Modulation of the tumour microenvironment in hepatocellular carcinoma by tyrosine kinase inhibitors: from modulation to combination therapy targeting the microenvironment

Since no effective therapy for advanced HCC was available before its application, sorafenib is regarded as the breakthrough point in the therapy of advanced HCC [24]. It has been approved by the FDA based on the results of a phase 3, double-blind clinical trial that involved 602 participants and generated a 3 month longer median survival (hazard ratio 0.69, 95% confidence intervals 0.55–0.87; P < 0.001) (NCT00105443) [24]. Sorafenib achieves its therapeutic effect by suppressing angiogenesis mainly through the inhibition of VEGFR1, 2, and 3. Meanwhile, it also inhibits the Ras/Raf/MEK/ERK signalling pathways, PDGFR-β, and hepatocyte factor receptor (c-KIT) (Fig. 1) [49]. The adverse effects of sorafenib include constitutional symptoms, dermatological events, gastrointestinal symptoms, liver dysfunction, and pain [24, 50]. The most frequent side effects observed in the clinic are diarrhoea and hand-foot skin reaction [18, 24, 50]. In addition to the aforementioned effects, an understanding of its effect on TME modulation might provide some extra insights for future research.

Sorafenib has been reported to interact with immune cells and induce an immunomodulatory effect, yet the modulatory effect of sorafenib remains to be comprehensively clarified (Fig. 3). Among them, the interaction between sorafenib and NK cells was proposed to be significant. NK cells serve as the key effectors in tumour immunosurveillance. Strategies targeting NK cells include natural-killer group 2, member D (NKG2D) ligands, NK cell engagers, NKG2A, and adoptive NK cell strategies [51]. Among them, NKG2D is expressed by NK cells and activates subsequent cytotoxic lymphocytes when binding to ligands, including major histocompatibility complex class I chain-related protein A and B (MICA, and MICB), on target cells [52]. Shedding of endogenous MICA and MICB, which plays an important role in immune evasion, largely depends on a disintegrin and metalloproteinase 9, 10 and 17 (ADAM9, ADAM10, and ADAM17), as well as matrix metalloprotease (MMP) 14 [53‐55].

Sorafenib was reported to activate NK cells mainly through regulating the shedding of MICA, interacting with macrophages, and inhibiting androgen receptors. Sorafenib could inhibit the expression of ADAM9, a metalloproteinase that was shown to inhibit the shedding of MICA, subsequently increasing the NK sensitivity of HCC cells [56]. Activation of NK cells following sorafenib treatment was also suggested to be associated with the interaction between macrophages and NK cells. Sorafenib administration activated NK cells in C57BL/6 wildtype and tumour-bearing mice. In an in vitro model, NK cells were activated by sorafenib-treated macrophages in a dose-dependent manner and showed increased degranulation and secretion of interferon (IFN)-γ. The activation of NK cells by sorafenib-treated macrophages was shown to depend on nuclear factor-kappa B (NF-κB), IL12, and IL18 [57]. A subsequent study elucidated that sorafenib administration led to the pyroptosis of macrophages, which then activated NK cells and resulted in HCC cell death [58]. Consistently, decreased expression of the major histocompatibility complex class I has been reported in the sorafenib-treated HCC cells. This chance hinders recognition by T cells and favours the NK cells-mediated response [58]. Moreover, the androgen receptor was documented to suppress IL-12A expression and inhibit the cytotoxic effect of NK cells against cancer cells. Sorafenib could enhance IL-12A signals through androgen receptor inhibition [59]. These results indicated that the therapeutic efficacy partially depended on the crosstalk between sorafenib and NK cells and provide a rationale to improve drug efficacy via combing NK cell-based therapy with sorafenib. In contrast, some studies reported that sorafenib could inhibit the proliferation and function of NK cells [60]. For instance, sorafenib was stated to dose- and time-dependently decrease the number of NK cells, downregulate the stimulatory receptor CD69 in NK cells, inhibit NK cell proliferation, decrease NK cell cytotoxicity by suppressing the pERK1/2 pathway and blocking the PI3K/AKT pathway [61‐63]. Therefore, the optimized time and dosage must be determined when combining NK cell therapy with sorafenib.

The immunomodulatory effect of sorafenib on TAMs is quite controversial. Accumulating evidence has strengthened the idea that TAMs are generally classified into the pro-inflammatory subset (M1) and the anti-inflammatory subset (M2). Macrophage polarization is the activation state of macrophages tightly regulated by the TME and metabolism [64‐66]. The drug efficacy of sorafenib was shown to partially depend on its effect on TAMs through miR-101 inhibition, reduced VEGF expression, macrophage sensitization, and pyroptosis induction [57, 67, 68]. Transfection of miR-101 has been shown to reduce dual specificity phosphatase 1 expression and cause subsequent down-regulation of ERK1/2, p38, and JNK. Sorafenib was reported to inhibit miR-101 expression and reduce the secretion of TGF-β and CD206 in M2, indicating a shift to the M1 polarized state [67]. As aforesaid, sorafenib could also sensitize macrophages and induce pyroptosis of macrophages and pro-inflammatory cytokines (IL6, TNF-α, and IL12) secretion, leading to subsequent NK cell activation [57, 58]. However, an in vitro study revealed that sorafenib could increase PD-L1 expression in tumour stroma and induce M2 accumulation through upregulating stromal cell-derived 1 alpha (SDF1α) [33]. Sorafenib was subsequently shown to induce M2 polarization in vivo but not in vitro. IFN-α was also proposed to shift the M2-like polarization of TAMs [69]. Additionally, sorafenib was shown to increase the infiltration of macrophages and Treg cells through the HIF-α/NF-κB pathway [70]. A review article compared the results of different studies examining the impact of sorafenib on macrophage in the TME. It raised the consideration that the unclear functional definition of TAMs, different dosages of sorafenib, and the impact of in vivo and in vitro studies may be responsible for discrepancy, as macrophage may experience dynamic changes after exposure to cytokines [23, 71, 72]. Therefore, the precent review described the effect of sorafenib on TAM based on the dosage and the model used (Table 2). In contrast to the results of the published review mentioned above, the cancer cell type is restricted to HCC. These results mainly suggested that sorafenib promoted the pro-inflammatory cytokines secretion in macrophages, which partially explained its antitumour effect, but its effect on altering polarization remained controversial.

The interplay between TANs and sorafenib is also worth noting. TANs have been suggested to exert protumour effect, and TANs recruit Treg cells and macrophages via CCL2 and CCL17 and their receptors. Sorafenib was shown to increase the numbers of TANs and the levels of CCL2 and CCL17 in a mouse HCC model [73]. Regarding Treg cells, on the one hand, recruitment of Treg cells was observed through an increased number of TANs and increased expression of CCL2 and CCL17 [73]. On the other hand, other studies conducted in HCC murine models revealed that sorafenib inhibited the recruitment of Treg cells in HCC [74, 75]. In the case of T cells, although a study showed an increase of CD4+ and CD8+ T cell infiltration in murine models in other malignancies after treatment with sorafenib, a recent study reported that the infiltration of CD4+ and CD8+ T-cells remained constant in HCC models [58, 76]. This finding is consistent with the observation that T cells play no role in the therapeutic effect of sorafenib on HCC [57].

Despite the immunomodulatory effect, sorafenib also induces TME remodelling through interacting with hepatic stellate cells (HSCs). HSCs are mainly responsible for extracellular matrix production and deposition, and have been reported to stimulate HCC cell proliferation and vascularization [77]. In human HSCs treated with different concentrations of sorafenib, sorafenib was shown to dosage dependently inhibit ERK1/ERK2 and Akt phosphorylation, suppress the PDGF signalling pathway, and down-regulated the secretion of PDGF-BB and TGF-β1 from HSCs. As a result, the invasion and proliferation of HCC cells were inhibited [78]. Nonetheless, after sorafenib administration in mouse model of HCC, the survival of HSCs increased through the upregulation of stromal-derived factor 1 alpha (SDF-1α) expression in both tumour and stromal cells and activation of the SDF-1α/C-X-C receptor type 4 (CXCR4) pathway [79]. Meanwhile, MAPK activation was reported in HSCs and promoted the myofibroblast differentiation [80]. More evidence is still needed to verify the effect.

It is well established that the cellular interaction within the TME plays a pivotal role in the EMT. The EMT, the process characterized by a loss of apical-basal polarity and cell–cell adhesion in epithelial cells and a transition to mesenchymal cells, plays a pivotal role in malignant progression and drug resistance [81]. It is triggered by mitogen-activated protein kinase (MAPK), mTOR, and Wnt signalling [82]. Snail homologue 1 (SNAI1) and snail homologue 2 (SNAI2) have also been recognized as key inducers of the EMT, which predict a poor prognosis of HCC [83]. TGF-β, FGF, IL-6, hepatocyte growth factor, insulin-like growth factor-1, and epidermal growth factor derived from CAFs and immune cells in the TME were also identified as EMT inducers [84‐86]. TGF-β, hepatocyte growth factor, epidermal growth factor, and FGF was shown to induce SNAIL production [84, 87, 88]. Specifically, TGF-β contributes to sorafenib resistance in HCC [89]. Sorafenib was postulated to inhibit the TGF-mediated EMT possibly via inhibition of TGF-β and MAPK signalling, and SNAI1 expression in HCC, thus inducing antitumour effects [83, 90]. Blockade of signal transducer and activator of transcription 3 (STAT3) by sorafenib was also reported (Fig. 4) [91, 92]. According to one study, the inhibition of TGF-β was achieved by inducing degradation of cell-surface TβR-II and caveolae/lipid raft-mediated internalization [93]. The underlying mechanism of the induction of the EMT during resistance acquisition following sorafenib administration was also suspected to be associated with abnormal miRNA expression [94]. However, long-term exposure to sorafenib was shown to cause drug resistance with EMT [95]. Sorafenib was also shown to promote the EMT, upregulate snail expression, and activate PI3K/AKT signalling in vitro [96]. Moreover, sorafenib increased the IL-6 expression in HCC cells and promoted metastasis and EMT progression in HCC cells. Knockdown of IL-6 could significantly decrease sorafenib resistance in HCC cells [97]. IL-6 is one of the pro-inflammatory cytokines secreted by TAMs, and sorafenib was shown to induce IL-6 secretion by TAMs [57, 58]. A reasonable hypothesis is that the regulation of sorafenib on TAM contributes to induction of EMT through IL-6 expression. The results mentioned above may suggest that the impact of sorafenib on EMT experience changes with different regimens and dosing. Current evidence has suggested the possibility of overcoming sorafenib sensitivity by targeting processes involved in the EMT.

An important mechanism accounting for the resistance-inducing effect described above is hypoxia caused by decreased vascularization. Hypoxia-mediated TME remodelling usually depends on HIF-driven transcriptional responses [98]. The HIF family comprises of HIF-1, HIF-2, and HIF-3. The HIF-α subunits of HIF-1 and HIF-2 are oxygen-sensitive and correlate with tumour progression [99]. Hypoxia was shown to correlate with the EMT and is related to the formation of the immunosuppressive microenvironment through not only homing bone marrow-derived cells (BMDCs), TAMs, and Tregs but also shaping and inducing specific macrophage phenotypes. Specifically, hypoxia was shown to induce M2 polarization and educate PBNs into TANs to promote malignant progression [33, 100, 101]. Several studies concluded that sorafenib could induce HIF-α accumulation and cause subsequent activation of resistant-inducing pathways. NF-κB activation and increased stromal-derived factor 1 alpha (SDF-1α) expression were also observed [79]. Further research based on photoacoustic imaging showed that sorafenib induced a decrease in oxygen saturation and an increase in HIF-α levels [102]. Although a previous study demonstrated that HIF-α was dose-dependently inhibited by sorafenib, recent evidence supports the hypothesis that sorafenib induces HIF-α accumulation, which partially accounts for the anti-angiogenic effect and resistance-inducing mechanism of sorafenib [79, 102, 103].

Lenvatinib is approved as a first-line TKI based on a phase 3 clinical trial conducted among 954 patients with unresectable HCC. With a Median OS of 13.6 months for patients in the lenvatinib arm (compared to 12.3 months for patients in the sorafenib arm) (hazard ratio 0.92, 95% confidence intervals 0.79–1.06), lenvatinib was determined to provide an overall survival benefit that is not inferior to sorafenib in patients with HCC (NCT01761266) [18]. The most common adverse effects were hypertension, diarrhea, appetite decline, and loss of weight in patients treated with lenvatinib [18]. Similar to sorafenib, lenvatinib is also a multitarget TKI that targets VEGFR1-3, TIE2, KIT, RET, RAF-1, BRAF, BRAFV600E, PDGFR, and FGFR1-4 [104‐107]. Compared to sorafenib, lenvatinib exerts a more potent effect on VEGFRs and FGFRs, the inhibition of which might improve immunity due to the immunosuppressive role of VEGFRs and FGFRs [108]. Hence, the current investigations of the effect of lenvatinib on the TME in HCC mainly focus on the immune microenvironment (Fig. 5).

In addition to its antiangiogenic effect, the antitumour effect of lenvatinib was observed to partially depend on the existence of CD8+ T cells, TAM modulation, a reduced Treg proportion, and increased IFN-γ secretion by cytotoxic T cells in murine HCC models [108, 109]. Notably, a study conducted in a murine HCC model reported that CD8+ T cell depletion significantly decreased the potency of lenvatinib, but induced no significant change in the efficacy of sorafenib [108]. Moreover, lenvatinib was also observed to decrease the fractions of macrophage and monocyte populations but increase CD8+ T cell populations [108]. A recent study further revealed that the antitumour effect of lenvatinib in HCC mainly depended on its antiangiogenic effect and its modulation of TAMs. Lenvatinib was shown to decrease the population of TAMs in HCC while increasing the populations of NK and cytotoxic cells in a murine HCC BNL model [109]. Additionally, IFN-γ and granzyme B secretion from cytotoxic T cells increased [109]. In consistence, a study obtaining peripheral immune cells and cytokine profiles of patients with HCC reported that the administration of lenvatinib reduced the numbers of T helper cells and Treg cells, significantly increased the numbers of cytotoxic T cells, increased IL-2, IFN-γ, and IL-5 levels, and decreased IL-6, IL-10, TNF-α, and TNF-β levels, suggesting that further investigations into the immunomodulatory effect of lenvatinib on TME are needed [110].

Interestingly, overexpression of PD-1 induced exhaustion of activated CD8+ cells, and inhibition of PD-1 has been approved as a third-line therapy for advanced HCC in 2017 [111, 112]. The reliance of lenvatinib on the existence of CD8+ T cells has shed light on combining pembrolizumab or nivolumab with lenvatinib [108, 113]. Several animal experiments revealed that the combination of lenvatinib and anti-PD-1 therapy amplified the antitumour effect on HCC, and amplification was presumed to be achieved by exerting immunomodulation through a reduction in TAM infiltration, synergistic modulation on T cells, reversion of immunosuppressive effect caused by anti-PD-1 therapy, and vascular normalization [109, 114, 115]. Long-term administration of PD-1 in patients was reported to increase of VEGF and FGF expression, and lenvatinib exerts a more potent effect on VEGFRs and FGFRs compared to sorafenib as mentioned [108, 114]. In this context, reversion of the immunosuppressive effect caused by anti-PD-1 was proposed to be achieved by upregulating PD-1, cytotoxic T lymphocyte-associated protein-4, and TIM-3 on T cells, downregulating IFN-γ and granzyme B, and inhibiting T cell cytotoxicity [114]. The immunomodulation exerted by the combination of lenvatinib and anti-PD-1 was recently proven to be associated with TGF-β inhibition [115].

Based on these results, clinical trials combing anti-PD-1 therapy with lenvatinib are ongoing. In patients with advanced gastric cancer, a single arm, phase 2 trial reported an ORR of 69% (NCT03609359), indicating potential in HCC therapy [116]. For patients with unresectable HCC, an open-label, multicentre, phase Ib trial has shown that combining lenvatinib with pembrolizumab provided promising antitumour activity with a tolerable safety profile. The median overall survival was 22 months and the objective response rate was reported to be 44.8% in the group receiving lenvatinib and pembrolizumab (compared with 24.1% for patients in the lenvatinib arm of the REFLECT trial) (NCT03006926) [117]. A multicentre, double-blind, phase 3 trial that further investigates this combination is ongoing (NCT03713593).

Regorafenib, a recently approved second-line TKI for patients with HCC who previously received sorafenib, has been proven to significantly prolong the time to progression in a clinical trial engaging 573 patients with HCC progression following sorafenib treatment (NCT01774344) [19]. The adverse events were reported to be hypertension, hand-foot skin reaction, fatigue, and diarrhoea [19]. With one hydrogen atom replaced by a fluorine atom, regorafenib was reported to have higher potency when compared to sorafenib [118, 119]. Targeting VEGFR, FGFR, PDGFRA, KIT, and RET, regorafenib also inhibits the RAF/MEK/ERK pathway [120]. ERK and STAT3 signalling are EMT-inducing pathways in HCC. The anti-HCC effect of regorafenib was shown to be partially induced by p-STAT3-related signaling inhibition through apoptosis [121]. Compared to sorafenib, regorafenib was also demonstrated to exhibit more potent efficacy in STAT3 inhibition through a mechanism mediated by SHP-1 [71].

Regorafenib was shown to modulate macrophage polarization, induce T cell activation, and mediate NK cell function, which enhanced its antitumour effects (Fig. 5). Reversion of M2 polarization in multiple syngeneic liver cancer models was achieved by suppressing the p38 kinase/Creb1/Klf4 axis. Interestingly, the efficacy of adoptively transferred antigen-specific T cells increased after treatment of regorafenib, suggesting the potential for combination therapy [122]. Another recent finding demonstrated that regorafenib treatment inhibited STAT3 and mediated a subsequent increase in CXCL10 expression at both the transcript and protein levels in the murine model and human peripheral blood. CXCL10 is a ligand for CXCR3 that is expressed on tumour-infiltrating lymphocytes and mediates CD8 T cell infiltration within tumour [123]. Moreover, the NK cell function is mediated by CD24 level, which is controlled by p-STAT3. As regorafenib downregulates the p-STAT3 signalling as mentioned above, it might enhance the antitumour effect of NK cells [124]. Regorafenib was also shown to inhibit the expression of MMP 9 as sorafenib did in liver cancer murine model, and thus inhibiting the shedding of MICA [125]. Additionally, in melanoma, the antitumour effect of NK cells is limited by the expression of HLA Class I. Regorafenib was reported to suppress HLA Class I-mediated tumour progression, suggesting the possibility of a similar effect on HCC [126, 127]. Accumulating evidence suggested that regorafenib augments the efficacy of NK cells, but further studies are still needed to verify this effect.

Apart from the immune microenvironment, regorafenib has been shown to obstruct the progression of the EMT by inhibiting of ERK and STAT3 (Fig. 4) [128]. Another study proposed that regorafenib inhibits the EMT by suppressing inhibitor of differentiation 1 (ID1) expression, which downregulates SNAI1 and promotes the EMT [129]. Research in colorectal cancer has described the inhibitory effect of regorafenib on the TGF-β1-induced EMT via enhancement of SHP-1 activity [130]. Interestingly, the effect of regorafenib on the EMT counteracts with sorafenib resistance induced by hepatocyte growth factor, suggesting the rationale for its application in sorafenib resistant patients [131].

Approved by FDA based on a clinical trial, cabozantinib achieved a median OS of 10.2 months in 470 patients with HCC (compared to 8 months in 237 patients who received the placebo) (hazard ratio 0.76, 95% confidence intervals 0.36–0.52; P = 0.0005) (NCT01908426) [20]. The side effects were reported to be hand-foot skin reaction, hypertension, increased level of aspartate aminotransferase, fatigue, and diarrohea [20]. Cabozantinib is a second-line TKI for HCC that targets VEGFR1-3, MET, AXL, and c-KIT. It was reported to exert its antitumour effect mainly by inhibiting VEGFR and cellular-mesenchymal epithelial transition factor (c-MET). C-MET was proven to induce hepatocarcinoma initiation in mice in cooperation with Wnt/β-catenin or Akt/mTOR cascades [132, 133].

Cabozantinib was proposed to exert an immunomodulatory effect by stimulating the neutrophil infiltration and reducing the macrophage infiltration [134, 135]. Previous results from a PTEN/p53-deficient murine prostatic carcinoma model showed that the administration of cabozantinib amplified the secretion of neutrophil chemotactic factors, including CXCL12 and HMGB1, leading to neutrophil infiltration [134]. It was also observed to downregulate M1 macrophages to prevent bone metastasis in prostate cancer cells, whereas it potentiated the growth of prostate carcinoma-associated fibroblasts at the same time [136]. In MC-38-CEA tumour cells derived from murine colon adenocarcinoma, cabozantinib was observed to alter the cell subception in the immune microenvironment, reducing macrophage infiltration [135]. Although evidence has accumulated for other malignancies, a recent study of HCC indicated no significant alteration in immune cells was observed in both c-Met/β-catenin or Akt/c-Met murine models. Interestingly, the combination of cabozantinib with a PD-1 inhibitor increased the numbers of spleen CD3+ CD8+ and CD3+ CD4+ PD-1+ cells in Akt/c-Met mouse [137]. Notably, cabozantinib was shown to have a limited effect on the macrophages, tumour-infiltrating T cells, CAFs, and fibrosis, which differs from the hypothesis proposed before the research [137]. The expression of PD-1 remained unchanged after cabozantinib treatment, and combining cabozantinib with an anti-PD-L1 antibody showed no increase in survival benefit in a nonclinical study [137]. However, a recent clinical trial reported that the combination of nivolumab, ipilimumab, and cabozantinib generated an ORR of 26% for patients with HCC (ORR = 17% in the nivolumab + cabozantinib arm), but the adverse events increased [138].