Introduction

Hepatocellular carcinoma (HCC) is commonly associated with liver cirrhosis and therapeutic decisions should consider both tumour burden and the severity of underlying liver disease. In early stages of HCC (Barcelona Clinic Liver Cancer [BCLC] stages 0 and A), surgical and percutaneous therapies such as liver transplantation, resection or radiofrequency ablation (RFA) can be applied, although they are associated with substantial tumour recurrence rates.1 Multinodular intermediate-stage HCC (BCLC stage B) can be treated by transarterial chemoembolization (TACE) or radioembolization, but overall survival is usually <20 months.2,3 Tumours in the advanced stage (BCLC stage C) have a dismal prognosis if untreated, with a median overall survival of 7 months.4 Survival can be increased by 3 months with the multikinase inhibitor sorafenib, the only systemic therapy approved for HCC.1 Clearly, great need still exists for novel approaches to treat HCC.

The field of cancer immunotherapy is moving fast because of encouraging clinical results obtained with monoclonal antibodies (mAbs) directed to block molecules that negatively regulate T-cell responses, in particular cytotoxic T-lymphocyte protein 4 (CTLA-4, also known as CD152), programmed cell death protein 1 (PD-1) and its ligand PD-L1.5,6,7 Such molecules inhibit T-cell activation and promote a state of T-cell dysfunction known as exhaustion.8 Immune checkpoint inhibitors, such as ipilimumab (anti-CTLA-4), nivolumab (anti-PD-1) and pembrolizumab (anti-PD-1), which have already received approval from regulatory agencies, have changed the landscape of cancer therapy. In HCC, immunotherapy is an appealing option to reduce the risk of relapse by eliminating micrometastatic residual disease after surgical or percutaneous ablation. In addition, immunotherapy, alone or in combination with transarterial or systemic therapies, might prolong survival by providing an effective control of tumour cell growth when ablation is not feasible. In this Review, we will discuss the antigenicity of HCC, the interaction of HCC with the immune system, current immunotherapeutic approaches and future directions.

Carcinogenesis and HCC antigenicity

In chronically inflamed livers, genetic and epigenetic changes underlie oncogenic transformation. These events also deregulate the expression of oncofetal, and cancer/testis proteins,8 which constitute tumour-associated antigens (TAAs) capable of eliciting defensive immune responses. TAAs include the oncofetal antigens α-fetoprotein and glypican-3 (GPC3). Cancer/testis antigens comprise cancer/testis antigen 1 (NY-ESO-1),8 melanoma-associated antigen 1 (MAGEA),8 synovial sarcoma, X breakpoint 2 (SSX2)9 and the 'stemness' molecule telomerase reverse transcriptase (hTERT).10 In addition, genome-wide sequencing has revealed that HCC, similar to other common solid tumours, exhibits nonsynonymous, somatic mutations in 30–50 genes11 that can confer antigenicity to the encoded protein. Some of the mutations might lead to growth advantages and are called 'driver' mutations, whereas other genetic variations have unknown oncogenic roles and are termed 'passenger' mutations.11 Contrary to TAAs, which are shared antigens, neo-antigens that develop from nonsynonymous mutations in the tumour are private antigens that have never been perceived by the immune system before oncogenesis. Neo-antigens are authentic tumour-specific antigens that can be exploited to induce therapeutic antitumour immune responses.

Hepatocarcinogenesis is considered a multistep process evolving on a background of liver cirrhosis, a preneoplastic condition. Successive acquisition of driver mutations over the years runs in parallel with the progression of nontransformed cells to low-grade dysplastic nodules, high-grade dysplastic nodules, early HCC and finally advanced HCC.12 The tumour mass is heterogeneous and passenger mutations can differ among cancer cells within the same tumour nodule or between the initial HCC and metastatic nodules. However, initial 'founder', driver mutations are common in the cells of primary and metastatic lesions.11 The targeting of these mutations, if they are antigenic, would therefore be an efficient way to control tumour growth.

HCC: immune recognition and escape

HCC immunogenicity is indicated by the presence of tumour-infiltrating lymphocytes13 and an evident reduction in relapse rates after resection and transplantation in patients with dense lymphocytic infiltration.14,15 Discernible anti-TAA CD8+ T-cell responses in the peripheral blood of 50% of patients with HCC are further indicators of the immunogenicity of this cancer type.8 The extent and frequency of these responses are higher in early HCC than in later stages and are associated with patient survival.8 Interestingly, TAA-specific CD8+ lymphocytes from peripheral blood produce IFN-γ upon stimulation, but tumour-infiltrating lymphocytes fail to do so, indicating the exhaustion of intratumour CD8+ T cells.8 Both clinical and preclinical data have revealed a highly immunosuppressive intratumour environment alongside defective effector T-cell recruitment in advanced HCC.16

Immune rejection

Our current understanding of the initial steps of the adaptive immune response include antigen uptake by dendritic cells (DCs) and their migration to the regional lymph nodes, where DCs present the processed antigen to CD4+ T cells on major histocompatibility complex (MHC) class II molecules.17 Antigen recognition stimulates CD4+ T cells to proliferate and produce IFN-γ (a process called TH1 [type 1 T helper cell] polarization) in the presence of DC-derived type I interferon and IL-12, and upon receiving co-stimulatory signals emanating from the binding of CD28 on the lymphocyte membrane to CD80 and CD86 on the DC surface.17,18 Once activated, CD4+ T cells express CD40L which interacts with CD40 on antigen-presenting cells (APC), further promoting IL-12 production and TH1 polarization.19 TH1 cells licence DCs for cross-presentation of the antigenic peptides to CD8+ T cells, which facilitates the development of CD8+ cytotoxic T lymphocytes (CTLs). CTLs enter the circulation and migrate to inflamed tissue, where they interact with their cognate MHC class I-peptide complex on the membrane of the target cells. Antigen-specific CD8+ lymphocytes exert effector functions by producing IFN-γ, which inhibits tumour cell growth, and by displaying cytotoxic activity through the release of granzyme B and perforin, and interaction with FAS (also known as TNF receptor superfamily member 6) and TRAIL (also known as TNF ligand superfamily member 10) receptors on tumour cells.18 Thus, immune rejection of cells expressing foreign antigens involves diverse cells and soluble mediators, with type I interferons, IFN-γ and IL-12 being the central drivers of this type of response. These cytokines also induce potent anti-angiogenic effects.20 In HCC, the concept that tumour elimination requires a TH1 type of response is supported by clinical findings showing that the expression of TH1 cytokines (IL-1α, IL-1β, IL-2 and IFN-γ) in tumour tissue is associated with good prognosis, whereas TH2 cytokines (IL-4, IL-5 and IL-10) are upregulated in advanced HCC with vascular invasion and metastasis.21

Growth promotion and immunosuppression

However, while a major role of immune cells is to counteract invading pathogens, and recognize and control abnormal cell proliferation, they also contribute to tissue regeneration and proliferation. During tissue injury, neutrophils and macrophages are among the first cells attracted to the wound, where they are involved in the elimination of pathogens and the removal of necrotic and apoptotic cells and other debris. As wound healing progresses, alternatively activated macrophages (so-called M2) predominate in the area and produce anti-inflammatory and immunosuppressive cytokines (such as IL-10 and transforming growth factor [TGF]-β), as well as growth factors that stimulate cell proliferation and tissue regeneration (epidermal growth factor [EGF] and insulin-like growth factor [IGF]), angiogenesis (vascular endothelial growth factor [VEGF] and platelet-derived growth factor [PDGF]) and the formation of extracellular matrix (TGF-β).22 Regulatory T cells (TREG cells), which are mainly known for their immunosuppressive functions, also stimulate cell proliferation and extracellular matrix formation in wound repair by producing EGF receptor (EGFR) ligands.23 Besides wound healing, all these cells and factors have also been found in cancer environments.24 In HCC, effector T cells are intermingled with a variety of stromal cells possessing immunosuppressive, cytoprotective, stromagenic and proangiogenic properties. For this reason, continuing progressive tumour growth and invasion despite recognition by the immune system seemingly reflects the preponderance of a persistent abnormal, 'wound healing-like' process over immune-mediated cancer rejection. In this scenario, immune-inhibitory forces and the permanent exposure to tumour antigens cause T-cell exhaustion, a process partly mediated by intratumour expression of immune-inhibitory checkpoint molecules and immunosuppressive factors. Accordingly, it has been proposed that a modification of the tumour microenvironment might potentially restore successful antitumour immunity.24

The players of immunosuppression

Both tumour and stromal cells orchestrate a strongly immunosuppressive tumour milieu, which opposes both the priming of T cells (Figure 1) and immune effector functions (Figure 2). A multiplicity of membrane-linked immune-inhibitory proteins and soluble factors mediate these effects.

Figure 1: Inhibition of DC immunogenic functions by the HCC microenvironment.
figure 1

HCC cells express TAAs and neo-antigens arising from private driver or passenger gene mutations. Following capture of cell debris, DCs process and present tumour antigens to T cells. In the HCC microenvironment, TREG cells block the immunostimulatory functions of DCs through various mechanisms: CTLA-4, constitutively expressed on TREG cells and inducibly expressed on activated effector T cells, prevents CD28 binding to CD80/CD86, thus inhibiting priming and expansion of CD4+ and CD8+ lymphocytes. CTLA-4 also induces reverse CD80/CD86 signalling in DCs, leading to the induction of the immunosuppressant molecules IL-10 and IDO. Intratumour DCs express both PD-1 and PD-L1. PD-1–PD-L1 ligation induces the production of IL-10 and IDO by DCs. Hypoxia enhances the production of proangiogenic molecules including adenosine (generated from ATP by the hypoxia-induced ectonucleases CD39 and CD73), PDGF, VEGF and lactic acid. The latter stimulates TAMs to release arginase and VEGF. Adenosine and VEGF inhibit APC co-stimulatory functions and promote angiogenesis. Abbreviations: A2AAR, human adenosine receptor A2A; APC, antigen-presenting cell; ARG, arginase; CTLA-4, cytotoxic T-lymphocyte protein 4; CXCL12, stromal cell-derived factor α; CXCR4, CXC chemokine receptor type 4; DC, dendritic cell; FoxP3, forkhead box protein P3; HCC, hepatocellular carcinoma; HIF-1α, hypoxia-inducible factor-1 α; IDO, indoleamine 2,3-dioxygenase; MHC, major histocompatibility complex; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1; PDGF, platelet-derived growth factor; TAA, tumour-associated antigen; TAM, tumour-associated macrophage; TCR, T-cell receptor; TGF-β, transforming growth factor β; TREG cell, regulatory T cell; VEGF, vascular endothelial growth factor.

PowerPoint slide

Figure 2: Inhibition of NK and T cell effector functions by the HCC microenvironment.
figure 2

Upon antigen recognition, IFN-γ released by CD4+ and CD8+ T cells induces PD-L1 on both APCs and tumour cells. PD-1–PD-L1 interactions result in T-cell exhaustion and prevent tumour-cell killing. Besides PD-1–PD-L1, CTLA-4, TIM-3, LAG-3 and BTLA are other membrane-linked T-cell inhibitory molecules. Both HCC cells and TREG cells secrete and respond to the EGFR ligand amphiregulin, which stimulates HCC cell growth and TREG cell activity. Tumour cells release CXCL12, which attracts myeloid and lymphoid cells via CXCR4. MDSCs abrogate NK and T-cell activity via TGF- β and other mechanisms. Myeloid cells mediate immunosuppression by releasing ROS, TGF- β, IL-10, PGE2, MMPs and via the enzymatic actions of arginase and IDO. CAFs generate ECM and maintain inflammation by producing PGE2 and MMPs. Hypoxia induces PD-L1 in HCC and myeloid cells in the tumour microenvironment. In myeloid cells PD-1–PD-L1 gives rise to the release of IL-10, TGF-β and arginase. Adenosine acting via its receptor blocks CD4 and CD8 T cell effector functions and inhibits macrophage activation. IFN-γ stimulates TAMs to secrete galectin-9, which upon binding TIM-3 enhances TREG cell activity and promotes IL-6 secretion by TAMs, which in turn induces IL-10 production by MDSCs. Abbreviations: A2AAR, human adenosine receptor A2A; APC, antigen-presenting cell; BTLA, B and T lymphocyte attenuator; CAF, cancer-associated fibroblast; CTLA-4, cytotoxic T-lymphocyte protein 4; CXCL12, stromal cell-derived factor α; CXCR4, C-X-C chemokine receptor type 4; DC, dendritic cell; ECM, extracellular matrix; EGFR, epidermal growth factor receptor; HCC, hepatocellular carcinoma; HVEM, herpesvirus entry mediator; IDO, indoleamine 2,3-dioxygenase; LAG-3, lymphocyte activation gene 3; NK, natural killer; MDSC, myeloid-derived suppressor cell; MMP, matrix metalloprotease; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1; PDGF, platelet-derived growth factor; PGE2, prostaglandin E2; ROS, reactive oxygen species; TAA, tumour-associated antigen; TAM, tumour-associated macrophage; TCR, T-cell receptor; TGF-β, transforming growth factor β; TIM-3, T-cell immunoglobulin and mucin domain-containing protein-3; TREG cell, regulatory T cell; VEGF, vascular endothelial growth factor.

PowerPoint slide

Tumour and stromal cells

Tumour cells

Cancer cells use cell-autonomous and non-cell-autonomous mechanisms to escape immune responses. Evidence in mouse sarcomas indicates that the immune system exerts a selective pressure on transformed cells, which results in a survival advantage of those cells that are less immunogenic or produce immunosuppressive factors (a phenomenon known as 'immunoediting').18 Tumour sculpting by the immune response is mediated by epigenetic and genetic changes facilitated by the intrinsic genetic instability of malignant cells. Three aspects of tumour biology are conceivably important in tumour evasion from the immune system: the silencing of tumour neo-antigens; altered expression of genes involved in antigen processing and presentation; and abrogation of IFN-γ receptor signalling pathways.18,25 Repressed expression of antigenic molecules probably affects developmental antigens (such as NY-ESO-1) and passenger mutations rather than driver mutations, as these are essential to sustain transformation. However, cancer cells might still evade the immune system despite persisting expression of antigenic molecules owing to defects in antigen processing and presentation. This pathway involves proteasomal protein fragmentation to yield peptides that are transported to the endoplasmic reticulum, loaded onto HLA class I heavy chains and transferred to the cell membrane where they are recognized by the T-cell receptor (TCR) of CD8+ T cells. Defects in proteasomal function or antigen peptide transporter 1 and 2 cause HLA class I downregulation, whereas a complete loss of HLA class I results from mutation or deletion of β2 microglobulin.18,26 Absent or reduced HLA class I expression impairs tumour cell recognition by CTLs, but hinders HLA class I recognition by killer inhibitory receptors on natural killer (NK) cells, thereby unleashing their lytic antitumour activity.27 NKG2D activating receptors on NK cells interact with stress-induced ligands, such as MHC class I polypeptide-related sequence A and B (MIC-A/MIC-B) and the unique long 16 binding protein family (ULBPs). These ligands are upregulated in many tumours in response to genetic stress and DNA damage,28 favouring tumour elimination by innate immunity. However, in progressed tumours the interaction between NKG2D expressed by NK cells and its ligands on HCC cells is disturbed in two ways.29 Firstly, soluble MIC-A is detected in serum from patients with advanced HCC,30 along with an upregulation of disintegrin and metalloproteinase domain-containing protein 9 (ADAM9) in tumour tissue.31 Interaction of soluble MIC-A with its receptor causes downmodulation of NKG2D and impairment of NK cell activity. Secondly, diminished expression of ULBP1 has been found in poorly differentiated HCC. This finding, which prevents killing of HCC cells mediated by NK cells, is associated with tumour recurrence after resection.32 HCC cells also escape innate and adaptive immunity by producing a large number of immunosuppressive molecules including TGF-β, IL-10, IL-8, indoleamine 2,3-dioxygenase (IDO), arginase, adenosine, lactic acid, VEGF, PDGF, EGFR ligands, TREG-cell-attracting chemokines and immunoinhibitory checkpoint molecules.9,33,34,35,36,37,38,39,40,41,42,43 Epigenetic changes, hypoxia and nuclear factor kappa B (NFκB) activation are involved in the induction of some of these immunoinhibitory factors.34,36

MDSCs and M2 macrophages

Myeloid-derived suppressor cells (MDSCs) are a heterogeneous population of immature and immunosuppressive myeloid cells, which display a variety of pro-tumoural effects. They promote tumour angiogenesis through VEGF production44 and subvert both innate and adaptive antitumour immunity.45 MDSCs impair CD4+ and CD8+ T-cell responses via increased arginase activity, leading to arginine depletion,43 and through the production of reactive oxygen and nitrogen species that disrupt TCR signalling.46 MDSCs abrogate hepatic NK-cell activity via membrane-bound TGF-β47 and facilitate the expansion of TREG cells and the induction of TREG cells through IL-10 and TGF-β production. Increased numbers of CD14+ HLA-DR−/low MDSCs have been found in both tumour tissue and peripheral blood from patients with HCC, and elevated cell counts were related to tumour progression.43,44

Macrophages form a continuous spectrum from classically activated (so-called M1) macrophages, which produce high IL-12 and low IL-10 levels, to alternatively activated M2 macrophages characterized by low IL-12 and high IL-10 production. The HCC microenvironment stimulates M2 polarization, the characteristic phenotype of the so-called tumour-associated macrophages (TAMs). Besides their immunosuppressive functions, TAMs support tumour progression by promoting angiogenesis, tumour cell invasion and metastasis.48,49

An intense crosstalk between MDSCs and TAMs takes place in the tumour microenvironment. The former release IL-10, which downregulates IL-12 in TAMs. High IL-10 and low IL-12 levels foster differentiation of CD4+ T cells to a TH2 phenotype with production of IL-4 that in turn induces the development of M2 macrophages. Inflammation, which frequently provides the niche in which HCC progresses, intensifies mutual MDSC–TAM interactions. In this setting, IL-10 production by MDSCs is enhanced as a result of increased IL-6 production by TAMs. High IL-10 levels downregulate HLA class II expression by macrophages (thus impairing antigen presentation), stimulate TREG-cell expansion and block NK-cell activation.45 TGF-β induces the expression of T-cell immunoglobulin and mucin domain-containing protein 3 (TIM-3, also known as hepatitis A virus cellular receptor 2, HAVcr-2) on TAMs, thereby fostering M2 polarization, IL-6 production and tumour promotion.33 CXCL12 (also known as stromal cell-derived factor 1 or SDF-1α), produced by endothelial cells and hepatic stellate cells (HSCs), is instrumental in the chemoattraction of myeloid cells to the tumour in a CXCR4 (CXC chemokine receptor type 4)-dependent manner.50

Other stromal cells

Cancer-associated fibroblasts (CAFs) are essential components of the HCC microenvironment. In other solid tumours CAFs display proangiogenic activities: they recruit endothelial progenitor cells and myeloid cells via CXCL12 and participate in tissue remodelling by producing extracellular matrix and secreting matrix metalloproteinases.51 CAFs produce cyclooxygenase-2, IL-8 and secreted protein acidic and rich in cysteine (SPARC) that stimulates TAMs to release TNF and PDGF, which further promote CAF activation.52 HCC-associated CAFs inhibit NK-cell function by releasing the immunosuppressive molecules prostaglandin E2 (PGE2) and IDO.53

HSCs and myofibroblasts, the main producers of extracellular matrix in the liver, have a key role in HCC progression through hepatocyte growth factor release and induction of both MDSCs and TREG cells.54,55,56 HCC cells have been shown to activate HSCs via secretion of amphiregulin, a pro-oncogenic EGFR ligand,38,57 which is also involved in the induction of TREG cells.58

Endothelial cells express various receptors for angiogenic factors including VEGFR1, VEGFR2, Tie-2 (angiopoietin-1 receptor) and PDGF receptor. The interaction of receptors with corresponding ligands induces proliferation and migration of endothelial cells.59,60 Endothelial cell activation facilitates immunosubversion via TGF-β-dependent TREG-cell induction.61 Tumour-associated endothelial cells express FasL (TNF ligand superfamily member 6), which participates in tumour invasion via initiation of apoptosis and extinction of the surrounding parenchymal tissue and contributes to immune evasion by deleting infiltrating CD8+ T cells without affecting TREG cells.62

DCs have pro-immunogenic functions that are severely deregulated in the tumour microenvironment owing to multiple factors including TREG cells, hypoxia, lactic acid, VEGF, immunosuppressive cytokines and adenosine accumulation (Figure 1).45,59 A subpopulation of CD14+ DCs that express high levels of CTLA-4 have been described in patients with HCC. These DCs are postulated to mediate robust tolerogenic effects through CTLA-4-dependent production of IL-10 and IDO.63

TREG cells are CD4+ T cells characterized by membrane molecules CD25, CTLA-4, and CD62L and they typically express the transcription factor FoxP3. TREG-cell activation is induced by TCR engagement concurrent with IL-10 and TGF-β signalling.22 These cells inhibit immune responses through various mechanisms including: depletion of IL-2 in the extracellular space by CD25 (also known as interleukin-2 receptor subunit α); competition with co-stimulatory CD28 via membrane-bound CTLA-4; downregulation of CD80 and CD86 through CTLA-4-mediated transendocytosis;64 suppression of effector T cells via membrane-bound TGF-β; secretion of TGF-β and IL-10 to the medium;65,66 and generation of adenosine through CD39 and CD73 ectoenzymes expressed on their membrane. Notably, TREG cells secrete and respond to the pro-oncogenic EGFR ligand amphiregulin,58,67 which contributes to optimal TREG-cell function. Evidence indicates that FoxP3+ TREG cells are increased in peripheral blood from patients with HCC and also markedly infiltrate the tumour.39 The abundance of tumour-infiltrating TREG cells positively correlates with the number of intratumoural macrophages68 and is an independent prognostic factor for overall survival.39 TREG-cell accumulation in tumours can be a result of cell recruitment from peripheral blood, proliferation of the resident TREG cell populations or conversion of CD4+ FoxP3 lymphocytes into CD4+ FoxP3+ cells. TREG-cell recruitment to tumours has been found to occur via the CCR6 (CC chemokine receptor type 6)–CCL20 (CC motif chemokine 20) axis.39 Conversely, suboptimal stimulation of naive CD4+ T cells in combination with tumour-derived TGF-β signalling promotes FoxP3 upregulation and subsequent conversion into TREG cells, according to findings in a mouse model of pancreatic cancer.69

Immunosuppressive molecules

Immune checkpoint molecules

Immune checkpoints are coinhibitory molecules that interrupt the immune response to avoid overactivation of T cells and collateral tissue damage. Members of this group include CTLA-4, PD-1, TIM-3, LAG-3 (lymphocyte activation gene 3 protein) and BTLA (B and T lymphocyte attenuator).70,71

CTLA-4 is expressed by activated T cells and is constitutively present on TREG cells (Figure 3). It outcompetes the stimulatory protein CD28 by binding to its ligands CD80 and CD86 on the APC membrane.72 In addition, by engaging its ligands, CTLA-4 conveys inhibitory signals to the T cell that oppose TCR signalling. CTLA-4 is critical for the control of CD4+ T-cell function73 and primarily involved in the priming phase of the immune response. Within the tumour, CTLA-4 further promotes immunosuppression by inducing TREG-cell activity and differentiation9,74 and also by upregulating IDO and IL-10 in DCs through reverse CD80 and CD86 signalling.63

Figure 3: Activation of antitumour responses by the checkpoint-blocking antibody anti-CTLA-4.
figure 3

Upper panel: CTLA-4 is transported to the cell membrane of activated T cells and outcompetes CD28 for binding to CD80/86 expressed on the membrane of APCs, depriving T cells of essential co-stimulatory signals. CTLA-4 blockade reconstitutes productive T-cell responses. Lower panel: CTLA-4 is expressed by TREG cells, which sequester CD80/86 on the APC membrane. CD80/86–CTLA-4 ligation induces APCs to secrete IDO and IL-10, which display strong immunosuppressive activities. CTLA-4-blocking antibodies reinvigorate T-cell responses and alleviate the immunosuppressive tumour environment via ADCC-mediated TREG cell elimination. Abbreviations: ADCC, antibody-dependent direct cytotoxicity; APC, antigen-presenting cell; CTLA-4, cytotoxic T-lymphocyte protein 4; IDO, indoleamine 2,3-dioxygenase; NK, natural killer; TCR, T-cell receptor; TREG cell, regulatory T cell.

PowerPoint slide

PD-1 is expressed by activated CD8+ and CD4+ lymphocytes and also by B cells and NK cells (Figure 2).70 In addition, PD-1 is found on TREG cells, MDSCs, monocytes and DCs.9,75,76 To date, two ligands have been identified for PD-1, named PD-L1 (B7-H1) and PD-L2 (B7-DC).77,78 PD-L1 is expressed by haematopoietic cells, including APCs and MDSCs, and by nonhaematopoietic cells, such as cells of the microvascular endothelium and parenchymal cells of different organs.65 PD-L2 has a more restricted expression profile, which is limited to the haematopoietic compartment.70 The distribution pattern of PD-1 and PD-L1 suggests that this system works as a rheostat of effector T-cell responses in peripheral tissues.65 PD-L1 is upregulated by various cytokines, with IFN-γ being the most potent inducer.65,79,80 PD-L1 is also a direct target of HIF-1α (hypoxia-inducible factor-1α) and hypoxic conditions cause a rapid elevation of PD-L1 expression in MDSCs, macrophages and tumour cells.81 In a setting of chronic antigen exposure, such as the tumour microenvironment, IFN-γ produced by antigen-specific T cells induces PD-1 expression on the cell surface of reactive T lymphocytes and upregulates PD-L1 in APC and tumour cells.9,79 PD-1–PD-L1 engagement blocks TCR signalling, inhibits T-cell proliferation and secretion of cytotoxic mediators, which leads to T-cell exhaustion.82 Moreover, the binding of PD-L1 (expressed on other cells) to its receptor PD-1 on macrophages promotes IL-10 release and thereby CD4+ T-cell repression.76 Effector CD8+ cells within HCC tumours show intense PD-1 expression and the number of PD-1+ CD8+ cells was found to be related to disease progression and postoperative recurrence.83 Similarly, high PD-L1 expression in HCC is associated with tumour aggressiveness and recurrence after resection.84,85 PD-1 and PD-L1 are coexpressed in a subset of tumour-infiltrating DCs, which are characterized by poor response to danger signals and potent immunosuppressive activities.84

TIM-3 is a transmembrane protein expressed on cells of the innate and adaptive immune system that interacts with several ligands including phosphatidylserine on the membrane of apoptotic cells,86 the alarmin HMGB1 (high mobility group protein B1) and galectin-9.87 Galectin-9 is a soluble protein produced by cells from many different tissue types (such as liver, small intestine and thymus) that regulates cell differentiation, adhesion and cell death.88 Evidence indicates that galectin-9 suppresses T-cell responses, which supports the concept that TIM-3 acts as an inhibitory receptor for T cells. Administration of TIM-3 blocking antibodies or the fusion protein TIM-3-Fc increases TH1-driven immune responses.89 TIM-3 is expressed on activated TREG cells and TIM-3+ TREG cells exhibit enhanced suppressor activity.90 In HCC, Li et al.91 observed increased TIM-3 expression in CD4 and CD8 lymphocytes that infiltrated the tumour and found that these cells were replicative senescent. Interestingly, IFN-γ derived from tumour-infiltrating lymphocytes induced galecting-9 expression by Kupffer cells and TIM-3+ T cells co-localized with galectin-9+ Kupffer cells within the tumour.91

LAG-3 is a membrane protein that binds MHC class II molecules with high affinity,70,92 thus reducing co-stimulatory functions of DCs. LAG-3 is not expressed on resting T cells but is upregulated upon activation.93 It is a marker of exhausted T cells,94 acting synergistically with PD-1 to promote cancer evasion from immunity.95 Accordingly, LAG-3 blockade enhances antitumour T-cell responses,96 and dual inhibition of PD-1 and LAG-3 results in synergistic restoration of T cell immunity.97

BTLA is another co-inhibitory molecule that is upregulated on lymphocytes upon activation, and overexpressed by tumour-specific CD8+ T cells in patients with cancer.98 BTLA+ T cells are inhibited in the presence of its ligand HVEM (herpesvirus entry mediator), which is expressed by a variety of tumours including HCC.99 HCCs with high HVEM expression (40% of patients) show reduced lymphocyte infiltration, diminished levels of effector T-cell mediators, more advanced disease, poorer overall survival and increased recurrence rates after resection.99

Other immune-inhibitory factors

IDO depletes the milieu of tryptophan and generates diverse kynurenine metabolites. Through these two effects IDO inhibits T-cell activation and proliferation, promotes TREG-cell function and induces naive CD4+ T cells to become FoxP3+ inducible TREG cells.100,101 IDO is upregulated by IFN-γ and other proinflammatory cytokines and is expressed by macrophages, DCs, endothelial cells, CAFs and cancer cells including HCC.35,53,102 IDO seems to be one of the mechanisms by which activation of antitumour immunity is rerouted to favour tumour promotion. In HCC tumours, activated T cells upregulate IDO in macrophages via IFN-γ production and conditioned media from these cells suppressed T-cell proliferation and functionality by a mechanism that was blocked upon addition of the IDO inhibitor 1-methyl-tryptophan.103

Arginase-1 (ARG1) converts L-arginine into L-ornithine and urea. In myeloid cells, ARG1 mediates inflammation-induced immune suppression. This effect is mainly achieved by depleting the extracellular medium of L-arginine.104 In the hypoxic environment of solid tumours, including HCCs,43 TAMs and MDSCs express high levels of ARG1.

Lactic acid is generated by tumour cells as a by-product of aerobic and anaerobic glycolysis. This factor is a potent immunosuppressant within the tumour microenvironment that acts through HIF-1α to increase ARG1 and VEGF expression in TAMs.40

Adenosine is released by almost all cell types and is also generated extracellularly via ATP breakdown executed by two ectoenzymes that act sequentially, the apyrase CD39 (which hydrolyses ATP or ADP to form AMP) and the 5′-nucleotidase CD73 (converting AMP to adenosine). Extracellular adenosine is maintained in equilibrium by reuptake mechanisms and by degradation to inosine through adenosine deaminase. Adenosine stimulates wound healing, promotes angiogenesis and regulates inflammation by binding to its receptor A2AAR (human adenosine receptor A2a).105 A2AAR signalling inhibits macrophage activation42 and modulates adaptive immunity through inhibition of CD4+ and CD8+ T-cell responses and induction of TREG cells.42,106 TREG cells express high levels of both CD39 and CD73 and therefore also contribute to the generation of adenosine in the tumour microenvironment.107 Targeting A2AAR has emerged as valuable adjunct to cancer immunotherapy.

Galectins comprise a family of 15 lectins, which bind the β-galactoside moiety of glycoconjugates on the cell membrane. Galectin-1, 3 and 9 contribute to tumour immune escape108 and are recognized as indicators of poor prognosis in several cancers including HCC.109,110 Extracellular dimeric galectin-1 is a driver of TH2 polarization and an inducer of IL-10+ FoxP3+ TREG cells.111 Secreted monomeric, homodimeric or tetrameric galectin-3 bind CD29 and CD7 on activated T cells to cause anergy (unresponsiveness) and apoptosis.111 Galectin-3 can abolish the antitumour activity of adoptively transferred tumour-specific CD8+ T cells112 and inhibits NK cell antitumour cytotoxicity.109 The interaction between galectin-3 and LAG-3 also seems to be involved in tumour immunosuppression.113 Galectin-9 is upregulated by proinflammatory cytokines and, when bound to TIM-3, stimulates TREG cells while promoting the exhaustion of effector T cells, as described above. Thus, disruption of galectin–ligand interactions constitutes a potential strategy to enhance antitumour immunity.109

A diverse set of immunomodulatory cytokines and growth factors, including TGF-β, CSF-1 (macrophage colony-stimulating factor 1), VEGF, amphiregulin and the TH2 cytokines IL-4, IL-8 and IL-10, are produced by tumour and stromal cells, generating a strong immunosuppressive HCC microenvironment.21,66 TGF-β and IL-10 are instrumental for the induction of FoxP3+ TREG cells, inhibition of DC immunogenic functions, suppression of TH1 responses and abolition of NK-cell activity.114 IL-4 and CSF-1 are important in the M2 polarization of TAMs.21 IL-8 is a potent protumorigenic factor upregulated by proinflammatory cytokines and hypoxia34,115 that acts directly on malignant cells116 and also through stimulation of angiogenesis and recruitment of myeloid cells to the tumour. VEGF overexpression in HCC is associated with poor prognosis.117 In addition to its angiogenic activity, VEGF exerts immunosuppressive functions by promoting MDSC accumulation, inhibiting DC maturation, inducing TREG cells and enhancing the expression of the immune checkpoint molecules PD-1, TIM-3 and CTLA-4 on CD8+ T cells.118 Amphiregulin is an EGFR ligand highly upregulated in HCC38 that combines direct pro-oncogenic effects on HCC cells with immunosuppressive activity by engaging with the EGFR on the surface of TREG cells.58

HCC immunotherapy

Traditional immunotherapy includes the use of vaccines and adoptive cell therapy (ACT) with cytokine-induced killer cells (CIKs), but three main advances have changed the landscape of cancer immunotherapy during the past decade. First, a better understanding has been gained of tumour biology and immune checkpoint molecules were recognized as the most critical determinants of tumour evasion to immunity.119,120 Second, progress in sequencing technologies now enables fast and affordable whole-genome analysis and the characterization of specific mutations present in individual tumours.11,121 Third, advances in clinical applications of gene therapy enable either ex vivo genetic modification of transferred cells or in vivo transduction of the tumour with vectors carrying genes meant to reinvigorate immune responses.122 These developments have greatly modified our views of cancer immunotherapy.

The antitumour effect of chemotherapeutic agents has usually been assessed morphologically by measuring changes in tumour volume as in the Response Evaluation Criteria in Solid Tumours.123 Immunotherapy acts indirectly and immune responses might take longer to develop, but antitumour effects tend to be more durable than with chemotherapy. An effective immunotherapy provides long-lasting antitumour activity with retardation of tumour growth and longer overall survival than chemotherapy alone. Regretfully, some immunotherapy trials were interrupted prematurely due to a lack of effect on time to progression, although, if continued, we speculate they might have confirmed a survival benefit. Successful trials with vaccines and immune checkpoint inhibitors have clearly shown that immunotherapy can improve overall survival with little or no effect on tumour burden or time to progression.5,124 In some patients, immune activation might increase tumour size owing to inflammation and infiltration by immune cells. These findings have led to the development of new 'immune response criteria' for immunotherapy that await prospective validation.125 To date, various immunotherapeutic modalities are used to treat HCC and a list of immunotherapeutic agents that have received FDA approval is presented in Supplementary Table 1 online.

Cytokines

Recombinant human IFN-α was the first immunotherapy to undergo substantial clinical development for the treatment of HCC, based on its immunostimulatory and anti-angiogenic properties and on the extensive experience gained in the treatment of chronic viral hepatitis. A small randomized trial in patients with advanced-stage tumours showed a statistically nonsignificant prolongation of overall survival.126 A second, larger trial studied the effect of IFN-α as an adjuvant treatment to resection, but failed to show any difference in recurrence-free survival (median: 42.2 months for the treated group compared with 48.6 months for the control group) (Table 1).127

Table 1 Clinical trials on non-cell-based immunotherapies in HCC

Vaccines

Vaccination can either target TAAs or tumour neo-antigens. Among TAAs, the oncofetal antigens α-fetoprotein and GPC3, the cancer/testis antigens NY-ESO-1, MAGEA1 and SSX2 and hTERT are candidate immunogens. Tumour neo-antigens can be identified by means of whole-exome analysis with next-generation sequencing technologies. The sequencing data combined with computational predictions about antigen presentation that take into account individual HLA allelic variants in patients enable the design of customized vaccines that are based on poly-neo-epitope peptides or RNA. Different anticancer vaccine platforms have been tested in the clinic as indicated below.

Vaccines based on RNA, peptides or proteins

Increased efforts are being made in the development of vaccines that utilize RNA, which encodes multiple neo-epitopes. The vaccine is administered via intradermal or intranodal injection to encourage local DC delivery, or intravenously as liposomal formulation for systemic RNA vaccination.128 Peptide-based vaccines are often delivered via intradermal injection with incomplete Freund's adjuvant with or without an immunomodulator.128 For optimal efficacy this type of vaccine should contain epitopes to be presented to both CD8+ T and CD4+ TH cells.124 Protein vaccines are more expensive, but they provide epitopes for both TH cells and cytotoxic T cells. Whole-tumour-cell vaccines contain all the relevant tumour antigens, but are presented in a diluted manner that fails to trigger efficient antitumour immunity. Two peptide vaccines have been explored in HCC as summarized in Table 1. A mix of two GPC3 peptides that were restricted to HLA-A*24:02 and HLA-A*02:01 was found safe and able to induce tumour infiltration by CD8+ cells in most patients.129 Only one objective tumour response was observed among 33 treated patients, although overall survival was longer in those patients with a higher frequency of peptide-specific CTL responses than those without. On the other hand, vaccination with a single 16 amino acid, hTERT-derived peptide that binds multiple HLA class II molecules resulted in no signs of clinical activity and no evident peptide-specific CTL responses were detected.10

Vaccines based on DCs

Vaccine platforms that are based on DCs have been widely used in solid tumours including prostate cancer, melanoma, renal cancer and HCC.130 Whereas mature DCs prime T cells and boost memory T cells very efficiently, antigen presentation by immature DCs promotes tolerance because of insufficient co-stimulatory signals.130 Thus, most DC-based vaccination trials incorporate Toll-like receptor ligands and/or cytokines to induce DC maturation. Sipuleucel-T (Provenge®, Dendreon, USA), a vaccine employing intravenous injections of peripheral blood mononuclear cells cultured ex vivo with a fusion protein formed by prostatic acid phosphatase and GM-CSF (granulocyte-macrophage colony-stimulating factor), was the first DC-like anticancer vaccine to receive FDA approval after demonstration of improved survival in a phase III study comprising 512 patients with prostate cancer.131 DC networks encompass different cell subsets, namely monocyte-derived DCs, plasmacytoid DCs (which produce high titres of type I interferon) and conventional DCs (which can be further divided into lymphoid-resident and migratory DCs).132 In clinical trials, DCs were differentiated in vitro with IL-4 and GM-CSF, yielding a cell population with features similar to monocyte-derived DCs.130 DCs that are used as vaccines should ideally be highly effective at presenting the captured antigen to CD8+ T cells via MHC class I. This process, which is known as cross-presentation, is thought to be critical for the induction of protective anticancer cytotoxic responses.130 In 2010, a new subset of DCs that express CD141 and the C-type lectin CLEC9A have been shown to excel at cross-presenting antigens to CD8+ T cells, thus representing a promising tool for DC-based antitumour vaccination.133 CD141+ DCs express Toll-like receptor 3 and produce high levels of type I interferon when stimulated with poly IC (polyinosinic-polycytidylic acid), resulting in a robust TH1 response and protective antitumour immunity.134 Rintatolimod and poly ICLC (poly IC stabilized with poly L-lysine and carboxymethylcellulose) are derivatives of poly IC that are increasingly being used as adjuvants for DC-based vaccines. An ongoing clinical trial135 in patients with various malignancies combines intradermal injection of a DC-targeted antibody linked to NY-ESO-1 with subcutaneous injection of poly ICLC.

In patients with HCC, DC-based vaccination has been tested in several small prospective studies using a wide variety of maturation cocktails, antigen sources, treatment schedules and routes of administration (Supplementary Table 2 online).136,137,138,139,140,141,142,143 Patients usually had advanced tumours and DC vaccination was applied alone or in combination with other therapies that might induce immunogenic tumour cell death, such as radiation, arterial embolization or in combination with ACT.140,144 Globally, DC-based vaccination has proven to be safe and capable of generating immune responses against tumour-specific antigens in a substantial proportion of patients.136,137,140,141,144,145 Direct antitumour activity was shown only in a minority of patients with HCC with no objective tumour remissions observed in most studies, the majority with <20 patients.136,137,140,144,145 Interestingly, response rates in those trials in which DCs were injected intravenously ranged from 4% to 33%.139,141,142

Adoptive cell therapy

Cell types

Three cell types have been commonly used for ACT in patients with cancer: CIKs, tumour-infiltrating lymphocytes and genetically modified T cells.

CIKs are generated ex vivo from peripheral blood mononuclear cells cultured in the presence of cytokines (IFN-γ, IL-1, IL-2) and anti-CD3 antibodies, and they consist of activated NKG2Dhigh T cells and activated NK cells and NK T cells.146 CIKs actively proliferate and can be easily obtained in great numbers from the patients. They display cytolytic activity against a wide range of cancer cells, independently of TCR–MHC class I interaction. The intensity of the antitumour response seems to be determined by the expression level of the activating receptor NKG2D on the surface of immune cells and the presence of its ligands MIC-A/MIC-B on the tumour cell membrane.146 These molecules are upregulated by cell stressors such as radiation or chemotherapy.

Tumour-infiltrating lymphocytes are obtained from surgical specimens of the tumour after expansion with anti-CD3 antibodies yielding billions of cells that can be transferred back into the patient. The cells are infused concurrently with IL-2 to promote T-cell growth. A lymphodepletion treatment with cyclophosphamide and fludarabine is administered before the infusion with tumour-infiltrating lymphocytes to enhance the production of endogenous homeostatic lymphokines IL-7 and IL-15, which are important for the in vivo expansion of adoptively transferred cells and eliminate suppressor myeloid and lymphoid cells.147 Impressive data were obtained with unselected tumour-infiltrating lymphocytes in patients with metastatic melanoma with 48% overall response and 13% of durable complete response.148

Genetically modified T cells for cancer therapy can be prepared in two ways: T cells equipped with a cloned TCR, and T cells engineered to express a chimeric antigen receptor (CAR). The use of either of these two forms of engineered T cells circumvents technical obstacles encountered in the expansion of tumour-infiltrating lymphocytes. T cells possess a heterodimer αβ receptor that recognizes antigenic peptides bound to MHC molecules. To prepare TCRs that recognize TAAs with high affinity, highly reactive T-cell clones obtained from the patients are used to determine the sequence of the TCR α and β chains, which are then inserted into a retroviral or lentiviral vector for transduction of autologous T cells.149,150 Improvements in vector design and the introduction of cysteines to form interchain disulphide bonds enable increased surface expression of the transgenic receptor and prevent 'mispairing' with the endogenous α and β chains.149 Infusion of TCR-modified T cells redirected against MART-1 (melanoma antigen recognized by T-cells 1) and administered together with IL-2 to patients with melanoma and prior lymphodepletion induced an overall response in a substantial proportion of patients.150 The treatment also resulted in on-target–off-tumour toxicity with occurrences of erythematous rash, hearing loss and uveitis due to low-level expression of TAAs in normal tissues.150 Similarly, antitumour effects accompanied by severe colitis were observed in patients with colorectal cancer, who had been treated with engineered T cells with a TCR reactive to carcinoembryonic antigen.151 However, no TCR-mediated toxicity was found in patients with melanoma or synovial sarcoma, who were treated with T-cells expressing TCRs against NY-ESO-1.152

As TCRs are restricted to recognizing peptides presented on specific MHC molecules, T cells expressing a transgenic TCR can only be used in those patients who possess the appropriate HLA molecule. This limitation is avoided by using a CAR, a fusion molecule formed by an ectodomain, a transmembrane region and a cytoplasmic tail.153 The ectodomain is a single-chain variable fragment constructed from the variable heavy chain and variable light chain of a mAb with specificity against a defined TAA. The endomain contains the transmembrane adaptor signalling protein CD3ζ and one or more co-stimulatory modules (CD28, CD137 or OX40) for effective transfer of activating signals into the cell.153 A number of strategies are being explored to improve in vivo survival, expansion and functionality of CAR-T cells. For instance, transduction of naive T cells or central memory T cells enables better in vivo survival of the engineered lymphocytes.153 Substitution of the single-chain variable fragment ectodomain by protein ligands or receptors interacting with receptors or ligands present on the surface of tumour cells have also been considered in the preparation of CARs.149 Whereas clinical trials in patients with CD20+ lymphoma using CD20-specific CAR-T cells have demonstrated antitumour activity and good tolerance,154 patients with solid tumours who are treated with TAA-specific CAR-T cells experienced serious toxicities, probably due to low-level expression of the targeted TAAs in normal tissues.155

ACT in primary liver cancer

ACT with CIKs has been extensively studied in a wide spectrum of patients with HCC in the Asia–Pacific region (Supplementary Table 3 online). CIK therapy resulted in an improvement in the profile of circulating lymphocytes in two small, single-arm, prospective studies (increased proportion of CD8+ and CD56+ cells, and type I and II DCs).156,157 In the adjuvant setting, an improvement in recurrence-free survival, but no difference in overall survival was observed in two randomized controlled trials involving >200 patients who had undergone liver resections.158,159 By contrast, a retrospective analysis in which >200 patients with resected livers plus ACT treatment were compared with an untreated control group, reported an increase in overall survival in the CIK-treated patients (5-year overall survival: 65.9% compared with 50.2%).160 A randomized controlled trial comparing CIK treatment with best supportive care after combined TACE and RFA showed a reduction in the 1-year recurrence rate in the CIK-treated group (9% compared with 30%).161 In keeping with these results, a retrospective study in patients treated with TACE plus RFA with or without adjuvant CIKs showed increased overall survival (56 compared with 31 months) and progression-free survival (17 compared with 10 months) in the group that received TACE, RFA and CIK treatments.162 In addition, a randomized study in which patients with a wide range of HCC tumours (from early to advanced) were recruited and treated with different standard therapies (from resection to TACE) reported a substantial increase in overall survival in CIK-treated patients (median: 24.9 compared with 11.3 months), but no differences in the recurrence rate after resection.163 Tran et al.164 characterized the mutational changes within tumours in a patient with metastatic cholangiocarcinoma and identified three CD4+ T-cell clones in tumour-infiltrating lymphocytes, which were reactive to a mutation present in the Erbb2-interacting protein (ERBB2IP, also known as Protein LAP2). Notably, treatment of the patient with previously expanded, mutation-reactive TH1 cells resulted in a slow and sustained decrease in tumour size. This study reflects the important contribution of next-generation sequencing technologies to the development of personalized and highly effective ACT. Thus, characterizing mutations and proteomic analysis of tumour-associated HLA-binding proteins, followed by computational predictions of their HLA-binding properties and immunogenicity will probably contribute to an enhanced effectiveness of individualized ACT approaches.121,165

Immune checkpoint inhibitors

Anti-CTLA-4 mAbs

To date, two anti-CTLA-4 mAbs are in advanced stages of clinical development and have demonstrated promising potential in other cancer types, such as melanoma. The basic regulatory function of CTLA-4 takes place at the priming of both naive and memory T cells upon interaction with APCs.65 Research in mice indicates that the anti-CTLA-4 mAb derives its antitumour activity from CTLA-4 blockade on both effector T cells and TREG cells (Figure 3).74 TREG cells are depleted within the tumour microenvironment by an antibody-dependent cell-mediated cytotoxicity (ADCC) mechanism that involves binding of the mAb to FcγRIII on the membrane of macrophages and NK cells.74

Ipilimumab is a fully human IgG1, anti-CTLA-4 mAb with a half-life of 12–14 days, which has shown remarkable effects on the long-term survival of patients with metastatic melanoma (2-year survival: 18% of treated patients compared with 5% in the control group).5 Grade 3–4 immune-related adverse events (such as pneumonitis, colitis, thyroiditis and hepatitis) occurred in 15% of patients. Genome-wide neo-epitome analysis showed that the mutational burden was higher in patients who benefited from ipilimubab than in those who did not.166

Tremelimumab is another anti-CTLA-4 mAb, which belongs to the IgG2 subclass (endowed with less ADCC-promoting activity than IgG1) and possesses a long half-life (22 days). Despite encouraging clinical activity, the primary end point of improved progression-free survival was not met in a phase III trial of tremelimumab in patients with metastatic melanoma.167 Tremelimumab was the first checkpoint-blocking mAb tested in patients with HCC and chronic HCV infection (Table 1). It yielded signs of antitumour activity, including objective remissions in 17% of patients, disease control in 76% of patients and a median time to cancer progression of nearly 7 months.168 The antitumour response rate was 42% among patients showing no or minor reduction in serum IFN-γ during therapy, whereas it was 0% among those who showed a substantial decrease in the level of this cytokine. The antitumour activity of tremelimumab was independent from its antiviral effects, which manifested by a notable drop in viral load, the appearance of HCV-specific immune responses and the emergence of new genomic variants of the virus that replaced the predominant variants present before therapy. The results of this pilot trial showed no definitive proof of efficacy, but provided evidence that immune checkpoint inhibitors are worth exploring in the treatment of HCC.

Anti-PD-1 mAbs

PD-1 is a potent inhibitor of T-cell responses in peripheral tissues at sites of inflammation and persisting antigen stimulation.119 Hence, the blockade of this molecule constitutes an appealing therapeutic option for HCC (Figure 4). Nivolumab is a fully human IgG4 anti-PD-1 mAb, which was tested in a phase I dose-escalation trial in patients with various forms of solid tumours and yielded positive responses in patients with metastatic melanoma (28%), renal cell cancer (27%) and non-small-cell lung cancer (18%), demonstrating grade 3–4 immune-related adverse events in only 6% of patients.6 In 2012, an international multicentre phase I/II trial began to investigate the safety and antitumour effect of nivolumab (0.1–10 mg/kg for up to 2 years) in patients with virus-related or non-virus-related, advanced HCC (73% of patients had extrahepatic metastasis and/or portal vein invasion), who had or had not been exposed to the kinase inhibitor sorafenib.169 The safety profile was acceptable and durable responses were observed across all dose levels. Responses were evaluable in 39 patients, of whom two showed a complete response and seven a partial response with overall survival of 72% of patients at 6 months.

Figure 4: Activation of antitumour responses by checkpoint-blocking antibodies against PD-1 or PD-L1.
figure 4

Upper panel: T-cell activation promotes PD-1 expression by T cells and the secreted IFN-γ induces PD-L1 expression by the target cell. Binding of PD-1 to PD-L1 blocks TCR signalling and induces T-cell exhaustion. Anti-PD-1 or anti-PD-L1-blocking antibodies reinvigorate T cells and restore their cytolytic functions (mediated by the release of soluble factors and by ligands of death receptors FasL and TRAIL). Lower panel:hypoxic conditions and proinflammatory cytokines promote the expression of PD-L1 both by tumour cells and MDSCs and of PD-1 by TAMs, which release IL-10 upon PD-1–PD-L1 interaction. Anti-PD-1 or anti-PD-L1-blocking mAbs revert T-cell exhaustion and attenuate the immunosuppressive intratumour environment. Abbreviations: ADCC, antibody-dependent direct cytotoxicity; APC, antigen-presenting cell; CD40L, CD40 ligand; CTL, cytotoxic T lymphocyte; CTLA-4, cytotoxic T-lymphocyte protein 4; FasL, tumour necrosis factor ligand superfamily member 6; mAb, monoclonal antibody; MDSC, myeloid-derived suppressor cell; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1; TCR, T-cell receptor; TRAIL, tumour necrosis factor ligand superfamily member 10; TRAILR, tumor necrosis factor receptor superfamily member 10; TREG cell, regulatory T cell.

PowerPoint slide

Another human IgG1 mAb targeting PD-L1, MED14736, has demonstrated activity in solid tumours including HCC, pancreatic and gastric carcinoma with related ≥grade 3 adverse events in 7%, which was considered to be an acceptable tolerance and did not lead to drug discontinuation.170 Other relevant immune checkpoint inhibitors, such as an anti-LAG-3 mAb (BMS-986016) have entered their clinical development phase in several solid tumour types.171

Other mAbs

Anti-GPC3 mAbs

GPC3 is a heparan-sulphate proteoglycan linked to the cell membrane through a glycosylphosphatidylinositol anchor.172 It is overexpressed in a high proportion of HCCs, and probably participates in tumour progression by stimulating Wnt, IGF and fibroblast growth factor signalling pathways.173 GC33 is a humanized antibody against GPC3, which has been tested in a phase I clinical trial in 20 patients with advanced HCC. No dose-limiting toxicity was reached and four patients corresponding to the group with high GPC3 expression showed stable disease.173 The antitumour effect of GC33 is potentially exerted via antibody-dependent direct toxicity and increased tumour infiltration with CTLs.174 A phase II clinical trial with GC33 as a second-line therapy for patients with advanced HCC has been completed.175

Immunostimulatory mAbs

In cancer immunotherapy, mAbs can be used either to block immune checkpoints or to agonistically bind co-stimulatory receptors on immune cells. Most immunostimulatory mAbs tested so far have been selected to target co-stimulatory molecules of the TNF receptor superfamily, namely CD40, CD137, OX40 and GITR (TNF receptor superfamily member 18).176 These mAbs are being developed clinically for application in solid tumours,177,178,179,180 but no data on HCC is as yet available. Liver toxicity stopped the clinical development of urelumab (BMS-663513), a fully human anti-CD137 agonist antibody, but trials were subsequently reinitiated with lower doses.178

Gene therapy

Gene therapy is a therapeutic modality based on the transfer of genetic material (transgenes) into cells to modify their gene-expression profiles. Transgenes can be natural or artificial genes encoding natural or chimeric proteins or subgenomic sequences directed to modify the expression of endogenous genes.181 Most gene-therapy vectors used for clinical application are of viral origin.181 The most widely used viral vectors are adenoviruses, adeno-associated viruses (AAVs) and retroviruses or lentiviruses. Retroviruses and lentiviruses integrate into the host genome and can be used for ex vivo transduction of replicative cells such as lymphocytes or progenitor cells. AAVs and adenoviruses possess a DNA genome that remains largely in episomal form and can be used for in vivo transduction of different tissues including the liver and tumour nodules. AAV vectors possess a linear single-stranded DNA genome of 4.7 kb. Despite their small cloning capacity (4.7 kb), AAVs have been widely used in the clinic owing to excellent tolerance and long duration of transgene expression.181 Adenoviruses have a linear double-stranded DNA genome, which is partly or entirely deleted to prepare first-generation or third-generation adenoviral vectors, respectively.181,182 The limitations of first-generation adenoviruses (low cloning capacity and short-term expression of the transgene) have been surpassed in later-generation adenoviruses that provide larger cloning capacity (36 kb), the ability to express the transgene for long periods of time and reduced proinflammatory effects upon in vivo administration.181,182

A different type of vector are the so-called oncolytic viruses, which are modified viral agents that selectively replicate in cancerous cells with resulting tumour cell lysis.183 Cancer selectivity is intrinsic to some viruses or can be conferred by deleting genes needed for replication in normal cells, but not in cancer cells. Tumour antigens released by dying cells are taken up and processed by APCs, which activates antitumour immunity and can be enhanced by arming oncolytic viruses with genes that encode cytokines, such as GM-CSF. Different types of oncolytic viruses encoding GM-CSF have been tested in clinical trials, including adenovirus,184 herpes simplex virus185 and vaccinia virus.186

Clinical experience indicates that the antitumour effect of oncolytic viruses depends more on the immunostimulatory activity of transgenic cytokines and virus-induced immunogenic cell death than on the oncolytic properties of the vector. The therapeutic efficacy of oncolytic viruses is limited, because vector replication and transgene expression is brief and repeated doses are inefficient, owing to the development of neutralizing antibodies after first vector administration. Oncolytic viruses are better envisioned as an internal vaccine to trigger adaptive antitumour responses that should be maintained and expanded with additional immunotherapies.187

Gene therapy for patients with cancer can also be performed by transducing the tumour or the tissue around it with non-replicative vectors designed to enforce the expression of immunostimulatory factors (cytokines and chemokines).182 This approach has been pursued in clinical trials using first-generation adenoviruses encoding CD40 ligand, a protein expressed on activated T cells (in melanoma and other solid tumours188), or IL-12 (expressed constitutively for the treatment of prostate cancer189 or expressed in an inducible manner for the treatment of melanoma190). Nowadays, the possibility of using long-term expression vectors for intratumour or peritumour production of immune-enhancing molecules, including mAbs,191 has opened promising horizons for gene therapy.

To date, immunogene therapy of HCC has been carried out using two different viral agents (Table 1). Repeated intratumoural injections of a nonreplicative adenovirus carrying the genes encoding human IL-12 were shown to be safe.192 Increased tumour infiltration by CD4+ and CD8+ T cells was observed in two out of four patients and one objective tumour remission was reported among 10 patients (Table 1). In another study, an oncolytic poxvirus carrying the genes encoding human GM-CSF produced one objective remission among three treated patients with HCC.193 However, similarly to the adenovirus, the armed poxvirus was not able to produce tumour remissions in non-injected nodules. In 2013, a phase II randomized trial comparing two doses of the viral agent showed that the treatment was generally well-tolerated and achieved a disease control rate of 46% at 8 weeks after treatment with no difference between subgroups (Table 1).194 Unfortunately, a randomized phase IIb trial comparing this agent to the best supportive care failed to meet its main end point of improved overall survival in patients with advanced HCC.186

Combined therapies

Diverse immunotherapies are potentially applicable to HCC (Figure 5). Owing to the multiplicity of mechanisms used by tumours to evade the immune response, combining different immunotherapeutic agents is an appealing approach to treat HCC. Some of these combinations are already being explored in other tumour types. Inhibition of two checkpoint molecules, PD-1 and CTLA-4,195 has been reported to substantially improve the objective response rate in melanoma, but at the expense of considerable but manageable immune-related toxicity. Notably, within the tumour microenvironment, tumour-reactive CD8+ T cells coexpress PD-1, TIM-3 and LAG-3 and, in mice, dual inhibition of PD-1 and LAG-3 has additive antitumour effects, providing a rationale for combining PD-1 and LAG-3 inhibitors in patients with cancer.196 A combination of blockers of co-inhibitory molecules with agonists of co-stimulatory receptors (urelumab with nivolumab) is currently being attempted in patients with solid tumours.197 Interestingly, CD137 agonists have been shown to enhance ADCC elicited by other mAbs. Thus, urelumab might be combined with ipilimumab or with other mAbs, which act via antibody-dependent direct toxicity, such as GC33, to achieve better antitumour responses. As indicated before, CD73 expressed by tumour cells and the endothelium mediates the production of adenosine, a key immunosuppressor within the tumour microenvironment. CD73 inhibition increases tumour infiltration by T cells198 and an anti-CD73 mAb substantially enhances the antitumour response elicited by PD-1 blockade in different experimental models of cancer.199 In patients with HCC, who were vaccinated against GPC3, antigen-specific CTLs express high levels of PD-1 and in vitro blockade of PD-1 enhances the degranulation of GPC3-specific cells upon exposure to the antigen.200 Similarly, promising results were reported after combining ipilimumab with a GM-CSF-based tumour vaccine in patients with pancreatic cancer.201 In mouse models of breast cancer, PD-1 blockade in association with a multipeptide vaccine prolonged progression-free survival together with increased accumulation of CD8+ T cells in the regressing tumours.202 Selective blockade of co-inhibitory molecules within the tumour microenvironment might preserve intratumour effector functions of tumour-specific T cells primed and expanded by the vaccine. Local intratumour administration or gene transfer approaches might limit unnecessary systemic exposure.

Figure 5: Classification of immunotherapies.
figure 5

The diverse immunotherapeutic interventions have been classified according to their modality (pink, orange, yellow, green, grey, blue or violet) and primary effect (class 1 to 9). *High-capacity adenoviruses might incorporate several transgenes for intratumour expression of one, two or three different mAbs, or one cytokine plus one or two mAbs. Transgenes encoding potentially toxic molecules should preferably be placed under the control of an inducible promoter. §Deliverable by intratumour gene transfer. Abbreviations: CIK, cytokine-induced killer cell; CTLA-4, cytotoxic T-lymphocyte protein 4; CXCL10, C-X-C motif chemokine 10; GITR, TNF receptor superfamily member 18; GPC3, glypican-3; LAG-3, lymphocyte activation gene 3; mAb, monoclonal antibody; NG-CAR T cell, new-generation chimeric antigen receptor T cell; NK, natural killer; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1; RFA, radiofrequency ablation; TACE, transarterial chemoembolization; TARE, transarterial radioembolization; TCR, T-cell receptor; TCR-mod, modified TCR; TH1, type 1 T helper cell; TIL, tumour-infiltrating lymphocyte; TIM-3, T-cell immunoglobulin and mucin domain-containing protein-3; VEGF, vascular endothelial growth factor.

PowerPoint slide

Other options for combined immunotherapy have not yet been tested in patients with cancer, but evidence from animal studies is certainly encouraging. Agonist OX40 mAbs induce a potent tumouricidal programme in CD4+ T cells,203 which raises hopes that OX40 agonists could potentiate the antitumour effects of adoptive CD4+ T-cell therapy. ACT (either tumour-infiltrating leukocytes or CIKs) might benefit from concurrent immune checkpoint blockade to prevent intratumour inhibition of effector T-cell functions. Accordingly, tri-therapy with ACT, OX40 agonists and anti-PD-1 could be an option. Anti-CD137 mAbs also potentiate ATC.204

A study from 2014, in which an oncolytic measles virus was engineered to locally express anti-PD-1 or anti-CTLA-4, showed improved antitumour efficacy of these recombinant vectors compared with the parental measles virus.205 These data support the clinical development of strategies based on combinations of virotherapy with checkpoint inhibitors. Gene therapy, either with oncolytic viruses or conventional adenoviruses that encode IL-12, induced T-cell activation, but this effect was counterbalanced by an increased number of intratumour TREG cells and vigorous IFN-γ-mediated PD-L1 upregulation on both malignant and immune cells.80 Thus, combining IL-12 immunogene therapy with strategies aimed at depleting intratumour TREG cells (for example, ipilimumab) or to block PD-L1–PD-1 interactions could result in potent antitumour synergy. Indeed a strong synergistic effect against mouse tumours was found by associating intratumour injection of a Semliki forest virus that encodes IL-12 with systemic administration of anti-PD1 or anti-PD-L1.187 In addition, combining this Semliki forest virus with agonist CD137 mAbs amplified the response to weak tumour antigens, leading to tumour rejection.206

Finally, immunotherapy can certainly be combined with those therapies of proven value in the treatment of HCC, from surgical or locoregional therapies to sorafenib. Embolization and ablative therapies, such as RFA, cause a release of tumour antigens and stimulation of antitumour immunity,207 which might be enhanced by simultaneous administration of checkpoint inhibitors. In mouse models, impressive data were obtained by combining radiotherapy and dual checkpoint blockade.208 Radiation broadens the TCR repertoire among intratumour T cells, anti-CTLA-4 eliminates intratumour TREG cells and promotes T-cell expansion and anti-PD1–PD-L1 interaction reverses T-cell exhaustion. In advanced HCC, radioembolization in combination with checkpoint inhibitors therefore seems a promising approach. A pilot study explored the safety and feasibility of tremelimumab in combination with locoregional therapies directed to promote immunogenic tumour cell death in 18 patients (subtotal TACE in eight and RFA in 10) with advanced-stage HCC exposed to sorafenib.209 This combined therapy proved to be safe and of the 10 patients evaluable for response outside TACE or RFA treated lesions, four achieved confirmed partial objective response and tumour biopsies showed immune cell infiltration in all evaluable patients.

Cancer chemotherapy induces tumour cell death, but might also stimulate antitumour immunity by inducing 'immunogenic cell death' and by reducing immunosuppressive cell populations, such as MDSCs and TREG cells, in the tumour microenvironment.210 Thus, although not effective alone, chemotherapy can be a good adjuvant to immunotherapy. In line with this finding, oxaliplatin combined with adenoviral-mediated intrahepatic expression of IL-12 resulted in a marked improvement of antitumour responses in a model of metastatic colon cancer to the liver. The adjuvant effect of oxaliplatin was associated with a decreased number of MDSCs and an increased CD8+ T cell:TREG cell ratio within the tumour.211 In addition, mAbs against pro-oncogenic proteins such as EGFR (for example, cetuximab) might contribute to tumour immuno-rejection via ADCC.212 Whether the combination of tyrosine kinase inhibitors, such as sorafenib, with HCC immunotherapies is beneficial, is still unclear. Although sorafenib reduces intratumour TREG cells in patients with renal cancer,213 it might also decrease IL-12 expression214 and impair DC function.215

Biomarkers

An unmet clinical need exists for predictive biomarkers in the entire field of cancer immunotherapy and particularly in HCC. A general trend, which was first discovered in melanoma, points towards increased therapeutic benefits (with vaccines or immunostimulatory mAbs) in those tumours that show a denser infiltrate of CD8+ T cells and a gene signature of T-cell infiltration and interferon induction before therapy.216 In the specific case of PD-1–PD-L1 blockade, it is interesting that PD-L1 expression on tumour cells217 or on tumour-infiltrating immune cells correlates with objective responses.218 However, the predictive value of negative PD-L1 staining is not sufficiently accurate to exclude these patients from treatment with PD-1–PD-L1 inhibitors and the predictive value is completely lost upon combined anti-CTLA-4 and anti-PD-1 blockade treatments.219 Thus, current efforts focus on discovery and validation of multiparameter histology scores in tumour biopsy samples that are more predictive, taking into account not only PD-L1 staining, but also the density and localization of activated T cells. In addition, genetic markers that predict the response to immune checkpoint inhibitors might prove to be helpful in the clinical setting. In this context, it was found that tumours (both colorectal and non-colorectal) with defective DNA mismatch-repair (an alteration associated with an increased number of somatic mutations) are more likely to benefit from PD-1 blockade.220

Future prospects

Combined immunotherapy is a promising option in HCC treatment (Figure 6). Concurrent systemic administration of various checkpoint inhibitors and immunostimulatory antibodies, however, might be hampered by potential toxicity. Transduction of the tumour with gene therapy vectors enables local expression of one or several immunotherapeutic molecules, resulting in a high intratumour concentration of these agents without reaching toxic systemic levels. Hence, multitargeted therapy can be pursued by associating systemic mAbs or ACT with internal gene therapy. In this aspect, new-generation vectors are highly promising tools, as they might allocate genes encoding two (or more) synergistic mAbs (such as anti-CTLA-4 inhibitor and agonistic anti-CD137) together with genes expressing immunostimulatory cytokines, such as IL-12, under the control of inducible promoters. This therapy could be combined with systemic blockade of PD-1–PD-L1, which would synergize with locally expressed anti-CTLA-4 and IL-12 (the latter induces PD-L1 upregulation). Similarly, gene therapy can be used to enhance the efficacy of ACT by manipulating the tumour to express chemokines and checkpoint blockers to attract effector T cells to the tumour tissue and prevent their inhibition once they are inside the tumour microenvironment (Figure 7).

Figure 6: Combination therapy for HCC.
figure 6

Combination of internal and systemic approaches avoids the toxicity risk of concurrent intravenous administration of several mAbs. Abbreviations: CTLA-4, cytotoxic T-lymphocyte protein 4; HCC, hepatocellular carcinoma; mAb, monoclonal antibody; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1.

PowerPoint slide

Figure 7: Clinical options for HCC therapy.
figure 7

a | Combined therapy with anti-CTLA-4 and anti-PD-1 antibodies acts via different mechanisms with synergistic effects, but also with a higher risk of systemic adverse effects. The triple combination of local anti-CTLA-4 mAbs, IL-12 and a systemic PD-1–PD-L1 inhibitor might result in vigorous antitumour responses. b | ACT with either tumour-infiltrating leucocytes, cytokine-induced killer cells or chimeric antigen receptor T cells can be difficult to deliver inside the tumour and even if cells reach their target, their effector functions might be dampened by the immunosuppressive TME. Gene therapy can be used to enhance the efficacy of ACT by manipulating the tumour to express chemokines and checkpoint blockers to attract effector T cells to the tumour tissue and prevent their inhibition. Abbreviations: ACT, adoptive cell transfer; CTLA-4, cytotoxic T-lymphocyte protein 4; CXCL10, CXC motif chemokine 10; HCC, hepatocellular carcinoma; mAb, monoclonal antibody; NK, natural killer; PD-1, programmed cell death protein 1; PD-L1, programmed cell death 1 ligand 1; TME, tumour microenvironment.

PowerPoint slide

Conclusions

The manifold ways in which cancerous cells can evade immunity in HCC calls for a combined therapeutic approach. HCC immunotherapy is a fast-moving field in need of active translational research to define the tolerance and efficacy of combinatorial strategies. Academic medicine should have, in collaboration with industry, a pivotal role in this formidable and urgent task.