cell processes, including promotion cells; characterized by surface
of cell differentiation and proliferation, control of cell apoptosis and cell cycle, regulation of the expression of the CD4 protein.
CD4+ T cells help to regulate the
epithelial–mesenchymal transition (EMT), angiogenesis, extracellular matrix (ECM) formation, sup- activity of other immune cells by
pression of immune response, and maintenance of genomic stability and stem cell homeostasis. In releasing T cell cytokines. The
addition, TGF-b signaling acts as a master regulator for immune cell proliferation, differentiation, major subtypes include Th1, Th2,
development, and survival [14,15]. Th17, and Treg cells.
CD8+ T cells: also known as cyto-
toxic T cells; characterized by sur-
TGF-b Effects on the Liver Immune Milieu face expression of the CD8 pro-
tein. CD8+ T cells recognize, bind,
The liver is the largest peripheral immunomodulatory organ and is filled with a multitude of innate and and kill pathogen-infected cells
adaptive immune cells, including the largest population of resident macrophages (Kupffer cells) and a and cancer cells.
high density of natural killer (NK) cells, NK T (NKT) cells, and liver-transiting and/or -resident CD8+ T Chimeric antigen receptor (CAR)-
transduced T cells: engineered
cells and CD4+ T cells [16]. TGF-b critically regulates immune cells in the liver to maintain a balance
T cells expressing a CAR on their
between immune tolerance and activation. TGF-b plays a central role in regulating immune cells surface. They are designed to
including tumor-associated macrophages (TAMs), myeloid-derived suppressor cells (MDSCs), NK facilitate both antigen presenta-
cells, and dendritic cells (DCs) (Box 2 and Figure 1). tion and T cell activation.
Immune-checkpoint inhibitors:
drugs targeting immune-check-
Chronic inflammation plays a well-established role in the development of HCC, often in association
point molecules including CTLA-
with liver fibrosis and cirrhosis. The activity of TGF-b has been established as essential for HCC path- 4, PD-1, and PD-L1, the key
ogenesis, including the activation of cancer-associated fibroblasts (CAFs) [17–19]. Liver sinusoidal inhibitory regulators of the im-
endothelial cells (LSECs) and hepatic stellate cells (HSCs) are major producers of TGF-b, which mune system that allow many tu-
mors to blunt the development of
contribute to hepatic regulatory T (Treg) cell induction [20,21]. These data suggest an immunosup-
an antitumor immune response.
pressive role for TGF-b in HCC and imply that targeting the TGF-b pathway in hepatic immune cells Immune exclusion class: a molec-
might enhance antitumor immunity in HCC patients. Although current HCC immunotherapies modu- ular immune classification subtype
late immune effector cell functions rather than the tumor immune microenvironment, clinical and characterized by T cell exclusion
from the tumor microenviron-
ment. It mostly overlaps with the
Wnt/CTNNB1 molecular classifi-
Box 1. The TGF-b Superfamily in HCC cation in HCC [48,72].
Immune exhaustion subtype:
TGF-b signal transduction is one of the major signaling pathways regulating cell homeostasis and embryo
molecular immune classification
development [76,77]. The TGF-b pathway plays a complex role in liver disease and liver cancer, on which it ex-
subtype characterized by the
erts fibrogenic/proinflammatory, tumor-suppressive, and/or prometastatic effects [78]. Clinically, high levels of presence of immune cells with low
TGF-b are associated with a favorable prognosis in early-stage cancers but are associated with increased tumor inflammatory activity. It typically
invasiveness and dedifferentiation in advanced tumors, suggesting that TGF-b initially restrains liver cancers presents with many dysfunctional
through its tumor-suppressor functions but may later exacerbate the malignancy through its pro-oncogenic ac- tumor-infiltrating T cells, M2-type
tivities [17,19]. macrophages, and higher expres-
sion levels of PD-1/PD-L1 and
The TGF-b superfamily proteins mainly include TGF-bs, activins, BMPs, growth and differentiation factors, and inhibitory receptors. It overlaps
cofactors/adaptors. The ligands, TGF-bs, activins, and BMPs act as dimers to bind with cell surface type 1 and with the activated stromal and
type 2 receptors, leading to the activation of gene responses by SMAD-dependent and SMAD-independent TGF-b signatures in HCC [48].
cascades (Tables S1 and S2 in the supplemental information online). The details of the canonical SMAD-depen- Immune-related adverse events
dent pathway and noncanonical SMAD-independent pathway were reviewed recently [44,77]. (irAEs): the immunotherapy-
induced side effects caused by
The complex set of interactions between these signaling pathways imposes significant challenges for attempts overactivation of the immune
at cancer therapeutics targeting the TGF-b superfamily. It remains unknown whether blocking the canonical system.
SMAD-dependent pathway will lead to a reduction or an increase in signaling via the noncanonical SMAD-in- Myeloid-derived suppressor cells
(MDSCs): a heterogeneous popu-
dependent pathway. Likewise, it is difficult to anticipate the effects of targeted inhibitors against crosstalk
lation of immature and immuno-
pathways on TGF-b signaling. The development of comprehensive, dynamic, and individualized genomic an-
suppressive myeloid cells that
alyses will be essential to define the activities of TGF-b signaling for use in future HCC therapeutics. display a variety of pro-oncogenic
effects.

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1011

Trends in Molecular Medicine

Regulatory T (Treg) cells: inhibi-

Box 2. Regulation of TGF-b in Immune Cells
tory CD4+ T cells characterized by
+
TGF-b activates SMAD2/3 and cooperates with IL-21 activity to promote naı̈ve CD4 T cell generation of Th17 surface expression of CD25,
cells, a subtype of proinflammatory Th cells that expresses IL-17 and contributes to NAFLD-associated liver CTLA-4, CD62L, and FoxP3 mole-
inflammation and development of HCC [43,79]. TGF-b also inhibits Th1 and Th2 lineages by downregulating cules. Treg cells suppress or
T-bet and Sox4 expression, respectively, and probably by promoting a shift of Th1 towards Th2 cell differen- downregulate the induction and
tiation [77,80]. In CD8+ cytotoxic T cells, TGF-b cooperates with the transcription factor ATF1 to suppress the proliferation of effector T cells.
Single-cell transcriptome
expression of IFN-g to inhibit its antitumor activity [77]. Intratumoral TGF-b suppresses NKT cells, a population
sequencing: individual cells are
responsible for recruiting effector immune cells to the tumor through the production of large amounts of IFN-g.
isolated from a sample and the
During B cell maturation, TGF-b cooperates with runt-related transcription factor 3 (RUNX3) to promote naı̈ve B expression of mRNAs from each is
cell class switching to IgA-producing (IgA+) cells, which are involved in NAFLD-associated HCC, by coexpres- characterized. Compared with the
sequencing of bulk cells, this
sion of PD-L1 and IL-10 and by suppressing liver cytotoxic CD8+ T lymphocytes [81]. TGF-b promotes the
technology can identify genes
differentiation of M2-type macrophages, which suppress the activity of CD8+ T cells, NK cells, and DCs and differentially expressed in small
increase the activity of CD4+ Treg cells [82]. cell populations.
T cell receptor (TCR)-transduced
Although TGF-b promotes chemotaxis in eosinophils and mast cells, it inhibits cytotoxicity and cytokine pro-
T cell: a type of engineered T cell
duction in NK cells and inhibits antigen presentation on DCs [15]. TGF-b1 enhances antigen-induced PD-1
designed to produce TCR a and b
expression through SMAD3-dependent transcriptional activation in antigen-specific T cells, suggesting that chains to recognize antigens on
the TGF-b pathway directly participates in immune-checkpoint regulation [29]. tumor cell surfaces in an MHC-
dependent manner [75].

preclinical data highlight that the highly immunosuppressive tumor environment in advanced HCC
contributes to impaired effector T cell recruitment and immunotherapy response [22]. The inflamma-
tion and tumor microenvironment that are regulated by TGF-b signaling are therefore critical to HCC
progression.

TGF-b/Treg Cell-Mediated Immunosuppression in HCC

The potent immune inhibitory function of Treg cells is a major obstacle to generating an effective
antitumor response in HCC [23]. Tumor cells and all stromal cells, including TAMs, HSCs, LSECs,
and MDSCs, modulate Treg cell activity in HCC via the TGF-b pathway (Figure 1, top center). The
Smad2/3-dependent TGF-b pathway upregulates the transcription factor FoxP3, which is typically
expressed on Treg cells [24]. HCC patients demonstrate high levels of FoxP3+ Treg cells in peripheral
blood and markedly increased tumor-infiltrating Treg cells [25]. Treg CD4+ cells also secrete TGF-b,
as well as adenosine and IL-10, to suppress effector T cells such as CD8+ cytotoxic T lymphocytes
(CTLs) [22], which are the main subset of tumor-infiltrating lymphocytes in HCC responsible for
secreting antitumor effector molecules such as interferon-gamma (IFN-g), IL-2, and TNFa [26]. Treg
cells also inhibit immune responses through their membrane-bound CTLA-4 immune-checkpoint
molecules to compete with costimulatory CD28, which interacts with the CD80/CD86 complex on
antigen-presenting cells (APCs), to facilitate downregulation of the CD80/CD86 complex via
CTLA-4-mediated transendocytosis [27,28]. Therefore, targeting Treg cells – for example, by block-
ing the TGF-b pathway or the membrane-bound receptor CTLA-4 – may be a particularly effective im-
mune-based approach for improving the immune response against HCC.

Upregulation of PD-1 by TGF-b

TGF-b secreted by tumor cells directly upregulates the transcriptional expression of PD-1 in cancer
cells (Figure 1, bottom center) [29]. PD-1 is expressed by activated CD8+ cells and binds with its li-
gands, programmed death ligand 1 and 2 (PD-L1/2), on APCs or tumor cells. The interaction of
PD-1 and PD-L1 is well known as one of major immunosuppressive mechanisms exploited by virally
induced cancers. PD-1 engagement by PD-L1 blocks T cell antigen receptor (TCR) signaling and
inhibits T cell proliferation and secretion of cytotoxic mediators, including Granzyme A (GZMA),
Granzyme B (GZMB), TNFa, and IFN-g, which collectively lead to T cell exhaustion (Figure 1, up-
per-right corner) [30,31]. Thus, increased TGF-b signaling may cause T cell exhaustion by upregula-
tion of PD-1 signaling and inhibition of TGF-b signaling may directly enhance antitumor immunity
in HCC [29].

1012 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

Trends in Molecular Medicine

Figure 1. Immunosuppressive Roles of TGF-b in Hepatocellular Carcinoma (HCC).

TGF-b directly modulates most immune cells, most notably regulatory T (Treg) cells, upregulating cell activity and promoting proliferation, leading to
cytotoxic T lymphocyte (CTL) exhaustion (top right). Myeloid-derived suppressor cells (MDSCs) abrogate hepatic natural killer (NK) cell activity via
membrane-bound TGF-b and facilitate the expansion of Treg cells. Dendritic cells (DCs) both secrete and respond to TGF-b and promote the induction
of Treg cells (top center). TGF-b induces T cell immunoglobulin mucin receptor 3 (TIM-3) on tumor-associated macrophages (TAMs) to stimulate
M2-type macrophage polarization and IL-6 production (left). Cancer-associated fibroblasts (CAFs) are essential components of the HCC
microenvironment and produce cyclooxygenase-2 (COX2), IL-8, and other cytokines that collectively stimulate TAMs to release tumor necrosis factor
alpha (TNFa) and platelet-derived growth factor (PDGF), which further promote CAF activation (bottom). HSCs (hepatic stellate cells) are the main
producer of extracellular matrix (ECM) and TGF-b and induce the proliferation and activation of both MDSCs and Treg cells via secretion of
amphiregulin (top-left corner). Liver sinusoidal endothelial cells (LSECs) are the major liver cell type responsible for TGF-b-dependent hepatic Treg cell
induction, and generate the chemokine C-X-C motif chemokine ligand 16 (CXCL16) and control the accumulation of CXCR6+ hepatic NKT cells (bottom
center). CTLA-4, cytotoxic T lymphocyte-associated protein 4; GZMA/B, Granzyme A/B; IgA+, IgA antigen-expressing B cell; LAYN, layilin; PD-1,
programmed cell death 1; PD-L1, programmed cell death 1 ligand 1.

The TGF-b Genomic Profile and Immune Landscape in Hepatocellular Carcinoma

Genomic Alterations in the TGF-b Superfamily in HCC
Genomic alterations in the TGF-b signaling pathway, including somatic mutation, homozygous
deletion, or amplification, were found in 39% of 9125 tumor samples across 33 cancer types in The
Cancer Genome Atlas (TCGA) [32]. Alterations in the TGF-b superfamily correlated with decreased
survival [17]. Although the frequency and type of genomic alteration in the TGF-b superfamily varied
widely across tumor types, alterations affecting TGF-b signaling were most common in gastrointes-
tinal cancers, including colorectal, esophageal/stomach, pancreatic, and liver. The most common
somatic mutations in HCC were summarized and reviewed recently [33,34].

While mutations in the TGF-b superfamily of 43 genes are also common overall in TCGA HCC
datasets, estimated at 38% overall in a cohort of 202 HCC samples [17], somatic mutations in
each individual TGF-b signaling pathway gene remain relatively infrequent (Table S1 in the supple-
mental information online and Figure 2) (http://www.cbioportal.org/) [35–37]. The TGF-b adaptor
SPTBN1, a key adaptor for SMAD3 nuclear translocation, is the most frequently mutated TGF-b

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1013

Trends in Molecular Medicine

Trends in Molecular Medicine

Figure 2. Genomic Alterations in the TGF-b Signaling Pathway in Hepatocellular Carcinoma (HCC).
Genomic alterations in the TGF-b superfamily, canonical SMAD-dependent pathway and noncanonical SMAD-independent pathway were systematically
characterized in HCC (Tables S1 and S2 in the supplemental information online). The frequencies (%) of gene amplification (dark-red boxes),
homozygous gene deletion (gray boxes), increased gene expression (light-red boxes), decreased gene expression (blue boxes), and somatic gene
mutation (brown boxes) in HCC were based on TCGA datasets (http://www.cbioportal.org/). Genomic alterations with 0% frequency or missing data
were not shown. R-SMADs, receptor-specific SMADs; Co-SMAD, common mediator SMAD.
Genomic alterations in the TGF-b superfamily, canonical SMAD-dependent pathway and noncanonical SMAD-independent pathway were systematically
characterized in HCC (Tables S1 and S2 in the supplemental information online). The frequencies (%) of gene amplification (dark-red boxes), homozygous
gene deletion (gray boxes), increased gene expression (light-red boxes), decreased gene expression (blue boxes), and somatic gene mutation (brown
boxes) in HCC were based on TCGA datasets (http://www.cbioportal.org/). Genomic alterations with 0% frequency or missing data were not shown.
R-SMADs, receptor-specific SMADs; Co-SMAD, common mediator SMAD.

gene in HCC at 3.9–6% and is also frequently methylated at its promoter, suggesting a tumor-sup-
pressor role [17,32]. Functional and mechanistic studies also support that SPTBN1 plays a tumor-
suppressor role through SMAD3-dependent pathways in HCC tumorigenesis [17,38–40]. While de-
letions in 51% of various TGF-b superfamily genes (22 of 43 genes) have been identified in a small
percentage (<2.1%) of HCC samples, 77% (33 of 43 genes) are also amplified, particularly TGFB2,
BMP6, and BMP5, at rates of 8.6%, 5.1%, and 3.9%, respectively. In addition, the mRNA levels of
58% (25 of 43) of TGF-b genes are significantly increased (by >4% from baseline) in HCC, suggest-
ing that the TGF-b signaling pathway is activated in the majority of HCCs at the transcriptome level.
Consistent with TCGA data, proteomic and phosphoproteomic profiling of a clinical early-stage

1014 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

HBV-associated HCC cohort containing 110 paired tumor and nontumor tissues demonstrated
increased TGF-b genes and increased cholesterol metabolism in HCC compared with non-tumor
tissue [41].

Genomic Alterations in TGF-b Targets in HCC

Expression of many TGF-b targets, themselves with both pro- and antioncogenic effects, is also
markedly increased in the TCGA database (Figure 2). Increased VEGFA (6.9% of cases) and
COL4A1 (5.6% of cases) mRNA levels are observed in HCC, driving increased angiogenesis and
accumulation of ECM, respectively. Increased SNAI2, which promotes the EMT, is observed in
8.1% of HCCs [42]. Increased expression of TGF-b targets SOX4 (8.1% of cases), NANOG (6.1%
of cases), and MET (6.9% of cases), suggesting that TGF-b signaling plays an important role in
the maintenance of HCC cancer stem cell homeostasis. However, increased levels of other TGF-
b targets, including the apoptotic genes DAPK2/3 (5.9% of cases) and the cell cycle inhibitors
CDKN1A (6.5% of cases) and CDKN1C (5.6% of cases), indicate that TGF-b signaling in HCC also
drives multiple tumor-suppressor pathways. The levels of the DNA repair genes FANCD2 and
FANCM are increased in 7.3% and 5.4% of HCC cases, respectively, suggesting an upregulated
DNA damage response [39].

Although TGF-b has potent growth-suppressing activity on most precursor cells of the immune sys-
tem, including T cells, B cells, NK cells, monocytes, macrophages, neutrophils and eosinophils, some
immunomodulatory effects of TGF-b signaling in HCC appear to be growth promoting. The expres-
sion of all SMAD3-dependent immune response genes is upregulated in HCC, suggesting a cell type-
dependent response to TGF-b signaling. For example, increased expression of IL-17A is observed in
6.9% of HCCs, suggesting that activated TGF-b signaling may promote an IL-17A-mediated inflam-
matory response in HCC [43].

In the noncanonical SMAD-independent pathway, the TGF-b superfamily receptor complex activates
other intracellular pathways, either through phosphorylation of or direct interaction with critical
signaling intermediates [44]. Targets include the PI3K/AKT pathway, the ERK/MAPK cascade, JNK/
MAPK/IKK pathways, and RHO–ROCK signaling intermediates (Figure 2). The extensive crosstalk be-
tween these signaling pathways permits genomic alterations in noncanonical TGF-b targets to lead to
complex responses. These pathways also directly or indirectly regulate R-SMAD or its cofactors in the
cell nucleus, leading to a highly controlled and context-dependent TGF-b response affecting a variety
of cellular functions in HCC [45].

Immune Landscape of HCC

Specific inflammatory signals and the expression of different immune cell subsets confer therapeutic
resistance and significantly predict HCC patient survival [46,47]. Recent studies have systematically
characterized these inflammatory signals to survey the global genome-wide immunological land-
scape in HCC by combining TCGA data with independent analyses [2,17,31,34,48]. Measurement
of the transcriptional expression of several immune-cell signatures has been used to characterize tu-
mor immune infiltration in liver cancers and predict the response. Multiple groups have reported im-
mune-cell signatures corresponding to tumor infiltration with CD8-positive cells [49], cytotoxic cells
(CD8+ T cells, T-gd cells, and NK cells) [49], T/NK cells (T cells and NK cells) [50], T helper-1 (Th1) cells
[49], and macrophages [49] (Figure 3).

An analysis of the HCC immune microenvironment in a collection of 196 TCGA samples, each of
which came from a different patient with HCC, found that 22% (43 of 196) of HCCs displayed
high levels of immune cell infiltration with high expression levels of the immune checkpoint genes
CTLA-4, PD-1, and PD-L1. These HCC samples were relatively depleted of naı̈ve B cells, activated
mast cells, neutrophils, monocytes, gamma delta T cells, and activated M2 macrophages and en-
riched with memory B cells, suppressive Treg cells, resting mast and DCs, and undifferentiated
M0 macrophages compared with normal livers. These data clearly indicate the broadly immunosup-
pressed microenvironment in a subset of HCCs, in which tumors induce tolerance in the immune

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1015

Trends in Molecular Medicine

Trends in Molecular Medicine

Figure 3. Summary of the TGF-b Signature in Hepatocellular Carcinoma (HCC) and Its Overlap with Immunological and Other Molecular
Classifications.
Four distinct clusters with unique TGF-b signatures were identified based on mRNA expression profiling of TGF-b superfamily genes using the TCGA HCC
cohort (http://www.cbioportal.org/) [17]. Other TGF-b molecular classifications and the stromal/fibrosis signature were derived from the late TGF-b and
TGF-b response signatures (TBRSs), the activated-stroma signature and the fibrosis gene signature. Activated tumor-infiltrating immune cells were
characterized by the presence of CD8+ T cells, cytotoxic cells, T/NK metagenes, T helper-1 (Th1) cells, and macrophages. Overlap with the immune class
signature was assessed. Overlap with IFN subclass was calculated and listed for the prediction of success with PD-1 blockage therapy. The DNA repair
signature and increase in oncogenic gene expression were identified from previous studies [17]. Integrating the TGF-b signature with immune
classifications and other molecular characteristics of HCC may help to individualize treatment options for biomarker-derived single or combination therapies.

cells that have already infiltrated and limit the invasion of the primary immune cell types with potent
antitumor activity [34].

To dissect the detailed characteristics of HCC-infiltrating lymphocytes in the highly complex liver tu-
mor microenvironment, single-cell transcriptome sequencing was performed on more than 5000 in-
dividual T cells isolated from blood, tumor, and adjacent normal tissues obtained from six individual
treatment-naı̈ve HCC patients [31]. This study demonstrated a differential distribution of T cell sub-
types in the HCC microenvironment compared with blood or adjacent tissue and identified the
markers CTLA-4, PD-1, LAYN, and HAVCR2 (TIM3) as associated with the tumor-infiltrating Treg cells
and exhausted CD8+ T cells.

1016 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

Understanding of the HCC immune landscape was further refined in a study that found that
approximately 25% of HCCs demonstrated a high degree of immune infiltration (so-called ‘im-
mune class’ of HCC), with high expression levels of PD-1/PD-L1 confirmed in an independent
HCC cohort [48]. This subset of HCC cases is likely to be well suited for PD-1-blocking immuno-
therapy. The immune class was further subdivided into an active immune response subtype (65%
of immune class samples), characterized by overexpression of adaptive immune response genes,
and an immune exhaustion subtype (35%), characterized by the presence of immunosuppressive
signals and cells (TGF-b and M2 macrophages). Importantly, patients in the active immune
response cluster showed lower rates of tumor recurrence and improved survival compared with
patients in the exhausted immune response cluster. This study provided a framework for the
development of a diagnostic tool to classify HCC patients and prediction of their response to
immunotherapy.

In summary, these comprehensive genomic analyses provide potential biomarkers and therapeutic
targets to augment HCC immunotherapy responses. A strong association between the TGF-b signa-
ture and the exhausted immune signature in HCC was identified [17,48], suggesting that the TGF-b
pathway is an important immune regulator and biomarker for HCC.

The TGF-b Signature as a Biomarker to Guide HCC Immunotherapy

To dissect the mechanisms by which dysregulated TGF-b signaling in liver cancers facilitates immu-
notherapy escape, comprehensive integrated analyses of TGF-b superfamily genes using the TCGA
HCC cohort were performed [17]. Unsupervised hierarchical clustering of these TGF-b genes in
HCC revealed four distinct clusters with unique TGF-b signatures: highly activated (14.5%), acti-
vated (45.0%), normal (30.6%), and inactivated (9.9%) (Figure 3) (http://www.cbioportal.org/)
[17,35–37].

Highly Activated TGF-b Signature

Gene expression from samples classified as having the highly activated TGF-b signature nearly
perfectly overlapped with the expression pattern observed in cases of the exhausted immune
class of HCC, a particularly immunoresistant subtype with a poor prognosis [48], and was most
distinct from the gene expression typical for the so-called IFN subclass, which is associated
with smaller liver tumor size and is a predictor of favorable response to PD-1 blockade therapy
[51]. These findings suggest that HCCs with the highly activated TGF-b signature as well as
those in the strongly related exhausted immune class may be resistant to PD-1 blockade ther-
apy. The highly activated and normal TGF-b signature clusters, in particular the highly activated
cluster, were strongly associated with other TGF-b molecular classifications, such as the late
TGF-b [52] and TGF-b response signatures (TBRSs) [53], and fibrosis/stromal signatures, such
as the ‘activated stroma’ [54] and ‘fibrosis gene’ signatures [17], suggesting the potential of
antifibrotic/anti-inflammatory agents to modulate immunoresistance in this subset of HCC
patients.

Activated TGF-b Signature

The activated TGF-b signature was associated with low levels of activated immune cells and immune
response genes and intermediate levels of the IFN subclass. HCCs with the activated TGF-b signature
would be expected to respond better to the combination of PD-1 blockade and inhibition of TGF-b
than to PD-1 blockade therapy alone.

Normal TGF-b Signature

The normal TGF-b signature was significantly associated with the active immune class (73% of sam-
ples). It also was most strongly associated with the IFN subclass and with higher levels of activated
immune cell and immune response genes, indicating the potential for effective response to immuno-
therapy in this subset without additional TGF-b inhibition.

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1017

Trends in Molecular Medicine

Inactivated TGF-b Signature

Clinician’s Corner
The inactivated TGF-b signature was associated with the ‘immune exclusion’ subtype and the lowest
Although multiple combination
levels of immune cell activation, immune response gene expression, and IFN subclass gene expres-
immunotherapies for HCC are in
sion [48,55]. These findings suggest that immune therapy and/or TGF-b inhibition in this group is un-
Phase II and Phase III trials,
likely to be efficacious. Interestingly, Chen et al. found that HCC cases with an inactivated TGF-b including combinations of
signature (10–20% of patients) experienced a lower survival rate than either the normal group or different immune-checkpoint in-
the activated TGF-b group [17]. However, alternative therapeutic strategies may be possible. hibitors or combinations of an im-
HCCs characterized by the inactivated TGF-b signature demonstrated loss of tumor-suppressor mune-checkpoint inhibitor with
function and decreased DNA repair activity, highlighting the potential of DNA damage-inducing ra- small molecule inhibitors and/or
diation therapies in treating these HCC patients [17,56]. locoregional therapies, under-
standing the detailed mecha-
The specific TGF-b signatures in HCC patients can also be tied to the activation of distinct nisms of resistance to current ther-
apies and the means by which new
oncogenic pathways. These oncogenes may help to explain the distinct responses of HCC sub-
drug combinations may overcome
types to therapies and offer directions for future therapy. HCC tumors of the highly activated
these roadblocks should remain a
TGF-b/exhausted immune class were characterized by elevated levels of KRAS, VEGFA, insulin- priority. Rigorous basic and/or
like growth factor 2 (IGF2), mechanistic target of rapamycin kinase (MTOR), the MDM2 proto-onco- translational research and valida-
gene, and STAT3/4/5A [57], leading several groups to speculate that these pathways are respon- tion will be key to developing
sible for the marked immunoresistance of the exhausted immune class HCC subtypes [13]. This effective new combination thera-
understanding spurred Phase I clinical trials in which a combination of an anti-VEGF inhibitor (bev- pies while mitigating the very
acizumab; ClinicalTrials.gov NCT03382886) or a multikinase inhibitor (lenvatinib; ClinicalTrials.gov real risk of inducing liver failure
NCT03006926) was administered together with immune-checkpoint inhibitors. This combination and death.
therapy led to significantly increased partial response rates, up to 62% or 46%, respectively, for The identification and validation
bevacizumab and lenvatinib. of biomarkers predictive of immu-
notherapy response remains an
In summary, findings from recent studies strongly suggest that the TGF-b immune signature unmet clinical need in HCC. PD-
may be used as a predictive biomarker for response to immunotherapy. Inhibition of TGF-b activity 1/PD-L1 histological expression
may improve the responses of certain immunoresistant HCC patients to immune-checkpoint is not useful in predicting the
inhibitors. treatment response to immune-
checkpoint inhibitors. Specific
HCC subtypes can be identified
HCC Immunotherapies in Clinical Use and Under Development by integrating multiple genomic
Single-Agent Immunotherapies immune-profiles derived from tu-
The best-studied HCC immunotherapeutic approaches to date have focused on the blockade of mor biopsy, including immune
class signatures, TGF-b signa-
immune checkpoints using antibodies to target PD-1, PD-L1, and CTLA-4 [10–12,58]. Other
tures, and Wnt/CTNNB1 molecu-
approaches under investigation include: chimeric antigen receptor (CAR)-transduced or T cell
lar classification. These subtypes
receptor (TCR)-transduced T cell therapy, in which genetically engineered T cells are trained have the potential to predict
to target specific tumor antigens such as glypican-3 (GPC3) or alpha-fetoprotein (AFP) on tumor prognosis and guide therapeutic
cells [59–61]; cytokine-induced killer cell (CIK) therapy, a technique in which T cells, NK cells, or strategies.
NKT cells are activated and expanded ex vivo and injected back into affected cancer patients
Several immune-checkpoint in-
[62]; and DC-based ‘vaccinations’, in which patient DCs are pulsed with tumor lysate or AFP
hibitors have been approved for
peptides from HCC to boost an immune response (Figure 4) [63,64]. Although early data have
use in advanced HCC. However,
demonstrated basic safety and tolerability for these therapies in HCC, the immune response their mechanism of action sug-
rates to date have been low and have failed to demonstrate a robust improvement in overall sur- gests that they are likely to also
vival [13]. be effective against early-stage
disease. For example, preopera-
Combination TGF-b Immunotherapies tive immunotherapy may poten-
tially be used to downstage a
The combination of targeted ‘conventional’ therapy and immunotherapy is emerging as one of the
lesion to make it surgically resect-
most promising tools to boost the responses of the immune system against HCC-derived neoanti-
able or administered with chemo-
gens [13]. Combination therapies that have been reported in Phase II and Phase III trials all pair therapy to exploit the increased
blockage of an immune checkpoint (using inhibitors against PD-1, PD-L1, or CTLA-4) with other first- tumor lysis and antigen release
or second-line therapies under investigation, including an oncolytic virus, small molecule inhibitors, that may provoke a more robust
and ablative therapies [13]. For example, a recent clinical trial (ClinicalTrials.gov NCT01853618) immune response. The optimal
demonstrated that administration of tremelimumab (anti-CTLA-4) in combination with traditional lo- timing of immunotherapy and/or
coregional therapies including radiofrequency ablation or chemoablation in advanced HCC patients combination therapies must be
leads to the accumulation of activated CD8+ T cells in tumors. These results suggest that local tumor better investigated.

1018 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

The majority of HCCs develop due

to underlying chronic liver dis-
ease; in particular, liver fibrosis
and cirrhosis. However, the lack
of a detailed mechanistic under-
standing of how cirrhosis de-
velops into HCC has complicated
the development of effective
treatments. Existing screening
tools are limited and no effective
treatments for liver cirrhosis exist.
According to genomic data, TGF-
b signaling is significantly associ-
ated with liver fibrosis and inflam-
mation and plays an important
role in controlling the liver inflam-
matory immune microenviron-
ment. A better understanding of
the role of this signaling pathway
in the development of cirrhosis
may guide future HCC prevention
strategies.

Trends in Molecular Medicine

Figure 4. Inhibitory Effects of TGF-b Signaling on Existing Immunotherapeutic Approaches for

Hepatocellular Carcinoma (HCC).
The TGF-b signaling pathway has the potential to frustrate current immunotherapeutic approaches for HCC. TGF-b
may inhibit PD-1/PD-L1 antibody-based immunotherapy by upregulation of PD-1. TGF-b facilitates the expansion
of Treg cells to interrupt CTLA-4 antibody-based immunotherapy. TGF-b may directly inhibit or cooperate with
Treg and other factors to inhibit CTL activity, resulting in an inhibitory effect on chimeric antigen receptor-
transduced T cell (CAR-T)/T cell receptor-transduced T cell (TCR-T) therapy. The inhibitory effect of TGF-b on
dendritic cells (DCs), natural killer (NK) cells, or natural killer T (NKT) cells may interrupt cytokine-induced killer
(CIK) cell therapies and DC-based ‘vaccinations’. GPC3, glypican-3; AFP, alpha-fetoprotein.

ablation may efficiently disrupt the immunosuppressive microenvironment, release tumor-associated

antigens from apoptotic or necrotic HCC tissue, and synergistically enhance the immune-enhancing
effect of anti-CTLA-4 treatment in HCC [65]. Combining cytotoxic therapies with immunotherapies
may create a more favorable microenvironment for the interaction of immune cells with the tumor
antigenic environment.

Immunotherapies against HCC integrating TGF-b blockade have earned particular attention
because of the critical immunosuppressive role of TGF-b, detailed above, in promoting Treg
cells in the tumor microenvironment and limiting immune activation (Figure 1). TGF-b also
directly suppresses the activation of DCs, tumor-infiltrating T cells, and NK and NKT cells
and antagonizes all existing immune-based strategies against HCC [15] (Figure 4). Pharmaco-
logical TGF-b-targeting agents include receptor kinase inhibitors and neutralizing antibodies
that inhibit the interactions of TGF-b ligands with their receptors, and early clinical trials

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1019

Trends in Molecular Medicine

have already demonstrated potent antitumor effects for many cancers, including HCC, with an
Outstanding Questions
acceptable safety profile [66].
Which HCC patients will benefit
from immunotherapy? Can immu-
Recently, the combination of anti-TGF-b antibodies or a systemic TGF-b receptor 1 inhibitor (galu-
notherapy be used for early- or in-
nisertib) with an immune-checkpoint inhibitor against PD-L1 successfully reduced TGF-b signaling termediate-stage disease? What is
in stromal cells, facilitated T cell penetration into the center of tumors, and provoked vigorous the optimal timing of administra-
antitumor immunity and tumor regression in mouse models [67,68]. Moreover, a bifunctional fusion tion of immunotherapy against
protein (M7824), combining a monoclonal antibody against PD-L1 with the extracellular domain of HCC? How can these therapies be
TGF-b receptor II, was recently developed. This fusion protein blocked signaling from the immune combined with other traditional
checkpoint PD-L1 surface protein and decreased TGF-b signaling within the tumor microenviron- therapies (surgical resection, lo-
ment by binding all three TGF-b isoforms. Its dual anti-immunosuppressive function resulted in coregional therapy, chemotherapy)
the activation of both the innate and the adaptive immune system, upregulation of tumor cell to improve response?

PD-L1 levels, and the induction of tumor regression in mouse models [69,70]. These encouraging Can inhibition of TGF-b, which reg-
results highlight the centrality of TGF-b signaling in regulation of the antitumor immune response ulates both growth and differentia-
and offer a glimpse of the translational potential of immunotherapeutics targeting TGF-b signaling tion, block the tumor-promoting ef-
for HCC. fects of this pathway without also
blocking its tumor-suppressive
functions? What are the effects of
Concluding Remarks TGF-b inhibition on noncanonical
SMAD-independent oncogenic
Immune-checkpoint inhibitors have revolutionized the treatment of multiple tumors, including mel-
signaling in HCC? Does crosstalk
anoma, lung cancer, and urothelial carcinoma [71]. However, tumor response rates for these im-
from pathways downstream of
mune-checkpoint inhibitors are less than 20% in HCC [10–12]. Tumor immunogenicity and features TGF-b signaling impair HCC tumor
of the tumor microenvironment are well-established determinants of liver cancer responses to im- growth? Can TGF-b inhibition un-
mune-checkpoint inhibitors, yet the underlying mechanisms are not fully understood. Thus, there lock a more robust response from
remains an urgent need for an improved understanding of the critical signaling pathways facili- the tumor immune
tating the immune escape process, the development of novel approaches for modeling liver cancer microenvironment?
immunity, and the establishment of biomarkers to evaluate the immune response against liver
How can levels of TGF-b signaling
cancers. be measured in the tumor microen-
vironment? Do the serum levels of
Recent studies have begun to clarify the basis of immunotherapy resistance in HCC. Single-cell RNA- TGF-b ligands accurately reflect
seq analyses have demonstrated the enrichment of exhausted tumor-infiltrating T cells in HCC [31]. their systemic levels? How can
Immune classification studies demonstrated the existence of an immune exclusion class representing TGF-b signaling be assessed in
about 30% of HCCs and characterized by Wnt/CTNNB1 mutations [48,72], although it remains unclear the clinic and reliably titrated?
how Wnt signaling drives this phenotype in HCCs. Last, the recognition of the TGF-b signature has Is there a common mechanism un-
spawned intense efforts to treat certain classes of HCC with a combination of immunotherapeutic ap- derlying immunotherapy resistance
proaches and TGF-b pathway inhibition [17]. among all HCC etiologies? Do un-
derlying liver diseases such as
Although TGF-b signaling pathways have a strong suppressive function on immune cells in the tu- cirrhosis, NAFLD/NASH, chronic
mor inflammatory microenvironment, they are also critical for immune homeostasis. Systemic or hepatitis, or alcohol-associated
T cell-conditional deficiency of TGF-b1, its receptors, or SMAD2/3 in mice led to the development liver disease disrupt drug delivery
of multifocal inflammatory disease, autoimmunity, and death soon after birth [15]. Deficiency of hu- against HCC?

man TGF-b1 is known to cause severe inflammatory bowel disease [73], raising concern for the pros- How do oncologists best classify a
pect of combination therapy inhibiting both TGF-b and immune checkpoints to cause severe im- particular HCC into any of the exist-
mune-related adverse events (irAEs). Given this concern, clinical experimentation with a ing nonoverlapping genomic pro-
combination of TGF-b inhibition and immune checkpoint blockade must be undertaken with care files? How can this classification
(see Outstanding Questions). Importantly, HCC patients in the ‘normal’ TGF-b signature classifica- be used to determine an optimal
tion have the highest survival rate among all of the TGF-b signatures, suggesting that it may be bet- immunotherapeutic approach?
Can artificial intelligence (AI) sys-
ter to normalize the TGF-b level in the tumor microenvironment than to completely suppress it
tems be developed to help clini-
[17,74].
cians manage HCC classification
and therapeutic selection?
A better understanding of the crosstalk between cancer cells and their immune microenvironment,
early recognition and treatment of side effects, and biomarker-based stratification of patients who
are likely to respond poorly or favorably will be critical for the development of safe combination ther-
apeutic strategies for HCC. Rigorous basic and/or translational research and validation will be key to
developing effective new combination patterns. We anticipate inhibition of TGF-b pathway signaling

1020 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

to be a central component of many future immunotherapeutic strategies against refractory malig-

nancies and look forwards to the possibility of dramatically improved survival among affected
patients.

Acknowledgments
We thank Dr John R. Stroehlein for critical review and helpful suggestions on the manuscript. We
thank Dr Xiaofan Lu for technical support. J.C. is an awardee supported by the AASLD Foundation
Pinnacle Research Award in Liver Disease, the Texas Medical Center Digestive Diseases Center
Pilot/Feasibility Project, and the Internal Research Grant (IRG) Program at The University of Texas
MD Anderson Cancer Center.

Supplemental Information
Supplemental information associated with this article can be found online at https://doi.org/10.1016/
j.molmed.2019.06.007.

References
1. Kulik, L. and El-Serag, H.B. (2019) Epidemiology and 13. Greten, T.F. et al. (2019) Targeted and immune-
management of hepatocellular carcinoma. based therapies for hepatocellular carcinoma.
Gastroenterology 156, 477–491.e1 Gastroenterology 156, 510–524
2. Llovet, J.M. et al. (2018) Molecular therapies and 14. Li, M.O. and Flavell, R.A. (2008) TGF-b: a master of all
precision medicine for hepatocellular carcinoma. T cell trades. Cell 134, 392–404
Nat. Rev. Clin. Oncol. 15, 599–616 15. Chen, W. and Ten Dijke, P. (2016) Immunoregulation
3. European Association for the Study of the Liver. by members of the TGFb superfamily. Nat. Rev.
(2017) EASL 2017 Clinical Practice Guidelines on the Immunol. 16, 723–740
management of hepatitis B virus infection. 16. Jenne, C.N. and Kubes, P. (2013) Immune
J. Hepatol. 67, 370–398 surveillance by the liver. Nat. Immunol. 14, 996–1006
4. Marrero, J.A. et al. (2018) Diagnosis, staging, and 17. Chen, J. et al. (2018) Analysis of genomes and
management of hepatocellular carcinoma: 2018 transcriptomes of hepatocellular carcinomas
Practice Guidance by the American Association identifies mutations and gene expression changes in
for the Study of Liver Diseases. Hepatology 68, the transforming growth factor-beta pathway.
723–750 Gastroenterology 154, 195–210
5. Llovet, J.M. et al. (2008) Sorafenib in advanced 18. Achyut, B.R. and Yang, L. (2011) Transforming growth
hepatocellular carcinoma. N. Engl. J. Med. 359, factor-beta in the gastrointestinal and hepatic tumor
378–390 microenvironment. Gastroenterology 141, 1167–
6. Kudo, M. et al. (2018) Lenvatinib versus sorafenib 1178
in first-line treatment of patients with 19. Majumdar, A. et al. (2012) Hepatic stem cells and
unresectable hepatocellular carcinoma: a transforming growth factor beta in hepatocellular
randomised Phase 3 non-inferiority trial. Lancet 391, carcinoma. Nat. Rev. Gastroenterol. Hepatol. 9,
1163–1173 530–538
7. Bruix, J. et al. (2017) Regorafenib for patients with 20. Carambia, A. et al. (2014) TGF-b-dependent
hepatocellular carcinoma who progressed on induction of CD4+CD25+Foxp3+ Tregs by liver
sorafenib treatment (RESORCE): a randomised, sinusoidal endothelial cells. J. Hepatol. 61,
double-blind, placebo-controlled, Phase 3 trial. 594–599
Lancet 389, 56–66 21. Zaiss, D.M. et al. (2013) Amphiregulin enhances
8. Abou-Alfa, G.K. et al. (2018) Cabozantinib in patients regulatory T cell-suppressive function via the
with advanced and progressing hepatocellular epidermal growth factor receptor. Immunity 38,
carcinoma. N. Engl. J. Med. 379, 54–63 275–284
9. Zhu, A.X. et al. (2019) Ramucirumab after sorafenib in 22. Prieto, J. et al. (2015) Immunological landscape and
patients with advanced hepatocellular carcinoma immunotherapy of hepatocellular carcinoma. Nat.
and increased alpha-fetoprotein concentrations Rev. Gastroenterol. Hepatol. 12, 681–700
(REACH-2): a randomised, double-blind, 23. Ringelhan, M. et al. (2018) The immunology of
placebo-controlled, Phase 3 trial. Lancet Oncol. 20, hepatocellular carcinoma. Nat. Immunol. 19, 222–232
282–296 24. Chen, W. et al. (2003) Conversion of peripheral
10. El-Khoueiry, A.B. et al. (2017) Nivolumab in patients CD4+CD25- naı̈ve T cells to CD4+CD25+ regulatory
with advanced hepatocellular carcinoma (CheckMate T cells by TGF-b induction of transcription factor
040): an open-label, non-comparative, Phase 1/2 Foxp3. J. Exp. Med. 198, 1875–1886
dose escalation and expansion trial. Lancet 389, 25. Chen, K.J. et al. (2011) Selective recruitment of
2492–2502 regulatory T cell through CCR6–CCL20 in
11. Zhu, A.X. et al. (2018) Pembrolizumab in patients with hepatocellular carcinoma fosters tumor progression
advanced hepatocellular carcinoma previously and predicts poor prognosis. PLoS One 6, e24671
treated with sorafenib (KEYNOTE-224): a non- 26. Flecken, T. et al. (2014) Immunodominance and
randomised, open-label Phase 2 trial. Lancet Oncol. functional alterations of tumor-associated antigen-
19, 940–952 specific CD8+ T-cell responses in hepatocellular
12. Sangro, B. et al. (2013) A clinical trial of CTLA-4 carcinoma. Hepatology 59, 1415–1426
blockade with tremelimumab in patients with 27. Josefowicz, S.Z. et al. (2012) Regulatory T cells:
hepatocellular carcinoma and chronic hepatitis C. mechanisms of differentiation and function. Annu.
J. Hepatol. 59, 81–88 Rev. Immunol. 30, 531–564

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1021

Trends in Molecular Medicine

28. Sharma, P. and Allison, J.P. (2015) The future of 51. Chiang, D.Y. et al. (2008) Focal gains of VEGFA and
immune checkpoint therapy. Science 348, 56–61 molecular classification of hepatocellular carcinoma.
29. Park, B.V. et al. (2016) TGFb1-mediated SMAD3 Cancer Res. 68, 6779–6788
enhances PD-1 expression on antigen-specific T cells 52. Coulouarn, C. et al. (2008) Transforming growth
in cancer. Cancer Discov. 6, 1366–1381 factor-beta gene expression signature in mouse
30. Barber, D.L. et al. (2006) Restoring function in hepatocytes predicts clinical outcome in human
exhausted CD8 T cells during chronic viral infection. cancer. Hepatology 47, 2059–2067
Nature 439, 682–687 53. Calon, A. et al. (2012) Dependency of colorectal
31. Zheng, C. et al. (2017) Landscape of infiltrating T cells cancer on a TGF-b-driven program in stromal cells for
in liver cancer revealed by single-cell sequencing. metastasis initiation. Cancer Cell 22, 571–584
Cell 169, 1342–1356.e16 54. Moffitt, R.A. et al. (2015) Virtual microdissection
32. Korkut, A. et al. (2018) A pan-cancer analysis reveals identifies distinct tumor- and stroma-specific
high-frequency genetic alterations in mediators of subtypes of pancreatic ductal adenocarcinoma. Nat.
signaling by the TGF-b superfamily. Cell Syst. 7, 422– Genet. 47, 1168–1178
437.e7 55. Pinyol, R. et al. (2019) Immune exclusion-Wnt/
33. Gingold, J.A. et al. (2018) Genomic profiling and CTNNB1 class predicts resistance to
metabolic homeostasis in primary liver cancers. immunotherapies in HCC. Clin. Cancer Res. 25, 2021–
Trends Mol. Med. 24, 395–411 2023
34. Cancer Genome Atlas Research Network. (2017) 56. Marquardt, J.U. (2018) The role of transforming
Comprehensive and integrative genomic growth factor-beta in human hepatocarcinogenesis:
characterization of hepatocellular carcinoma. Cell mechanistic and therapeutic implications from an
169, 1327–1341.e23 integrative multiomics approach. Gastroenterology
35. Chen, J. et al. (2016) Mutations of chromatin structure 154, 17–20
regulating genes in human malignancies. Curr. 57. Heppler, L.N. and Frank, D.A. (2017) Targeting
Protein Pept. Sci. 17, 411–437 oncogenic transcription factors: therapeutic
36. Gao, J. et al. (2013) Integrative analysis of complex implications of endogenous STAT inhibitors. Trends
cancer genomics and clinical profiles using the Cancer 3, 816–827
cBioPortal. Sci. Signal. 6, pl1 58. Greten, T.F. and Sangro, B. (2017) Targets for
37. Cerami, E. et al. (2012) The cBio cancer genomics immunotherapy of liver cancer. J. Hepatol. 68,
portal: an open platform for exploring 157–166
multidimensional cancer genomics data. Cancer 59. Gao, H. et al. (2014) Development of T cells
Discov. 2, 401–404 redirected to glypican-3 for the treatment of
38. Yao, Z.X. et al. (2010) Epigenetic silencing of beta- hepatocellular carcinoma. Clin. Cancer Res. 20,
spectrin, a TGF-b signaling/scaffolding protein in a 6418–6428
human cancer stem cell disorder: Beckwith– 60. Dargel, C. et al. (2015) T cells engineered to express a
Wiedemann syndrome. J. Biol. Chem. 285, 36112– T-cell receptor specific for glypican-3 to recognize
36120 and kill hepatoma cells in vitro and in mice.
39. Chen, J. et al. (2017) Loss of the transforming growth Gastroenterology 149, 1042–1052
factor-beta effector b2-spectrin promotes genomic 61. Sun, L. et al. (2016) Engineered cytotoxic T
instability. Hepatology 65, 678–693 lymphocytes with AFP-specific TCR gene for
40. Chen, J. et al. (2016) TGF-b/b2-spectrin/CTCF- adoptive immunotherapy in hepatocellular
regulated tumor suppression in human stem cell carcinoma. Tumour Biol. 37, 799–806
disorder Beckwith–Wiedemann syndrome. J. Clin. 62. Lee, J.H. et al. (2015) Adjuvant immunotherapy with
Invest. 126, 527–542 autologous cytokine-induced killer cells for
41. Jiang, Y. et al. (2019) Proteomics identifies new hepatocellular carcinoma. Gastroenterology 148,
therapeutic targets of early-stage hepatocellular 1383–1391.e6
carcinoma. Nature 567, 257–261 63. Palmer, D.H. et al. (2009) A Phase II study of adoptive
42. David, C.J. et al. (2016) TGF-b tumor suppression immunotherapy using dendritic cells pulsed with
through a lethal EMT. Cell 164, 1015–1030 tumor lysate in patients with hepatocellular
43. Gomes, A.L. et al. (2016) Metabolic inflammation- carcinoma. Hepatology 49, 124–132
associated IL-17A causes non-alcoholic 64. Butterfield, L.H. et al. (2006) A Phase I/II trial testing
steatohepatitis and hepatocellular carcinoma. immunization of hepatocellular carcinoma patients
Cancer Cell 30, 161–175 with dendritic cells pulsed with four alpha-
44. Zhang, Y.E. (2017) Non-Smad signaling pathways of fetoprotein peptides. Clin. Cancer Res. 12, 2817–
the TGF-b family. Cold Spring Harb. Perspect. Biol. 9, 2825
a022129 65. Duffy, A.G. et al. (2017) Tremelimumab in
45. Guo, X. and Wang, X.F. (2009) Signaling cross-talk combination with ablation in patients with advanced
between TGF-b/BMP and other pathways. Cell Res. hepatocellular carcinoma. J. Hepatol. 66, 545–551
19, 71–88 66. Colak, S. and Ten Dijke, P. (2017) Targeting TGF-b
46. Hoshida, Y. et al. (2008) Gene expression in fixed signaling in cancer. Trends Cancer 3, 56–71
tissues and outcome in hepatocellular carcinoma. 67. Tauriello, D.V.F. et al. (2018) TGFb drives immune
N. Engl. J. Med. 359, 1995–2004 evasion in genetically reconstituted colon cancer
47. Hou, J. et al. (2014) Hepatic RIG-I predicts survival metastasis. Nature 554, 538–543
and interferon-alpha therapeutic response in 68. Mariathasan, S. et al. (2018) TGFb attenuates tumour
hepatocellular carcinoma. Cancer Cell 25, 49–63 response to PD-L1 blockade by contributing to
48. Sia, D. et al. (2017) Identification of an immune- exclusion of T cells. Nature 554, 544–548
specific class of hepatocellular carcinoma, based on 69. Lan, Y. et al. (2018) Enhanced preclinical antitumor
molecular features. Gastroenterology 153, 812–826 activity of M7824, a bifunctional fusion protein
49. Bindea, G. et al. (2013) Spatiotemporal dynamics of simultaneously targeting PD-L1 and TGF-b. Sci.
intratumoral immune cells reveal the immune Transl. Med. 10, eaan5488
landscape in human cancer. Immunity 39, 782–795 70. David, J.M. et al. (2017) A novel bifunctional anti-PD-
50. Alistar, A. et al. (2014) Dual roles for immune L1/TGF-b Trap fusion protein (M7824) efficiently
metagenes in breast cancer prognosis and therapy reverts mesenchymalization of human lung cancer
prediction. Genome Med. 6, 80 cells. Oncoimmunology 6, e1349589

1022 Trends in Molecular Medicine, November 2019, Vol. 25, No. 11

Trends in Molecular Medicine

71. Ribas, A. and Wolchok, J.D. (2018) Cancer 77. David, C.J. and Massague, J. (2018) Contextual
immunotherapy using checkpoint blockade. Science determinants of TGFb action in development,
359, 1350–1355 immunity and cancer. Nat. Rev. Mol. Cell Biol. 19,
72. Harding, J.J. et al. (2018) Prospective genotyping of 419–435
hepatocellular carcinoma: clinical implications of 78. Principe, D.R. et al. (2014) TGF-b: duality of function
next-generation sequencing for matching patients to between tumor prevention and carcinogenesis.
targeted and immune therapies. Clin. Cancer Res. 25, J. Natl. Cancer Inst. 106, djt369
2116–2126 79. Martinez, G.J. et al. (2010) Smad2 positively regulates
73. Kotlarz, D. et al. (2018) Human TGF-b1 deficiency the generation of Th17 cells. J. Biol. Chem. 285,
causes severe inflammatory bowel disease and 29039–29043
encephalopathy. Nat. Genet. 50, 344–348 80. Flavell, R.A. et al. (2010) The polarization of immune
74. Sanmamed, M.F. and Chen, L. (2018) A paradigm cells in the tumour environment by TGFb. Nat. Rev.
shift in cancer immunotherapy: from enhancement to Immunol. 10, 554–567
normalization. Cell 175, 313–326 81. Shalapour, S. et al. (2017) Inflammation-induced IgA+
75. Barrett, D.M. et al. (2015) Chimeric antigen receptor- cells dismantle anti-liver cancer immunity. Nature
and TCR-modified T cells enter Main Street and Wall 551, 340–345
Street. J. Immunol. 195, 755–761 82. de Gramont, A. et al. (2017) Novel TGF-b inhibitors
76. Massague, J. (2012) TGFb signalling in context. Nat. ready for prime time in onco-immunology.
Rev. Mol. Cell Biol. 13, 616–630 Oncoimmunology 6, e1257453

Trends in Molecular Medicine, November 2019, Vol. 25, No. 11 1023