C57BL/6 mice and BALB/c athymic nude mice (male, 4 weeks old) were obtained from Shanghai Slac Laboratory Animal (Shanghai, China) and fed in a pathogen-free vivarium under standard conditions. The principles and experimental protocols of animals used were approved by the Animal Care and Use Committee of Shanghai Institute of Nutrition and Health, SIBS, CAS.

To establish luciferase-labeled orthotopic HCC model, 1×10⁶ Hepa1-6 tumor cells stably expressing firefly luciferase were suspended in 50 µL Dulbecco’s Modified Eagle Medium (DMEM) (containing 20% Matrigel) and orthotopically injected into the liver of C57BL/6 mice under anesthesia with tribromoethanol (240 mg/kg, Sigma, Massachusetts, USA). After 7 days, tumor volumes were evaluated by bioluminescent imaging with the Xenogen IVIS imaging system (Perkin-Elmer, Fremont, California, USA). Mice were injected with 10 mg/mL D-luciferin (Alameda, California, USA) intraperitoneally at 150 µL per mouse 2 min before imaging. Then tumor-bearing mice were randomized into vehicle and different treatment groups, and the tumor growths were monitored every 5 days.

To establish the subcutaneous HCC model, Hepa1-6 or LPC-H12 cells (2×10⁶ in 100 µL (Fatal Bovine Serum) FBS-free medium containing 20% Matrigel) were injected into the left flanks of recipient mice. Tumor-bearing (about 0.5 cm in diameter) mice were randomized into control and different treatment groups. Tumors were measured by using vernier caliper and the volumes were calculated according to the formula 1/2a×b ², where a was the long diameter and b was the short diameter.

For macrophage depletion, mice bearing LPC-H12 tumors were injected intraperitoneally with 150 µL clodronate liposomes as the first dose and 100 µL clodronate liposomes every 3 days subsequently for longer depletion according to the manufacturer’s instructions. For CD8⁺ T cell depletion, mice bearing LPC-H12 tumors were injected intraperitoneally with 200 µg neutralizing anti-CD8 antibody every 4 days. Vehicle groups received an equivalent amount of saline or in vivo MAb rat IgG2a isotype (BioXcell, West Lebanon, USA).

Bone marrow-derived macrophages (BMDMs, M0) were isolated by pestling the femurs of 8-week-old C57BL/6 mice and cultured for 7 days in Iscove's Modified Dulbecco's Medium (IMDM) containing 10% FBS and 20 ng/mL M-CSF after removal of red blood cells. The medium was changed every 3 days. To mimic tumor microenvironment, BMDM culture medium was changed to condition medium (CM) from Hepa1-6 for 24, 48 or 72 hours (M_hepa1-6) on day 5, 6 or 7, respectively. The CM was collected from Hepa1-6 tumor cells incubated in serum-free medium for 48 hours and added to fresh BMDM culture medium (with the ratio of 1:1). For M1 polarization, matured BMDMs incubated in IMDM containing 10% FBS with 100 ng/mL LPS and 50 ng/mL IFN-γ for 12 hours. For M2 polarization, matured BMDMs incubated in IMDM containing 10% FBS with 10 ng/mL IL-4 and 10 ng/mL IL-13 for 24 hours. For CKI incubation, CKI was added to the culture medium (CKI concentrations: 0.43, 0.66, 1.32 mg/mL, based on the total alkaloid concentration in CKI; incubation time: 12 hours for M_hepa1-6 and M1, 24 hours for M2). All cytokines were purchased from Peprotech (Rocky Hill, USA). The contents of the identified primary bioactive alkaloids in CKI were provided in online supplementary table 1.

Fresh mouse spleen tissue, isolated from 8-week-old C57BL/6 mice, was passed through a 70 µm Strainer (BD Falcon) to obtain single-cell suspensions, and then red blood cells were removed by lysis. CD8⁺ T cells were isolated from single-cell suspensions by using EasySep Mouse CD8⁺ T cell Isolation Kit (Stemcell Technology, Vancouver, Canada) according to the manufacturer’s instructions. For CD8⁺ T cell proliferation assay, CD8⁺ T cells were labeled with carboxyfluorescein succinimidyl amino ester (CFSE, Life Technologies, San Diego, California, USA) and cocultured with CKI-primed macrophages (CD8⁺T: macrophage=5:1, CD8⁺ T number: 2×10⁵) in l640 medium supplemented with 10% FBS, anti-CD3 (2.5 µg/mL, eBioscience) and anti-CD28 (5 µg/mL, eBioscience) for 72 hours. Then, the proliferation rate of CD8⁺ T cells was measured by flow cytometry and quantified by FlowJo V.10 software.

For tumor cell viability assays, Hepa1-6 cells were plated in 96-well plates at 5000 cells per well in DMEM with 10% FBS. After incubated overnight at the incubator, CKI (0.43 mg/mL) and sorafenib (0, 10, 15, or 20 µM) were added into the medium. After 3 days, the cell number was quantitated using Cell TiterGlo ATP Luminescent assay kit (Promega, Madison, USA). For tumor cell lysis assay, Hepa1-6 cells stably expressing firefly luciferase were planted with CD8⁺ T cells, which were cocultured with CKI-primed macrophages before, in 96-well plates in 1640 medium with 10% FBS (CD8⁺T: Hepa1−6=12:1, total cell number: 5000), and incubated at 37℃ for 24 hours. To detect the lysis level of tumor cells, D-luciferin (10 µL per well) was added into the 96-well plates. The concrete fluorescence values of each well were measured by Fluoroskan FL Microplate Fluorometer and Luminometer (Thermo Fisher Scientific).

The data were analyzed by using GraphPad Prism V.7. The statistical significance of differences was examined with Student’s two-tailed t-test. Kaplan-Meier survival curves were plotted and were compared by log-rank test. Flow cytometry data were analyzed by FlowJo V.10. P values <0.05 were considered statistically significant.

To investigate the combined effect of CKI and sorafenib, we determined the anti-HCC effect of sorafenib at a subclinical dose (10 mg/kg) and a clinical dose (30 mg/kg) in syngeneic mouse models. Although tumor growth was efficiently suppressed at 30 mg/kg (online supplementary figure 1A), sorafenib led to body weight loss (online supplementary figure 1B) and a significant increase in aspartate aminotransferase (ALT) and alanine aminotransferase (AST) (online supplementary figure 1C), suggesting that side effects, especially those related to liver toxicity, were induced. In contrast, a subclinical dose of sorafenib (10 mg/kg) resulted in compromised antitumor effects, along with no changes in body weight or liver and renal function (online supplementary figure 1B and C).

To develop a safer but more efficient therapeutic strategy, we assessed the effect of combined treatment of CKI (150 µL per mouse) and subclinical dose of sorafenib in an orthotopic liver tumor model and two subcutaneous models in C57BL/6 mice. Here, CKI alone inhibited tumor growth; however, the combined treatment induced the most marked regression in both orthotopic and subcutaneous tumors (figure 1A-C). As expected, the combined therapy did not cause obvious weight loss in mice, hepatotoxicity or nephrotoxicity (online supplementary figure 2A, B and C), suggesting that this strategy might be potent in the treatment of HCC. Meanwhile, we also compared the therapeutic effect of CKI, as well as the combination treatment with different doses of sorafenib in LPC-H12 subcutaneous tumor model (online supplementary figure 3A). CKI combined with 10 mg/kg sorafenib showed an equivalent anti-HCC effect of clinical 30 mg/kg sorafenib (online supplementary figure 3A). Although CKI combined with 30 mg/kg dose sorafenib possessed the most effective anti-HCC effect (online supplementary figure 3A), this treatment induced a most significant increase in ALT and AST index (online supplementary figure 3E and F), suggesting that CKI could not reduce 30 mg/kg sorafenib-induced side effects. Taking account of the safety and antitumor efficiency, we chose CKI and 10 mg/kg sorafenib for the in vivo combined treatment dose in the following study.

To explore the influence of CKI and sorafenib combined treatment on the HCC recurrence, two surgical subcutaneous HCC resection and treatment models were established (figure 1D,E). Most mice in the vehicle group had relapses after tumor resection, whereas combined therapy, regardless of starting before or after surgery, significantly reduced postsurgical recurrence (figure 1D,E). Further, we explored whether the mice that received combined treatment had anamnestic responses to the rechallenged tumors. The tumor-bearing mice were operated to resect the tumors, and tumor cells were reimplanted (figure 1F). Although no treatment was performed after the second tumor challenge, the growth of the replanted tumors was inhibited in mice that had received CKI and combined treatment beforehand (figure 1F), suggesting that combined therapy induced the long-lived immunologic memory against HCC in mice.

To investigate the influence of CKI combined with low-dose sorafenib on the HCC microenvironment, immune cells infiltrating into the tumors or circulating in the blood were examined by flow cytometry. In both orthotopic and subcutaneous tumors, the amount of total infiltrated immune cells was increased after CKI alone or combined treatment (online supplementary figure 4A). Meanwhile, CKI alone and the combined treatment induced a significant increase in the proportions of M1-TAMs and a reduction of M2-TAMs (figure 2A,B). M0 macrophages in Hepa1-6 subcutaneous tumors were also reduced (online supplementary figure 4B). CD8⁺ cytotoxic T cells were significantly increased in HCC tumors (figure 2A,B). Other immune cells such as CD4⁺ T cells, myeloid-derived suppressor cells (MDSCs), neutrophils, DC cells, and natural killer (NK) cells were not influenced significantly in the blood or tumor tissues, except that the combination therapy diminished MDSCs in the blood of Hepa1-6-bearing mice (online supplementary figure 4B and C). These phenomena were also observed in CKI and 30 mg/kg sorafenib combined treatment (online supplementary figure 3B-D). Besides the influence of treatment on immune cell populations, the expression of function-inhibitory receptors, including lymphocyte-activation gene 3 (Lag-3), programmed cell death protein 1 (PD-1), T-cell immunoreceptor with Ig and ITIM domains (TIGIT) and T-cell immunoglobulin and mucin-domain containing-3 (Tim-3), on CD8⁺ T cells were also determined (figure 2C). The combined treatment effectively reduced their surface levels on CD8⁺ T cells compared with other treatments (figure 2C). These results suggest that CKI and combined therapy resulted in fundamental remodeling of the tumor microenvironment through the action of macrophages and CD8⁺ T cells. It is notable that immunoregulatory effects are not observed in sorafenib-treated mice, which is consistent with the fact that sorafenib directly targets cancer cells. In immunohistochemical assays, tumor tissues in CKI or combination therapy groups showed a decline in M2-TAMs (Arg-1) as well as an increase in M1-TAMs (INOS) and CD8⁺ T cells (figure 2D). Besides this, cleaved caspase-3 and terminaldeoxynucleotidyl transferase mediated dUTP nick-end labeling (TUNEL) staining were enhanced (figure 2D), suggesting that the CKI combined treatment promoted the apoptosis of HCC cells in vivo.

To explore the role of immune microenvironment, we examined the therapeutic effect of CKI and combined treatment on HCC in nude mice. No tumor suppression was observed in any groups (online supplementary figure 5A). In vitro, we investigated whether CKI treatment had a distinct influence on CD8⁺ T cell proliferation and found that CKI had no effect on the proliferation of CD8⁺ T cells (online supplementary figure 5D). Moreover, CKI did not enhance the cytotoxicity of sorafenib to HCC cells in vitro (online supplementary figure 5E). Although nude mice lack T cells, there was still a significant increase in M1-TAMs and reduction of M2-TAMs in tumors after combined treatment (online supplementary figure 5B). CKI alone did not increase the M1-TAMs ratio; however, the M1/M2 fold change was still upregulated by CKI treatment, indicating the tumour microenvironment was still improved by CKI or combined treatment (online supplementary figure 5B and C). These results suggest that CKI improves the therapeutic efficiency of sorafenib through an immunomodulatory rather than cytotoxic mechanism.

We further used the CD8-neutralizing antibody to block CD8⁺ T cells in C57BL/6 mice bearing HCC tumors. Here, we first assessed the CD8⁺ T cell depletion efficiency of CD8a-neutralizng antibody (online supplementary figure 6A). As expected, the Fluorescence Activated Cell Sorting (FACS) results showed that CD8⁺ T cells were efficiently depleted by anti-CD8 antibody, whereas the proportion of monocytes or macrophages was not changed in murine blood and spleen (online supplementary figure 6B). Then, we found that depletion of CD8⁺ T cells abolished the anti-HCC effect of CKI and sorafenib combined treatment (figure 3A), confirming the role of CD8⁺ T cells in combination therapy. After CD8⁺ T cell blockade, although the amount of intratumoral and blood CD8⁺ T cells was lower, there was still a significant increase in M1-TAMs and a decrease in M2-TAMs by combined treatment in the tumor tissues (figure 3B and online supplementary figure 7B), whereas the other types immune cells were not changed both in murine tumor or blood (online supplementary figure 7A and B). The M1/M2 fold change was still upregulated after combined treatment with CD8⁺ T cell depletion (online supplementary figure 7C), indicating that CD8⁺ T cells mediated the antitumor effect but not the macrophage immunomodulation by combined treatment. Moreover, immunohistochemical assays also showed that CD8⁺ T cell depletion did not influence the effect of combined treatment on M2-TAMs and M1-TAMs (figure 3H). Based on these results, we speculated that CKI indirectly enhanced CD8⁺ T cells’ function via activating macrophages in HCC microenvironment.

Next, we explored the role of macrophages in C57BL/6 mice by using clodronate liposomes in C57BL/6 mice bearing subcutaneous LPC-H12 tumor. Both in murine blood and spleen, clodronate liposomes had an efficient depletion on monocytes or macrophages without effect on CD8⁺ T cells (online supplementary figure 6A and B). Macrophage depletion inhibited the tumor growth, supporting the facilitative role of tumor-infiltrated macrophages in cancer development (figure 3C). However, in mice with macrophage depletion, CKI and sorafenib combined treatment no longer suppressed the growth of tumors (figure 3C). Similarly, after depletion, the proportion of M1-TAMs, M2-TAMs or monocytes in the tumor microenvironment or murine blood remained low in both vehicle and treatment groups, while the proportion of CD8⁺ T cells was not elevated by CKI and sorafenib combined treatment (figure 3D and online supplementary figure 7B). In macrophage-depleted tumors, the intratumoral CD8⁺ T cells could not be induced by combined treatment (figure 3D,I), confirming the indispensability of macrophages for CD8⁺ T cell expansion, whereas the other types of immune cells were not changed in murine tumor or blood (online supplementary figure 7A and B). Here, we also found that the increased M1/M2 ratio by combined treatment was abolished by macrophage depletion (online supplementary figure 7C), indicating macrophage depletion also eliminated the combined treatment effect on M2 to M1 transition. To further confirm the role of macrophages in boosting CD8⁺ T cell antitumor function after combined treatment, we used hepa1-6 cells plus BMDMs to establish the macrophage replenishment tumor model in macrophage-depleted mice (figure 3E). Although the tumor growth in macrophage-depleted mice was slower, after replenishment of macrophages, the decreased growth by macrophage depletion was recovered (online supplementary figure 8), suggesting that the observed tumor growth in replenishment group which was comparable with that in control was mediated by macrophages. Importantly, the replenished macrophage-mediated tumor growth recovery from macrophage depletion was further inhibited by combined treatment (figure 3F). Correspondingly, combined treatment upregulated the infiltration of CD8⁺ T cells in macrophage depletion followed by replenishment group (figure 3G). Thus, CKI enhanced the anti-HCC effect of sorafenib by decreasing the M2-TAM distribution, elevating the level of M1-TAMs, and subsequently activating CD8⁺ T cells to enhance antitumor immunity.

In the HCC microenvironment, macrophages can be primed to different statuses, including tumor-killing M1 activation and tumor-promoting M2 activation.11 Since CKI modulated the proportions of M1-TAMs and M2-TAMs in the HCC microenvironment, we determined the effect of CKI treatment on the polarization of macrophages (figure 4A). First, we assessed whether Hepa1-6 CM could make BMDMs polarize to an immunosuppressive status to mimic HCC microenvironment through coculturing BMDMs with Hepa1-6 CM for 24, 24 and 72 hours (online supplementary figure 9A). The results showed that BMDMs were polarized to a more anti-inflammatory status and acquired a more immunosuppressive function on the proliferation of CFSE-labeled CD8⁺ T cells with extending the Hepa1-6 CM coculture time (online supplementary figure 9A-D). Then Hepa1-6 CM-educated BMDMs (M_hepa1-6) were incubated with different concentrations of CKI in vitro. Flow cytometric analysis showed that CKI shifted the peak toward M1 polarization and resulted in a significant increase in M1 macrophages and concomitant decrease in M2 macrophages in a dose-dependent manner (figure 4B). Also, mRNA levels of M1 markers, as well as proinflammatory factors, were increased, while expression levels of M2 markers (anti-inflammatory factors) were reduced after CKI treatment (figure 4C and online supplementary figure 10). The same trends were observed in the concentrations of M1 and M2-related inflammatory cytokines or proteins (figure 4D). These results indicate that CKI primes macrophages toward M1 differentiation.

Next, we examined the influence of CKI on polarized macrophages. BMDMs were induced to M1 or M2 status and then treated with different doses of CKI. The expression of typical inflammatory factors was determined by qPCR and ELISA assays. In both M1 and M2 macrophages, CKI enhanced the levels of proinflammatory genes and downregulated anti-inflammatory factors in a dose-dependent manner (figure 4E–H), suggesting that CKI promoted the macrophages to switch to a dominant proinflammatory polarization. To verify the findings in vivo, we separated the tumor-infiltrating macrophages from HCC tumor burdens in mice and determined the levels of inflammatory factors. In mice treated with CKI and combined therapy, proinflammatory factors in the tumor-infiltrating macrophages were significantly elevated, and anti-inflammatory factors were repressed (figure 4I).

To explore the influence of CKI-primed macrophages on CD8⁺ T cells, primary CD8⁺ T cells were cocultured with M_hepa1-6, M1, or M2 macrophages which had previously been incubated with CKI (figure 5A). We found that despite CKI having been removed, CKI-treated M_hepa1-6, M1, and M2 macrophages promoted the proliferation of CD8⁺ T cells (figure 5B and online supplementary figure 9C and E). Importantly, CKI incubation released the immunosuppressive effect of M_hepa1-6 and M2 macrophages on the proliferation of CD8⁺ T cells in a dose-dependent manner (figure 5B and online supplementary figure 9C and E). Next, we determined the cytotoxic activity of CD8⁺ T cells. After coculture with CKI-primed M_hepa1-6, M1 or M2 macrophages, both gene expression and protein secretion of typical cytotoxic cytokines in CD8⁺ T cells were upregulated (figure 5C,D). To investigate how CKI-primed macrophages act on CD8⁺ T cells (mediated by contact-dependent effect or macrophage-produced soluble molecules), the CKI-primed M_hepa1-6 cell culture supernatant was collected and cocultured with CD8⁺ T cells (online supplementary figure 11). The proliferation levels of cytotoxic cytokines in CD8⁺ T cells, induced by CKI-primed M_hepa1-6 supernatant, were similar to CKI-primed M_hepa1-6(online supplementary figure 11), indicating that macrophage-produced soluble molecules were responsible for the function of CKI-primed macrophages on CD8⁺ T cells. We further examined the tumor-cell-killing activity of CD8⁺ T cells, which were cocultured with CKI-primed macrophages (CKI had been removed in advance), separated, and then coincubated with Hepa1-6 cells. The percentage of lysis of Hepa1-6 cells was significantly increased in CKI treatment groups (figure 5E), suggesting that the cytotoxic function of CD8⁺ T cells was markedly improved by CKI-primed macrophages. Moreover, levels of typical exhaustion-specific genes were diminished in CD8⁺ T cells that had been cocultured with CKI-primed macrophages previously (figure 5C). Cytotoxicity elevation and exhaustion remission of CD8⁺ T cells were confirmed in vivo (figures 2C and 5F). Notably, although cytotoxic markers were increased in CD8⁺ T cells from both CKI and combined treatment HCC tumors, all the levels of CD8⁺ T cell exhaustion markers were reduced in combined treatment, Lag-3 and Tim-3 decreased in the CKI alone treatment group (figures 2C and 5F), suggesting that sorafenib may contribute to relieving the exhaustion of CD8⁺ T cells in vivo. These results could partly explain why CKI combined with sorafenib treatment reduced tumor recurrence and elicited a potent antitumor memory response against HCC.

Macrophages can polarize to proinflammatory M1 status on stimulation by GM-CSF, IFN-β, IFN-γ, LPS, or TNFα through the corresponding receptors CSF2Rα, IFN-α/βR, IFN-γR, TLR4, or TNFR1.26 27 To elucidate the attribution of the effects of CKI on macrophages, we detected the expression of these receptors in TAMs from HCC tumor tissues. In mice treated with CKI, TAMs showed upregulated expression of IFN-γR, TLR4 and TNFR1 (figure 6A). This finding was confirmed in CKI-primed M_hepa1-6, M1, and M2 macrophages (figure 6B and online supplementary figure 12A). To explore the receptor responsible for the education of CKI to macrophages, we blocked TNFR1, TLR4, or IFN-γR with R-7050, Sparstolonin B, and anti-IFN-γR antibody in M_hepa1-6 cells. TLR4 or IFN-γR blockade did not affect CKI-mediated proinflammatory function and TNFR1 upregulation in macrophages (online supplementary figure 12B and C), whereas CKI-induced elevation of proinflammatory cytokines and TNFR1 was prevented by TNFR1 blockade (figure 6C and online supplementary figure 12C), suggesting that CKI primed macrophages by targeting TNFR1.

On this basis, we determined the influence of CKI on TNFR1 activation and its downstream signaling. Flow cytometry analysis showed the expression of TNFR1 on M_hepa1-6 cell surface was increased by CKI incubation (figure 6D). Besides this, the association of TNFR1 with TRADD, RIP1, and TRAF2 was enhanced (figure 6E), suggesting that CKI facilitated the formation of protein complexes necessary for TNFR1 activation. The protein level of TRAF2, which eventually bridges the TNFR1 complex to trigger the downstream signaling, was also upregulated (figure 6E). These results indicate that CKI enhances the activation of TNFR1 signaling. Consistently, the amounts of p-TAK1, p-IKKα/β, p-IκBα, p-P65, and p-P38 were induced more intensely in M_hepa1-6 after CKI incubation (figure 6F), whereas CKI treatment had little effect on JNK and ERK (online supplementary figure 12D). When cell-surface TNFR1 was neutralized by anti-TNFR1 antibody or the formation of TNFR1/TRADD/RIP1 complex was inhibited by R-7050, the enhanced expression of TRAF2, p-TAK1, p-P65, and p-P38 induced by CKI was also blocked (figure 6G).

Mice bearing LPC-H12 subcutaneous tumors were injected with anti-TNFR1 antibody or R-7050 intraperitoneally during CKI and sorafenib combined treatment. Anti-TNFR1 antibody or R-7050 abolished the anti-HCC effect of CKI combined with sorafenib (figure 6H and online supplementary figure 12E), indicating that TNFR1 mediates the therapeutic effects of CKI. In summary, CKI acts on macrophages to facilitate the association of TNFR1 with TRADD/RIP1/TRAF2 and trigger downstream NF-κB and MAPK p38 pathways which plays proinflammatory functions, tunes TAM status toward M1 activation, and reverses the polarization of M2-TAMs. CKI-primed macrophages further promote the proliferation and tumor-killing activity of CD8⁺ T cells to induce HCC cells apoptosis, protect functional CD8⁺ T cells from exhaustion to maintain anti-HCC immunologic memory activity, and finally suppress HCC tumor growth and recurrence. Owing to the anti-tumor mechanisms of two drugs that mutually reinforce each other, combined treatment of CKI with a subclinical dose of sorafenib could greatly strengthen therapeutic efficacy against HCC while increasing safety (figure 7).