Hepatocellular carcinoma (HCC) is a disease with both a grim prognosis and rising incidence. The most up to date estimates demonstrate a median overall survival of 9 months for all stages of untreated HCC, a number that worsens with increasing Barcelona Clinic Liver Cancer (BCLC) stage [1]. In the world, liver cancer is the third leading cause of cancer mortality while in the United States, it is the fourth highest cause [2, 3]. The rising incidence of liver cancer in the United States is attributed to the epidemics of hepatitis C virus infection and non-alcoholic fatty liver disease [4, 5]. For early stage HCC, standard of care treatments include resection, localized therapies such as ablation and radiation, and liver transplantation [6]. Until recently, the only first line systemic therapy approved for advanced HCC was the anti-angiogenic multikinase inhibitor sorafenib, based upon prolongation of median survival by approximately 3 months with low rates of tumor radiographic response, attributed to a mechanism of disease stabilization [7].

Immunotherapy is a cancer treatment strategy that has been explored for many years but only recently has seen clinical success, mainly in the form of immune checkpoint inhibitors. Antibodies to the immune checkpoints PD-1, CTLA-4, and PD-L1, have proved to be relatively safe and beneficial in treating triple negative breast cancer, renal cell carcinoma, melanoma, urothelial carcinoma, squamous cell carcinomas of the head and neck, Merkel-cell carcinoma, and non-small cell lung cancer, among others [8–14]. Checkpoint inhibition (CPI) blocks the negative regulatory signals either directly on T cells or on cells that interact with T cells (such as tumor cells, stromal cells, and antigen-presenting cells), providing a stimulus to pre-existing anti-tumor immunity. Recently, two PD-1 inhibitory monoclonal antibodies, pembrolizumab and nivolumab, received regulatory approvals in the second-line setting for advanced HCC as monotherapy [15, 16]. There is also early phase clinical trial data demonstrating activity from anti-CTLA-4 inhibition alone and in combination with transcatheter arterial chemoembolization (TACE) or ablation in a subset of patients [17, 18]. Response rates range from 10 to 25% among the different checkpoint inhibitors used, and clinical data are reviewed more extensively elsewhere [19, 20]. Moreover, despite the potential concern for relatively worse toxicity related to CPI due to already poor liver function in the HCC population, overall clinical trials have shown an acceptable safety profile for HCC patients, with rates of immune-related toxicity similar to that in patients with other tumor types and without underlying hepatic dysfunction [21, 22].

The site of HCC development, the liver, makes immunotherapy a promising yet complicated strategy for treatment. First, the liver itself is an immune organ, with rich populations of immune cells, some of which are unique to the liver such as Kupffer cells [23]. As there are elements that can promote both tolerance and anti-tumor immunity within the liver, evidence for the use of CPI in HCC must be inferred from model systems and from the clinical data. In other solid tumor types, metastases to the liver portend a poor response to CPI and are associated with decreased tumor infiltration of CD8⁺ T cells, demonstrating the power of the liver to generate tolerance to tumors derived from other sites [24]. Multiple examples from mouse models further substantiate the induction of systemic tolerance when exogenous antigens are expressed in hepatocytes, an effect mediated by T regulatory cells (Tregs) [25, 26]. Conversely, NK and NK T cells are thought to be potent anti-cancer effector cells, of which the liver has a particular abundance [27–29]. Next, upwards of 80–90% of HCC arises in the context of underlying liver injury which can progress to fibrosis or cirrhosis; therefore, it is important to take into account the variable effects on the immune microenvironment in this state of fibrosis and chronic inflammation [30]. Lastly, the toxic and viral insults that promote carcinogenesis in the liver may drive immunosuppression directly through host/viral interactions or via chronic inflammation, although conversely, pathogen-associated molecules could serve as a source of neo-antigens to be recognized by effector T cells [31]. Thus, there is a tightly interwoven, exceedingly complex, relationship of chronic inflammation and the anti-cancer immune response in the liver which may represent an opportunity for CPI in HCC, but also demands thoughtfully designed treatment strategies to subvert suppressive mechanisms.

The liver is an immune organ made up in bulk by hepatic parenchymal cells. Besides the biliary epithelium, the majority of the remaining 20 % are non-parenchymal cells such as stellate cells, macrophages, NK, and T cells including TCRγδ T cells (Table 1, Fig. 1) [32, 33]. The unique anatomy of the liver puts lymphocytes in direct apposition to hepatocytes through the lack of a basement membrane in liver sinusoids [32]. Due to the chronic antigen load from the gastrointestinal tract, the liver needs to maintain a level of tolerance to balance elimination of gut bacterial pathogens while avoiding severe inflammation induced by non-pathogenic gut commensals. The liver also serves as a major producer of immune-related molecules like C-reactive protein (CRP) and soluble pattern recognition receptors (PRRs) for molecules derived from pathogenic organisms, thus playing a central role in systemic inflammation and immunity [33].Fig1Fig. 1

Liver immunobiology across a spectrum from healthy liver to inflammation and oncogenesis. Top panel: Viral and toxic insults drive inflammation in the liver and alter the normal baseline response to gut commensals. Chronic inflammation can lead to alteration of normal immunity to both commensal organisms and pathogens, and eventually, to oncogenesis. Bottom panel: General mechanisms underlying tolerance and immunity and interactions between various cell types are outlined in each of the following states: healthy liver (left), fibrosis and cirrhosis (middle), and hepatocellular carcinoma (right). Cells that generally maintain tolerance in healthy liver and promote immune suppression and oncogenesis are colored in red while cells that favor protective anti-microbial or anti-tumor immunity are colored in blue

Both the systemic and local immune systems, as well as components of innate and adaptive immune systems, are altered in the setting of liver fibrosis and cirrhosis which occurs due to chronic inflammation from toxins, infectious agents, or other insults such as steatosis (Table 1, Fig. 1). It is well-known clinically that cirrhotic patients are systemically immunocompromised and infections constitute a major source of mortality in end-stage liver disease [54]. Bacterial infection and sepsis occur in part due to increased bacterial translocation through a “leaky” gastrointestinal barrier created by portal hypertension, as well as due to weakened systemic and local immunity [54–57]. The exact mechanisms related to the initiation of inflammation from each type of insult are reviewed extensively elsewhere and therefore are not discussed here. Rather, we focus on the general mechanisms involved in fibrosis initiation and changes in the immune status during progression to cirrhosis, an irreversible state that is the end-stage of fibrosis [58].

The main cell types involved in liver fibrosis initiation appear to be HSCs and Kupffer cells. As a result of inflammation due to toxins such as alcohol, steatosis, or viral infection, inflammatory cytokines activate HSCs through TLR4, which then produce extracellular matrix proteins such as collagen [59–61]. The cytokine IL-17 can drive production of pro-fibrogenic IL-6, TNFα and TGFβ by HSCs and Kupffer cells [62]. Mouse models of liver fibrosis demonstrate that under inflammatory conditions, Kupffer cells no longer induce tolerance to experimental antigens as they do in normal liver [63]. During liver damage, Kupffer cells produce reactive oxygen species, TGFβ, and platelet-derived growth factor (PDGF) which activates HSCs [23, 33, 64]. Both Kupffer cells and HSCs secrete matrix metalloproteinases during chronic tissue injury which is mediated by TNFα and TGFβ, promoting remodeling of the extracellular matrix [65, 66]. Fibrosis and the buildup of extracellular matrix leads to a hypoxic environment which results in VEGF upregulation, which may later support tumor angiogenesis [64].

NK cells can protect against fibrosis via killing of activated HSCs, although can also drive inflammation [27, 28]. There is an inverse correlation between IFNγ-producing NKp46^high NK cells and the degree of fibrosis in HCV-infected patients [67]. NK cell killing of HSCs and production of IFNγ become suppressed over time with advancing fibrosis and can be further suppressed by alcohol consumption, as seen in a carbon tetrachloride-induced fibrosis mouse model [68]. STAT1 signaling is an important negative regulator of the fibrosis pathway, opposing the effects of TGFβ secreted by HSCs and supporting NK cell cytotoxicity [69]. HSCs become more resistant to NK cell killing in later stages of fibrosis due to SOCS1 upregulation by HSCs [70]. As far as NK T cell populations’ role, there may also be duality based on the particular type of NK T cells involved. Type I NK T cells are thought to be protective in acute liver injury but harmful in chronic inflammation as they activate HSCs and neutrophils whereas in the setting of liver tumors, type I NK T cells are associated with tumor control [45, 71, 72].

As liver injury and fibrosis progress, the extracellular matrix becomes stiffer and the normal anatomy of liver is altered which can then cause impaired production of key immune molecules normally supplied by the liver such as complement pathway proteins. Cirrhotic patients have lower levels of C3 and C4 proteins than healthy controls, associated with infections and mortality, while in contrast and perhaps surprisingly, cirrhotic patients had higher levels of mannan-binding protein and opsonization [73, 74]. Another group that found while mannose-binding lectin (MBL) is not necessarily lower in cirrhotic patients compared to healthy controls, lower levels of MBL in cirrhosis are associated with an increased risk of infections [75]. As fibrosis progresses, a dysfunctional immune response feeds forward the cycle of inflammation. For example, patients with cirrhosis have higher levels of TLR2 expression and circulating endotoxin leading to exaggerated responses to bacterial products [35]. However, the TLR signaling apparatus can become dysfunctional rather than protective against infection, as with more infections seen in cirrhotic patients with TLR4 polymorphisms and with TLR2 and TLR4 dysfunction [55, 76]. Higher circulating levels of endotoxin and IL-10 in cirrhotic patients compared to healthy controls has been associated with “immune paralysis” – an inability of APCs to upregulate MHC and present antigens effectively to T cells [41, 77]. Patients with primary biliary cirrhosis were found to have defective Fc-Receptor mediated clearance of pathogen/antibody complexes, one proposed mechanism of the impaired phagocytosis by APCs that is seen in liver disease, although this has not been seen in alcoholic cirrhotic patients [78]. Low albumin levels in cirrhotic patients can drive neutrophil dysfunction; as albumin binds excess endotoxin, elevated endotoxin levels lead to chronic signaling in innate immune cells as a consequence of hypoalbuminemia [79]. Other pro-inflammatory molecules such as soluble CD163 and MCP-1, activators of macrophages, also are increased in the serum of cirrhotic patients [80, 81].

The chronic high levels of pro-inflammatory chemokines and cytokines alters both systemic and local immune cell subsets from that seen in patients without liver disease. Compared to healthy controls, cirrhotic patients have an increased number of activated monocytes and specifically, more non-classical (CD16⁺) monocytes, which increase with progressive fibrosis and are capable of activating HSCs [82, 83]. Cirrhotic patients with ascites are found to have lower numbers of naïve CD4⁺ and CD8⁺ T cells and higher numbers of activated CD4⁺ T cells in the peripheral blood, as well as increased production of IL-10 and TGFβ by T cells [83, 84]. Suppressive Tregs expressing CD40 ligand occur in both mouse models of liver injury and explanted hepatitis C (HCV)-positive livers [85]. IL-17, capable of activating HSCs and Kupffer cells to produce collagen via activation of the STAT3 pathway, is mainly secreted by T cells, including TCRγδ T cells [62, 86]. Tregs and Th17 cells are both found to be increased in more advanced HBV-related fibrosis compared to earlier stage fibrotic livers; however, an elevated Th17/Treg ratio has been shown to correspond with higher liver stiffness measurement, a correlate of worsening liver fibrosis [87, 88]. Furthermore, chronic antigen stimulation can lead to T cell exhaustion, with upregulation of inhibitory receptors such as PD-1 and progressive loss of polyfunctional cytokine production [89]. Patients with chronic viral hepatitis have exhausted viral-specific T cells; blockade of the PD-1/PD-L1 pathway can partially reverse T cell dysfunction and has demonstrated some success in control of chronic viral infection [90, 91]. In autoimmune liver disease, MAIT cells are also found to be exhausted with less IFNγ production and upregulation of PD-1 and CTLA-4 is seen in autoimmune liver disease and hepatitis B infection [92, 93].

Effector cells that are found within tumors and peripheral blood of HCC patients are generally dysfunctional, although existence of certain effector cells and other immune mediators are shown to be associated with improved prognosis, such T and NK cells, suggesting a productive immune response to HCC is possible [116, 117]. TCRγδ T cells are expanded in the blood of liver cancer patients and show capability to kill tumor cells ex vivo [118]. A 14-gene panel of immune-related genes (including TNF, CD8A, IFNG, and various chemokines and TLRs) predicted prognosis in early stage but not late stage HCC, suggesting that a protective immune microenvironment can exist in early but not late stage HCC [119]. CXCL10, CCL5 and CCL2 correlated with infiltration of CD8⁺ T cells, Th1 CD4⁺ T cells, and NK cells [119]. Cytokines such as IFNγ, TNFα, and TLR3 ligands could induce production of these chemokines by cancer cells which then serve to recruit T and NK cells [119]. Myeloid cells can be induced via CpG oligonucleotides to stimulate CD8⁺ T cells, demonstrating the dichotomous nature of myeloid compartment under different conditions [120]. V-domain Ig suppressor of T cell activation (VISTA), while thought to be a negative regulator of T cells, is associated with better prognosis in HCC, in contrast to its association with worse outcomes in other tumor types [121–123]. The association of VISTA with tumor-infiltrating CD8⁺ T cells in HCC may be a signal of activated, albeit exhausted, effector cells that are protective against tumor progression whereas in melanoma and pancreatic cancer, VISTA was mainly expressed by myeloid subsets.

Tumor antigen-specific responses can occur in HCC, and the association of HCC with pathogens such as the hepatitis viruses may be an opportunity for targeting non-host antigens that will be recognized as foreign to the immune system. Spontaneous tumor antigen-specific T cell responses have been detected in HCC, including in a patient that had a complete response following treatment with sorafenib and in another patient cohort following local or systemic chemotherapy [124, 125]. TCR sequencing identified a concentration of shared TCR α and β chains in liver tumors compared to T cells in blood or adjacent non-tumor liver, implying clonal T cells within tumors [96]. Another group found that while there were detectable T cell responses to tumor-associated antigens, responses declined with advancing disease and tumor antigen-specific CD8⁺ T cells were dysfunctional with low production of IFNγ, Granzyme B, and perforin [126]. In a mouse model of HCC driven by virally-induced SV40 large T antigen, there is clearance of most virally-infected cells and in cells that persist, they retain expression of viral products [127]. However, despite the frequency of virally-associated HCC, the response to CPI does not occur in the same high proportion of patients as in other virally-associated cancers such as Merkel cell carcinoma, suggesting that anti-viral immune responses are not sufficient for a successful response to immunotherapy [128].

Due to limited clinical data overall thus far for the use of CPI in HCC, correlative studies using samples from CPI-treated liver cancer patients has lagged behind that in other cancer types. Therefore, most data we have regarding the changes in the liver post-CPI are derived from mouse models of HCC in which various checkpoint inhibitors have been used. In mouse models of HCC, anti-PD-1 has been shown to have activity both as monotherapy and in combination with other anti-cancer therapies. Due to the heterogeneity of models available, of which none wholly replicates the process of HCC initiation and progression in humans, the results are variable and based on the model used [129]. Importantly, anti-PD-1 has been shown to have activity in mouse models of HCC that incorporate a fibrotic liver microenvironment and that replicate findings seen in human tumors such as progressively exhausted PD-1⁺ CD8⁺ T cells and accumulation of Tregs, as well as in patients with Child Pugh B liver dysfunction [22, 130]. Sorafenib therapy upregulated PD-L1 in orthotopic liver tumors and caused the accumulation of suppressive macrophages and Tregs which could be mitigated with a CXCR4-antagonist [131]. Anti-PD-1 showed synergy with the combination of the CXCR4 antagonist and sorafenib although not with sorafenib alone, demonstrating that a multi-targeted approach may be needed to overcome a suppressive microenvironment [131]. This model is particularly clinically relevant as many HCC patients will have been treated with tyrosine kinase inhibitors prior to CPI which may alter the tumor microenvironment.

To date, the published clinical trials of CPI in HCC have reported relatively limited immune profiling analyses on blood and archival tumor samples in subsets of patients. In the CheckMate040 and KEYNOTE-224 clinical trials of anti-PD-1 therapy, there were no cases of HCV or HBV viral re-activation. In CheckMate040, there were transient decreases in HCV viral load in HCV-infected patients [15, 16]. In patients with HCV and HCC treated with anti-CTLA-4, the majority had a decrease in viral load, including three patients with a transient complete viral response; however, anti-viral T cell responses did not correlate with tumor response [18]. In another study combining anti-CTLA-4 therapy with ablation, anti-viral responses were again seen in HCV⁺ patients and those patients who did not have an anti-viral response also did not benefit in terms of tumor control [17]. This clinical trial included on-treatment biopsies at the time of ablation, which revealed that CD8⁺ T cell infiltration at six weeks after initiation of anti-CTLA-4 correlated with tumor response [17]. In other tumor types, PD-L1 has been used as a predictor of response to anti-PD-1 CPI. In CheckMate040, no association was found between radiographic response and tumor cell expression of PD-L1, whereas the KEYNOTE-224 trial of pembrolizumab, which used a combined score of the tumor and microenvironment immune cell PD-L1 expression, found a correlation between PD-L1-expression and response [15, 16].

The unique immunobiology of the liver promotes oncogenesis and tumor tolerance under conditions of fibrosis and chronic inflammation, while also presenting an opportunity for therapeutic targeting with immune checkpoint inhibitors. While toxic and pathogenic insults may provide neo-epitopes and pathways to target with anti-cancer agents, the background of chronic inflammation promotes immune suppression in an immune organ already predisposed towards tolerance. Beyond immune cell populations unique to the liver, other factors associated with chronic liver disease may also shape the response to immunotherapy. The microbiome has been demonstrated to predict response to CPI in other malignancies and is particularly relevant to HCC, due to the altered microbiome in the setting of gut translocation in chronic liver disease patients [56, 132]. The microbial contribution, including both gut commensals and pathogenic hepatitis viruses, to oncogenesis and response to CPI should be two key focus areas of future investigation. While mouse models cannot fully recapitulate the complex interaction of fibrosis caused by various toxic and pathogenic insults, the altered liver architecture seen in cirrhosis, and human immune cell populations unique to the liver, several relevant models have thus far demonstrated the benefits to using combination therapies to simultaneously stimulate effector T cells and inhibit suppressive immune populations [130, 131]. Compared to tumor types which are considered immunologically “cold” (having very little immune cell infiltration), the rich infiltrates of leukocytes within the liver present an opportunity to use novel immunotherapy combinatorial strategies to re-polarize immune cells to a productive anti-cancer response. Furthermore, strategies targeting suppressive populations such as HSCs and MDSC that worsen fibrosis and impair protective T cell function are a potential path forward to enhancing the efficacy of CPI. Additionally, tumor cell intrinsic mechanisms of resistance to CPI should be explored. Given that the majority of HCC patients have developed cancer in a background of impaired liver function and liver inflammation, the clinical need for strategies that are both effective and safe in this population is of great importance, as well as determining how best to sequence or combine available agents for HCC. Identification of biomarkers of immune response is also paramount in guiding choice of individual treatments and sequencing therapy. Further correlative and basic science studies should reveal the full potential of the immune system to re-shape the dysfunctional liver tumor microenvironment and overcome the barriers to successful anti-cancer immunotherapy.