Aberrant LYZ expression in tumor cells serves as the potential biomarker and target for HCC and promotes tumor progression via csGRP78

Significance

Hepatocellular carcinoma (HCC) is prevalent worldwide with high heterogeneity, and more effective targeted therapies remain urgent. In this study, based on the proteomic profiles of HCC in the patient cohort, we found that Lysozyme (LYZ) had opposite expression patterns between tumor and paracancerous tissues in patients of different subtypes of HCC and that high LYZ abundance in HCC tissues predicted poor prognosis in patients. More importantly, targeting LYZ inhibited HCC growth in both subcutaneous and orthotopic xenograft tumor models. Mechanistically, LYZ promoted HCC cell proliferation and migration in both autocrine and paracrine manners independent of its muramidase activity via cell surface GRP78. These results highlight the potential of LYZ in optimized treatments for HCC patients.

Abstract

Hepatocellular carcinoma (HCC) takes the predominant malignancy of hepatocytes with bleak outcomes owing to high heterogeneity among patients. Personalized treatments based on molecular profiles will better improve patients’ prognosis. Lysozyme (LYZ), a secretory protein with antibacterial function generally expressed in monocytes/macrophages, has been observed for the prognostic implications in different types of tumors. However, studies about the explicit applicative scenarios and mechanisms for tumor progression are still quite limited, especially for HCC. Here, based on the proteomic molecular classification data of early-stage HCC, we revealed that the LYZ level was elevated significantly in the most malignant HCC subtype and could serve as an independent prognostic predictor for HCC patients. Molecular profiles of LYZ-high HCCs were typical of those for the most malignant HCC subtype, with impaired metabolism, along with promoted proliferation and metastasis characteristics. Further studies demonstrated that LYZ tended to be aberrantly expressed in poorly differentiated HCC cells, which was regulated by STAT3 activation. LYZ promoted HCC proliferation and migration in both autocrine and paracrine manners independent of the muramidase activity through the activation of downstream protumoral signaling pathways via cell surface GRP78. Subcutaneous and orthotopic xenograft tumor models indicated that targeting LYZ inhibited HCC growth markedly in NOD/SCID mice. These results propose LYZ as a prognostic biomarker and therapeutic target for the subclass of HCC with an aggressive phenotype.

Liver cancer has become the third leading cause of cancer-related deaths worldwide according to the global cancer statistics 2020 released by International Agency for Research on Cancer (1). Hepatocellular carcinoma (HCC) takes the predominant malignancy of hepatocytes which is usually diagnosed at an advanced stage, making a bleak outcome of traditional curative treatments. Although systemic treatment with tyrosine kinase inhibitors or immune checkpoint blockades provides a benefit for advanced HCC patients, however, the response of patients to these therapies is quite limited with only a modest increase in overall survival (2–6). Tumor heterogeneity, originating from the sequential accumulation of mutated driver genes and interaction between genomic and epigenomic alterations, is a major issue in the field of the molecular biology of HCC that affects therapeutic response and contributes to drug resistance (7–9). Therefore, stratification of HCC patients based on the molecular profiles would better facilitate effective options for prognostic evaluation and personalized treatments.

With the rapid development of omics technology and analytic power, researchers have achieved much progress in classifying HCC by molecular profiles from both transcriptomic and proteomic layers (10, 11), to unveil the heterogeneity and further provide optimized treatment strategies. Commonly, HCC was divided roughly into two major classes on the basis of transcriptomic characteristics, the proliferation class and the nonproliferation class, with each encompassing ~50% of HCC patients (7, 11). The proliferation class is typical of the activation of signaling pathways involved in cell survival and proliferation, including rat sarcoma (Ras)-mitogen-activated protein kinase (MAPK), phosphatidylinositol 3-kinase (PI3K)-protein kinase B (AKT)-mammalian target of rapamycin (mTOR), and hepatocyte growth factor (HGF)-mesenchymal-epithelial transition factor (MET) cascades, which is more aggressive and generally associated with a poor prognosis. Our previous work based on the proteomic data has stratified 101 clinical early-stage HCCs into three subtypes (S-I, S-II, and S-III subtypes), each of which had a different clinical outcome. The S-III subtype, characterized by proliferation, immune infiltration, and disrupted cholesterol homeostasis, was associated with the lowest overall survival and the highest recurrence after first-line surgery (12, 13), and this was verified by an additional subsequent study (12, 13). The need to understand the molecular pathogenesis of HCC to accurately screen patients with poor prognosis and to develop more effective targeted therapies remains urgent.

Lysozyme (LYZ) was first found and named for its bacteriolytic activity in human nasal mucus by Alexander Fleming 100 y ago (14). Since its discovery, LYZ has been found to exist ubiquitously in human blood and secretions in abundance, including tears, saliva, urine, and mucosal surfaces, with both structure and function conserved across animal species (15). Through the secretion generally by monocytes/macrophages and directly hydrolyzing bacterial cell wall peptidoglycan at β-1,4 glycosidic linkages or binding to the lipopolysaccharide, LYZ is commonly recognized as a crucial component in the context of innate immunity (16, 17). Further, LYZ was recently studied to function in signaling transduction by directly activating Toll-like receptor 4 during neuroinflammation (18) and modulate insulin trafficking by liberating Nod1 ligands from microbes in pancreatic beta cells during host glucose tolerance (19).

The correlation between LYZ and human tumors was also noticed well before for the elevated level of LYZ in serum and urine in monocytic and monomyelocytic leukemia (20, 21), rendering LYZ the potential to serve as a biomarker for monocytic/monomyelocytic leukemia and related kidney injury (22, 23). More recently, bioinformatic studies exploring the immune signatures of melanoma, laryngeal squamous cell carcinoma, glioblastoma, and soft tissue sarcomas indicated the mRNA level of LYZ as an immune-related marker to predict the prognosis or response of these tumors to immune therapies (24–27). Except for myeloid and leukemic cells, solid tumor cells were also found to express LYZ, which correlated with tumor prognosis. In a study involving 177 breast cancer tissue sections, researchers disclosed that 126 of these sections showed positive staining of LYZ in tumor cells detected by immunohistochemistry (IHC). Importantly, a high score of LYZ staining correlated with better prognostic overall survival and relapse-free survival, implying that LYZ is a favorable prognostic factor in patients with breast cancer (28). In contrast, retrospective studies about human gastric carcinoma and oral squamous cell carcinoma indicated tumoral expression of LYZ, but not infiltrated stromal cells or normal tissue cells, associated with poor prognosis and carcinogenesis (29, 30). HCC cells were also found to express and secrete LYZ, which was considered to be modulated by cytokines-mediated signaling activation and Wingless-related integration site (WNT) signaling activation (31–33). Despite studies describing the relationship between LYZ and tumors, the prognostic value and roles of LYZ in tumors, especially for HCC, currently remain ambiguous without intensive investigations and therefore need to be further addressed.

In the present study, we found that the LYZ level was elevated significantly in the S-III HCCs, the most malignant subtype according to the proteomic molecular profiles, and proposed LYZ as a prognosticator for HCC. Further analysis identified that the molecular characteristics of LYZ-high HCCs resembled closely those of the S-III subtype of HCC, with promoted proliferation and metastasis. IHC staining results revealed that LYZ originated largely from tumoral expression by HCC cells, which was regulated by signal transducer and activator of transcription 3 (STAT3) activation. Both in vitro and in vivo models indicated that targeting LYZ inhibited HCC cell proliferation remarkably. Finally, we proposed that LYZ promoted HCC development in both autocrine and paracrine manners independent of its muramidase activity through the activation of downstream protumoral signaling pathways via cell surface glucose-regulated protein 78 (csGRP78). Taken together, these results reveal a mechanism through which intratumoral cell–cell interaction orchestrates HCC progression and potentially pave a way for personalized treatment of HCC by targeting LYZ.

Results

Proteome-Based Molecular Classification Identifies Tumor Cell-Originated LYZ as an Independent Prognostic Predictor for HCC Patients with Poor Outcome.

Mass spectrometry (MS) has identified the proteomic intensity of 101 sections from hepatitis B virus–related early-stage primary HCC in our previous work (SI Appendix, Table S1) (12). We referred to the 478 S-III up-regulated proteins (https://www.nature.com/articles/s41586-019-0987-8) and performed hierarchically progressive analysis to screen candidate proteins that are robust for both screening and targeted treatment for patients with poor outcomes. After stringent filtering for both prognostic and druggable potential, twelve candidate druggable proteins with a hazard ratio (HR) value over 1.5 were identified (SI Appendix, Table S2). Of these proteins, LYZ showed both markedly differential expression across proteomic subgroups and the highest protein abundance in the most malignant S-III subtype of HCC and therefore was focused (Fig. 1A). Although the LYZ intensity between 101 tumor tissues and paired paracancerous tissues tended to be similar (SI Appendix, Fig. S1A), however after molecular stratification, this trend for LYZ expression was quite different. As shown in Fig. 1B, in the S-I and S-II subtypes of HCC with relatively better prognosis, LYZ intensity in paracancerous tissues was higher, while in the S-III subtype with the worst prognosis, LYZ intensity was much higher in the tumor tissues. Further, the LYZ intensity in S-III HCCs was elevated strikingly compared with that in S-I/S-II HCCs (SI Appendix, Fig. S1B). Importantly, high LYZ expression levels notably correlated with the shorter overall survival and relapse-free survival of HCC patients (Fig. 1 C and D), indicating that LYZ is a poor prognosticator for HCC patients. Subsequent analysis revealed that LYZ intensity in patients with high levels of serum alpha-fetoprotein (AFP) or microvascular invasion (MVI) was much higher than that in patients with low levels of AFP or without MVI (SI Appendix, Fig. S1 C and D), which further confirmed the correlation of LYZ with the malignancy of HCC. Forest plots displaying the HR values of LYZ level for HCC patients under different clinical covariates, including AFP, MVI, and recurrence risk after resection (34), indicated that high levels of LYZ positively correlated with poor prognosis, especially for patients with high levels of serum AFP or MVI (SI Appendix, Fig. S1E).

Fig. 1.

LYZ is a prognosticator of both the overall survival and relapse-free survival for HCC patients. (A) Workflow for selecting potential prognostic and druggable proteins for S-III HCCs with dots showing the 12 candidate proteins. Red denotes secretory proteins; blue denotes nonsecretory proteins. HR, hazard ratio (an HR of greater than 1 indicates that the observed variable increases the risk of death); OS, overall survival; RFS, relapse-free survival. (B) Boxplot shows differential LYZ expression in HCC tumor tissues (T) and paired paracancerous tissues (P) of different subtypes of HCC detected by MS. Each dot represents an individual sample analyzed by MS. (C and D) Kaplan–Meier analysis of the OS and RFS probability for HCC patients with different LYZ expression intensities under the optimal cut-off value. The frequency of patients with S-I/II/III HCC in the LYZ-low group was 35, 30, and 14, respectively, and that of patients with S-I/II/III HCC in the LYZ-high group was 1, 2, and 19, respectively. (E) IHC staining of LYZ and CK8/CK18 expression in HCC and paired nontumor liver tissues. Representative graphs indicate positive staining of LYZ in HCC cells (Below) but not in normal hepatocytes (Above). (F) Kaplan–Meier analysis of the OS for HCC patients in relation to LYZ staining scores detected in (E). Patients with positive staining of LYZ in HCC cells (N = 24) or in over 30% stromal cells (N = 11) were classified into LYZ high group; the others (N = 143) were classified into the LYZ low group. (G) Upper HE staining indicates distinct differentiation degrees within a single HCC tissue. IHC analyses of LYZ, Ki-67, and CK8/18 expressions in the same HCC tissue are shown below. The results shown here are representative of three HCC patients. Data in B show upper quartile, median, and lower quartile. (Scale bars in E and G represent 50 μm.) **P < 0.01, ***P < 0.001 (two-tailed paired Wilcoxon test for B; Log-rank test for C, D, and F).

Since LYZ is generally recognized as a marker for myeloid cells, especially for monocytes and macrophages, we therefore analyzed the expression correlation between LYZ and other myeloid cell–associated markers, including CD14 (monocytes), CD68 (macrophages), and myeloperoxidase (neutrophils). As shown in SI Appendix, Fig. S2A, the correlation of LYZ with these common myeloid cell–associated markers varies in HCC. Meanwhile, we found an obvious correlation of LYZ with the stem-like HCC cell markers (KRT19, CD44) (35, 36), suggesting the ectopic expression of LYZ in HCC tissues. To examine this hypothesis, we determined the LYZ expression pattern in the HCC tissue microarray of an external cohort through IHC staining. As indicated in Fig. 1E and SI Appendix, Fig. S2B, positive staining of LYZ in stromal cells was observed in both tumor and paracancerous tissues; meanwhile, positive staining of LYZ was also detected in HCC cells of 24/35 LYZ-high HCC tissues, co-stained with the hepatocyte markers CK8/CK18 (markers used in clinic to distinguish the liver-origin cancer cells) but not in hepatocytes of the adjacent tissues. Importantly, high staining scores of LYZ correlated with shorter overall survival of HCC patients (Fig. 1F), implying that ectopic expression of LYZ in HCC cells promotes tumor progression. Further studies revealed that LYZ preferred to be expressed in the poorly differentiated HCC cells (Fig. 1G). Notably, HCC cells expressing LYZ exhibited higher proliferative potential than those that did not express LYZ, indicated by markedly increased Ki-67 expression in LYZ-positive HCC cells (Fig. 1G). Collectively, these results demonstrate that LYZ has the potential to serve as an independent prognostication for HCC patients, especially for these clinicopathologically malignant patients.

Tumoral Expression of LYZ Is Regulated by STAT3 Activation in HCC Cells.

Having shown that LYZ was aberrantly expressed by HCC cells, we next determined the expression of LYZ in different malignant cell lines of human liver preserved in our laboratory. As shown in SI Appendix, Fig. S3A, LYZ was in the highest abundance in HepG2 cells, followed by an intermediate expression in Huh-7 and PLC/PRF/5 cells, and modest expression in Hep3B cells, whereas LYZ was not detected in SNU-387 and SNU-475 cells, indicated by western blot (WB). The secretion of LYZ was consistent with the intracellular expression in those tested cells (SI Appendix, Fig. S3A). A previous study suggested that tumoral expression of LYZ by HCC cells was associated with WNT signaling activation (33). We then examined the correlation between LYZ expression and activation of several HCC-associated signaling pathways, including STAT3, WNT/β-catenin, AKT/mTOR, and extracellular signal-regulated kinase (ERK) signaling pathways (37, 38). It indicated that the LYZ expression levels increased within Hep3B, Huh-7, and HepG2 cells (SI Appendix, Fig. S3 A and B). Importantly, the phosphorylation of STAT3, but not phosphorylated AKT, ERK1/2, and active β-catenin, was increased in these tested cells as well, which was consistent with the expression tendency of LYZ in these cells (SI Appendix, Fig. S3B), suggesting the correlation of LYZ expression with STAT3 activation. To confirm the regulation of LYZ expression by STAT3 activation in HCC cells, either HepG2 or Huh-7 cells were treated with S3I-201 (the inhibitor of STAT3 phosphorylation) for different time durations, and LYZ expression was determined by WB. As shown in SI Appendix, Fig. S3C, S3I-201 inhibited STAT3 phosphorylation in both HepG2 and Huh-7 cells over time. Notably, LYZ expression was also compromised in both cells (SI Appendix, Fig. S3C). Similarly, S3I-201 inhibited LYZ expression by curbing the activation of STAT3 signaling in a dose-dependent manner (SI Appendix, Fig. S3D). Further, small interfering RNAs (siRNAs) targeting either STAT3 or CTNNB1 were introduced into HepG2 cells to knock down STAT3 or β-catenin expression, respectively. It revealed that STAT3 knockdown decreased LYZ expression as expected (SI Appendix, Fig. S3E). Notably, LYZ expression was also compromised in cells with β-catenin knockdown (SI Appendix, Fig. S3E), which was consistent with the previous study (33). To validate the results, we next stimulated STAT3 and WNT/β-catenin signaling activation, respectively, with IL-6 and ISX-9 to further confirm the codependent regulation of LYZ expression in HCC cells. As shown in SI Appendix, Fig. S3F, either STAT3 or WNT/β-catenin signaling activation induced LYZ expression in Hep3B cells. Strikingly, combined stimulation with both IL-6 and ISX-9 further enhanced LYZ expression compared with the monostimulation. Collectively, these results suggest that the tumoral expression of LYZ in HCC cells is regulated by collaborative signaling activation involving STAT3 and WNT/β-catenin signaling pathways.

Molecular Characteristics of LYZ-High HCCs Reveal Enhanced Malignancy.

Our previous work has defined the molecular profiles of HCCs based on the proteomic profiles and identified impaired metabolism and enhanced proliferation and metastasis in S-III HCCs (12). Given the elevated expression of LYZ in predominantly S-III HCCs, we next explored the molecular features of LYZ-high HCCs. Single-sample gene set enrichment analysis (ssGSEA) identified 23 core pathways associated with HCC carcinogenesis according to the proteomic profiles of HCC (Fig. 2A). It revealed that proliferation and metastasis-associated pathways were modestly increased in LYZ-low HCCs, together with partially retained liver-function-related metabolic signatures (such as fatty acid and amino acid metabolism), suggesting the retained hepatocyte-like characteristics for these tumors compared with the molecular features of paracancerous tissues. Notably, proliferation, metastasis, and stemness-related pathways (such as cell cycle, epithelial-mesenchymal transition, WNT, and STAT3) and proteins (such as CDK1, PCNA, MMP9, and KRT19) were strikingly up-regulated in LYZ-high HCCs, and meanwhile, liver-function-related metabolic pathways and proteins (such as CYP27A1, MAOA, and APOC3) were down-regulated significantly (Fig. 2A and SI Appendix, Fig. S4A), which resembles closely the molecular characteristics of S-III HCCs. The liver lobules are typically subdivided into three metabolic zones: zone 1, zone 2, and zone 3. Zone 1 (periportal zone) hepatocytes distribute around the nutrient and oxygen-rich periportal tracts, which were studied to be responsible for lipid oxidation, glycogen synthesis, and protein synthesis, while zone 3 (pericentral zone) hepatocytes distribute around the central veins, which were reported to be responsible for glycolysis and biotransformation reactions, and generally have a high neoplastic potential (39, 40). In our study, we found that the markedly down-regulated metabolic signatures in LYZ-high HCC, including fatty acid oxidation and amino acid metabolism (Fig. 2A), are typically zone 1-associated biological processes. Differential expression protein analysis for liver zonation markers indicated that zone 1-associated proteins (ARG1, PCK1, and MTHFS) were also significantly down-regulated, while the zone 3-associated protein ICAM1 was markedly up-regulated in LYZ-high HCC (SI Appendix, Fig. S4A), consistent with the notion that zone 3 hepatocytes generally have a high neoplastic potential (40). Further, gene set enrichment analysis (GSEA) revealed that proliferation and metastasis-associated pathways were markedly enriched in LYZ-high HCCs (Fig. 2B), suggesting more malignancy of these tumors. In contrast, liver-function-related metabolic pathways were significantly enriched in LYZ-low HCCs (Fig. 2C).

Fig. 2.

Molecular characteristics of LYZ-high HCCs reveal boosted proliferation and migration. (A) Heat map shows altered pathways at the proteomic layer in paracancerous tissues (P) and HCCs distinguished by LYZ level under the optimal cut-off value. The color of each cell represents the average ssGSEA enrichment score of the indicated pathway; red denotes activation and blue denotes suppression. (B and C) GSEA of proliferation, metastasis, and metabolism-related pathways in LYZ-high HCC versus LYZ low HCC. (D) WB analysis of LYZ expression in WT and different LYZ KO HepG2 cells. Cellular β-Actin expression served as the loading control. (E) Volcano plots indicate the differentially expressed proteins in LYZ KO versus WT HepG2 cells. Red and green bubble plots represent significantly up-regulated and down-regulated proteins, respectively. Representative signature proteins associated with metabolism, proliferation, and migration are indicated. (F) GO analysis shows the biological pathways enriched by differentially expressed proteins in LYZ KO versus WT HepG2 cells. Red and blue bars denote the indicated pathways enriched by, respectively, up-regulated and down-regulated proteins. (G) GSEA plots show the indicated pathways enriched in the LYZ KO HepG2 cells.

To validate the above findings, we additionally established LYZ knockout (KO) HepG2 cells through CRISPR/Cas9-mediated gene editing method, and two KO HepG2 cell strains were obtained and confirmed by sequencing of LYZ gene (SI Appendix, Fig. S4B) and WB analysis of LYZ protein expression (Fig. 2D), respectively. MS analysis of wild-type (WT) and LYZ KO HepG2 cells identified 715 differentially expressed proteins (SI Appendix, Table S3), including 486 up-regulated proteins and 229 down-regulated proteins in LYZ KO HepG2 cells (Fig. 2E). Gene ontology (GO) analysis of these differentially expressed proteins revealed that the up-regulated proteins (such as CYP27A1, MAOA, and CPT1A) enriched predominantly the liver-function-related metabolic pathways, such as fatty acid oxidation, amino acid metabolism, and epithelial cell differentiation as well, implying the partially rescued hepatocytic functions in LYZ KO cells (Fig. 2 E and F). Meanwhile, the down-regulated proteins in KO cells (such as CDC20, WDHD1, and FSCN1) were involved largely in proliferation and migration-related pathways, such as cell cycle, mitosis, cell adhesion, and migration (Fig. 2 E and F). Overlapped differentially expressed proteins from distinct LYZ KO cells versus WT cells enriched similar pathways indicated by GO analysis (SI Appendix, Fig. S4C). Further, Reactome pathway analysis revealed that tumor-related signaling pathways, such as IL6-STAT3 signaling, ERK signaling, and WNT signaling were also strikingly down-regulated (SI Appendix, Fig. S4D). Given that TBX3, a well-studied target downstream of the WNT/β-catenin pathway and generally identified as one of the markers for zone 3 hepatocytes (41–43), was also markedly down-regulated in LYZ KO cells (Fig. 2E), it is suggestive that the metabolic rescue in LYZ KO cells might attribute to the feedback of down-regulated WNT activation, which promoted liver tumorigenesis through up-regulation of zone 3 hepatic metabolism (40, 44). Consistently, GSEA revealed that pathways associated with cell cycle and mitosis were down-regulated, whereas fatty acid metabolism was up-regulated in LYZ KO cells (Fig. 2G), suggesting a reversal of molecular features in these cells. Together, these results demonstrate that LYZ-high HCCs share typical molecular characteristics with S-III HCCs, possessing strikingly enhanced proliferative and metastatic potentials.

Targeting LYZ Inhibits HCC Growth Both In Vitro and In Vivo.

Having shown the correlation of LYZ with the malignancy of HCC from proteomes of both the patients’ cohort and tumor cell line, we next explored the applicative potential of LYZ to serve as a therapeutic target for HCC treatment. Since LYZ is a secreted protein, we first used a specific antibody against LYZ or a competitive inhibitor (N, N′, N″-Triacetylchitotriose, TAG) to neutralize LYZ in the culture supernatant of HepG2 and PLC/PRF/5 cells, both of which expressed and secreted LYZ in considerable abundance (SI Appendix, Fig. S3A). Real-time cell analysis (RTCA) results indicated that neutralization of LYZ in the culture supernatant inhibited both cells’ proliferation significantly in a dose-dependent manner (Fig. 3A). The effects of LYZ on HCC cell migration were additionally studied through the transwell assay in HepG2 and Huh-7 cells, which are prone to migration in vitro system. It revealed that either LYZ-specific antibody or TAG treatment suppressed HCC cell migration markedly, in a dose-dependent manner (Fig. 3 B and C). Notably, the effective concentration of LYZ-specific antibody or TAG to inhibit proliferation and migration was much lower for PLC/PRF/5 or Huh-7 cells than that for HepG2 cells, which might be a link to the relatively low LYZ abundance in those cells compared with that in HepG2 cells (SI Appendix, Fig. S3A).

Fig. 3.

Neutralization of LYZ suppresses HCC cell proliferation and migration. (A) HepG2 or PLC/PRF/5 cells were cultured in the E-Plate 16 wells and treated with the indicated concentrations of LYZ-specific antibody (Ab5, 5 μg/mL; Ab10, 10 μg/mL) or TAG (0.1/0.5/1 mg/mL), or the isotype control IgG. Cell proliferation was monitored by RTCA system for the indicated time periods. Data are representative of three independent experiments. (B and C) Transwell assay for HepG2 (B) or Huh-7 (C) cells treated with the indicated concentrations of LYZ-specific antibody or control IgG. Representative fields of migrated cells and summarized data are shown for at least three independent experiments. (D–G) WT (N = 10) or LYZ KO (N = 7) HepG2 cells were inoculated subcutaneously into NOD/SCID mice as described in Materials and Methods. (D) Average tumor growth was measured by tumor size. (E and F) The image and weight of tumors from each group are shown. Each dot represents an individual tumor from each group. (G) IHC staining of Ki-67 expression in WT or LYZ KO tumors. Representative images and summarized data for Ki-67 staining scores are shown. Each dot represents the average Ki-67 staining score for at least three random fields of an individual tumor. (H–J) Huh-7-Luc cells were orthotopically inoculated into NOD/SCID mice as described in Materials and Methods, followed by intraperitoneal treatment with LYZ-specific antibody (200 μg/mouse, N = 10), Sorafenib (15 mg/kg, N = 8), or control isotype IgG (200 μg/mouse, N = 10) daily. (H) Scheme of the imaging and therapy schedule. (I) Average radiance of luciferase activity of Huh-7-Luc cells in mice treated as indicated. (J) IHC staining of Ki-67 expression in orthotopic tumors from mice treated as indicated. The results shown here are representative of at least three mice. Data in A–D and I are shown as the mean ± SEM. [Scale bars in B and C represent 200 μm (Upper) or 100 μm (Lower). Scale bars: G (50 μm); J (20 μm).] *P < 0.05, **P < 0.01, ***P < 0.001 (two-way ANOVA followed by Tukey’s multiple comparisons for A and I, and by Sidak’s multiple comparisons for D; one-way ANOVA followed by Tukey’s multiple comparisons for B and C; two-tailed unpaired Student’s t test for F and G).

Next, we examined the role of LYZ for HCC growth in vivo. WT or LYZ KO HepG2 cells were inoculated subcutaneously to nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice, and tumor growth was monitored every other day. As shown in Fig. 3D, WT HepG2 cells grew into palpable spheric tumors within 2 wk; by contrast, LYZ KO cells grew dramatically slowly. Differential sizes and weights between WT and LYZ KO tumors also illustrated that LYZ deficiency led to notably delayed tumor growth (Fig. 3 E and F). Consistently, IHC results indicated the staining scores of Ki-67, a nuclear marker for cell proliferation, were much lower for LYZ KO tumoral sections than those for WT sections (Fig. 3G). To further evaluate the therapeutic potential of the neutralizing antibody against LYZ to treat HCC in vivo, we next inoculated luciferase-tagged Huh-7 cells (Huh-7-Luc) into the subcapsular region of the liver lobe of NOD/SCID mice and treated tumor-bearing mice with LYZ-specific antibody daily after randomization based on the in vivo luciferase activity (Fig. 3H). It revealed that LYZ-specific antibody treatment markedly suppressed Huh-7-Luc cell growth, compared with control IgG treatment, indicated by in vivo imaging of luciferase activity (Fig. 3I and SI Appendix, Fig. S5A). Meanwhile, we did not observe obvious lesions in the indicated tissues from the treated mice, including the liver, kidney, lung, spleen, and heart (SI Appendix, Fig. S5B). Likewise, Sorafenib treatment daily also suppressed Huh-7-Luc cell growth as indicated (Fig. 3I), which is a first-line drug for systemic treatment of advanced HCC patients (3). Further, IHC staining of Ki-67 revealed decreased proliferation of LYZ-specific antibody or sorafenib-treated tumor cells, compared with the control tumor cells (Fig. 3J). Collectively, these results demonstrate that LYZ is a promising target for HCC treatment.

LYZ Promotes HCC Cell Proliferation and Migration in Both Autocrine and Paracrine Manners Independent of the Muramidase Activity.

The above results have indicated the potential of LYZ to serve as the therapeutic target for HCC, we next explored the mechanism through which LYZ affects HCC cell proliferation and migration. First, two LYZ-specific short hairpin RNAs (shRNAs) and a control shRNA were introduced into HepG2 cells, respectively, via lentivirus-mediated infection. The knockdown efficiency was examined through WB; it indicated that either shRNA, especially the LYZ-shRNA #2 (sh2), reduced both the exogenous LYZ protein expression in transfected HEK 293T cells and the endogenous LYZ protein expression and secretion in transfected and sorted HepG2 cells, compared with the control shRNA (Fig. 4A and SI Appendix, Fig. S6 A and B). These distinct cells were then monitored for cell proliferation by RTCA. It revealed that LYZ knockdown in HepG2 cells resulted in markedly reduced cell proliferation (Fig. 4B), and similar results were obtained in the LYZ KO HepG2 cell strains (Fig. 4C). To further testify the role of LYZ in cell proliferation, we rescued LYZ expression in the LYZ KO HepG2 cells through lentivirus-mediated gene transfer (Fig. 4D). As indicated in Fig. 4E, cell proliferation decelerated in LYZ KO HepG2 cells, compared with WT cells. Markedly, the reconstitution of LYZ rescued the proliferation of LYZ KO HepG2 cells up to the WT level. The effect of LYZ knockdown on cell apoptosis was also examined, and it revealed that the knockdown of LYZ had little effect on HepG2 cell apoptosis indicated by the flow cytometry (FCM) analysis (SI Appendix, Fig. S6C). Since LYZ correlated closely with the mitotic cell cycle (Fig. 2 B and F), we therefore examined the effect of LYZ on the HepG2 cell cycle process. Compared with WT HepG2 cells, LYZ deficiency resulted in a decreased cell population of the G2/M phase; meanwhile, the quiescent population of the G0 phase increased accordingly (Fig. 4F), implying a sort of arrest of the cell cycle by LYZ deficiency. Notably, the reconstitution of LYZ rescued the cell cycle disruption of LYZ KO cells (Fig. 4F). Together, these results suggest the requirement of LYZ for HCC cell proliferation by involving in the cell cycle process.

Fig. 4.

LYZ deficiency results in decelerated cell proliferation by disrupting the cell cycle process. (A) WB analysis of LYZ expression and secretion in cell extract (CE) and culture supernatant (Sup) of HepG2 cells infected with lentivirus expressing either control shRNA (shCtrl) or LYZ-specific shRNA (sh2/sh3). (B and C) Control or LYZ knockdown HepG2 cells (B) or WT or LYZ KO HepG2 cells (C) were cultured and cell proliferation was measured at the indicated time by CCK-8. Data shown are summarized by three independent experiments. (D) WT or LYZ KO HepG2 cells were infected with either the control vector (Vec) or WT LYZ, followed by FACS sorting and WB analysis of LYZ reconstitution in these cells. Cellular β-Actin expression served as the loading control. (E) The indicated cells were cultured and cell proliferation was measured at the indicated time by CCK-8. Data shown are representative of three independent experiments. (F) FCM analysis of the cell cycle process of the indicated cells. Data shown here are summarized for three independent experiments. Data in B, C, E, and F are shown as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, n.s., no significance (two-way ANOVA followed by Tukey’s multiple comparisons).

Since LYZ was not expressed constitutively in hepatocytes and not all HCC cells expressed LYZ intratumorally as studied above (Fig. 1E and SI Appendix, Fig. S3A), we asked whether LYZ-expressing HCC cells affect the performance of LYZ-non-expressing HCC cells through the paracrine of LYZ. Different concentrations of LYZ were supplied into the culture medium of HCC cells, including Huh-7, Hep3B, SUN-387, and SNU-475, all of which modestly or did not express LYZ (SI Appendix, Fig. S3A). The RTCA results revealed that recombinant LYZ promoted the proliferation of the indicated HCC cells, in a dose-dependent manner (SI Appendix, Fig. S7A). Importantly, either LYZ-specific antibody or TAG counteracted the accelerated proliferation of HCC cells induced by recombinant LYZ (Fig. 5A). Of note, the LYZ-specific antibody or TAG itself had little effect on the proliferation of Hep3B, SUN-387, and SNU-475 cells, other than Huh-7 cells. The possibility is the modest or no expression of LYZ in Hep3B, SUN-387, and SNU-475 cells, nevertheless, intermediate expression of LYZ in Huh-7 cells (SI Appendix, Fig. S3A). To further simulate cell–cell interaction via LYZ, we cultured LYZ-low-abundance HCC cells with the culture supernatant of HepG2 cells, which expressed and secreted LYZ in high abundance (SI Appendix, Fig. S3A). We found that neutralization of LYZ in HepG2 supernatant with either LYZ-specific antibody or TAG decelerated the proliferation of LYZ-low-abundance HCC cells, compared with control supernatant (Fig. 5B and SI Appendix, Fig. S7B). Likewise, HepG2 supernatant with LYZ knockdown decelerated proliferation of the indicated cells; more importantly, an exogenous supplement of recombinant LYZ markedly rescued proliferation of the LYZ-low-abundance HCC cells cultured in the supernatant lacking in LYZ (Fig. 5C and SI Appendix, Fig. S7C). The effects of LYZ on the migration of LYZ-low-abundance HCC cells were additionally studied. As shown in Fig. 5D and SI Appendix, Fig. S7D, recombinant LYZ supplement in the medium of the lower chamber significantly promoted migration of Huh-7 and Hep3B cells, whereas neutralization of exogenous LYZ with either LYZ-specific antibody or TAG dramatically counteracted the promigration effects of recombinant LYZ.

Fig. 5.

LYZ promotes HCC cell proliferation and migration in the paracrine manner. (A) The indicated cells were cultured in E-Plate 16 wells and treated with or without recombinant LYZ (1 mg/mL), together with LYZ-specific antibody (10 μg/mL), TAG (0.1 mg/mL), or control IgG (10 μg/mL). Cell proliferation was monitored by RTCA system for the indicated time periods. (B) Huh-7 or Hep3B cells were cultured in the HepG2 supernatant supplied with LYZ-specific antibody (10 μg/mL), TAG (0.1 mg/mL), or control IgG (10 μg/mL). Cell proliferation was monitored by RTCA. (C) The indicated cells were cultured in either control (shCtrl) or LYZ knockdown (sh2/sh3) HepG2 supernatant, supplied with or without recombinant LYZ (1 mg/mL). Cell proliferation was monitored by RTCA. (D) Transwell assay to examine the effect of LYZ on Huh-7 cell migration. Huh-7 cells were plated in the Boyden chamber inserts with or without treatment of recombinant LYZ (1 mg/mL) in the lower chambers, together with the LYZ-specific antibody (10 μg/mL), TAG (0.1 mg/mL), or control IgG (10 μg/mL). Representative fields of migrated cells and summarized data are shown. (E–G) Huh-7-Luc cells together with either WT (N = 7) or LYZ KO (N = 7) HepG2 cells were coinoculated into NOD/SCID mice as described in Materials and Methods. The in vivo growth of Huh-7-Luc cells was determined by IVIS. (E) Scheme of the coinoculation experiment. (F) Summarized results for the average radiance of luciferase activity in mice treated as indicated. (G) Images of the radiance of luciferase activity in each mouse treated as indicated. Data in A–D and F are shown as the mean ± SEM and are representative of at least three independent experiments. [Scale bars in D represent 200 μm (Upper) or 100 μm (Lower).] **P < 0.01, ***P < 0.001 (two-way ANOVA followed by Tukey’s multiple comparisons for A–C and F; one-way ANOVA followed by Tukey’s multiple comparisons for D).

Since macrophages also express high levels of LYZ, we next determined the effects of macrophage-secreted LYZ on HCC cell proliferation and migration with coculture experiments. WT and Lyz2 KO bone marrow-derived macrophages (BMDMs) were induced in vitro (SI Appendix, Fig. S8A) and then coincubated with SNU-475 or Hep3B cells, which did not express LYZ or had low LYZ abundance (SI Appendix, Fig. S3A). RTCA results indicated that treatment with the LYZ-specific antibody to block BMDM-secreted LYZ strikingly inhibited HCC cell proliferation when cocultured with WT BMDMs. Similarly, HCC cells coincubated with Lyz2 KO BMDMs exhibited marked inhibition of cell proliferation, compared with those coincubated with WT BMDMs. More importantly, an exogenous supplement of recombinant LYZ significantly rescued the inhibition of HCC cell proliferation resulting from coincubation with Lyz2 KO BMDMs (SI Appendix, Fig. S8B). Additionally, the Transwell assay revealed that coculture with BMDMs promoted SNU-475 cell migration through the secretion of LYZ, which could be counteracted by LYZ blockage or deficiency (SI Appendix, Fig. S8C). Collectively, these results suggest that macrophage-secreted LYZ promotes HCC cell proliferation and migration in a paracrine manner as well.

To further validate the paracrine effects of LYZ in vivo, Huh-7-Luc cells along with WT or LYZ KO HepG2 cells were coinoculated into NOD/SCID mice, and Huh-7-Luc cell growth was monitored by the in vivo imaging system (IVIS). In vivo luciferase activity revealed that Huh-7-Luc cells coinoculated with LYZ KO HepG2 cells grew more slowly than those coinoculated with WT HepG2 cells (Fig. 5 E–G), and similar results were obtained with LYZ-specific antibody treatment, suggesting that WT HepG2 cells supported the proliferation of Huh-7-Luc cells in vivo. Collectively, these results demonstrate that LYZ-high-abundance HCC cells promote the proliferation and migration of LYZ-low/non-expression HCC cells through the paracrine of LYZ intratumorally.

Previous studies have identified the glutamic acid 35 (Glu35, E35) and aspartic acid 53 (Asp53, D53) to be essential for the muramidase activity of LYZ to hydrolyze bacterial cell wall peptidoglycan (45, 46). We, therefore, sought to explore whether the muramidase activity of LYZ is critical for its proliferation-promoting effects. WT and two LYZ mutates (E35A, D53A) were, respectively, reconstituted into LYZ KO HepG2 cells and then verified for their expression and muramidase activity. As indicated in SI Appendix, Fig. S9 A and B, although substitution to alanine of either Glu35 or Asp53 did not affect the protein expression of LYZ, either mutation was sufficient to substantially abolish the muramidase activity of LYZ. We next tested the effects of the LYZ mutates on HepG2 cell proliferation and cell cycle process. As expected, LYZ deficiency decelerated HepG2 cell proliferation by disrupting the cell cycle process, which was rescued by reconstitution with WT LYZ (SI Appendix, Fig. S9 C and D). Notably, reconstitution with LYZ deficient of the muramidase activity also rescued cell proliferation and cell cycle process up to WT level (SI Appendix, Fig. S9 C and D). The paracrine effects of LYZ mutates on HCC cell proliferation were additionally examined. RTCA results revealed that supernatant containing the reconstituted WT LYZ markedly rescued the LYZ-low/non-expression HCC cell proliferation grown in LYZ-deficient HepG2 culture supernatant (SI Appendix, Fig. S9E). Likewise, the supernatant containing the reconstituted muramidase activity-deficient LYZ rescued the proliferation of these cells as well (SI Appendix, Fig. S9E). Together, these results suggest the muramidase activity of LYZ is dispensable for HCC cell proliferation and cell cycle process.

LYZ Promotes HCC Cell Proliferation through Targeting Cell Surface GRP78.

Having illustrated the roles of LYZ in promoting HCC cell proliferation and migration in both autocrine and paracrine manners above, we thusly sought to explore the molecular mechanism by which LYZ promotes HCC cell proliferation and hypothesized that LYZ functions as a mitogen by targeting certain molecular proteins existing on the plasma membrane (PM) of HCC cells. First, we examined the binding of LYZ onto the HCC cell membrane. As shown in SI Appendix, Fig. S10A, the immunofluorescence (IF) results indicated that LYZ existed on the surface of PLC/PRF/5 cells, which themselves could express and secrete LYZ, suggesting secreted LYZ bound to the outer surface of these cells. To identify the candidate PM proteins that interacted with LYZ, we transfected C-terminal Flag-tagged LYZ into SNU-475 cells and separated PM and cytosol fractions. WB confirmed the purity of these fractions without endoplasmic reticulum (ER) membrane contamination (SI Appendix, Fig. S10B). Fractions of PM were next immunoprecipitated with anti-Flag agarose beads, and the immunoprecipitates were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie blue staining. Compared with control immunoprecipitates, two specific protein bands were observed and detected by MS (Fig. 6A and SI Appendix, Table S4). The MS results identified the protein band near 45 KD to be actin with substantial intensity, which is consistent with a previously reported study (47). Meanwhile, the protein band near 75 KD was identified to be GRP78 with predominant hits. GRP78, also known as heat shock protein family A (Hsp70) member 5 or binding-immunoglobulin protein, is a stress-inducible chaperone protein, which belongs to the Hsp70 superfamily and is typically restricted in the ER lumen. Notably, aberrant translocation of GRP78 to the cell surface was also studied in multiple types of tumor cells, as well as angiogenic endothelial cells and human embryonic stem cells, but rarely in other normal cells (48, 49). In our study, we found that all the tested HCC cells expressed GRP78 in the cytosol with high abundance (SI Appendix, Fig. S10C). Importantly, considerable populations of the indicated cells expressed GRP78 on the cell surface (denoted as csGRP78) (SI Appendix, Fig. S10D), suggesting that csGRP78 could be a receptor-like molecule for mitogen on HCC cells.

Fig. 6.

LYZ promotes HCC cell proliferation through interacting with csGRP78. (A) MS identification of LYZ-interacting cell surface proteins. Control vector (Vec) or Flag-LYZ were transfected into SNU-475 cells, followed by separation of PM fractions and immunoprecipitation with anti-Flag agarose beads. The immunoprecipitates were separated by SDS-PAGE and Coomassie blue staining, followed by MS identification of specific bands. (B and C) IP-WB analysis of the LYZ-GRP78 interaction. (B) Flag-LYZ and HA-GRP78 were cotransfected into HEK293T cells, followed by IP-WB analysis. (C) IP with PM fractions from SNU-475 cells as described in Materials and Methods, followed by WB analysis of LYZ-csGRP78 interaction. GRP78 and ATP1A1 from PM fractions of SNU-475 cells served as the IP control. (D) Confocal analysis of the colocalization of LYZ with GRP78 on the cell surface. PLC/PRF/5 cells were treated with or without monensin, followed by analysis of the colocalization of secreted LYZ with csGRP78. Representative confocal images are shown, with arrows indicating the colocalization of LYZ with GRP78 on the cell surface. (E and F) IP-WB analysis of the effects of TAG or LYZ mutates on LYZ-csGRP78 interaction. (E) IP with PM fractions from SNU-475 cells in the absence or presence of increasing concentrations of TAG (0.1-1 mg/mL), followed by WB analysis of LYZ-csGRP78 interaction. (F) Control vector, Flag-LYZ WT, or the indicated mutations were transfected into HEK293T cells, followed by treatment as described in Materials and Methods and IP-WB analysis of LYZmut-csGRP78 interaction. (G) Huh-7 cells were cultured in E-Plate 16 wells and treated with or without recombinant LYZ (1 mg/mL), together with anti-LYZ (10 μg/mL), anti-GRP78 (10 μg/mL), or control IgG (10 μg/mL) antibodies. Cell proliferation was monitored by RTCA. (H and I) SNU-475 cells were transfected with control or GRP78-specific siRNAs, followed by treatment with or without recombinant LYZ (1 mg/mL). Cell proliferation and migration were determined by RTCA and Transwell, respectively. (J and K) WB analysis of the indicated signaling pathways activation in differential SNU-475 cells stimulated with recombinant LYZ (1 mg/mL). Data in G and H are shown as the mean ± SEM and are representative of three independent experiments. (Scale bars in D and I represent 20 μm and 100 μm, respectively.) ***P < 0.001 (two-way ANOVA followed by Tukey’s multiple comparisons for G to I).

To confirm the immunoprecipitation (IP)-MS results, we first coexpressed LYZ and GRP78 in HEK293T cells and found that LYZ coimmunoprecipitated with GRP78 (Fig. 6B). Importantly, incubation of PM fractions from SNU-475 cells with anti-Flag agarose beads, which were preincubated with cell lysate of Flag-LYZ overexpressed HEK293T cells, revealed that the beads-enriched LYZ also coimmunoprecipitated with GRP78 from PM fractions of SNU-475 cells (Fig. 6C), suggesting the interaction of LYZ with GRP78 on HCC cell membrane. The colocalization of LYZ with GRP78 on the cell membrane of HCC cells was further validated by confocal imaging. As shown in Fig. 6D, LYZ colocalized with GRP78 on the cell surface of PLC/PRF/5 cells, whereas monensin, an inhibitor of protein transport by Golgi apparatus, dramatically blocked the interaction of LYZ with csGRP78. As TAG inhibited HCC cell proliferation through the competitive binding to LYZ (Figs. 3A and 5A), we, therefore, examined the effect of TAG on LYZ-GRP78 interaction. As indicated in Fig. 6E, LYZ interacted with GRP78 from PM fractions of SNU-475 cells. Notably, treatment of these PM fractions with TAG interfered with the LYZ-csGRP78 interaction, in a dose-dependent manner (Fig. 6E). Similar results were obtained by FCM analysis. Incubation of recombinant LYZ resulted in significant adhesion to the GRP78-overexpressed cells, compared with control cells. Nevertheless, TAG suppressed the adhesion of recombinant LYZ to the GRP78-overexpressed cells nearly to the basal level (SI Appendix, Fig. S10E). These results imply that TAG inhibited HCC cell proliferation through competitively binding to LYZ and thusly disrupting the LYZ-csGRP78 interaction. Since LYZ promoted HCC cell proliferation independent of its muramidase activity (SI Appendix, Fig. S9 C and E). We additionally sought to determine the requirement of the muramidase activity of LYZ for LYZ-csGRP78 interaction. It revealed that incubation of PM fractions from SNU-475 cells with the beads-enriched LYZ mutates had little effect on the LYZ-csGRP78 interaction, compared with the beads-enriched WT LYZ (Fig. 6F).

Further, we attempted to clarify the functions of GRP78 in LYZ-augmented HCC cell proliferation. As shown in Fig. 6G, the LYZ supplement accelerated Huh-7 cell proliferation as expected, the effect of which was counteracted by neutralization with the LYZ-specific antibody. Strikingly, the LYZ-augmented cell proliferation was also compromised by blockade with GRP78-specific antibody, compared with that by the control IgG treatment (Fig. 6G). Likewise, consistent results were obtained with GRP78-specific siRNAs. Knockdown of GRP78 in SNU-475 cells markedly attenuated exogenous LYZ-augmented cell proliferation and migration (Fig. 6 H and I and SI Appendix, Fig. S10F). Since GRP78 was extensively studied to promote tumor progression through the activation of downstream signaling pathways, such as the STAT3, WNT, ERK, and AKT signaling pathways (50). We next examined the activation of these signaling pathways induced by LYZ via GRP78. As shown in Fig. 6J, LYZ stimulation activated the STAT3, WNT, ERK, and AKT signaling pathways in a time-dependent manner. More importantly, the knockdown of GRP78 dramatically impaired the activation of these signaling pathways (Fig. 6K). Collectively, these results elucidate that LYZ interacts with csGRP78 on HCC cells and promotes cell proliferation and migration through the activation of downstream protumoral signaling pathways via the LYZ-csGRP78 axis.

Discussion

HCC ranks as the predominant malignancy of hepatocytes, which is usually diagnosed with bleak outcomes. Owing to high heterogeneity among patients (8, 9), stratification of HCC based on the molecular profiles will definitely facilitate the development of more personalized therapeutic strategies and benefit patients from individualized treatments. Significant efforts have been made by researchers in classifying HCC on the basis of molecular signatures from both the transcriptional and proteomic perspectives. Meanwhile, promising prognostic and therapeutic targets have been identified to distinguish patients who prefer to have bleak outcomes and might benefit from further targeted treatments, such as sterol O-acyltransferase 1 and pyrroline-5-carboxylate reductase 2 (10–13). In the current study, we show the potential of LYZ to serve as a prognostic predictor to discriminate HCC patients with poor outcomes and further reveal the molecular characteristics of LYZ-high HCCs with typically malignant signatures. More importantly, we highlight the therapeutic potential of targeting LYZ for HCC treatment both in vitro and in vivo and additionally elucidate the underlying molecular mechanisms through which LYZ promotes HCC development.

With the exception of the well-known functions participating in antibacterial innate immunity through myeloid cells, the prognostic roles of serum and urine LYZ were observed initially in monocytic/monomyelocytic leukemia and related kidney injury (20, 21). Afterward, the expression of LYZ in solid tumor cells and its correlation with the prognosis of patients were also noted through IHC staining, including breast cancer, gastric carcinoma, oral squamous cell carcinoma, and HCC (28–31). However, studies on the prognostic significance of LYZ in solid tumors and the underlying mechanism for cancer progression have not yet been intensively elucidated, especially for HCC. In the present study, we addressed the heterogeneous expression pattern of LYZ in HCC from the proteomic perspective and revealed that the expression of LYZ in different subtypes of HCC was absolutely distinct. LYZ intensity increased strikingly in the most malignant subtype of HCC and correlated significantly with poor prognosis of HCC patients (Fig. 1 B–D), although there was no significant difference in LYZ intensity between tumor and paracancerous tissues among the whole cohort or even decrease in tumor tissues of the S-I/S-II subtypes of HCC (SI Appendix, Fig. S1A and Fig. 1B). Our findings would provide more objective understanding to scrutinize the roles of LYZ in HCC with a higher resolution based on the proteomic molecular classification and draw attention to taking the interpatient tumor heterogeneity into account to minimize the bias of patient enrollment, especially when analyzed in the small-sized cohort. In the present study, we also elucidated the therapeutic potential of targeting LYZ for HCC treatment by showing that the LYZ-specific antibody significantly suppressed HCC growth both in vitro and in vivo. Although LYZ is expected to serve as a promising target for LYZ-high HCC patients and no obvious tissue lesions were observed in treated mice (SI Appendix, Fig. S5B), the underlying side effects still need to be fully evaluated.

Of note, earlier studies suggested that in vitro incubation of LYZ with monocytes or tumor cells provided resistance to transplanted tumor cells, such as murine fibrosarcoma, possibly by recruiting and potentiating monocyte-mediated tumor cytotoxicity through the cationic property of LYZ (51, 52). More recent studies with functional annotation clustering of gene expression profiling indicated LYZ as an immune-related gene to predict the prognosis or response of tumors to immune therapies (24–27), suggesting an unelicited role of LYZ in the tumor microenvironment. In our work, we revealed that LYZ derived from macrophages also supported HCC proliferation and migration directly (SI Appendix, Fig. S8 B and C). Further genetic evidence from our ongoing studies would elucidate the in vivo roles of LYZ in the interaction between tumor cells and myeloid cells from the perspective of the HCC tumor microenvironment.

Albeit LYZ is known to be expressed predominantly in myeloid cells, especially in the monocytes/macrophages, we found that the correlation of LYZ expression with the expression of myeloid cell–associated markers (CD14, CD68, and MNDA) varies apparently (SI Appendix, Fig. S2); and meanwhile, obvious correlation was observed between LYZ and the stem-like HCC cell markers (KRT19, CD44) (35, 36), implying the heterogeneity of LYZ-expressing myeloid cells and probably an ectopic expression of LYZ in HCC tissues. Exactly, marked expression of LYZ in HCC cells, especially in these poorly differentiated cells, was detected by IHC staining of HCC tissue microarrays, with it detected in 24 out of 35 LYZ-high HCCs, whereas no positive staining of LYZ in hepatocytes was detected in paired nontumor liver tissues (Fig. 1E). Notably, HCC cells expressing LYZ exhibited higher proliferative potential and predicted poor prognosis (Fig. 1 F and G), suggesting that the aberrant LYZ expression in HCC cells correlated with the malignant progression of HCC. Previous studies indicated tumoral expression of LYZ in HCC cells was modulated by cytokines and WNT signaling activation (32, 33). The current study additionally revealed that STAT3 activation might be an alternative driver for the aberrant expression of LYZ in HCC cells (SI Appendix, Fig. S3 C–F). Strikingly, LYZ in turn stimulated the activation of STAT3 signaling pathway in HCC cells (Fig. 6J), suggesting a positive regulatory loop between LYZ and STAT3 in promoting HCC progression. Uncontrolled STAT3 activity in tumors is well recognized to propel mutagenesis and retain the stem cell-like properties of tumor cells (35, 36, 53). Consistently, our data indicated that stemness-related signatures (KRT19, STAT3, and WNT) were significantly enriched in LYZ-high HCC tumors (Fig. 2A and SI Appendix, Fig. S2A), emphasizing the malignant trait of these tumors.

Belonging to the heat shock protein 70 superfamily and generally residing in the ER lumen, GRP78 functions as a vital ER chaperone in a wide range of cellular processes, including protein folding and assembly of membrane or secreted proteins, quality control for misfolded proteins, and regulation of cellular calcium homeostasis (54). Additionally, aberrant translocation of a subfraction of GRP78 to the cell surface has attracted much attention in diverse types of cancer cells under ER stress or ectopic expression conditions (55, 56). Furthermore, the link of csGRP78 to tumor progression, drug resistance, and poor patient outcome has also been extensively studied in different cancer types, including astrocytoma, colorectal cancer, HCC, and breast cancer (57–60), rendering csGRP78 a promising target for cancer imaging and therapies (48, 50). In this study, we identified csGRP78 as a candidate target through which LYZ-high-abundance HCC cells promoted the proliferation of LYZ-low/non-expression HCC cells, indicating that csGRP78 served as a receptor-like protein for LYZ on HCC cells. However, GRP78 knockdown or blockage alone was sufficient to inhibit the proliferation of HCC cells (Fig. 6 G and H), even in SNU-475 cells which are absent of LYZ expression, suggesting an LYZ-independent mechanism for GRP78 to promote HCC proliferation. On the other hand, since GRP78 does not contain classical transmembrane domains, it is thought to exist preferentially as a peripheral protein on the PM and function with other cell surface–binding partners to activate downstream cellular signaling cascades (56). Previous studies reported that csGRP78 coordinated with Cripto, a developmental oncoprotein localized on the cell surface, to enhance cancer cell proliferation by regulating the downstream MAPK/PI3K and transforming growth factor β signaling pathways (61, 62). Blockade of the csGRP78/Cripto interaction on the cell surface disrupted these oncogenic signaling pathways and thusly inhibited cell proliferation. Additional cell surface–binding partners, including but not limited to alpha2-macroglobulin, CD109, prostate apoptosis response-4, and Kringle 5 were also studied for csGRP78 to modulate diverse downstream signaling cascades, dependent of the involved cofactors in specific cells (63–67). Here, we illustrated that LYZ promoted HCC proliferation and migration by targeting GRP78 to activate downstream protumoral signaling pathways, including the STAT3, WNT, ERK, and AKT signaling pathways (Fig. 6 J and K), which would improve our understanding of the mechanism through which LYZ-csGRP78 supports HCC progression.

Taken together, we identify that LYZ performs as a prognostic indicator to distinguish HCC patients with molecular malignancy and poor outcomes based on the proteome-directed classification and further illustrate that tumoral expression of LYZ supports HCC progression in both the autocrine and paracrine manners by interacting with csGRP78 to activate downstream STAT3 signaling pathway. Targeting LYZ displays marked antiproliferation potential against HCC growth both in vitro and in vivo (SI Appendix, Fig. S11). These results shed light on a synergistic mechanism in tumor environment to promote HCC progression through cell–cell interaction and confer LYZ the clinical prospect as a promising prognosticator and therapeutic target for personalized treatment of HCC.

Materials and Methods

Materials used can be found in SI Appendix, including sources of cell lines, sequences, and reagents. Methods and experimental details are also described in SI Appendix, including generation of HepG2-LYZ KO cells, lentiviral package, proliferation and Transwell assays, immunoprecipitation, MS, WB, IF, FCM, IHC, tumor xenografts, candidate proteins screening, differential analysis and functional enrichment analysis for proteomic data, survival analysis, and statistical analysis.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

The HCC proteome data that support the findings of this study were from the iProX database “Proteomics identifies new therapeutic targets of early-stage hepatocellular carcinoma” (12) and are accessible through iProX accession number IPX0000937000 (https://www.iprox.cn//page/project.html?id=IPX0000937000) (68). All study data are included in this article and SI Appendix.