Lysyl Oxidase-Like 4 Fosters an Immunosuppressive Microenvironment During Hepatocarcinogenesis

Human Samples

A tissue block containing 90 pairs of human HCC tissues and adjacent normal liver tissues was obtained from US Biomax, MD, USA (HLiv-HCC180Sur-03) along with survival data encompassing a follow-up time of 2.3-7 years. Tissue was collected from operations performed between January 2010 and September 2011. Patient informed consent and approval was obtained by US Biomax, Inc.

Animal Experiments

C57BL/6J mice were obtained from The Jackson Laboratory (USA). Four-week-old C57BL/6J mice were fed a choline-deficient, l-amino acid–defined (CDAA) diet (Dyets Inc., PA, USA) and given i.p. injections of carbon tetrachloride (CCl₄; 0.2 μg/g) twice weekly. BALB/cAnN-nu (nude) mice obtained from Charles River Laboratories (MA, USA) were used to establish the orthotopic model of liver cancer, as described.⁽¹⁵⁾ All mice were maintained on a 12-hour light and 12-hour dark cycle, and their body weight was measured weekly. All mice were housed in the pathogen-free laboratory animal unit of The University of Hong Kong and received humane care according to the criteria outlined in the NIH Guide for the Care and Use of Laboratory Animals. The animal procedures detailed here and in the Supporting Information were reviewed and approved by the Committee on the Use of Live Animals in Teaching and Research of The University of Hong Kong.

Statistical Analysis

All statistical analyses in this study were performed using GraphPad Prism 8 software (CA, USA). All data are presented as mean ± standard deviation. Two-group comparisons were performed using a two-tailed unpaired Student t test (data with normal distribution) and Mann-Whitney U test (data with non-normal distribution). For comparisons of more than two groups, both nonparametric test (Kruskal-Wallis test) and one-way ANOVA with Tukey’s multiple comparison test were performed. A P value of less than 0.05 was considered statistically significant.

LOXL4 Is Overexpressed During Hepatocarcinogenesis

CDAA diet-induced hepatic tumors in rodents have been indicated to accurately mimic carcinogenesis in the human liver by our and other studies.^{(16, 17)} CCl₄ was suggested to promote hepatocarcinogenesis by accelerating liver inflammation, fibrosis, and hepatocellular ballooning.⁽¹⁸⁾ In this study, we examined LOXL4 expression during hepatocarcinogenesis using a CDAA+CCl₄–induced liver tumor model in C57BL/6J mice, whereas the choline-sufficient, l-amino acid–defined (CSAA) diet-fed rodents served as control. The livers of mice receiving CDAA+CCl₄ exhibit extensive steatosis at 1-3 months and develop cirrhosis at 6 months and liver tumors at 6-9 months,⁽¹⁹⁾ and this pattern was consistently observed in our study (Supporting Fig. S1A-D). Infiltration of different types of immune cells was then examined in mice fed a CDAA diet for 3, 6, and 9 months. Increased infiltration of CD45⁺ leukocytes was observed at the first 6 months—the nontumoral stage of hepatocarcinogenesis, reflecting the sustained chronic inflammation in the human liver at the early stage of cancer-related liver diseases. Interestingly, the leukocyte infiltration became less pronounced at the tumorigenesis stage, suggesting that the hepatic microenvironment underwent a switch that led to a change in the immune profile in the liver. However, the subpopulation of F4/80⁺ macrophages, CD11C⁺ dendritic cells and CD3⁺ T lymphocytes were consistently increased during hepatocarcinogenesis (Fig. 1A, Supporting Fig. S1E). However, the subpopulation of F4/80⁺ mouse T cell immunoglobulin and mucin domain 4 Tim-4⁺ Kupffer cells was reduced at 9 months (Supporting Fig. S1F), suggesting the depletion of resident macrophages on tumorigenesis. To identify the role of LOXL proteins during hepatocarcinogenesis, we first assessed the expression of LOXL proteins in HCC tissues and adjacent liver tissues. Significant increases in the expression of LOXL4 but not LOXL1-3 were observed in HCC tissues (Supporting Fig. S2A) and were correlated with the poor survival of patients with HCC (Supporting Fig. S2B). In the rodent model of hepatocarcinogenesis, the mRNA level of LOXL1 and LOXL3 was increased at the inflammation stage but decreased at the tumorigenesis stage, whereas that of LOXL4 was dramatically increased at the tumorigenesis stage (Fig. 1B). LOXL2 was not detected in the rodent model. In addition, no marked change over time in the expression of LOXL4 was found in the liver of mice receiving CSAA, suggesting that LOXL4 up-regulation is a pathogenesis-related phenomenon and may not be related to aging. The significant increase in LOXL4 protein expression in the liver of tumorigenic mice at 9 months compared with nontumorigenic mice at 6 months further supported the hypothesis that LOXL4 may participate in shaping the carcinogenic hepatic microenvironment (Fig. 1C). Immunofluorescence staining indicated massive colocalization of LOXL4 protein with F4/80, a specific marker for murine macrophages, but not with cellular surface markers representing other immune cells (Fig. 1D). These observations suggest that LOXL4 is involved in shaping macrophages during hepatocarcinogenesis.

Several lines of evidence have shown that LOXL4 may play an oncogenic role in human HCC.^{(10, 20)} To further identify its role and regulation, we compared the mRNA expression of LOXL4 in tumorigenic mice. In contrast to its protein localization, LOXL4 mRNA was not detectable in F4/80⁺ tumor-associated macrophages (TAMs) but was highly expressed in F4/80^- parenchymal hepatic cells (Supporting Fig. S3A). On the other hand, the expression of LOXL4 in F4/80⁺ macrophages from tumor and nontumor region was minimally changed (Supporting Fig. S3B,C). This pattern was further proven by the observation that bone marrow–derived macrophages (BMDMs) expressed trace levels of LOXL4 mRNA, whereas immortalized murine hepatocytes and the hepatoma cell line Hepa 1-6 had higher levels of LOXL4 (Supporting Fig. S3D). Given the report of a study that liver cancer cells may secrete exosomes carrying LOXL4,⁽¹⁰⁾ we hypothesized that internalization of neoplastic cell-derived exosomes by macrophages leads to an increase in the intracellular LOXL4 level. Indeed, LOXL4 protein but not mRNA was observed in exosomes derived from Hepa 1-6 cells (Supporting Fig. S3E) and could be absorbed by BMDMs (Supporting Fig. S3F). Internalization of Hepa 1-6 cell-derived exosomes increased the LOXL4 protein level in BMDMs, whereas internalization of exosomes from LOXL4-knockdown Hepa 1-6 cells resulted in low LOXL4 protein level in BMDMs (Fig. 1E). We then explored the possible mechanism leading to LOXL4 overexpression in HCC. By retrieving data from The Cancer Genome Atlas, we identified a positive correlation between the copy number and the mRNA expression of LOXL4 in HCC tissues. In addition, the mRNA level of LOXL4 was negatively correlated with its DNA methylation status (Fig. 1F). The DNA demethylating agent 5-azacytidine markedly induced LOXL4 mRNA expression in hepatoma cells (Supporting Fig. S4A). Collectively, these observations suggest that hepatocarcinogenesis involves LOXL4 overexpression through epigenetic regulation in neoplastic cells.

The Protumorigenic Role of LOXL4 in the Liver Requires Macrophages

To study the role of LOXL4, we used the adenovirus (AAV) gene transfer technologies to conditionally knock down the expression of hepatic LOXL4 in an orthotopic murine model of liver cancer. AAV8-delivered short hairpin (sh) RNA against LOXL4 target specifically silenced the expression of LOXL4 in the mice liver (Fig. 2A). Consistent with another study,⁽¹⁰⁾ the hepatic LOXL4-ablated orthotopic mice showed decelerated weekly tumor growth, reduced end-point hepatic tumor volume, and microvascular density (Fig. 2A, Supporting Fig. S4B) compared with their littermates, the negative control AAV8-scrambled shRNA-treated mice. The hepatic infiltration of CD11b⁺ F4/80⁺ TAMs population was largely reduced in LOXL4 knockdown mice (Fig. 2B). LOXL4 silencing in HCC cells showed reduced in vitro survival and proliferation, whereas there was no growth difference in long-term culture (Supporting Fig. S4C). Nonetheless, supplementation of recombinant LOXL4 (rLOXL4) had minimal effect on the HCC cells proliferation and survival (Supporting Fig. S4D), suggesting that intracellular LOXL4 overexpression in HCC cells supports cellular proliferation, whereas secretory LOXL4 may act on tumor microenvironment.

To confirm this, rLOXL4 was intraperitoneally injected into the mice orthotopic model of liver cancer (Supporting Fig. S5A). Injection of murine rLOXL4 increased the LOXL4 protein level in sorted TAMs and resulted in accelerated growth of orthotopic tumors in the livers (Fig. 2C), whereas the food intake and body weight of the mice remained unaltered (Supporting Fig. S5A). In addition, rLOXL4 injection significantly increased the population of CD11b⁺ F4/80⁺ TAMs in the hepatic microenvironment (Fig. 2D, Supporting Fig. S5B). Expansion of TAMs results from monocyte infiltration or local proliferation.⁽²¹⁾ We intraperitoneally injected mice with fluorescent PKH26-stained bone marrow–derived monocytes and found that rLOXL4 injection potently increased the PKH26⁺ population in TAMs, whereas bromodeoxyuridine staining showed that minimal proliferation of resident macrophages was induced by LOXL4 (Fig. 2E). These data suggest that the LOXL4-induced increase in TAMs is primarily due to monocyte infiltration.

To further understand the role of LOXL4 in regulating macrophages, we supplemented BMDMs with rLOXL4 during in vitro maturation. rLOXL4 supplementation notably increased the F4/80⁺ population at the early stage of BMDM maturation; however, this effect became less pronounced at the later stage (Supporting Fig. S5C), suggesting that LOXL4 probably accelerated macrophage maturation. In addition, monocyte chemoattractant protein 1 expression was potently induced by rLOXL4 supplementation, which led to increased motility of BMDMs (Supporting Fig. S5D). To confirm the involvement of macrophages, liposomal clodronate was applied to deplete the TAMs from mice with orthotopic liver tumors. Injection of liposomal clodronate successfully depleted TAMs from mice harboring liver tumors regardless of rLOXL4 treatment, and the effect of rLOXL4 in promoting tumor growth became insignificant (Fig. 3A). As endogenous expression of LOXL4 in HCC cells directly induced cell proliferation, we performed the adoptive transfer of rLOXL4-treated BMDMs to LOXL4-ablated tumor-bearing mice to confirm the role of LOXL4-induced macrophages on HCC growth. Transfer of rLOXL4-treated BMDMs significantly accelerated growth of orthotopic tumors in the livers compared with vehicle-treated BMDMs (Supporting Fig. S5E), suggesting the predominant role of LOXL4 on macrophages in modulating HCC growth in vivo. These findings suggest that the tumorigenic action of LOXL4 in HCC requires the presence and modulation of macrophages.

LOXL4 Shapes the Immunosuppressive Phenotype of Macrophages

Macrophages are recruited to produce proinflammatory factors, which in turn also promote transformed epithelial cell growth.⁽²²⁾ A study postulated that these activated macrophages, which are supposed to kill the aberrant cells, are redifferentiated into cells with a tumor-promoting M2 phenotype in the tumor microenvironment.⁽²³⁾ However, we observed insignificant changes in the expression of the M2 macrophage marker CD206 in macrophages treated with rLOXL4 (Supporting Fig. S6A). This result was further supported by the unaltered expression of M2 cytokines, such as IL-10 and TGF-β (Supporting Fig. S6B). Interestingly, we then observed that rLOXL4 treatment substantially maintained the lymphocyte antigen 6 complex locus C (Ly6C) cell surface marker on differentiated BMDMs (Fig. 3B). Because a study suggested that macrophages expressing high levels of Ly6C exhibit immunosuppressive properties,⁽²⁴⁾ we sorted Ly6C⁺ and Ly6C^- populations from rLOXL4-treated BMDMs. Treatment with rLOXL4 promoted the expression of immunosuppressive genes in both Ly6C^- and Ly6C⁺ macrophages, and the effect was more prominent in Ly6C⁺ cells (Fig. 3C). The immunosuppressive property of rLOXL4-treated macrophages was further supported by the observation that coculture of rLOXL4-treated macrophages with CD8⁺ T cells significantly reduced T-cell proliferation. Intracellular staining of granzyme B and Ki-67, which is a feature of activated cytotoxic T cells, revealed that coculture with rLOXL4-treated macrophages blocked cytotoxic T-cell activation (Fig. 3D). Similarly, TAMs sorted from rLOXL4-treated mice with orthotopic liver tumors exhibited a significant inhibition on the T-cell proliferation and activation (Fig. 3E). rLOXL4 alone showed minimal direct activation on cytotoxic T cells (Supporting Fig. S6C).

Activation of T cells requires l-arginine, which is metabolized by arginase-1 and inducible nitric oxide synthase (iNOS) in macrophages.^{(25, 26)} We observed that treatment with rLOXL4 had a minimal effect on arginase-1 expression but markedly elevated iNOS in macrophages (Supporting Fig. S6D). However, neither supplementation with l-arginine nor inhibition of iNOS restored T-cell proliferation (Supporting Fig. S6E) or the expressions of intracellular Ki-67 and granzyme B (Supporting Fig. S6F), suggesting that the regulatory mechanism of LOXL4 on immunosuppressive macrophages is independent of arginine.

PD-L1 Mediates the Immunosuppression of Macrophages Induced by LOXL4

Macrophages promote immunosuppression in the tumor microenvironment through various mechanisms independent of l-arginine. Interaction of macrophages with tumoricidal T cells through cell antigens such as PD-L1/PD-L2 and CD86/CD80 impairs T-cell activation and induces functional exhaustion.^{(27, 28)} We therefore examined whether LOXL4 is involved in the immune checkpoint regulation of macrophages during HCC growth. The expression of PD-L1 and CD86, instead of PD-L2 or CD80, was significantly induced in vitro in BMDMs treated with rLOXL4 (Fig. 4A). Moreover, induction of PD-L1 and CD86 by rLOXL4 was dose independent in macrophages (Supporting Fig. S7A). Interestingly, rLOXL4 supplementation or LOXL4 depletion failed to induce PD-L1 and CD86 expression in hepatic tumor cells (Supporting Fig. S7B), suggesting that the action of LOXL4 was specific in macrophages. Treatment of BMDMs with rLOXL4 significantly increased PD-L1 mRNA expression (Fig. 4B), suggesting the transcriptional activation of PD-L1 by LOXL4. Blockade of PD-L1 in rLOXL4-treated macrophages using a neutralizing antibody restored the proliferation and activation of cocultured T cells (Fig. 4C). To identify whether PD-L1 presentation mediates the in vivo immunosuppressive and tumorigenic effect, we used an anti-PD-L1 neutralizing antibody to block TAM-expressed PD-L1 in mice with orthotopic liver tumors. Adoptive transfer of splenic T cells was performed to reconstruct the immunogenic microenvironment in tumor-bearing nude mice, and we found that carboxyfluorescein succinimidyl ester (CFSE)-labeled T cells homed to the orthotopic tumors 72 hours after adoptive transfer (Supporting Fig. S7C). rLOXL4 treatment reduced the CD3⁺ T-cell population in hepatic tumors of mice (Fig. 4D) and blunted the tumoricidal effect of the adoptive T-cell transfer. In addition, anti-PD-L1 neutralizing antibody significantly restored the T cell–associated tumor regression (Fig. 4E). Collectively, these findings demonstrate that the immunosuppressive action of LOXL4 on TAMs requires PD-L1 presentation.

To understand the mechanism underlying the regulation of LOXL4 on PD-L1 expression in macrophages, label-free proteomics analysis was performed to evaluate the intracellular protein expression changes in BMDMs exposed to rLOXL4. Analysis on protein expression revealed that expression of signal transducers and activators of transcription (STATs), including STAT1, STAT2, and STAT3, was up-regulated. Immunoblotting analysis, however, observed that rLOXL4 induced STAT1 and STAT3 activation, whereas STAT2 was barely detectable in BMDMs (Fig. 5A). A study observed that heterodimers of STAT1-STAT3 may act as transcription factors in facilitating expressions containing IRF responsive element, most of which are immunosuppressive.⁽²⁹⁾ Blockade of STAT3 or STAT1 by pharmacological inhibition (Supporting Fig. S7D) resulted in reduced expression and transcriptional activation of PD-L1 (Fig. 5B) in rLOXL4-treated BMDMs. In addition, inactivation of STATs-mediated PD-L1 expression in rLOXL4-treated macrophages abolished its suppressive effect on the proliferation and activation of cocultured cytotoxic T cells (Fig. 5C, Supporting Fig. S7D). Furthermore, gene set enrichment analysis revealed that genes associated with the interferon (IFN)-associated signaling pathway were positively regulated by LOXL4. Most of the genes related to IFN-associated signaling were significantly regulated by rLOXL4-exposed macrophages (Fig. 5D). Further validation confirmed the increase of type I IFN-stimulated genes, such as interferon-stimulated gene 15, chemokine (C-C motif) ligand 5, and intracellular adhesion molecule 1, in rLOXL4-treated BMDMs (Supporting Fig. S7E). Blockade of type I IFN-associated signaling by interferon alpha and beta receptor subunit 1 (IFNAR1) neutralizing antibody decreased the STAT3 and STAT1 activation induced by rLOXL4 (Fig. 5E) as well as the expressions of type I IFN-stimulated genes (Supporting Fig. S7E). Accordingly, the increased expression and transcriptional activation of PD-L1 induced by rLOXL4 were also attenuated by IFNAR1 neutralizing antibody (Fig. 5F).

LOXL4 catalyzes the conversion of the amine group and generates hydrogen peroxide (H₂O₂) and ammonia as byproducts in a copper-dependent manner. LOXL4 intervention significantly increased the H₂O₂ level in BMDMs in a time-dependent manner (Fig. 6A). Treatment with the copper chelator tetrathiomolybdate (TTM) blocked H₂O₂ production in rLOXL4-treated BMDMs (Fig. 6B), suggesting that increase of H₂O₂ by LOXL4 requires copper-mediated enzymatic reaction. Acquisition of immunosuppressive phenotype may be induced by H₂O₂-associated redox imbalance in macrophages.⁽³⁰⁾ Direct incubation of BMDMs with H₂O₂ significantly induced the PD-L1 expression (Fig. 6C), whereas blockade of H₂O₂ in rLOXL4-treated BMDMs repressed the induction of PD-L1 (Fig. 6D). Activation of STATs by LOXL4 was also blocked on H₂O₂ scavenge (Fig. 6D), suggesting that LOXL4-induced H₂O₂ activates the IFN-associated STATs-mediated PD-L1. Consistently, copper chelation by TTM decreased the STAT3 and STAT1 activation as well as the expression of PD-L1 induced by rLOXL4 (Fig. 6E). Interestingly, we observed that LOXL4 increased the H₂O₂ level in HCC cells, but the increased H₂O₂ failed to alter STAT1/3 expression in HCC cells (Fig. 6F), suggesting that the H₂O₂-mediated PD-L1 expression by LOXL4 is cell type specific. These observations suggest that the copper-dependent H₂O₂ production was responsible for the activation of IFN-associated STATs-mediated PD-L1 expression in LOXL4-treated macrophages.

LOXL4 Accelerates Tumorigenesis in the Liver

We reasoned that LOXL4 triggers the immunosuppressive function of macrophages through STAT1-STAT3-mediated PD-L1 activation and expected that this action promotes the liver carcinogenesis process. We then used CCl₄ in the CDAA diet-fed rodent model of liver tumorigenesis. rLOXL4 supplementation was provided at the beginning of cirrhosis initiation (month 7) and after tumor formation (month 9). Treatment with rLOXL4 significantly increased the LOXL4 level in the liver (Fig. 7A). Notably, rLOXL4 supplementation accelerated tumor formation when it was continuously administered for 3 months beginning at month 7. This acceleration was also reflected by the increased maximal tumor size, tumor lesion frequency, and liver/body weight ratio (Fig. 7B). However, the effect of rLOXL4 was less pronounced when it was administered after significant tumor formation (month 9), although this pattern could not exclude the possibility of an effect from relatively short-term rLOXL4 treatment. Histological analysis showed that rLOXL4 treatment potently accelerated the neoplastic transformation of the liver, as evidenced by features including solid tumor formation, cytoplasmic inclusions, and a trabecular morphology (Fig. 7C) as well as increased carcinogenic glypican-3⁺, proliferating Ki-67⁺, and CD31⁺CD105⁺ microvascular cells (Fig. 7D). This transformation was accompanied by an increase in TAMs infiltration and a reduction in CD8⁺ cytotoxic T cells in the liver of rLOXL4-treated mice (Fig. 7E), whereas CD4⁺ T cells were not significantly changed. Similarly, we observed potent overexpression of PD-L1 and alpha-fetoprotein in the livers of rLOXL4-treated mice (Fig. 7F). These findings systematically reveal that LOXL4 is an important factor for hepatocarcinogenesis promotion.

Up-Regulation of LOXL4 Correlates With a Dismal Prognosis in HCC

To determine whether LOXL4 correlates with the immune landscape and clinical outcome of HCC, we compared the expression of LOXL4 in liver cancer tissue and adjacent nontumoral tissue by assessing a tissue array containing sections from 90 patients (Supporting Fig. S8A, representative image in Fig. 8A, patient information in Supporting Table S1) and observed a significant increase in LOXL4 expression in tumors. Approximately 70% of LOXL4⁺ cells are CD68⁺ TAMs (67.87% ± 12.03%) (Supporting Fig. S8B), and LOXL4⁺CD68⁺ cells were significantly higher than LOXL4⁺CD68⁻ cells in HCC tissues (Fig. 8B), suggesting that LOXL4 is primarily expressed in TAMs of HCC. Expression of TAMs-specific LOXL4 was positively correlated with PD-L1 level in CD68⁺ TAMs (Fig. 8B). To determine if LOXL4 correlates with the immune landscape and clinical outcome of HCC, we assessed the correlation of LOXL4 expression with patients’ survival. Combined LOXL4 and PD-L1 expression in CD68⁺ TAMs predicted the survival of patients with HCC. Patients with LOXL4^lo PD-L1^lo in CD68⁺ cells of HCC tumors had better overall survival than those bearing LOXL4^hi PD-L1^hi (Fig. 8C). Similarly, CD8A^hi and TAMs-specific LOXL4^lo in CD68⁺ cells predicted better survival of patients with HCC than those of CD8A^lo and TAMs-specific LOXL4^hi (Fig. 8C). This indicates that LOXL4-mediated macrophages probably influence the function of CD8⁺ T cells and clinical outcomes of HCC. Although the expression of LOXL4 in CD68⁺ TAMs showed no significant difference among cohorts of patients with HCC at different stages (Fig. 8D), further subgroup analysis showed that patients with HCC and cirrhosis history expressed higher LOXL4 in CD68⁺ TAMs than participants without cirrhosis (Fig. 8E). LOX family proteins, including LOXL4, have been reported to catalyze the crosslinking of ECM components in a copper-dependent manner, which contributes to fibrogenesis through retarding matrix metalloproteinase–induced proteolytic degradation.⁽³¹⁾ The difference in LOXL4 expression between patients with and without cirrhosis led us to identify the patient’s subgroup based on viral hepatitis status. LOXL4 was significantly overexpressed in HCC tumors of patients who were HBV positive (Supporting Fig. S8C). Furthermore, patients with HBV infection showed increased hepatic LOXL4 expression (Supporting Fig. S8D). Nonetheless, the viral load type had no significant effect on LOXL4 expression, given that patients with HCC and either HBV or HCV background both highly expressed LOXL4 (Supporting Fig. S8E). These data suggest that hepatic LOXL4 expression is a common feature in patients with chronic hepatitis and HCC. Therefore, high expression of LOXL4 in HCC may signify an immunosuppressive microenvironment as well as a dismal clinical outcome.