Analysis and Validation of Human Targets and Treatments Using a Hepatocellular Carcinoma–Immune Humanized Mouse Model

MICE

NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (The Jackson Laboratory) and humanized mice were bred and generated, respectively, as described earlier.^{(37, 38)} All manipulations with mice were approved by Institutional Animal Care and Use Committee in Agency for Science, Technology and Research. All animal experimental procedures were conducted in accordance to the approved protocols and Guide for the Care and Use of Laboratory Animals. All human fetal liver and HCC Patient tumor tissues with written consent were obtained from guardians of donors, and in accordance with the ethical guidelines of Kandang Kerbau (KK) Women’s and Children’s Hospital (Singapore) and National University Hospital (Singapore). HCC humanized mice were generated by engrafting different HCC cell lines or PDXs subcutaneously or orthotopically (intrasplenic injection with 2 × 10⁶ purified and live human tumor cells in 50 µL of saline) according to HLA-A (HCC-PDX tumor 1# [A*02 and A*11] matched with hematopoietic stem cell [HSC] 1# [A*02 and A*11]; HCC-PDX tumor 2# [A*02 and A*11] matched with HSC 1# [A*02 and A*11]; HCC-PDX tumor 3# [A*02 and A*01] matched with HSC 2# [A*02 and A*01]; HCC-PDX tumor 4# [A*02 and A*11] matched with HSC 1# [A*02 and A*11]; and HepG2 and LM3 cell lines [A*02 and A*24] matched with HSC 3# [A*02 and A*24]) (Supporting Table S1). To deplete specific immune-cell subsets in immune humanized mice, anti–human cluster of differentiation 45 (anti-hCD45) (a human leukocyte marker), anti-hCD4 (a human T_h cell marker), anti-hCD8 (a human T_c cell maker), anti-hCD14 (a human monocyte marker), anti-hCD19 (a human B-cell marker), and anti-hCD56 (a human natural killer [NK] cell marker) (50 µg/mouse) were given i.v. weekly. For the evaluation of drug efficacy, small molecules or Abs were injected into the mice either alone or in combination. Briefly, C188-9 dissolved in DMSO was injected intraperitoneally (75 mg/kg) on a daily basis, whereas anti–human IL-33 Ab (50 µg/mouse), bevacizumab, and pembrolizumab (5 mg/kg) were injected i.v. on a weekly basis. Further methodological details can be found in the Supporting Information.

The Human Immune System Promotes HCC Proliferation and Angiogenesis

To explore the role of the humanized immune system with respect to HCC, we engrafted HCC-PDX tumor tissue into NSG humanized mice and hCD45-depleted humanized mice that were treated with anti-hCD45 Ab to remove the humanized immune system 2 weeks after engraftment. Interestingly, the size and weight of the HCC tumors and spleens from humanized mice increased compared with those from the NSG and hCD45-depleted humanized mice (Fig. 1A). Similar results were also observed in HCC-PDX tumors from different donors, in both subcutaneous and orthotopic models (Supporting Figs. S1A-S6A). The expression of MKI67 (Ki67) on the HCC-PDX tumor was quantified by using immunohistochemical analysis, which indicated a significant increase in the number of Ki67⁺ cells in the HCC tumor from humanized mice (Fig. 1B and Supporting Figs. S1B-S6B). Moreover, the 5-ethynyl-2′-deoxyuridine (EDU) levels in isolated and purified HCC-PDX tumor cells from the three groups were analyzed using flow cytometry. Similarly, we observed that the HCC-PDX tumor cells isolated from humanized mice had the highest EDU level among these groups (Fig. 1C). Furthermore, angiogenesis was quantified in the vascular endothelium of the HCC-PDX tumor by counting the vessels with a lumen (CD31 immunohistochemical staining that was both negative and positive) (Fig. 1D and Supporting Figs. S1C-S6C). Both the microvessel density (MVD) and the microvessel cross-sectional area (VCSA) were calculated (Fig. 1D and Supporting Figs. S1D-S6D). More vascular invasion was observed in HCC tumors in humanized mice than in both NSG and hCD45-depleted humanized mice, suggesting that the humanized immune system could augment HCC tumor proliferation and angiogenesis. To dissect the underlying molecular mechanisms that regulated the proliferation and angiogenesis affected by the humanized immune system, the LM3 (HCC) cell line was selected because of its sharing the HLA-A type with the HCC-PDX tumor (all are HLA-A type 2). In concordance with the results from HCC-PDX tumors, LM3 tumors in humanized mice were larger than those in NSG and humanized mice treated with anti-hCD45 Abs (Fig. 1E). After 3 weeks of engraftment, the LM3 tumors were harvested, and the purified LM3 HCC cells were subjected to RNA sequencing (Fig. 1F). Principal component analysis (PCA) of all expressed genes (Fig. 1G) and heatmap analysis of differentially expressed (DE) genes (Fig. 1H and Supporting Table S2) revealed that the transcriptomic profiles among these groups were distinct and well separated, indicating that LM3 cells were strongly influenced by the immune system. Results from the hCD45-depleted group indicated that the immune system had already exerted some impact on the tumors even though it had been removed 2 weeks after tumor implantation. Ingenuity pathway analysis revealed that the acute-phase-response signaling pathways were highly modulated in LM3 cells from humanized mice, including triggering receptor expressed on myeloid cells 1, IL-6, high mobility group box 1 (HMGB1), and IL-10 signaling pathways, compared with NSG and hCD45-depleted humanized mice (Fig. 1I). Immunohistochemistry staining of HCC-PDX tumors and western blot analysis of isolated HCC-PDX tumor cells indicated that the JAK2/STAT3 of the tumor was highly phosphorylated in humanized mice (Fig. 1J,K). Hence, we anticipated that the humanized immune system in our model may enhance HCC-PDX proliferation through the activation of the JAK2/STAT3 intracellular signaling pathways in the HCC cells.

Intratumor hCD14 Cells Contribute to Promoting HCC Proliferation and Angiogenesis

To validate the functions of human immune cells from different organs on HCC in vitro, hCD45⁺ cells were isolated from blood, spleen, bone marrow (BM), and tumor-infiltrating leukocytes (TILs) of humanized mice bearing HepG2 cells and LM3 cells (with HLA-A2, similar to HCC-PDX tumors 1#-4#), and the average composition of immune-cell subsets was analyzed (Fig. 2A). A colony formation assay was performed by two-dimensional co-culturing of hCD45⁺ cells and HepG2 cells or LM3 cells for 10 days. The results showed that only TILs could strongly stimulate HepG2 cell and LM3 cells proliferation when they were compared with immune cells from other origins (Fig. 2B). Subsequently, hCD45⁺ cells were isolated from blood and TILs from humanized mice bearing HCC-PDX tumor 1#. The mRNA expression levels of various chemokines, cytokines, and growth factors were analyzed by RT–quantitative real-time PCR; IL-33, IL-6, CCL15, VEGF, CCL16, PDGF, IL-8, IL-10, and CXCR7 were highly expressed in hCD45⁺ TILs (Fig. 2C and Supporting Table S3). To further identify the subtypes of hCD45⁺ TILs that influenced HCC proliferation and angiogenesis, different immune-cell subsets were depleted in vivo using humanized mice bearing HCC-PDX tumor 1# (Supporting Fig. S7). The weight of each HCC-PDX tumor 1# was measured 4 weeks after treatment, and the results revealed that hCD14⁺ or CD19⁺ cell depletion could reduce HCC tumor weight. In contrast, hCD8⁺ cell depletion increased tumor weight (Fig. 2D). Immunohistochemical staining of Ki67 and EDU flow cytometry assays showed a decrease in HCC proliferation after hCD14⁺ cell depletion (Fig. 2E,F and Supporting Fig. S8). Moreover, angiogenesis in the HCC tumor, indicated by the MVD and VCSA, was impaired by hCD14⁺ or hCD19⁺ cell depletion (Fig. 2G,H and Supporting Fig. S8). The presence and protein expression levels of phospho-JAK2 (p-JAK2) and p-STAT3 in the HCC tumor and the purified HCC cells were examined by using immunohistochemistry (Fig. 2I,J and Supporting Fig. S8) and western blotting (Fig. 2K), respectively. Our results showed that only hCD14⁺ cell depletion could significantly diminish the expression of JAK2 and STAT3 in the tumor. In addition, hCD14⁺ or CD19⁺ cell depletion also altered the composition of human immune-cell subsets of TILs (Fig. 2L,M). After the depletion of hCD14⁺ cells, the proportion of total CD8⁺ cells, central memory T_h cells (T_hCMs), effector memory cytotoxic T (T_c) cells (T_cEMs), and central memory T_c cells (T_cCMs) increased, whereas the proportion of effector memory T_h cells (T_hEMs), type 1 macrophages (MØ1s) or type 2 macrophages (MØ2s), and mononuclear MDSCs (M-MDSCs) decreased (Fig. 2L). The total number of CD8⁺ cells, CD19⁺ cells, T_hCMs, T_cEMs, and T_cCMs increased per gram of tumor tissue, whereas effector memory T_h cells (T_hEMs), MØ1s or MØ2s, and M-MDSCs decreased in hCD14⁺ cell–depleted tumor tissues (Supporting Fig. S9A). Similarly, a depletion in hCD19⁺ cells also led to an increase in the percentage of total CD8⁺ cells and T_cCMs and a decrease in the proportion of MØ2s (Fig. 2M). In terms of the cell number per gram of tumor tissue, total CD8⁺ cells, T_hCMs, T_cEMs, T_cCMs, hCD14⁺hCD16⁻ cells, and early-stage MDSCs increased after hCD19⁺-cell depletion (Supporting Fig. S9B). Taken together, the intratumor hCD14⁺ cells could stimulate HCC tumor proliferation and angiogenesis by modulating the profile of the immune cells in the tumor and plausibly through the activation of the IL6/JAK2/STAT3 signaling pathway.

Intratumor Human Immune Cells are Strongly Modulated by HCC Tumors

To characterize the phenotypes and functions of the humanized immune system in the absence or presence of an HCC tumor burden, we isolated multiple immune-cell subsets, including CD4⁺, CD8⁺, CD14⁺, CD19⁺, and CD56⁺ cells, from the spleen and from among the TILs of humanized mice bearing an HCC-PDX tumor and analyzed their RNA profiles (Fig. 3A). Through PCA and heatmap analysis, it was found that the profiles of CD4⁺ (Fig. 3B), CD8⁺ (Fig. 3C), CD14⁺ (Fig. 3D), CD19⁺ (Fig. 3E), and CD56⁺ (Fig. 3F) cells among the groups were distinct and well separated. The list of DE genes was detected and calculated for CD4⁺, CD8⁺, CD19⁺, CD14⁺, and CD56⁺ cells (Supporting Table S2). Despite minor differences between splenic immune cells from humanized mice with and without HCC, the intratumor human immune cells were significantly different, including in terms of the expression of surface markers, chemokines, cytokines, and functional proteins. For example, cell migration (CCL13 and CCL18), angiogenesis and invasion (CCL2, CCL7, CCL9, CCL10, IL-8, IL-10, PDGFA, VEGFA, EGF, MMP7, MMP9, and MMP19), and MØ2 type markers (CD163 and peroxisome proliferator–activated receptor gamma genes) were found to be up-regulated in hCD14⁺ TILs, whereas MØ1 type maker (CD86) and phagocytosis markers (neutrophil cytosol factor 1 [NCF-1], phosphatidylinositol 4-phosphate 5-kinase type-1 gamma [PIP5K1C], vasodilator-stimulated phosphoprotein [VASP], hematopoietic cell kinase [HCK], phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit gamma [PIK3CG], spleen tyrosine kinase [SYK], actin related protein 2/3 complex subunit 4 [ARPC4], AKT1, WASP family member 2 [WASF2], vav guanine nucleotide exchange factor 1 [VAV1], actin related protein 2/3 complex subunit 1B [ARPC1B], P21 (RAC1) activated kinase 1 [PAK1], cofilin 1 [CFL1] and AKT serine/threonine kinase 2 [AKT2]) were down-regulated. These results suggested that HCC tumors could alter human immune cells in humanized mice, particularly on intratumor human immune cells.

IL-33 Secreted From Intratumor hCD14⁺ Cells Up-Regulates IL-6 Expression in Multiple Intratumor Immune Cells and Enhances HCC Proliferation

Previous studies have shown that serum IL-33 levels are higher in patients with HCC than in healthy donors.⁽³⁹⁾ Consistent with these clinical findings, our results also verified the elevation of IL-6 and IL-33 expression in intratumor hCD45⁺ cells (Fig. 2C). In our HCC-PDX humanized mouse model, IL-33⁺ cells were only present in intratumor hCD14⁺ cells, and these HCC cells did not express IL-33 (data not shown), whereas IL-6⁺ cells were detected among the CD4⁺, CD14⁺, and lineage (Lin)⁻ TILs (Fig. 4A-C). Therefore, the correlation between the IL-6 level in the captioned immune cells and the IL-33 level in intratumor hCD14⁺ cells was analyzed in HCC samples from 10 patients. The results showed that the IL-33 level in intratumor hCD14⁺ cells was closely correlated with the IL-6 level in intratumor hCD4⁺ and hCD14⁺ cells but was weakly correlated with the intratumor hCD19⁺ cells and Lin⁻ cells (Fig. 4A). Interestingly, ILC groups from Lin⁻ cells were found in our model, particularly in BM and among TILs (Supporting Fig. S10A). The major ILC subtype in the BM and among TILs were ILC1 and ILC2, respectively (Supporting Fig. S10B). Meanwhile, the majority of the ILC2s could express IL-6, whereas other ILC subsets, such as ILC1, NCR⁺ILC3, and NCR⁻ILC3, did not express IL-6 (Fig. 4E). Interleukin 1 receptor-like 1 (ST2) is a member of the IL-1 receptor family, which is the receptor of IL-33. We found that ST2 was highly expressed in intratumor CD4⁺, CD14⁺, and Lin⁻ cells (Supporting Fig. S11A). Furthermore, the proportion of ST2⁺ IL-6⁺ cells was much higher than that of ST2⁻ IL-6⁺ cells in CD4⁺, CD14⁺, and Lin⁻ cells (Supporting Fig. S11B), suggesting that IL-33 could modulate IL-6 expression through the ST2 receptor. To confirm the regulatory effect of IL-33 on HCC in vivo, anti–human IL-33 Ab was injected into HCC-PDX humanized mice to neutralize IL-33. After 4 weeks of treatment, the tumor weight for these mice was drastically reduced compared with that of their saline-injected counterparts (Fig. 4F). Meanwhile, human IL-33 neutralization led to a decrease in the IL-6 level in intratumor CD4⁺, CD14⁺, and Lin⁻ cells (Fig. 4G) and led to lower activation of JAK2 and STAT3 in HCC cells (Fig. 4H). Furthermore, the MVD and VCSA suggested that IL-33 was also involved in angiogenesis (Fig. 4I,J). After human IL-33 neutralization, the proportions of intratumor Th2 and regulatory T cells decreased, whereas the proportion of T_cEMs increased, suggesting that CD14⁺ monocyte–derived IL-33 may also be involved in regulating other immune-cell types in the tumor environment (Fig. 4K). These results suggested that IL-33 from intratumor hCD14⁺ cells could up-regulate IL-6 expression in multiple cell types, thereby enhancing HCC proliferation.

Co-receptor CD14 Is Essential for IL-33 Production Through Damp (HMGB1)/TLR4/Activator Protein 1 Pathway

To clarify the mechanism of IL-33 production from intratumor hCD14⁺ cells only, we induced IL-33 expression using primary human immune cells and U937 cells in vitro. We attempted to detect IL-33 expression induced by using different reagents, including co-culture with HepG2, Hep3B, Huh7, and PLC5 cell lines and stimulation by lipopolysaccharide (LPS), IL-33, IL-6, HepG2 lysates, Hep3B lysates, Huh7 lysates, and PLC5 lysates. We observed that only HepG2 lysates and LPS could induce high levels of IL-33 in primary human immune cells (Fig. 5A). Meanwhile, we stimulated U937 cells using HepG2 lysates. However, no IL-33 expression caused by the direct addition of HepG2 lysates to U937 cells was observed (Fig. 5B). U937 is a monocytic cell line, and low levels of CD14 and 1,25-dihydroxyvitamin D3 (VD3) may induce CD14 expression in U937; these are results we validated (Supporting Fig. S12). We observed that VD3 could induce CD14 expression and that HepG2 lysates stimulated the induction of IL-33 in VD3-treated U937 cells (Fig. 5B). The flow cytometry plot also indicated that only CD14⁺ cells expressed IL-33 (Fig. 5B). In VD3-treated U937 cells, HepG2 lysates, Hep3B lysates, PLC5 lysates, and LPS could induce high levels of IL-33. We also performed specific targeted hCD14 gene knockdown in U937 cells using five short-hairpin RNAs (shRNAs) (Supporting Fig. S13). TRC1.5 vector controls and shRNAs expressing lentiviral particles were generated using third-generation lentivirus packaging systems. The results indicated that hCD14 shRNAs could significantly impair CD14 expression induced by VD3 in U937 cells (Supporting Fig. S13B). As a result, the intracellular level of IL-33 induced by VD3 and HepG2 lysates was down-regulated in hCD14-silenced U937 cells (Supporting Fig. S13C). In conclusion, CD14 is an essential molecule for IL-33 production. By analyzing the network of highly expressed genes in intratumor CD14⁺ cells, we found that activator protein 1 (AP-1) was as a key transcription factor because of its regulation of many of the up-regulated genes. Similar results, including activation of the AP-1 pathway, can also be observed in the RT–quantitative real-time PCR heatmap of CD45⁺ TILs (Supporting Fig. S14A), U937 cells treated with VD3, HepG2 lysates (Supporting Fig. S14B), and healthy donor CD45⁺ cells treated with HepG2 lysates (Supporting Fig. S14C). To further validate the role of AP-1 signaling in IL-33 expression, we treated primary hCD45⁺ cells with LPS, HCC lysates (HepG2), (R)-Ethyl 6-(N-(2-chloro-4 fluorophenyl) sulfamoyl) cyclohex-1-enecarboxylate (TAK-242) (inhibitor of TLR4) and (E,E,Z,E)-3-Methyl-7-(4-methylphenyl)-9-(2,6,6-trimethyl-1-cyclohexen-1-yl)-2,4,6,8-nonatetraenoic acid (SR-11302) (inhibitor of AP-1). The results showed that either blockage of TLR4 or inhibition of AP-1 can suppress LPS- or HCC lysate–induced IL-33 expression (Fig. 5D). Additionally, we found that LPS and HCC lysates could increase Jun family (JUNB, c-JUN, and JUND) protein levels in hCD14⁺ cells. However, TAK-242 and SR-11302 can mitigate the elevation of Jun Proto-Oncogene (JUN)-family proteins (Fig. 5E). We also confirmed that AP-1 can bind multiple motifs in the promoter region of IL-33 in U397 cells (Supporting Fig. S15A,B). HCC lysates can enhance the binding ability of AP-1 (JUN family) to the promoter of IL-33. TAK-242 and SR-11302 can mitigate the effect of HCC lysates (Fig. 5F). Moreover, HMGB1 protein is a type of DAMP that is typically located in the nucleus as a nuclear factor. However, it can also be translocated to the cytoplasm and released into the extracellular matrix, linking inflammation and HCC. HMGB1 may activate several intracellular signaling pathways (NF-κB or AP-1) that regulate monocyte-induced cytokine synthesis by binding TLRs 2, 4, 5, and 9. Earlier studies have also demonstrated that HMGB1 is highly expressed in HCC tissues compared with paratumor and normal tissues and that it also promotes HCC proliferation and angiogenesis. To determine the role of HMGB1 in IL-33 synthesis, we treated primary hCD45⁺ cells with HCC lysates and anti-HMGB1 neutralizing Abs. The results indicated that anti-HMGB1 neutralizing Abs could attenuate HCC lysate–induced IL-33 levels in CD14⁺ monocytes (Fig. 5G). Furthermore, increased supernatant HMGB1 levels were observed in interferon-gamma (IFN-gamma)–treated HCC cells (Fig. 5H). Overall, IL-33 could be synthesized by CD14⁺ cells through the DAMP(HMGB1)/TLR4/AP-1 pathway.

hCD14⁺ Monocytes Among TILs are Crucial for Combinatorial Immunotherapy on HCC

The above data proved that the presence of intratumor CD14⁺ cells enhances HCC proliferation and angiogenesis by activating the JAK2/STAT3 pathway or releasing angiogenic cytokines (VEGF and so on). HCC-PDX humanized mice were treated with C188-9 (STAT3 inhibitor) or bevacizumab (VEGF inhibitor) alone or in combination with pembrolizumab, and the additive effects were observed following treatment (Fig. 6A,B). Next, the efficacy and safety of triple-combination therapy using C188-9, bevacizumab, and pembrolizumab were evaluated and showed a more significant anti-HCC effect (Fig. 6C). Similar results were consistently observed from subcutaneous (Supporting Figs. S1G, S2G and S3G) and orthotopic mouse models reconstituted with PDXs from different donors (Supporting Figs. S4D, S5D and S6D). The Ki-67 immunohistochemical staining revealed that HCC tumors treated with triple-combination therapy had the lowest level of proliferation (Fig. 6D and Supporting Fig. S15). Besides this, MVD and VCSA results also indicated that triple-combination therapy could reduce angiogenesis (Fig. 6E,F and Supporting Fig. S15). Interestingly, the phosphorylation of STAT3 in HCC tumors was reduced after triple-combination therapy, albeit with an increase in the tumor size following pembrolizumab treatment (Fig 6G,I and Supporting Fig. S16). Meanwhile, the expression of JAK2 was unchanged after triple-combination therapy (Fig. 6H,I and Supporting Fig. S16). To further investigate the role of individual human immune-cell types in this triple-combination therapy, we compared the anti-HCC effect of triple-combination therapy when hCD14 and hCD8 cells were depleted separately in both subcutaneous (HCC-PDX tumor 1#) (Supporting Fig. S17) and orthotopic models (HCC-PDX tumors 1# and 2#) (Supporting Figs. S18 and S19). The results revealed that the anti-HCC effect of triple-combination therapy were greatly hampered without hCD8 or hCD14 cells in HCC humanized mice. It confirmed that these two human immune-cell types were both critical in this combination therapy. Taken together, the results showed that triple-combination therapy using C188-9, bevacizumab, and pembrolizumab could significantly increase the efficacy of the anti-HCC response in vivo compared with monotherapy or dual-therapy.

In summary, our humanized-immune-system HCC mouse model provides a useful in vivo platform for testing HCC combinational immunotherapy on the basis of the dissection of molecular features of the human immune microenvironment and human tumors. A general picture of tumor–immune-cell interaction is proposed in our study (Fig. 7). Initially, when the human immune system is exposed to an HCC tumor, it becomes activated and releases proinflammatory cytokines, such as IFN-gamma. The HCC tumor senses the immune responses and generates resistant reactions by releasing DAMPs, such as HMGB1, to immunomodulate CD14⁺ monocytes. For example, this occurs through the binding of DAMPs to TLR4 and the subsequent synthesis of IL-33 through the TLR4/AP-1 signaling pathway and VEGF, and so on. IL-33 stimulates CD4⁺ cells, CD14⁺ cells, and ILC2s to release IL-6, stimulating HCC proliferation through the JAK2/STAT3 pathway and VEGF-related angiogenesis to overcome the killing activity of the immune system. In this context, we demonstrated that a triple therapy comprising C188-9 to inhibit STAT3-related HCC proliferation induced by IL-6, bevacizumab to block VEGF-induced angiogenesis from intratumor monocytes, and pembrolizumab to block PD-1–triggered T-cell exhaustion induced by PD-L1/2 has a combinational effect.