The Transcription Factor SALL4 Regulates Stemness of EpCAM-positive Hepatocellular Carcinoma

Background & Aims :Recent evidence suggests that hepatocellular carcinoma can be classiﬁed into certain molecular subtypes with distinct prognoses based on the stem/maturational status of the tumor. We investigated the transcription program deregulated in hepatocellular carcinomas with stem cell features. Methods :Gene and protein expression proﬁles were obtained from 238 (analyzed by microarray), 144 (analyzed by immunohistochemistry), and 61 (analyzed by qRT-PCR) hepatocellular carcinoma cases. Activation/suppression of an identiﬁed transcription factor was used to evaluate its role in cell lines. The relationship of the transcription factor and prognosis was statistically examined. Results :The transcription factor SALL4, known to regulate stemness in embryonic and hematopoietic stem cells, was found to be activated in a hepatocellular carcinoma subtype with stem cell features. SALL4-positive hepatocellular carcinoma patients were associated with high values of serum alpha fetoprotein, high frequency of hepatitis B virus infection, and poor prognosis after surgery compared with SALL4-negative patients. Activation of SALL4 enhanced spheroid formation and invasion capacities, key characteristics of cancer stem cells, and up-regulated the hepatic stem cell markers KRT19 , EPCAM , and CD44 in cell lines. Knockdown of SALL4 resulted in the down-regulation of these stem cell markers, together with attenuation of the invasion capacity. The SALL4 expression status was associated with histone deacetylase activity in cell lines, and the histone deacetylase inhibitor successfully suppressed proliferation of SALL4-positive hepatocellular carcinoma cells. Conclusions :SALL4 is a valuable biomarker and therapeutic target for the diagnosis and treatment of hepatocellular carcinoma with stem cell features.


Introduction
Cancer is a heterogeneous disease in terms of morphology and clinical behavior. This heterogeneity has traditionally been explained by the clonal evolution of cancer cells and the accumulation of serial stochastic genetic/epigenetic changes [1]. The alteration of the microenvironment by tumor stromal cells is also considered to contribute to the development of the heterogeneous nature of the tumor through the activation of various signaling pathways in cancer cells, including epithelial mesenchymal transition programs [2].
Recent evidence suggests that a subset of tumor cells with stem cell features, known as cancer stem cells (CSCs), are capable of self-renewal and can give rise to relatively differentiated cells, thereby forming heterogeneous tumor cell populations [3]. CSCs were also found to generate tumors more efficiently in immunodeficient mice than non-cancer stem cells in various solid tumors as well as hematological malignancies [4]. CSCs are also more metastatic and chemo/radiation-resistant than non-CSCs and are therefore considered to be a pivotal target for tumor eradication [5,6].
Hepatocellular carcinoma (HCC) is a leading cause of cancer death worldwide [7]. Recently, we proposed a novel HCC classification system based on the expression status of the hepatic stem/ progenitor markers epithelial cell adhesion molecule (EpCAM) and alpha fetoprotein (AFP) [8]. EpCAM-positive ( + ) AFP + HCC (hepatic stem cell-like HCC; HpSC-HCC) is characterized by an onset of disease at younger ages, activation of Wnt/b-catenin signaling, a high frequency of portal vein invasion and poor prognosis after radical resection, compared with EpCAM À AFP À HCC (mature hepatocyte-like HCC; MH-HCC) [9]. EPCAM is a target gene of Wnt/b-catenin signaling, and EpCAM + HCC cells isolated from primary HCC and cell lines show CSC features including tumorigenicity, invasiveness, and resistance to fluorouracil [9,10]. Thus, EpCAM appears to be a potentially useful marker for the isolation of liver CSCs in HpSC-HCC. However, key transcriptional programs responsible for the maintenance of EpCAM + CSCs are still unclear.
In this study, we aimed to clarify the transcriptional programs deregulated in HpSC-HCC using a gene expression profiling approach. We found that the SALL4 gene encoding Sal-like 4 (Drosophila) (SALL4), a zinc finger transcriptional activator and vertebrate orthologue of the Drosophila gene spalt (sal) [11], was upregulated in HpSC-HCC. In adults, SALL4 is known to be expressed in hematopoietic stem cells and their malignancies, but its role in HCC has not yet been fully elucidated [12][13][14]. We therefore investigated the role of SALL4 in the regulation and maintenance of EpCAM + HCC.

Clinical HCC specimens
A total of 144 HCC tissues and adjacent non-cancerous liver tissues were obtained from patients who underwent hepatectomy for HCC treatment from 2002 to 2010 at Kanazawa University Hospital, Kanazawa, Japan. These samples were formalinfixed and paraffin-embedded, and used for immunohistochemistry (IHC). A further 61 HCC samples were obtained from patients who underwent hepatectomy from 2008 to 2011; these were freshly snap-frozen in liquid nitrogen and used for RNA analysis. Of these 61 HCCs, 8 and 36 cases were defined as HpSC-HCC and MH-HCC, respectively, according to previously described criteria [8].
27 HCC cases were included in both the IHC cohort (n = 144) and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) cohort (n = 61), and SALL4 gene and protein expression were compared between these cases. An additional fresh HpSC-HCC sample was obtained from a surgically resected specimen and immediately used for preparation of a single-cell suspension. All experimental and tissue acquisition procedures were approved by the Ethics Committee and the Institutional Review Board of Kanazawa University Hospital. All patients provided written informed consent.

Microarray analysis
Detailed information on microarray analysis is available in the Supplementary Materials and methods.

Cell culture and reagents
Human liver cancer cell lines HuH1, HuH7, HLE, and HLF were obtained from the Japanese Collection of Research Bioresources (JCRB), and Hep3B and SK-Hep-1 were obtained from the American Type Culture Collection (ATCC). Single-cell suspensions of primary HCC tissue were prepared as described previously [15]. Detailed information is available in the Supplementary Materials and methods. The histone deacetylase (HDAC) inhibitor suberic bis-hydroxamic acid (SBHA) and suberoylanilide hydroxamic acid (SAHA) were obtained from Cayman Chemical (Ann Arbor, MI). Plasmid constructs pCMV6-SALL4 (encoding SALL4A), pCMV6-SALL4-GFP, and 29mer shRNA constructs against human SALL4 (No. 7412) were obtained from OriGene Technologies, Inc. (Rockville, MD). These constructs were transfected using Lipofectamine 2000 (Life Technologies, Carlsbad, CA) according to the manufacturer's protocol.

Western blotting
Whole cell lysates were prepared using RIPA lysis buffer. Nuclear and cytoplasmic proteins were extracted using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Pierce Biotechnology Inc., Rockford, IL). Mouse monoclonal antibody to human Sall4 clone 6E3 (Abnova, Walnut, CA), rabbit polyclonal antibodies to human Lamin B1 (Cell Signaling Technology Inc., Danvers, MA), and mouse monoclonal anti-b-actin antibody (Sigma-Aldrich, St. Louis, MO) were used. Immune complexes were visualized by enhanced chemiluminescence (Amersham Biosciences Corp., Piscataway, NJ) as described previously [15,16].

Quantitative reverse transcription-polymerase chain reaction (qRT-PCR)
Detailed information on qRT-PCR is available in the Supplementary Materials and methods.

IHC and immunofluorescence (IF) analyses
IHC was performed using an Envision+ kit (Dako, Carpinteria, CA) according to the manufacturer's instructions. Anti-SALL4 monoclonal antibody 6E3 (Abnova, Walnut, CA), anti-EpCAM monoclonal antibody VU-1D9 (Oncogene Research Products, San Diego, CA), and anti-CK19 monoclonal antibody RCK108 (Dako Japan, Tokyo, Japan) were used for detecting SALL4, EpCAM, and CK19, respectively. Anti-Sall4 rabbit polyclonal antibodies (ab29112) (Abnova) and vector red (Vector Laboratories Inc., Burlingame, CA) were used for double color IHC analysis. Samples with >5% positive staining in a given area were considered to be positive for a particular antibody. For IF analyses, Alexa 488 fluorescein isothiocyanate (FITC)-conjugated anti-mouse immunoglobulin G (IgG) (Life Technologies) was used as a secondary antibody.
Cell proliferation, spheroid formation, invasion, and HDAC activity assay Detailed information on this topic is available in the Supplementary Materials and methods.

Statistical analyses
Student's t tests were performed with GraphPad Prism software 5.0 (GraphPad Software, San Diego, CA) to compare various test groups assayed by cell proliferation assays and qRT-PCR analysis. Spearman's correlation analysis and Kaplan-Meier survival analysis were also performed with GraphPad Prism software 5.0 (GraphPad Software).

Activation of SALL4 in HpSC-HCC
To elucidate the transcriptional programs deregulated in HpSC-HCC, we performed class-comparison analyses and identified 793 genes showing significant differences in differential expression between HpSC-HCC (n = 60) and MH-HCC (n = 96) (p <0.001), as previously described [9]. Of them, 455 genes were specifically up-regulated in HpSC-HCC, and we performed transcription factor analysis using this gene set to identify their transcriptional regulators by MetaCore software. We identified four transcription factor genes, SALL4, NFYA, TP53, and SP1, that were potentially activated in HpSC-HCC (Fig. 1A). Involvement of TP53 and SP1 in the stemness of HCC has previously been described [17,18], but the roles of SALL4 and NFYA were unclear.
We investigated the interaction networks affected by SALL4 and NFYA using the MetaCore dataset. We showed that SALL4 might be a regulator of Akt signaling (SP1), Wnt signaling (TCF7L2), and epigenetic modification (JARID2, DMRT1, DNMT3B) [19], and could potentially regulate two other transcriptional regulators, SP1 and NFYA, through Akt and Myb signaling pathways (Fig. 1B). As a recent study indicated that SALL4 is a direct target of the Wnt signaling pathway [20], which is dominantly activated in HpSC-HCC [9], we focused on the expression of SALL4 in HpSC-HCC, and confirmed its up-regulation in HpSC-HCC compared with MH-HCC as evaluated by microarray data (Fig. 1C). We validated this using an independent HCC cohort evaluated by qRT-PCR (Fig. 1D). We further examined the expression of SALL4, EPCAM, and AFP using microarray data of 238 HCC cases (Fig. 1E) and qRT-PCR data of 61 HCC cases (Fig. 1F). For the tumor/non-tumor ratios, we identified a weak positive correlation between SALL4 and EPCAM (r = 0.31, p <0.0001) and between SALL4 and AFP (r = 0.31, p = 0.0003) in the microarray cohort. We further evaluated expression of these genes in HCC tissues by qRT-PCR, and we validated the strong positive correlation between SALL4 and EPCAM (r = 0.70, p <0.0001) and between SALL4 and AFP (r = 0.66, p <0.0001) in the independent cohort.
Next we performed IHC analysis of 144 HCC cases surgically resected at Kanazawa University Hospital. We first confirmed the nuclear accumulation of SALL4 stained by an anti-human SALL4 antibody ( Fig. 2A). We further confirmed the concordance of SALL4 protein expression evaluated by IHC, and SALL4 gene expression evaluated by qRT-PCR using the same samples (Fig. 2B). We detected the nuclear expression of SALL4 in 43 of 144 HCC cases (Table 1). After evaluating the clinicopathological characteristics of SALL4-positive and -negative HCC cases, we identified that SALL4-positive HCCs were associated with a significantly high frequency of hepatitis B virus (HBV) infection and significantly high serum AFP values. We further identified that SALL4-positive HCCs were associated with expression of the hepatic stem cell markers EpCAM and CK19. Co-expression of SALL4, EpCAM, and CK19 was confirmed by double color IHC analysis (Fig. 2C). Evaluation of the survival outcome of these surgically resected HCC cases by Kaplan-Meier survival analysis indicated that SALL4-positive HCCs were associated with significantly lower recurrence-free survival outcomes within one year compared with SALL4-negative HCCs (p = 0.0049) (Fig. 2D). Because SALL4 expression was positively correlated with EpCAM and AFP expression in primary HCC cases, we evaluated the expression of SALL4 in EpCAM + AFP + and EpCAM À AFP À HCC cell lines. Consistent with the primary HCC data, two of three EpCAM + AFP + HCC cell lines (Hep3B and HuH7) abundantly expressed SALL4, as shown by qRT-PCR (Fig. 2E) and Western blotting (Fig. 2F). We identified the expression of two isoforms of SALL4 proteins with molecular weights of 165 kDa (SALL4A) and 115 kDa (SALL4B), and SALL4B was found to be the dominant endogenous isoform in HCC cell lines. All EpCAM À AFP À HCC cell lines (SK-Hep-1, HLE, and HLF) and one EpCAM + AFP + cell line (HuH1) did not express SALL4. Nuclear accumulation of SALL4 in Hep3B and HuH7 cells was confirmed by IHC using subcutaneous tumors developed in xenotransplanted NOD/SCID mice (Fig. 2G). We further evaluated the expression of EPCAM and SALL4 using single cell suspensions derived from a surgically resected primary HCC. EpCAM + and EpCAM À cells were separated by magnetic beads, and we revealed a strong spheroid formation capacity of sorted EpCAM + cells compared with EpCAM À cells (Fig. 2H, left panel). Interestingly, when comparing the expression of SALL4 in these sorted cells, we identified a high expression of SALL4 in sorted EpCAM + cells compared with EpCAM À cells (Fig. 2H, right panel)

SALL4 regulates stemness of HpSC-HCC
To explore the role of SALL4 in HpSC-HCC, we evaluated the effect of its overexpression in HuH1 cells which showed little expression of SALL4 irrespective of EpCAM + and AFP + HpSC-HCC phenotype. We transfected plasmid constructs encoding SALL4 (pCMV6-SALL4) or control (pCMV7), and we similarly identified the expression of two isoforms by using this construct (Fig. 3A). Evaluation of the subcellular localization of GFP-tagged SALL4 (pCMV6-SALL4-GFP) showed that it could be detected in both the cytoplasm and nucleus (Fig. 3B). We observed strong up-regulation of the hepatic stem cell marker KRT19, modest up-regulation of EPCAM and CD44, and down-regulation of the mature hepatocyte marker ALB in HuH1 cells transfected with pCMV6-SALL4 compared with the control (Fig. 3C). Up-regulation of CK19 by SALL4 overexpression was also confirmed at the protein level by IF analysis (Fig. 3D). Phenotypically, SALL4 overexpression in HuH1 cells resulted in the significant activation of spheroid formation and invasion capacities with activation of SNAI1, which induces epithelial-mesenchymal transition, compared with the control (Fig. 3E and F, Supplementary Fig. 1A). We further investigated the effect of SALL4 knockdown in HuH7 cells, which intrinsically expressed high levels of SALL4. Expression of SALL4 was decreased to 50% in HuH7 cells transfected with SALL4 sh-RNA compared with the control when evaluated by qRT-PCR (Fig. 4A). However, the reduction of SALL4 protein was more evident when evaluated by Western blotting, suggesting that this sh-RNA construct might work at the translational as well as the transcriptional level (Fig. 4B). Knock down of SALL4 resulted in a compromised invasion capacity and spheroid formation capacity with decreased expression of EPCAM and CD44 in HuH7 cells ( Fig. 4C and D, Supplementary Fig. 1B and C).

SALL4 and HDAC activity in HpSC-HCC
The above data suggested that SALL4 is a good target and biomarker for the diagnosis and treatment of HpSC-HCCs. However, it is difficult to directly target SALL4 as no studies have investigated the inhibition of its transcription using chemical or other approaches [21]. We therefore re-investigated the interaction networks associated with SALL4, and found that SALL4 activation appeared to induce epigenetic modification (Fig. 1B). In particular, a recent study suggested that SALL4 forms a nucleosome remodeling and deacetylase (NuRD) complex with HDACs and potentially regulates HDAC activity [22]. We therefore confirmed that SALL4 knock down resulted in the reduced activity of total HDAC in HuH7 cells (Fig. 4E). We also evaluated the effect of the overexpression of SALL4 in HuH1 and HLE cells, which do not express SALL4 endogenously, and SALL4 overexpression was found to result in a modest increase of HDAC activity and mild enhancement of chemosensitivity to an HDAC inhibitor SBHA in both cell lines (Supplementary Fig. 2A and B). We further investigated HDAC activity in two SALL4-positive (Hep3B, HuH7) and two SALL4-negative (HLE, HLF) HCC cell lines. Interestingly, high HDAC activities were detected in SALL4-positive compared with SALL4-negative HCC cell lines (Fig. 4F). The HDAC inhibitor SBHA was found to inhibit proliferation of SALL4-positive HCC cell lines at a concentration of 10 lM. By contrast, SBHA had little effect on the proliferation of SALL4-negative HCC cell lines at this concentration (Fig. 4G). SBHA treatment suppressed the expression of SALL4 gene/protein expression in SALL4-positive HuH7 and Hep3B cell lines ( Supplementary Fig. 3A and B). We further investigated the effect of SAHA, an additional HDAC inhibitor, in these HCC cell lines, and SAHA was found to more efficiently suppress the cell proliferation of SALL4-positive cell lines compared with SALL4-negative cell lines (Supplementary Fig. 3C).
Taken together, our data suggest a pivotal role for the transcription factor SALL4 in regulating the stemness of HpSC-HCC. SALL4 was detected in HpSC-HCCs with poor prognosis, and inactivation of SALL4 resulted in a reduced invasion/spheroid formation capacity and decreased expression of hepatic stem cell markers. The HDAC inhibitors inhibited proliferation of SALL4positive HCC cell lines with a reduction of SALL4 gene/protein expression, suggesting their potential in the treatment of SALL4-positive HpSC-HCC.

Discussion
Stemness traits in cancer cells are currently of great interest because they may explain the clinical outcome of patients according to the malignant nature of their tumor. Recently, we

JOURNAL OF HEPATOLOGY
proposed an HCC classification system based on the stem/maturation status of the tumor by EpCAM and AFP expression status [8]. These HCC subtypes showed distinct gene expression patterns with features resembling particular stages of liver lineages. Among them, HpSC-HCC was characterized by a highly invasive nature, chemoresistance to fluorouracil, and poor prognosis after radical resection, warranting the development of a novel therapeutic approach against this HCC subtype [9]. In this study, we showed that the transcription factor SALL4 was activated in HpSC-HCC and that SALL4 might regulate HCC stemness, as characterized by the activation of EpCAM, CK19, and CD44 with highly tumorigenic and invasive natures. Furthermore, we identified that SALL4-positive HCC cell lines tended to show high HDAC activity and chemosensitivity to the HDAC inhibitors SBHA and SAHA. This study reveals for the first time the utility of SBHA for the treatment of HCC with stem cell features.
SALL4 is a zinc finger transcription factor originally cloned based on sequence homology to Drosophila sal [11]. SALL4 mutations are associated with the Okihiro syndrome, a human disease involving multiple organ defects [23,24]. SALL4 plays a fundamental role in the maintenance of embryonic stem cells, potentially through interaction with Oct4, Sox2, and Nanog [25][26][27][28][29][30]. Furthermore, knockdown of SALL4 significantly reduces the efficiency of induced pluripotent stem cell generation [31]. SALL4 is also expressed in hematopoietic stem cells and leukemia cells, where it regulates their maintenance [14,32]. SALL4 is known to encode two isoforms (SALL4A and SALL4B), and a recent study  suggested the important role of SALL4B on maintaining the stemness of embryonic stem cells [25]. Interestingly, our data indicated that SALL4B is also a dominant form in HpSC-HCC cell lines. It is unclear how SALL4 isoform expression is regulated in cancer, and future studies are required to explore the mechanisms of SALL4 isoform regulation.
In the liver, SALL4 is expressed in fetal hepatic stem/progenitors but not in adult hepatocytes, and a mouse study demonstrated that inhibition of SALL4 in hepatic stem/progenitors contributes to their differentiation [33]. Interestingly, recent studies indicated that AFP-producing gastric cancer expresses SALL4, suggesting that SALL4 might play a role in the hepatoid differentiation of gastric cancer [34]. Consistently, our data indicated a positive correlation between SALL4, AFP, and EPCAM expression in two independent HCC cohorts. Strikingly, SALL4 was recently shown to be expressed in a subset of human liver cancers with poor prognoses, while modification of SALL4 expression resulted in the alteration of cell proliferation in vitro and tumor growth in vivo, consistent with our current study [35]. A recent study reported the expression of SALL4 in 46% of HCC cases, which is almost comparable to our present study [36]. Furthermore, a very recent study of two independent large cohorts demonstrated that SALL4 is a marker for a progenitor subclass of HCC with an aggressive phenotype [37]. It is still unclear how SALL4 expression is regulated and which target genes are directly activated by SALL4 binding. Future studies using next generation sequencing are required to fully understand the mechanisms of SALL4 regulation of HCC stemness.
In this study, we demonstrated that SALL4-positive HCC cell lines have high HDAC activity and chemosensitivity against the HDAC inhibitors SBHA and SAHA compared with SALL4-negative HCC cell lines. SALL4 was recently found to directly connect with the epigenetic modulator NuRD complex [22], thereby possibly affecting the histone modification associated with stemness. The NuRD complex is a multiunit chromatin remodeling complex containing chromodomain-helicase-DNA-binding proteins and HDACs that regulate histone deacetylation [38]. Its role in cancer is still controversial, while its function in HCC has not yet been determined.
Our data suggest that SALL4 plays a role in controlling HDAC activity and contributing to the maintenance of HCC with stem cell features. Consistently, HDAC inhibitors might be useful for the eradication of SALL4-positive HCC cells through their inhibitory effects on histone deacetylation by NuRD [39]. Encouragingly, a recent study demonstrated the utility of a SALL4binding peptide to inhibit its binding to phosphatase and tensin homolog deleted on chromosome 10 (PTEN) through interaction with HDAC, thereby targeting leukemia cells [21]. Further studies are required to understand the relationship between SALL4, the NuRD complex, and the maintenance of stemness in HCC.

Financial support
This study was supported by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology, Japan (23590967), a grant from the Japanese Society of Gastroenterology, a grant from the Ministry of Health, Labour and Welfare, and a grant from the National Cancer Center Research and Development Fund (23-B-5), Japan.