TROY is specifically expressed in liver progenitors
Recently, we established an in vitro hepatocyte differentiation model, in which human embryonic stem cells (ES) were induced to differentiate into endoderm (DE), liver progenitor (LP), and premature hepatocytes (PH) [10]. Deep RNA-sequencing was conducted for cells in these four developmental stages, as well as normal liver specimens (N) and HCC clinical samples (T). The genes which mark different hepatocyte developmental stages, including ES markers (OCT4 and SOX2), EN markers (FOXA2 and SOX17), LP markers (CK19 and CK7), PH markers (AFP and GPC3) and mature hepatocyte markers (ALB and CYP3A4) showed their unique expression pattern in this model, indicating the validity of this model (Fig. 1A). To identify genes involving stemness regulation, we focused on a cluster of genes with specific expression in liver progenitors (Supplementary Table 1). KEGG pathway and Gene Ontology (GO) enrichment analysis revealed that this cluster of genes was closely associated with the Wnt signaling pathway, the pluripotency of stem cells, and developmental signaling (Supplementary Fig. 1A). Among these genes, TROY was selected for further study with its distinct expression pattern. Both RNA-sequencing data and qPCR results revealed that TROY reached its peak expression in the liver progenitor stage and dramatically decreased following differentiation but increased in HCC (Fig. 1B). Double staining showed that TROY+ cells were co-expressed with hepatic progenitor marker CK19 (Fig. 1C). These findings suggested that TROY may represent and mark specific HCC CSC subsets.
Overexpression of TROY is correlated with poor outcome
To validate the expression and clinical association of TROY in HCCs, a tissue microarray containing 148 pairs of primary HCCs (tumor vs non-tumor tissues) was applied to analyze the association of TROY expression with clinicopathological features. Informative immunohistochemistry (IHC) results were obtained in 130 HCCs. Non-informative samples included lost samples and samples with too few cells. IHC results found that TROY-expressing cells were detected in most HCC tumor tissues with expression percentage around 0–5%, while the expression of TROY is rarely detected in non-tumor tissues (Fig. 1D, 1E). Based on the frequency of TROY positive cells, the HCC patients were divided into high-frequency group (TROY+, > 0.5%, n = 53) and low-frequency group (TROY−, ≤ 0.5%, n = 77, Supplementary Table 2). Kaplan–Meier survival analysis found that the disease-free survival (DFS) rate was lower in the TROY+ group compared with the TROY− group (P = 0.02; Fig. 1F). In addition, clinical association study found that TROY expression was significantly correlated with age, tumor size and metastasis (Fig. 1G, Supplementary Table 3).
Since TROY, like other CSCs markers expressed in a minor subpopulation of HCC malignant cells, high mRNA expression of TROY represented a high percentage of TROY expression in that patient. Here, bioinformatics-aided analysis based on RNA-sequencing data from the HCC TCGA database to get an in-depth look at whether the expression of TROY was associated with pluripotency signaling in HCC patients. HCC samples were ranked by their expression of TROY, and the differentially expressed genes were identified between the top 20 and the last 20 cases with the highest (TROYhi) and lowest (TROYlo) expression (Fig. 1H). Gene set enrichment analysis (GSEA) results showed that high expression of TROY is significantly associated with the established gene sets “stem cell proliferation” and “regulation of establishment or maintenance of cell polarity” (Fig. 1I). By using Deseq2 (log2 FC ≥ 4, P < 0.01) [22], 823 up-regulated genes and 370 down-regulated genes were found. The expression profile of these genes was able to effectively segregate TROYhi patients and TROYlo patients in unsupervised clustering analysis. TROYhi patients exhibit increased expression of various stemness-related genes (Fig. 1J). Collectively, these findings strongly suggest that TROY might be involved in regulating stemness-related signaling in HCC.
TROY enhances stemness propertises of HCC cells
TROY has 2 major transcripts: TROY isoform 1(#1) encodes a 423 amino acid polypeptide and TROY isoform 2(#2) encodes a 417 amino acid polypeptide. The amino acid sequence of isoform 2 differs from isoform 1 in position 416–423. To investigate the potential role of TROY in cancer stemness maintenance, we firstly checked the expression level of TROY in all HCC and immortalized liver cell lines by flow cytometry, qPCR, and western blot analysis. Compared with immortalized liver cell line LO2, all the other HCC cell lines showed upregulated TROY expression (Supplementary Fig. 1B, C, D). Then the two major isoforms of TROY were separately cloned into a lentiviral vector and stably transfected into Hep3B, PLC8024, and LO2 cells. The ectopic expression of TROY was confirmed in both protein and mRNA levels by qPCR (Supplementary Fig. 2A), immunofluorescence (Supplementary Fig. 2B), and western blot (Fig. 2A, Supplementary Fig. 2C). In contrast to control cells, TROY-overexpressing cells showed marked upregulation of pluripotency markers such as OCT4, NANOG, and SOX2 expression by western blot (Fig. 2B, Supplementary Fig. 2C), qPCR (Supplementary Fig. 2D), immunofluorescence (Supplementary Fig. 2E). Functionally, the introduction of TROY enhanced capabilities of spheroid formation (Fig. 2C, Supplementary Fig. 3A), colony formation (Supplementary Fig. 3B). Also, it accelerated cell cycle progress (Supplementary Fig. 3C, 3D) and cell proliferation (Supplementary Fig. 3E). Since resistance to chemotherapy is an important hallmark of CSCs, we then investigated whether TROY confers chemo-resistance features to HCC cells. TROY-overexpressing cells demonstrated a lower apoptotic rate (Fig. 2D, Supplementary Fig. 3F) and higher cell viability (Supplementary Fig. 4A-C) in the presence of cisplatin, 5-Fu, or sorafenib treatment. As high ALDH activity leads to several types of malignancies, serves as a cancer stem cell marker, and correlates with poor prognosis [23], we then tested ALDH activity in cells transfected with TROY. Results showed TROY overexpression dramatically enhanced ALDH activity in HCC cells (Fig. 2E), indicating the cancer stemness maintenance function of TROY in HCC.
To further investigate the role of TROY in vivo, we established a xenograft mice model via subcutaneous injection of empty vector- and TROY-transfected HCC cells into the left and right dorsal flanks of nude mice, respectively. TROY-transfected cells were found to overtly increase xenograft tumor growth (Fig. 2F, G, Supplementary Fig. 4D-F).
Silencing of TROY Impairs cancer stemness in HCC cells
To confirm whether TROY is required for the cancer stemness maintenance of HCC cells, we silenced TROY expression with two short hairpin RNAs (shRNA) against TROY in HCC cell lines Huh7 and HepG2. The silencing effect was detected by qPCR (Supplementary Fig. 5A), immunofluorescence (Supplementary Fig. 51B), and western blot analysis (Fig. 3A). The results found that TROY silencing could decrease the expressions of NANOG, OCT4, and SOX2 by qPCR (Supplementary Fig. 5C), immunofluorescence (Supplementary Fig. 5D), and western blot analysis (Fig. 3B). Next, functional assays were performed in TROY-silenced HCC cells, and results showed knockdown of TROY dramatically suppressed the abilities of spheroid formation (Fig. 3C), foci formation (Supplementary Fig. 6A), and cell proliferation (Supplementary Fig. 6B). Chemo-drug sensitivity was investigated by XTT and flow cytometry assays. Results demonstrated that TROY silencing impaired the cell cycle process (Supplementary Fig. 6C, 6D), the cell viability (Fig. 3D, Supplementary Fig. 6E) and increased the apoptotic index (Fig. 3E) of HCC cells in the presence of sorafenib, cisplatin, and 5-Fu. Furthermore, the in vivo tumor growth assay showed that knockdown of TROY significantly decreased the tumor growth (Fig. 3F, G), tumor-initiating capacity, and liver CSCs ratio (Fig. 3H).
TROY promotes cell migration and metastasis by inducing EMT
As a high expression of TROY was closely associated with metastatic status in HCC patients (Fig. 1H), the effects of TROY on cell motility and metastasis were studied by both in vitro and in vivo assays. Cell migration and invasion assays showed that overexpression of TROY significantly enhanced HCC cell motility (Fig. 4A, Supplementary 7A). Conversely, the migrative and invasive abilities of HCC cells were impaired when TROY was silenced by shRNAs (Fig. 4B, Supplementary 7B). For in vivo lung metastatic mouse model, 10 weeks after tail vein injection of HCC cells, only a few nodules were observed in 1/4 of mice induced by shTROY-transfected Huh7 cells, whereas metastatic nodules on lung surfaces were detected in 4/4 of tested mice injected with shNTC-transfected cells (Fig. 4C). In addition, 4/4 of mice injected with TROY-overexpressing PLC8024 cells formed multiple metastatic nodules, however, only 2/4 of mice were found small metastatic nodules on the lung surfaces. H&E staining was used to further confirm the lung metastasis lesions (Fig. 4D). Taken together, these findings strongly suggested that TROY could promote HCC metastasis.
To investigate whether this metastatic ability was caused by epithelial-mesenchymal transition (EMT), immunofluorescence staining (Fig. 4E), and western blot (Fig. 4F) analysis was performed to investigate the expression pattern of representative markers. The results revealed that TROY could downregulate epithelial marker E-cadherin and increase mesenchymal markers N-cadherin and Vimentin, indicating that TROY could promote EMT in HCC cells.
TROY activates PI3K/AKT/TBX3 axis and upregulates TBX3
To characterize the underlying molecular mechanism of TROY in stemness regulation, we analyzed the 823 up-regulated genes in TROYhi HCC patients. The KEGG pathway result revealed that these genes were enriched in “Pathways in cancer”, “PI3K-AKT signaling” and “Signaling pathways regulating pluripotency of stem cells” (Fig. 5A, Supplementary Table 4). Among these signaling, PI3K/AKT signaling was involved in all three pathways. Furthermore, PI3K/AKT/TBX3 axis which was included in “Signaling pathways regulating pluripotency of stem cells” plays an important role in maintaining pluripotency of mouse ES cells and it has been reported as a treatment target in multi-type of embryonal cancers (Fig. 5B). Gene co-expression analysis using 366 pairs of HCC samples found that the expression of TROY was positively correlated with TBX3 expression (Fig. 5C), suggesting that TROY may be involved in stemness regulation via the control of PI3K/AKT/TBX3 signaling. To validate the result, western blotting (Fig. 5D) and qPCR (Supplementary Fig. 7C) were performed to examine the expression of AKT/TBX3 signaling. As expected, TBX3 expression and phosphorylation of AKT were induced in TROY-overexpressing HCC cells, while decreased upon TROY silencing. Evidence had already demonstrated that TBX3 regulated stem cell maintenance via controlling stem cell self-renewal and differentiation [24]. It has no known function in adult tissues but is frequently overexpressed in a wide range of epithelial and mesenchymal derived cancers [25]. Interestingly, we found that TBX3 protein is upregulated and translocated from cytoplasm into the nucleus upon TROY overexpression (Fig. 5E), demonstrating that TROY’s overexpression activated the transcriptional activator function of TBX3 in the nucleus. Collectively, these results demonstrated that TROY activates PI3K/AKT signaling and upregulates TBX3 expression in HCC. To further validate the role of TBX3 in TROY-activated PI3K/AKT pathway, we knockout of TBX3 in TROY overexpression HCC cell lines. Western blotting results revealed that expressions of NANOG, SOX2, and OCT4 were dramatically downregulated upon TBX3 knockout in TROY overexpression HCC cell lines (Supplementary Fig. 7D). Functional studies showed that drug resistance (Supplementary Fig. 7E) and spheroid formation ability (Supplementary Fig. 7F) were decreased upon TBX3 knockout in TROY overexpression HCC cells.
TROY interacts with P85α and induces its polyubiquitylation
Next, we investigated the molecular mechanism of how TROY activates the PI3K/AKT/TBX3 signaling. By examining the interactive proteins of TROY via the online protein–protein interaction database [26], we found that Phosphoinositide-3-Kinase Regulatory Subunit 1 (PIK3R1, also named p85α), a predominant regulatory subunit of PI3K has been reported to be able to bind TROY (Supplementary Fig. 8A). Immunoprecipitation analysis was then conducted to confirm the interaction between TROY and p85α, and the result showed that p85α was bound with TROY in both HepG2 and Huh7 cells (Fig. 5F). Immunofluorescent staining further confirmed the colocalization of TROY and p85α in HCC cells (Fig. 5G). An important mechanism of the regulation of signaling pathways was posttranscriptional modifications such as phosphorylation or ubiquitylation. We then asked whether the interaction of TROY with p85α was able to induce p85α phosphorylation or ubiquitylation. GSEA analysis result showed that the GO term “K63 linked polyubiquitin modification dependent protein binding” was skewed toward the TROYhi group (Supplementary Fig. 8B). The western blotting results demonstrated that the Lys63-linked polyubiquitylation was highly upregulated in TROY-overexpressing PLC8024 cells, accompanied by p85α protein degradation (Fig. 5H). Knockdown of TROY could effectively inhibit the degradation rate of p85α in cycloheximide (CHX)-treated Huh7 cells, and markedly decrease the half-life of p85α degradation from 3 to 6 h (Fig. 5I). Consistently, western blot results showed the expression of p85α was evidently decreased upon TROY overexpression in HepG2 and PLC8024 cells, while increased when silencing of TROY in Huh7 and Hep3B cells (Fig. 5J), suggesting that TROY plays an important role in p85α degradation via polyubiquitylation.
Loss of P85α is required for PI3K-AKT Activation and TROY-mediated cancer stemness maintenance
Given that p85α has been well characterized to negatively regulate PI3K pathway [27], we explored whether loss of p85α could lead to PI3K-AKT activation. CRISPR-Cas9 was applied to knock out p85α expression in TROY-silenced HCC cells. Western blotting results revealed that expressions of p-AKT and TBX3 were re-activated upon p85α knockout (Fig. 6A), indicating p85α acting upstream of PI3K. Functional studies including ALDH, apoptosis, and sphere formation assays were performed to evaluate the role of p85α in TROY-mediated cancer stemness maintenance. The results showed that the ALDH activity (Fig. 6B) and drug resistance (Fig. 6C) were regained upon p85α knockout in TROY-silenced HCC cells.
To further explore the role of PI3K-AKT signaling in TROY-mediated cancer stemness maintenance, PI3K inhibitor wortmannin was applied in TROY-expressing HCC cells. The result showed that wortmannin treatment abrogates TROY-induced ALDH activity (Fig. 6D) and spheroid formation (Fig. 6E, Supplementary Fig. 8C).
In the clinic, sorafenib, an FDA-approved tyrosine kinase inhibitor for the first-line therapy of advanced HCC patients, significantly against protein kinases including VEGFR, PDGFR, and RAF kinases [28]. Increasing evidence has demonstrated that drug resistance of sorafenib can be acquired by cancer cells by activating signaling pathways including PI3K/AKT signaling [29]; nevertheless, the detailed mechanism for the activation is not fully understood. Here, we treated TROY-expressing HCC cells with both sorafenib and wortmannin. Interestingly, the apoptosis experiment result showed that the combination of wortmannin and sorafenib can enhance the inhibitory of HCC cell lines (Fig. 6F). Furthermore, mice with tumors induced by TROY-expressed HCC cells were given either vehicle control, wortmannin, sorafenib, or a combination of both for 16 days. Compared with the vehicle control group, wortmannin, sorafenib and combined treatment groups all demonstrated reduced tumor size and weight, and the combined treatment group showed the maximal suppression of tumors (Fig. 6G). The bodyweight of mice was measured as an indication of drug toxicity, and no significant difference among the four treatment groups was observed (Supplementary Fig. 8D).
CAF-derived TGF-β1 upregulates TROY and activates PI3K/AKT/TBX3 signaling
To explore the mechanism of TROY upregulation in HCC, we used EPIC [30] (http://epic.gfellerlab.org) to predict the tumor microenvironment in TROYhi and TROYlo patient groups. Interestingly, the number of cancer associate fibroblasts (CAFs) was found significantly higher in TROYhi patient samples (Fig. 7A and B). GSEA results showed that high expression of TROY is significantly associated with the established gene sets “TGF-β receptor binding” (Fig. 7C). TGF-β1, mainly secreted by CAF, has been well characterized as a messenger between tumors and fibroblasts and played a significant role in tumor migration and stemnesss [31, 32]. IHC staining confirmed that TGF-β was secreted by CAFs rather than other cells in HCC specimens (Supplementary Fig. 8E). Hence, we hypothesized that CAF-derived TGF-β1 upregulated TROY expression in tumor cells. Disitertide, a TGF-β1 inhibitor, reduced the spheroid formation ability (Fig. 7D, Supplementary Fig. 8F) and ALDH expression (Fig. 7E) in HCC cell lines co-cultured with CAF conditional medium (CM). Consequently, depletion of TGF-β1 abolished the stemness induced by CAF conditional medium. Next, functional assays were conducted to assess the effect of TGF-β1 on cancer stemness maintenance. Indeed, the results showed that TGF-β1 treatment promoted both primary and secondary spheroid formation ability (Fig. 7F), drug resistance ability (Fig. 7G, Supplementary Fig. 8G), and the ALDH activity (Fig. 7H) in HCC cells. Western Blot results showed that TGF-β1 treatment could effectively upregulate TROY expression and activate PI3K/AKT/TBX3 signaling in HCC cells (Fig. 7I, Supplementary Fig. 8H). Collectively, our findings suggested that CAF-derived TGF-β1 upregulates TROY and activates PI3K/AKT/TBX3 signaling in HCC cell lines.