Targeting TROY-mediated P85a/AKT/TBX3 signaling attenuates tumor stemness and elevates treatment response in hepatocellular carcinoma

Background Previous in vitro hepatocyte differentiation model showed that TROY was specifically expressed in liver progenitor cells and a small proportion of hepatocellular carcinoma cells, suggesting that TROY may participate in hepatocellular carcinoma (HCC) stemness regulation. Here, we aim to investigate the role and mechanism of TROY in HCC pathogenesis. Method Bioinformatics analysis of the TCGA dataset has been used to identify the function and mechanism of TROY. Spheroid, apoptosis, and ALDH assay were performed to evaluate the stemness functions. Validation of the downstream pathway was based on Western blot, co-immunoprecipitation, and double immunofluorescence. Results HCC tissue microarray study found that a high frequency of TROY-positive cells was detected in 53/130 (40.8%) of HCC cases, which was significantly associated with poor prognosis and tumor metastasis. Functional studies revealed that TROY could promote self-renewal, drug resistance, tumorigenicity, and metastasis of HCC cells. Mechanism study found that TROY could interact with PI3K subunit p85α, inducing its polyubiquitylation and degradation. The degradation of p85α subsequently activate PI3K/AKT/TBX3 signaling and upregulated pluripotent genes expression including SOX2, NANOG, and OCT4, and promoted EMT in HCC cells. Interestingly, immune cell infiltration analysis found that upregulation of TROY in HCC tissues was induced by TGF-β1 secreted from CAFs. PI3K inhibitor wortmannin could effectively impair tumor stemness to sorafenib. Conclusion We demonstrated that TROY is an HCC CSC marker and plays an important role in HCC stemness regulation. Targeting TROY-positive CSCs with PI3K inhibitor wortmannin combined with chemo- or targeted drugs might be a novel therapeutic strategy for HCC patients. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1186/s13046-022-02401-6.


Background
Hepatocellular carcinoma (HCC) is the sixth malignant cancer and the fourth cause of cancer death worldwide in 2018 [1]. Lack of suitable biomarkers for early diagnosis and limited treatment strategies for cancer recurrence, metastasis, and chemoresistance are the main reasons for high mortality in liver cancer. Cancer stem cells (CSCs) are a minor subpopulation of cancer cells sharing similar characteristics with tissue progenitor cells such as capabilities of self-renewal and multi-lineage differentiation, which play important roles in cancer recurrence and metastasis [2][3][4]. Signaling pathways responsible for normal stem cells development were reported to be hijacked by tumor cells, such as c-MYC, TGF-β, Notch, Hippo, Wnt, and Hedgehog signaling [5]. Clinically, poor-differentiated HCCs with high recurrence rates always expressed markers of liver progenitors including alphafetoprotein (AFP), cytokeratin 19 (CK19), cytokeratin 7 (CK7), and SRY-Box Transcription Factor 9 (SOX9) [6][7][8][9]. All these findings suggest that understanding the regulatory network of liver progenitor cells is important to identify CSCs targets in HCC.
We previously established an in vitro hepatocyte differentiation model, of which human embryonic stem cells (ES) were induced to be differentiated into definitive endoderm (DE), liver progenitor (LP) cells, and premature hepatocytes (PH) [10]. By deep RNAsequencing, we identified a previously undescribed gene expression pattern of liver progenitor cells, which might be important to HCC progression. Bioinformatic-aided network analysis has identified several novels oncofetal drivers of HCC, of which, a liver progenitor cell-specific gene, TNF Receptor Superfamily Member 19 (TNFRSF19), also known as TROY, was investigated in this study. TROY belongs to the TNF receptor superfamily, which normally transduces signals upon ligand interaction via specific adaptor proteins bound to their intracellular domain [11]. However, TROY is different because no such ligands have been identified to interact with TROY, and its intracellular adaptor proteins remain to be elusive [12,13]. Functionally, TROY was reported to be a stem cell marker in kidney [14], brain [11], and gastric units [15]. High expression of TROY is associated with poor prognosis in colorectal cancer [16], glioblastoma [17], and nasopharyngeal carcinomas [18]. However, the precise role and underlying mechanism of TROY in liver CSCs have not been explored.

Conclusion:
We demonstrated that TROY is an HCC CSC marker and plays an important role in HCC stemness regulation. Targeting TROY-positive CSCs with PI3K inhibitor wortmannin combined with chemo-or targeted drugs might be a novel therapeutic strategy for HCC patients. Keywords: Oncofetal properties, Pluripotency signaling pathway, K63 polyubiquitin modification, Cancer-associated fibroblast

Graphical abstract
In this study, we characterized TROY as a novel CSC marker of HCC. Overexpression of TROY could increase HCC stemness both in vitro and in vivo. Mechanically, TROY was found to activate PI3K/AKT/TBX3 pathway via direct protein-protein interaction with PI3K subunit p85α. PI3K/AKT/TBX3 axis plays an important role in maintaining pluripotency of mouse ES cells [19] and it has been reported as a treatment target in multi type of embryonal cancers [20]. As an essential protein for pluripotency maintenance, TBX3 transcriptionally upregulates expressions of core pluripotent genes SOX2, NANOG, OCT4 and it can promote cell metastasis via activating the EMT pathway in HCC cells. The roles of TROY in liver CSCs might provide potential therapeutic targets for HCC treatment.

HCC samples and cell lines
HCC tissue microarray (TMA) containing 148 pairs of primary HCC specimens and their corresponding nontumor tissues were constructed as described previously [21]. All samples were collected from Sun Yat-sen University Cancer Center (Guangzhou, China) between 2003 and 2010. Human HCC cell lines Huh7, HepG2, SNU398, SNU182, SNU878, and human embryonic kidney cell line 293FT were purchased from American Type Culture Collection (ATCC, Manassas, Virginia, USA). Immortalized human hepatocyte cell line LO2 was obtained from the Institute of Virology. Cell lines Huh7, HepG2, LO2, and 293FT were cultured in DMEM (high glucose) supplemented with 10% fetal bovine serum (FBS). SNU398, SNU182, SNU878 were maintained in RPMI-1640 medium with 10% FBS. HepG2 was cultured in MEM containing 10% FBS. All cell lines were incubated at 37 °C and 5% CO 2 .

Construction of TROY Overexpression and knockdown cells
Briefly, the full-length TROY isoform1 (NM_018647.4) and isoform2 (NM_148957.3) were cloned into the pLenti6 expression vector (Invitrogen) separately. Lentiviral shRNA clones for TROY were obtained from GeneCopoeia (HSH106281-LVRU6GP for TROY and CSHCTR001-LVRU6GP for a scramble control) targeted both isoform 1 and 2. Stable knockout p85α HCC cell line by lentiviral transduction of p85α single guide RNA (Supplementary Table 4) lentiCRISPR v2. For lentivirus production, HEK293T packaging cells were transfected with lentiviral vectors using Lipofectamine 2000 Reagent (Invitrogen). After 72 h of incubation, the viral supernatant was collected and filtered. HCC cells were incubated overnight with the viral supernatant and supplemented with 10 μg/ml polybrene. Stable clones were selected using puromycin.

QPCR and western blotting
Total RNA was extracted via TRIZOL Reagent (Invitrogen), and cDNA was synthesized by a reverse transcription PCR kit (TAKARA). Quantitative realtime PCR (qPCR) was performed using the TB Green Premix Ex Taq II (TAKARA) and a LightCycler ® 480 System (ROCHE). The mRNA expression of TROY, NANOG, SOX2, OCT4, TBX3, and β-Actin was assessed by qPCR with the primers (Supplementary Table 5). Total RNA was extracted using TRIZOL. The cDNA was synthesized from 0.5 μg of total RNA by PrimeScript RT Master Mix (Takara Bio). The protein expression of genes in this research was assessed by western blotting (Supplementary Table 5).

In vitro cell proliferation, drug resistance and apoptosis assays
In vitro cell proliferation rate was assessed by monitoring the cell number for 5-7 days in cell culture. The cells in each condition were seeded at 1,000 cells/well for proliferation assay or 8,000 cells/well for drug resistance assay in 96-well in triplicate and incubated in a 37 °C humidified CO 2 incubator with the drug or vehicle containing medium refreshed every other day. The number of cells was determined by XTT kit (Roche Diagnostics, Indianapolis, IN). For the apoptosis assay, cells were seeded at 5 × 10 5 cells/well into a six-well plate in triplicate. After 24 h, cells were subjected to the drug or vehicle containing medium for 24-48 h and then followed by flow cytometry analysis. The cell apoptosis assay was

ALDH measurement
To determine the intracellular aldehyde dehydrogenase (ALDH) concentration, an equal number of HCC cells were seeded onto six-well dishes. The attached cells were trypsinized and washed with PBS. Cells were stained with AldeRed ™ ALDH Detection Assay (Sigma-Aldrich SCR150) for analyzing the ALDH level. Stained cells were analyzed using the FACSCanto II Analyzer (BD Biosciences). The results were performed with FlowJo software.

In vitro spheroid formation, migration, and invasion assay
Sphere-formation assay is an in vitro method commonly used to identify CSCs and study their self-renewal properties. For spheroid formation, a total number of 3,000 cells were seeded in Corning ® Costar ® Ultra-Low Attachment Multiple 6-Well Plate (Merck, CLS3471) with DMEM/F12 medium (Life Technologies), supplemented with 20 ng/mL human recombinant EGF (Life Technologies), 10 ng/mL human recombinant FGF (Life Technologies), B27 (1: 50, Invitrogen), and 4ug/mL insulin (BIOIND, Kibbutz Beit Haemek, Israel) in triplicate and incubated in a 37 °C humidified CO2 incubator with 30 uL medium refreshed every second day. For serial passage of spheroids, the primary spheroids were centrifuged and dissociated to single cells using TrypLE Expression (Invitrogen) and neutralized the reaction using trypsin inhibitor (Invitrogen). Migration and invasion assays were determined to evaluate cell motility (BD Biosciences, San Jose, CA). Approximately 8 × 10 4 cells were seeded in the medium without FBS on transwell upper chambers, and the lower chamber was supplied with a medium of complete 10% FBS. After 24-48 h, the lower membrane surface was stained using crystal violet to count the cell number under a microscope.

In vivo functional assay
For the subcutaneous injection model, 5 × 10 6 TROYtransfected cells, shTROY-transfected cells, and their corresponding control cells were injected subcutaneously into different sides of 4-week-old BABL/c nude mice or NOD/SCID mice. The tumor size was measured every 5-7 days with a caliper and calculated with the formula 0.5 × length × width × width. Mice were sacrificed 4-6 weeks. For the metastasis mice model, 1 × 10 6 TROY-transfected cells, shTROY-transfected cells, and their corresponding control cells were injected through the tail vein of BABL/c nude mice 6-week-old. Mice were sacrificed around 70 days. For drug treatment, 2 weeks after tumor implantation, mice were administered with (1) vehicle with DMSO, (2) 20 mg/kg sorafenib every day via oral gavage, (3) 1 mg/kg wortmannin every day via oral gavage, (4) combine treatment. After 5 weeks, mice were sacrificed.

Isolation of cancer-associated fibroblasts
Briefly, freshly collected HCC tumour tissue were cut into as small pieces as possible in sterile phosphatebuffered saline (PBS) solution, followed by collagenase digestion. The suspension was filtered through 20 mm stainless steel wire mesh to collect a single cell suspension. The filtrate was centrifuged and washed before being finally plated on 6 cm tissue culture dishes in 5 ml DMEM medium supplemented with 20% fetal bovine serum (FBS). After culturing for 30 min at 37℃, nonadherent cells (mainly tumor cells) were removed to obtain pure fibroblasts. The adherent fibroblasts were subcultured for further study.

Conditioned medium
Cancer-associated fibroblasts were cultured in DMEM medium with 10% FBS until 70% confluence. After complete removal of the normal culture medium, CAF cells were continually cultured in DMEM medium with 3% FBS for 24 h before medium collection. The culture medium was then centrifuged at 1000 g for 30 min and the supernatant was collected as the conditioned medium for further study.

Statistical analysis
Statistical analysis was performed using GraphPad Prism V.8 software. Independent student's t-test was used to compare parametric continuous variables. Pearson's χ2 test was performed to compare the correlation between genes and clinicopathological parameters. Kaplan-Meier plots and log-rank tests were used to analyze HCC patient survival. For limiting dilution assay, the tumorinitiating frequency and statistics were calculated with ELDA software. The values represent the mean ± SD of three independent experiments. The P-values were marked as *P < 0.05, **P < 0.001, and ***P < 0.0001 in all figures.

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 nontumor 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 lowfrequency 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 (TROY hi ) and lowest (TROY lo ) 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 (log 2 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 TROY hi patients and TROY lo patients in unsupervised clustering analysis. TROY hi patients exhibit increased expression of various stemness-related genes (Fig. 1J). Collectively, these findings strongly suggest that TROY might be involved in regulating stemnessrelated signaling in HCC.  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.

TROY enhances stemness propertises of HCC cells
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 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 TROYoverexpressing 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 upregulated genes in TROY hi 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

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 TROY hi 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 TROYsilenced 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. gfell erlab. org) to predict the tumor microenvironment in TROY hi and TROY lo patient groups. Interestingly, the number of cancer associate fibroblasts (CAFs) was found significantly higher in TROY hi 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.

Discussion
The existence of CSCs is one of the major contributors of malignant phenotypes of HCC such as high tumor recurrence, metastasis, and resistance to conventional chemotherapy or radiotherapy. Therefore, the development of therapeutic strategies targeting CSCs will be promising in HCC treatment. Emerging evidence suggests that CSCs share similar phenotypes and regulatory networks with normal tissue progenitor cells. In addition, poor-differentiated HCCs always demonstrated lineage-reversed phenotype with unsatisfactory prognosis [33]. To explore key signaling involved in the CSC regulation in HCC, we established an in vitro hepatocyte differentiation model, which was induced from ES cells to mimic liver development and focused on genes specifically expressed in liver progenitors. Among them, TROY was specifically expressed in the liver progenitor stage, shut downed in mature hepatocytes, and re-activated in HCC. TROY belongs to the TNF receptor superfamily consisting of 19 TNFRSF ligands and 29 different TNFRSF receptor [12]. Interactions between the TNFRSF ligands and TNFRSF receptors activate downstream signaling and regulate multiple biological processes such as cell survival, proliferation, and differentiation [34]. However, the ligand of TROY has not been found yet. Previous studies found that TROY does not bind to any known TNF ligands, and its intracellular domain is distinct from other characterized members of the TNF receptor superfamily. In the present study, we firstly reported that TROY could interact with p85α and play an important role in its k63 ubiquitination and degradation. The molecular mechanism of how TROY induced k63 ubiquitination was not explored in the present work. However, we found that TROY could recruit TRAF family members (TRAF1, TRAF2, TRAF3, and TRAF5) and E3 ubiquitin-protein ligases (ITCH, NEDD4, NEDD4L, WWP1, and FBXW7) through direct protein-protein interaction (Supplementary Fig. 5C), suggesting its role in regulating k63 ubiquitination of p85α.
Previous studies reported that TROY is a marker of normal tissue stem cells in adult kidney [14], brain [11], and stomach [15]. In mouse intestinal tissue, TROY was found to interact with similar signaling pathways activated in ESs or cell lineage differentiation [13]. In the context of cancers, TROY was found to promote tumor progression via multiple signaling pathways. For example, TROY exerted its oncogenic function in glioma through activating nuclear factor kappa B (NF-κB) signaling [17] and increased tumorigenesis in nasopharyngeal carcinoma by suppressing TGFβ signaling [18]. In the present study, we found that TROY-expressing cells could not be detected in normal liver tissues, while a small number of TROY-expressing cells could be observed in HCC cells, which is consistent with the characteristic of CSCs. Interestingly, the high-frequency of TROY-expressing cells was negatively associated with poor HCC outcome (P = 0.02) as well as tumor metastasis.
To explore the oncogenic role of TROY in HCC development and progression, both in vitro and in vivo functional assays were applied in TROY-expressing cells and TROYsilencing cells. The results found that TROY had a strong oncogenic ability by promoting cell growth in vitro and tumor formation in vivo. The effect of TROY on stemness regulation was also addressed in the present study, and experimental data suggested that TROY played important role in stemness regulation. Overexpression of TROY could increase the abilities of sphere formation and chemoresistance, which could be effectively abrogated in TROYsilencing cells. In addition, overexpression of TROY could upregulate stemness-related genes such as SOX2 and OCT4 in HCC cells. Taken together, these findings revealed that TROY plays key roles in HCC progression including metastasis through upregulating HCC stemness.
Next, we characterized the molecular mechanism of TROY in stemness regulation in HCC as a liver CSC marker and found TROY could promote the stemness by activating p85a/AKT/TBX3 pathway. Importantly, p85α acted as a tumor brake in different types of cancers, and the mutations in the i-SH2 region of p85α have been frequently detected in tumor tissues, blocking p110α activity and thus activating PI3K/AKT signaling [35]. Interestingly, the deletion of p85α in the mouse liver could lead to gradual morphological changes of hepatocytes and result in the development of HCC [36]. These observations indicated the essential role of p85α in suppressing cancer progression. In the present study, the IHC staining demonstrated a decreased expression of the p85α in HCC clinical samples compared with para-tumor tissues. The TROY overexpression in HCC cells could downregulate p85 expression and conversely, the silencing of TROY upregulated p85α expression, indicating that the TROY-mediated p85α downregulation was essential for HCC development and progression.
As a subunit of PI3K, the loss of p85α function contributed to the activation of downstream PI3K/AKT/TBX3 signaling in CSCs. Targeting the activated TROY/ p85α/ TBX3 axis using the small-molecule PI3K inhibitor wortmannin impairs the tumor stemness properties of HCC cells. Notably, the application of wortmannin increased the sensitivity of HCC cells to sorafenib in both in vitro and in vivo experiments, suggesting that wortmannin might be used as a novel therapeutic strategy to augment the efficacy of sorafenib by targeting liver CSCs.
The crosstalk between tumor cells and stromal cells in the tumor microenvironment has been widely studied and played important roles in tumor progression [37]. Among all the stromal cells that constitute the tumor microenvironment, CAFs were regarded as the most abundant populations that have a great impact on tumor behavior [38]. CAFs exerted their influence on tumor or other stromal cells via direct cell-cell contact or releasing regulatory factors that remodel the microenvironment, such as TGFβ, IL-6, and PDGFs [39]. Importantly, TGFβ is a well-characterized immunosuppressive cytokine mainly secreted by CAFs that is responsible for transforming innate or adaptive immune cells into immunosuppressive cells [40]. In the present study, we found that CAFs secreted TGFβ could increase TROY expression of cancer cells and activate the TROY-mediated PI3K/AKT/TBX3 signaling pathway, promoting stemness properties of HCCs. These findings uncovered the undescribed role of TGFβ in enhancing liver CSCs features and highlighted the importance of targeting TGFβ-induced TROY upregulation and subsequent PI3K/AKT/TBX3 activation.

Conclusions
Collectively, we defined the role of TROY as a liver CSCs marker that regulates the self-renewal, drug resistance, tumorigenicity, and metastasis of HCC cells. The expression of TROY could be upregulated in HCC tissues by CAFs-derived TGF-β1. Furthermore, we clarified that the interaction between TROY and p85α could lead to p85α degradation and subsequently activate PI3K/AKT/TBX3 signaling, which helps to maintain the pluripotency of liver CSCs by upregulating the expression of SOX2, NANOG, and OCT4 and promote cell motility via activating EMT pathway in HCC. Targeting liver CSCs with PI3K inhibitor wortmannin combined with sorafenib might be a novel therapeutic strategy for HCC patients.