Pseudogene-mediated DNA demethylation leads to oncogene activation

A pseudogene epigenetically regulates an oncogene by interacting with DNMT1 and affecting its methylation in a specific region.

Abstract

Pseudogenes, noncoding homologs of protein-coding genes, once considered nonfunctional evolutionary relics, have recently been linked to patient prognoses and cancer subtypes. Despite this potential clinical importance, only a handful of >12,000 pseudogenes in humans have been characterized in cancers to date. Here, we describe a previously unrecognized role for pseudogenes as potent epigenetic regulators that can demethylate and activate oncogenes. We focused on SALL4, a known oncogene in hepatocellular carcinoma (HCC) with eight pseudogenes. Using a locus-specific demethylating technology, we identified the critical CpG region for SALL4 expression. We demonstrated that SALL4 pseudogene 5 hypomethylates this region through interaction with DNMT1, resulting in SALL4 up-regulation. Intriguingly, pseudogene 5 is significantly up-regulated in a hepatitis B virus model before SALL4 induction, and both are increased in patients with HBV-HCC. Our results suggest that pseudogene-mediated demethylation represents a novel mechanism of oncogene activation in cancer.

INTRODUCTION

Hepatocellular carcinoma (HCC) is one of leading causes of cancer-related deaths globally, with more than 700,000 new cases and 600,000 estimated HCC deaths each year. Hepatitis B virus (HBV) infection is one of the main causes of HCC, particularly in Asia. While surgery, liver transplantation, or radiological intervention may be a viable option for early stage disease, prognosis for advanced stage HCC remains bleak, with most patients eventually dying within 20 months after diagnosis. Sorafenib, an oral multikinase inhibitor, is one of the few approved agents for patients with advanced HCC (1, 2). However, the effectiveness of sorafenib for advanced HCC is debatable (2). There is an unmet clinical need for the development of more effective therapies for the treatment of HCC. The lack of effective treatment options for HCC is at least, in part, due to our lack of understanding the pathogenesis of this disease. Identifying previously unknown pathway(s) that are responsible for HCC could be translated into targeted therapy and improve the outcomes of these patients.

SALL4 (Sal-like protein 4) is a potent stem cell factor for self-renewal and pluripotency of embryonic stem cells (3, 4). During development, SALL4 expression diminishes gradually and is eventually silenced in most normal tissues. Notably, high SALL4 expression levels have been observed in many malignancies such as liver cancer, acute myeloid leukemia, breast cancer, and lung cancer (5–8). Reexpression of SALL4 in cancers is associated with a more aggressive cancer phenotype, drug resistance, and reduced patient survival (5, 6, 9–11). Patients with HCC with detectable SALL4 expression have enriched hepatic progenitor-like gene signatures and poorer prognoses (10). Furthermore, targeting SALL4 in HCC cell lines by knocking down or using inhibitory peptides resulted in cellular death (12), suggesting that SALL4 plays a crucial role in hepatocarcinogenesis and that targeting SALL4 may provide an innovative therapeutic approach for this disease. However, mechanistically, how SALL4 is reactivated in HCC is still unclear, although it has been reported that aberrant methylation could be a contributing factor (13, 14). By defining the mechanism of SALL4 reactivation in HCC, we can better treat HCC.

DNA methylation is a frequently studied mechanism of epigenetic regulation in humans that is mediated by DNA methyltransferases (DNMTs); of which, DNMT1 has a structural binding preference (15). Research by multiple groups including ours has demonstrated that noncoding RNAs (ncRNAs) such as ecCEBPα, Dali, Dum, and Dacor1 can interact with DNMT1 to inhibit its methylation activity. These ncRNAs thus indirectly alter local methylation states in different cancers, acting as key tissue-specific epigenetic regulators of gene expression (15–18). It was also reported that the exon 1–intron 1 region of the SALL4 gene locus is hypermethylated in non–SALL4-expressing K562 leukemic cells. Reprogramming of these cells resulted in demethylation of this region and a subsequent increase in SALL4 expression (14). Recently, a report described demethylation of specific CpG sites downstream of the SALL4 transcriptional start site in HBV-related HCC, which could contribute to SALL4 reactivation in HCC (13). However, it is still unclear how HBV infection could initiate the demethylation and reactivation of specific oncogenes.

Pseudogenes are a class of ncRNAs once regarded as unimportant “junk” DNA relics due to their lack of coding potential. However, studies have demonstrated that pseudogene transcripts can regulate gene expression of oncogenes and tumor suppressors by acting as antisense transcripts, processed small interfering RNAs (siRNAs) and competing endogenous RNAs (ceRNAs) (19–21). Recent pseudogene expression analysis in over 2800 patient samples showed strong concordance between pseudogene expression and tumor subtypes, as well as patient prognoses, highlighting the clinical importance of pseudogenes (22). Our group focused on characterizing regulatory functions of SALL4 pseudogenes. SALL4, a well-studied oncogene with high expression levels in several hematological malignancies and solid tumors, has eight pseudogenes of different lengths varying from 500 nucleotides to 6000 nucleotides, and yet there have been no studies investigating SALL4 pseudogenes (5–8).

As many pseudogenes are actively transcribed in cells, we postulated that they could interact with RNA-binding proteins such as DNMT1 via highly homologous RNA motifs and exert regulatory functions. We therefore tested the hypothesis as to whether pseudogenes are involved in DNA methylation as DNMT1-interacting lncRNAs in an HBV⁺ HCC model.

RESULTS

SALL4 expression is negatively correlated with methylation of the 5′ untranslated region–exon 1–intron 1 region

As it has been reported that aberrant methylation could lead to increased SALL4 expression (13, 14), we first examined patients with HBV⁺ HCC. Using a publicly available dataset (23), the overall transcript levels and methylation status of SALL4 were analyzed using probes covered the entire SALL4 gene locus. A substantial negative correlation between SALL4 expression and methylation was observed in primary tumors at the probe 1 (Fig. 1, A and B) which was only observed in the 5′ untranslated region (5′UTR)–exon 1 region. Sites located either proximal or distal (probe 2) to the 5′UTR–exon 1 locus showed poor to no correlation with SALL4 expression (Fig. 1C).

Fig. 1. SALL4 expression is negatively correlated with methylation of the 5′UTR–exon 1–intron 1 region.

(A) Schematic representation of the methylation probes. The numbers refer to each CpG dinucleotide. The probe in the 5′UTR–exon 1 junction, “probe 1,” assesses the methylation status of the CpG dinucleotide no. 11, and the intronic probe, “probe 2,” assesses the CpG dinucleotide no. 68 in fig. S1. (B and C) SALL4 expression and methylation correlation analysis in 19 patients with HBV⁺. Compared to paired adjacent nontransformed liver tissue, there is a negative correlation between SALL4 expression and probe 1 methylation, which is not observed using probe 2. (D) Bisulfite sequence of the 5′UTR–exon 1–intron 1 region in wild-type SNU398 and SNU387 HCC cell lines. White color represents hypomethylation, while black represents hypermethylation of the individual CpG dinucleotide. Degree of methylation was determined as a proportion of methylated cytosine residue at a position out of 10 clones. Only CpG dinucleotides 1 to 35 are represented as the sequencing efficiency was poor for dinucleotides 36 to 39. (E) Absolute quantification of SALL4 mRNA expression in wild-type SNU398 and SNU387. β-Actin was used as a positive control for the assay. Complementary DNA for β-actin quantification was diluted 10 times and back-calculated accordingly later. The levels of β-actin were comparable between SNU398 and SNU387 at about 400 to 600 copies of transcripts per cell. However, SNU398 cells expressed more than 150 copies of SALL4 mRNAs, while SNU387 expressed less than 10 copies on average. (F) SALL4 protein levels in wild-type SNU398 and SNU387. β-Actin was used as a loading control for immunoblotting. *P < 0.05; **P < 0.01; ***P < 0.001.

To investigate the mechanisms leading to the negative correlation between SALL4 methylation and expression, we used HCC cell lines. The 5′UTR–exon 1–intron 1 region was first inspected and found to have over 30 CpG dinucleotides (fig. S1). Bisulfite sequencing in the HCC cell lines SNU398 and SNU387 revealed distinct and unique methylation profiles for the two cell lines (Fig. 1D). Within the profiled region, SNU387 showed a near universal, methylated profile in stark contrast to SNU398 that showed a completely demethylated profile, with the exception of four CpG dinucleotides. The result was consistent with previous reports observing that methylation of the SALL4 5′UTR–exon 1–intron 1 region is differentially methylated in K562-induced pluripotency reprogrammed cells and patients with HBV-related HCC. (13, 14). To investigate the relationship between the observed methylation profiles and gene expression, we examined SALL4 expression in both SNU398 and SNU387 (Fig. 1, E and F). The level of SALL4 transcription was substantive as more than 100 copies of SALL4 mRNAs per cell were detected in SNU398, while SNU387 cells only expressed about 10 copies per cell. It was also evident that SALL4 was expressed at much higher magnitude in SNU398, in which the SALL4 loci was hypomethylated. Together, both the cell line and patient data suggest that SALL4 methylation and expression are negatively correlated. We therefore confirmed that DNA methylation could be a potential regulatory mechanism for SALL4 expression in HCC.

CRISPR–DNMT1-interacting RNA demethylates and activates SALL4

To further investigate the correlation between methylation of the SALL4 5′UTR–exon 1– intron 1 region and SALL4 expression, the CRISPR–DNMT1-interacting RNA (CRISPR-DiR) technique (24) was used to induce locus-specific demethylation by blocking DNMT1 activity in SNU387 cells. Briefly, the single-guide RNA (sgRNA) was constructed to contain a SALL4 exon 1 targeting sequence, two RNA loops of ecCEBPα with validated DNMT1 inhibitory function (15), and the dCAs9-interacting domain (Fig. 2A). Numerous sgRNAs targeting the SALL4 5′UTR–exon 1–intron 1 locus were designed, with the five most efficient candidates shortlisted (sgSALL4_1 to sgSALL4_5) via in vitro sgRNA selection (fig. S2A). Transduced cells were also validated to express the dCas9-mCherry through fluorescence-activated cell sorting (FACS) (fig. S2B). Among the five selected sgRNAs for SALL4, sgSALL4_1 showed the most significant up-regulation of SALL4 mRNA and protein levels, as well as induction of DNA demethylation within the SALL4 5′UTR–exon 1–intron 1 CpG island (fig. S3, A to C). Hence, we focused on sgSALL4_1 for further investigations.

Fig. 2. CRISPR-DiR demethylates and activates SALL4.

(A) sgRNA design for the CRISPR-DiR. The red region targets and interacts with the SALL4 5′UTR. The black region interacts with dCas9. The blue regions are the two segments of ecCEBPα that interact with DNMT1 (15). (B) Bisulfite sequence of CRISPR-DiR transduced SNU387 cells. The data represent the methylation profile 14 days after CRISPR-DiR for SALL4. The numbers indicate each of the CpG dinucleotides in the 5′UTR–exon 1–intron 1 junction. The white color represents hypomethylation, while the black hypermethylation of the individual CpG dinucleotides. Only CpG dinucleotides 1 to 35 are represented as the sequencing efficiency was poor for dinucleotides 36 to 39. (C and D) SALL4 transcript and protein levels after CRISPR-DiR in SNU387. The Western blot image is cropped as there were multiple lanes in between “D21” and “5-aza”. However, they are from the same blot and exposed for the same duration. (E) Soft agar growth assay for CRISPR-DiR in SNU387. (F) Growth curve assay for CRISPR-DiR for SALL4 in SNU387. 5-Aza-2-deoxycytidine(decitabine) was used as a positive control. NT denotes nontargeting negative control. Means ± SD, n ≥ 3; **P < 0.01; ***P < 0.001.

Methylation of the SALL4 5′UTR–exon 1–intron 1 CpG island was monitored in SNU387 cells with four independent CRISPR-DiR inductions, one for each shortlisted sgRNA. Of these inductions, sgSALL4_1 was the most potent sgRNA tested. Upon transduction of SNU387 cells with sgSALL4_1, significant demethylation changes were observed after 14 days, which continued for over seven additional days (Fig. 2B). Conversely, no change in methylation was observed in nontargeting, negative control transduced cells. To examine potential off-target effects of CRISPR-DiR, we concurrently monitored methylation of a region in SALL4 exon 4 and confirmed that demethylation was localized to only the targeted 5′UTR–exon 1–intron 1 CpG island (fig. S2C).

Following CRISPR-DiR targeted demethylation of the SALL4 5′UTR–exon 1–intron 1 CpG island, both SALL4 transcript and protein levels increased as predicted (Fig. 2, C and D). The magnitude of SALL4 up-regulation observed in these cells was comparable to that of treatment with 5-aza-2′-deoxycytidine, a global demethylating agent. As SALL4 overexpression promotes cancer cell growth, we performed growth assays on sgSALL4-transduced SNU387 cells and observed increased anchorage-independent and anchorage-dependent growth compared to negative control (Fig. 2, E and F), suggesting that targeted demethylation of the SALL4 locus leads to up-regulated expression of SALL4, with concomitant enhanced cellular growth.

SALL4P5 demethylates and activates SALL4 and associates with DNMT1

There are eight SALL4 pseudogenes, and because none are located on the same chromosome as SALL4, it is unlikely that they will be transcribed as siRNAs or antisense transcripts to deregulate SALL4. However, the identified pseudogenes do share high sequence homology with the paralogous coding SALL4 gene. It is therefore possible that the SALL4 pseudogenes could bind to other proteins with either matching DNA/RNA motifs or with comparable secondary and tertiary structures owing to their high sequence homology. As previously reported, ecCEBPα, a ncRNA that overlaps with and thus has regions of identity with its paralog, CEBPα, can interact with DNMT1 and affect CEBPα gene expression. We therefore postulated that SALL4 pseudogenes, which are highly homologous to SALL4, could potentially mediate SALL4 demethylation (fig. S4A).

Each SALL4 pseudogene was transiently overexpressed in SNU387 cells, and the methylation profile of the 5′UTR–exon 1–intron 1 CpG island was assessed (Fig. 3A). Only SALL4 pseudogene 5 (SALL4P5) overexpression resulted in a demethylation pattern comparable to that seen using CRISPR-DiR. Consistently, SALL4P5 knockdown in hypomethylated SNU398 cells led to increased methylation of the locus as predicted (Fig. 3B).

Fig. 3. SALL4P5 demethylates and activates SALL4 and associates with DNMT1.

(A) Bisulfite sequencing after transient overexpression of individual SALL4 pseudogenes in SNU387. The methylation status of CpG dinucleotides in SALL4 5′UTR–exon 1 intron is shown. (B) Bisulfite sequencing after SALL4P5 knockdown in SNU398. ShSCR denotes scrambled short hairpin RNA. shSALL4P7 was used as a negative control. (C) Transcript localization in SNU398. Cells were fractionated into nuclear and cytoplasmic fractions, and transcript expression was quantified. β-Tubulin was used as a cytoplasmic fraction control, 18S ribosomal RNA (rRNA) as the nuclear fraction control. (D) Biotin-labeled pulldown of DNMT1 in SNU398. Full-length SALL4P5 was used as a bait to pull-down complexes, and DNMT1 presence was probed using immunoblotting. “as P5” denotes the negative control, antisense-SALL4P5, and “ecCEBPα” denotes the positive control. Full-length SALL4P7 was used as a pseudogene negative control as well. (E and F) SALL4 transcript and protein expression after pseudogene overexpression in SNU387. EV, empty vector. (G) Soft agar growth assay for pseudogene overexpression in SNU387. Means ± SD, n ≥ 3; ***P < 0.001.

Additional evidence to suggest direct SALL4P5-DNMT1 interaction can be seen by their matched cellular localization. Cellular localization of pseudogene transcripts is a critical factor in determining their function, as they must be localized in the same cellular compartment as their binding partners to exert specific biological functions. It is known that DNMT1 facilitates methylation exclusively in the nucleus. Critically, SALL4P5 is also primarily localized to the nucleus. This contrasts with SALL4P7, which has a predominant cytoplasmic localization in SNU398 (Fig. 3C). Therefore, although SALL4P7 shares sequence homology with SALL4P5 and SALL4, its primary localization in cytoplasm could, in part, account for its inability to demethylate the SALL4 locus.

As DNMT1 is known as a maintenance DNA methylator, we therefore investigated whether our observed SALL4P5 demethylation phenotype is due to a SALL4P5-DNMT1 interaction. As there are no known interacting SALL4P5-DNMT1 binding regions or pockets, we performed an unbiased biotinylated pull-down assay using full-length SALL4P5. First, to validate the efficacy of the pulldown, DNMT1 protein could successfully pull-down ecCEPBα (Fig. 3D). Similarly, SALL4P5 was able to successfully pull down DNMT1, whereas SALL4P7 and the antisense negative control did not.

Having shown an association between SALL4 exon 1–intron 1 demethylation and SALL4 expression up-regulation, we next investigated the effect of transiently overexpressing SALL4P5 on SALL4 levels. SALL4P5 overexpression significantly up-regulated SALL4 transcript (Fig. 3E) and protein levels, the latter of which was more notable and equivalent to the overexpression of SALL4 itself (Fig. 3F). SALL4P7 overexpression also increased SALL4 protein levels. Although SALL4P7 may not play a role in SALL4 demethylation, it could still contribute to gene regulation as homologous pseudogenes could also function as ceRNAs (21) by sequestering bioavailable microRNAs that target and repress SALL4.

Consistent with the phenotype of elevated SALL4 levels, SALL4P5 overexpression also significantly increased colony formation of SNU387 cells (Fig. 3G). In addition, wound healing migration assays demonstrated that overexpression of SALL4P5 and SALL4 enhanced migration rates of SNU387 cells compared to SALL4P7, a SALL4 pseudogene that does not deregulate SALL4 expression or methylation profiles, and the negative control (fig. S4B). The opposite is true for the knockdown experiment, in which SNU398 cells with down-regulated SALL4P5 and SALL4 expression show less wound healing migration than shSCR and SALL4P7-tranduced cells (fig. S4C). The data suggest that SALL4P5 could have oncogenic effects as it can directly up-regulate SALL4 expression and cell growth.

SALL4P5 is up-regulated in patients with HCC and in a model of hepatitis B induction

The aforementioned results demonstrate that SALL4P5 up-regulation can reactivate SALL4 expression in cell lines. We next sought to validate these findings in primary patient samples and measured SALL4P5 expression in patients with HCC, who frequently have elevated levels of SALL4 (10). Twenty patients with HCC with paired nondisease samples were screened from a Hong Kong cohort. Among these 20 patients, 19 of them were HBV⁺ and 7 had increased SALL4 levels of over 1.5-fold (Fig. 4A). Within these seven patients, only SALL4P5 expression was concomitantly up-regulated, while SALL4P7 showed little to no change. For patients, such as patient no. 1, with no SALL4 level change, SALL4P5 expression was also unaltered.

Fig. 4. SALL4P5 is up-regulated in patients with HCC and during hepatitis B induction.

(A) Relative SALL4 transcript expression in paired HCC patient samples. All expression data are normalized against adjacent nontransformed tissues. Patient no. 1 was used as a negative control with unaltered SALL4 and SALL4P5 levels. The levels of SALL4, SALL4P5, and SALL4P7 were assessed for the other six patients, patient nos. 2 to 7, as they had more than 1.5-fold elevation in SALL4 levels compared to adjacent nontransformed tissue. (B) Absolute quantification (RNA copies per cell) of hepatitis B antigen X (HBx), and SALL4 transcripts during hepatitis B induction in HepAD38B. Transcript levels of HBx and SALL4 transcripts were monitored every 6 to 12 hours after HBV induction (C) Methylation profile of SALL4 5′UTR–exon 1–intron 1 region in HBV-induced (tet-off) and HBV-uninduced (tet-on) HepAD38B. Each box describes the degree of methylation at a specific CpG dinucleotide, ranging from 0, denoted by white color, to 100, denoted by black color, percentage.

HBV infection is the single most common risk factor of HCC, as more than 50% of patients contract hepatitis B before HCC (25). We therefore sought to investigate whether SALL4P5-mediated demethylation and subsequent reactivation of SALL4 during hepatitis B infection could drive oncogenesis. The HepAD38B model was used, as it allows controlled induction of HBV production using the tet-off system (13). Using digital droplet polymerase chain reaction (ddPCR), we validated that the HepAD38B cells produced the major hepatitis B viral transcripts such as core, surface, hepatitis B antigen X (HBx), and polymerase transcripts (fig. S5A). Upon hepatitis B induction, SALL4 and SALL4P5 transcript levels also increased (Fig. 4B and fig. S5, B and C). The expression level for HBx increased first, SALL4P5 expression then followed at 54 hours, and lastly SALL4 expression at 84 hours in stepwise manner. When performing bisulfite sequencing of these critical time points, it was found that the methylation across the CpG island in the 5′UTR–exon 1–intron 1 junction decreased upon hepatitis B induction in a time-dependent manner, suggesting that the infection-induced up-regulation of SALL4P5 demethylates and reactivates SALL4 (Fig. 4C).

DISCUSSION

This work highlights the previously undescribed capability of a pseudogene to epigenetically regulate an oncogene by interacting with DNMT1 and affecting its methylation in a specific region. It is reported that there are at least 12,000 pseudogenes (26) in the human genome. A recent pan-cancer analysis of pseudogene expression in different cancers demonstrated that pseudogene expression alone can serve as a molecular and prognostic factor for patients with different cancer subtypes, highlighting the functional and clinical importance of pseudogenes (22). There may be other pseudogenes capable of mediating similar DNMT1 interactions and exerting the demethylating function on oncogenes and tumor suppressors in different cancers. Monitoring expression levels of oncogene-demethylating pseudogenes could enable predicting oncogene activation and disease progression to improve patient outcomes. As an example, in this study, we provide insights into SALL4 reexpression in HCC, which is of therapeutic value, owing to SALL4’s functional and prognostic significance in this disease (7). In addition, pseudogenes could exert noncanonical functions by interacting with other RNA-binding proteins, with potential wide-ranging implications in gene regulation and function.

Previous studies from Di Ruscio et al. (15) demonstrated that RNAs require distinct secondary structures to associate with DNMT1. This preferential interaction through structure could potentially explain why DNMT1 interacts with SALL4P5, but not SALL4P7, even though the two pseudogenes are highly homologous. Another notable aspect of SALL4P5-SALL4 regulation is that unlike ecCEBPα, which resides in the CEBPα locus, the SALL4P5 locus is on chromosome 3, while its paralogous coding gene SALL4 is on chromosome 20. This trans regulation implies that it could be sequence homology and perhaps secondary structure that plays a more critical role than chromosomal location for a ncRNA to work as a DiR. Moreover, as DNMT1 plays a crucial role in de novo methylation, the SALL4P5-DNMT1 interaction could contribute to SALL4 reactivation and other oncogene activation in cancers.

These studies represent one of the first examples of gene locus specific demethylation resulting in the activation of an oncogene. DNA hypermethylation of CpG islands in gene promoter regions has consistently correlated with inactivation of tumor suppressors in cancers (27). Conversely, demethylation of oncogene promoters leads to increased gene expression (28). By an innovative strategy, we found that the methylation profile of the 5′UTR–exon 1–intron 1 region of SALL4 was critical in SALL4 up-regulation leading to cell growth. Further investigation will be needed to determine whether this noncanonical methylation site, downstream from the promoter, is only significantly affected in the context of pseudogene-mediated demethylation.

MATERIALS AND METHODS

Experimental design

Using publicly available methylation and expression profiles of patients with HBV⁺ HCC, we aimed to investigate whether SALL4 expression is correlated with the methylation levels of the 5′UTR–exon 1–intron 1. Upon validating the negative correlation, we performed CRISPR-DiR (24) to locus-specifically alter the methylation profile and observed SALL4 up-regulation. To elucidate endogenous ncRNA responsible for SALL4 demethylation, we investigated homologous SALL4 pseudogenes and their effect on SALL4 expression and methylation. Last, we probed the SALL4 pseudogenes’ expression in patients with HCC and HBV induction model to seek biological relevance.

Cell culture

All HCC cell lines were obtained from American Type Culture Collection and grown according to the manufacturer’s instructions in the absence of antibiotics. Human HCC cell lines (SNU398, SNU387, and HepAD38B) were maintained in Dulbecco’s modified Eagle’s medium and RPMI 1640 (Life Technologies, Carlsbad, CA) with 10% fetal bovine serum (Invitrogen) and 2 mM l-glutamine (Invitrogen). These cell lines were cultured at 37°C in a humidified incubator with 5% CO₂. SNU398 was derived in 1990 from a 42-year-old, Asian male patient with HCC. SNU387 was derived in 1990 from a 41-year-old, Asian female patient with HCC. HepAD38B cell line was derived from a 15-year-old, male patient with hepatoblastoma.

Patient samples

A total of 20 paired human primary HCC and adjacent nontumor liver tissues that were surgical removed and snap-frozen in liquid nitrogen were obtained from the Sun Yat-Sen University Cancer Centre (Guangzhou, China). All patients gave written informed consent for the use of their clinical specimens for medical research. All samples used in this study were approved by the committees for ethics review for research involving human participants at the Sun Yat-Sen University, the University of Hong Kong, and the Institutional Review Board of National University of Singapore (reference code: B-14-239E).

RNA extraction and gene expression analysis

Total RNA was extracted from cells using the TRIzol reagent (Invitrogen) and purified using the RNeasy Mini kit from Qiagen. Purified RNA (1 μg) was used for complementary DNA (cDNA) synthesis using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific) was used to assess the expression levels of the mRNAs, microRNAs, and pseudogenes of interest. GoTaq qPCR Master Mix (Promega) was used as a SYBR master mix reagent for the quantitative PCR procedures. The quantitative real-time PCR data were analyzed using the QuantStudio Design and Analysis Software version 1.2 (Thermo Fisher Scientific) and represented as relative expression (ΔΔCt), normalized against either glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or β-actin. The primer sequences used for the quantitative real-time PCR are provided in table S1.

Genomic DNA extraction

Genomic DNA was extracted from HCC cell cultures using a DNeasy Blood & Tissue kit (Qiagen) for bisulfite-sequencing assay according to the manufacturer’s protocols.

Bisulfite treatment and sequencing

SALL4 5′UTR–exon 1–3′UTR region methylation status was assessed using bisulfite sequencing. In brief, 1 μg of genomic DNA extracted using the DNeasy Blood & Tissue kit (Qiagen) was bisulfite-converted by using the EZ DNA Methylation kit (Zymo Research). PCR products were gel-purified (Qiagen) from the 1.5% tris-acetate-EDTA gel and cloned into the pGEM-T Easy Vector System (Promega) for transformation. The cloned vectors were transformed into ECOS 101 DH5α cells, and miniprep was performed to extract plasmids. Sequencing results were analyzed using BiQ analyzer software. Samples with more than 90% conversion rate and 70% sequences identity were analyzed. The minimum number of clones for each sequenced condition was 6. Primers used for the bisulfite sequencing of SALL4 5′UTR–exon 1–intron 1 are listed in table S2. Methylation percentages were subsequently plotted using R (29) and the pheatmap package (30).

Protein extraction and immunoblotting

Total cell lysates in protein lysis buffer (PLB) [100 mM KCl (Ambion), 5 mM MgCl₂ (Ambion), 25 mM EDTA (pH 8.0) (Life Technologies), 10 mM Hepes (Life Technologies), 0.5% NP-40 (Roche), 20 mM dithiothreitol (Fermentas), proteinase inhibitor tablet (Roche)]. PLB was added to the cell pellets and incubated for 15 min on ice. The lysates were centrifuged for 10 min at 15,000g at 4°C. Protein concentrations were measured using the Bradford Protein Assay (Bio-Rad Laboratories), and absorbance was measured at 595 nm on the Tecan Infinite 2000 PRO plate reader (Tecan, Seestrasse, Switzerland). Equal amounts of protein for each sample were diluted with 4× sample buffer (Thermo Fisher Scientific) and heated at 95°C for 5 min. The proteins were resolved by SDS–polyacrylamide gel electrophoresis 12% self-cast gel and transferred onto polyvinylidene difluoride membrane using the Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad) in transfer buffer [25 mM tris, 192 mM glycine, and 20% (v/v) methanol (Fisher Chemical)]. After blocking with 5% milk in tris-NaCl buffer (tris-buffered saline) (Thermo Fisher Scientific) with 0.1% Tween 20 (Sinopharm Chemical Reagent Co. Ltd), membranes were incubated with the appropriate primary and secondary antibodies. The immune-reactive proteins were detected using the protein bands were visualized using SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and visualized on Image Quant LAS 500 machine (GE Healthcare) according to the manufacturer’s instructions. SALL4 (Santa Cruz Biotechnology, EE30, no. sc-101147), GAPDH (Cell Signaling Technology, no. 5174), and DNMT1 (Abcam, no. ab87656) antibodies were used for immunoblotting as per the manufacturer’s instructions.

Bacterial transformation

ECOS 101 competent cells (DH5α) from Yeastern Biotech Co. Ltd. were used for transformation following the manufacturer’s instructions. Cells (50 μl) were thawed at room temperature in a water bath. Prechilled DNA (2 μl) was immediately added. The tubes were kept on ice for 5 min to increase the transformation efficiency. The cells went through heat shock in a 42°C water bath for 30 s. The cells were kept on ice again for 5 min and plated on LB plates. The plates were incubated at 37°C for 16 hours.

Plasmid extraction

Plasmid was extracted from 1.5 ml of bacterial culture with the QIAamp DNA Mini Kit (Qiagen) and purified for sequencing validation (1st BASE). For transfected cell line DNA extraction, plasmid was extracted from the cell pellet that is suspended in 100 μl of phosphate-buffered saline (PBS).

Plasmid transfection

SNU398 cells were seeded at a density of 75,000 cells per well in 12-well plates 24 hours before transfection. SNU387 and SNU182 cells were seeded at higher density of 100,000 cells per well in 12-well plates. Plasmid (500 ng) was added to 3 μl of P3000 reagent (Life Technologies) in 75 μl of Opti-MEM before mixing with 2 μl of Lipofectamine 3000 (Life Technologies). The reagent mixture was incubated at room temperature for 10 min before adding to each well.

Cell viability assay

Twenty-four hours after transfection, cells were trypsinized and split into 5 individual wells of five separate 12-well plates. Upon adherence, cells were fixed using 10% neutral buffered formalin solution (Sigma-Aldrich, HT501128-4L) and labeled as day 0. Subsequently, the remaining plates were fixed daily from days 2 to 5 (excluding day 1) before staining with crystal violet (Sigma-Aldrich, C0775-100G) for 3 to 5 min at room temperature. Stained wells were washed three times with Milli-Q water and left to dry. Crystal violet stain was solubilized using 10% acetic acid (Sigma-Aldrich, A6283-2.5L). The plates were left on a shaker at room temperature for at least 20 min. The absorbance reading for each well was measured at 595 nm using the Tecan Infinite 2000 PRO plate reader (Tecan, Seestrasse, Switzerland).

Soft agar assay

A 0.6% agarose base was prepared by mixing 3.9 ml of 2% UltraPure agarose (Invitrogen) with 9.1 ml of cell culture medium. The mixture (2 ml) was added to individual wells of six-well plates. Twenty-four hours after transfection, cells were trypsinized, counted, and diluted to a concentration of 15,000 cells per well. Four hundred fifty microliters of the 2% agarose was added to 2.55 ml of cells for a final agarose concentration of 0.3%, and 1 ml of the mixture was added to each well containing the solidified 0.6% agarose base. Culture medium (1 ml) was added to the top agar layer upon solidification, and the cells were incubated at 37°C in a humidified incubator. Culture medium was changed every 2 days. Images of the colonies were taken under ×4 magnification every 4 to 5 days for a period of up to 14 days and quantified using ImageJ.

Caspase-Glo 3/7 activity assay

Apoptotic activity was measured using the Caspase-Glo 3/7 activity assay (Promega). The assay was performed following the manufacturer’s instructions. The substrate for caspase 3/7 was added 48 hours after transfection to detect caspase activity.

Wound healing migration assay

Wound healing ability was measured using a CytoSelect 24-Well Wound Healing Assay Kit (Cell Biolabs). The experiment was performed according to the manufacturer’s instructions. SNU387 and SNU398 cells were cultured to full confluence. At the start of the experiment, the inserts were removed and the wells were washed with PBS to removed unattached cells and debris. RPMI 1640 culture medium (500 μl) was added to the wells. Cells were incubated at 37°C to allow the closure of the induced wound. After incubation for 24 to 48 hours, migrated cells were visualized using phase-contrast microscopy.

In vitro transcription and biotinylated RNA pulldown

The DNA template was first amplified by PCR with primers containing a 5′ T7 tag for in vitro transcription (IVT). Antisense SALL4P5 control was also amplified by having the reverse primers carrying the T7 tag. The IVT was performed as per the manufacturer’s guidelines. Purified PCR product (1 μg) was incubated with the transcription mix, which was composed of 10× transcription buffer, 400 mM NTP (nucleotide triphosphate) mix, and 200 U of T7 RNA polymerase for 5 hours at 37°C. Ribonuclease-free water (140 μl) and 1000 μl of 100% ethanol were added to the transcription product and incubated for at least 30 min at −20°C. The reaction mixture was centrifuged for 1 hour at 4°C to precipitate the RNA. The RNA pellet was collected and dissolved in 100 μl of ultrapure water. The RNA was further purified using RNeasy Mini 250 columns (Qiagen) according to the manufacturer’s instructions. The purified RNA obtained from IVT was labeled with biotin at the 3′ end using the Pierce RNA 3′ End Desthiobiotinylation Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. Biotin labeling efficiency of the RNA probes was determined using the Chemiluminescent Nucleic Acid Detection Module Kit (Thermo Fisher Scientific) following the manufacturer’s protocol. Biotin labeling efficiency was normalized against the efficiency of antisense transcript to determine amount of initial RNA bait used for the subsequent pull-down experiment. Pulldown using these labeled RNA probes was carried out using the Pierce Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Protein lysates eluted from the pulldowns were used for immunoblotting and other downstream analysis. The primer sequences with the T7 tag for the PCR are provided in table S3.

In vitro generation of sgRNA transcripts

Approximately 1.4 kb of genomic fragment spanning SALL4 5′UTR–exon 1–intron 1 was PCR-amplified (Zymo Research) and cloned into the pGEM-T Easy vector. The vector was linearized with Bam HI restriction enzyme (New England Biolabs). SALL4-targeting sgRNA candidates were transcribed with the HiScribe Quick T7 High Yield RNA Synthesis Kit (New England Biolabs) following the manufacturer’s instructions. The sgRNA target sequences within SALL4 locus are listed in table S4.

In vitro cleavage and selection of sgRNA transcripts

In vitro cleavage assay was performed using purified Cas9 nuclease from Streptococcus pyogenes (New England Biolabs) to select SALL4-specific sgRNA among candidates. The experiment was performed according to the manufacturer’s protocol. The sgRNAs were denatured at 95°C for 3 min, and then Cas9 protein and sgRNAs were incubated for 10 min at 25°C to form a complex. Last, a linearized SALL4 DNA fragment was added to the mixture, and the entire reaction was incubated at 37°C for 1 hour. The reaction mixture was composed of purified Cas9 protein, individual sgRNA, and linearized SALL4 genomic fragment in a ratio of 10:10:1. Proteinase K (1 μl) was added to each sample after the cleavage reaction, and it was then incubated at room temperature for 10 min. The result was analyzed with a 1% agarose gel.

Lentiviral transduction of sgSALL4_1 and dCas9

Lentiviruses expressing dCas9 or sgRNA were packaged in 293 T cells and the plasmids psPAX2 and pMD2.G. TransIT-LT1 Transfection Reagent (Mirus) was used for transfection into 293 T cells. Virus was collected at 48 and 72 hours after transfection. The collected virus was filtered through 0.45-μm microfilters and stored at −80°C. Transduction of SNU-387 cells was performed by mixing virus and polybrene (4 μg/ml; Santa Cruz Biotechnology) together to add to the cells seeded in T75 flasks 24 hours before the transduction. Twenty-four hours after the transduction, the medium was replenished with normal RPMI 1640 culture medium. Transduction efficiency was determined by green fluorescent protein (for sgRNA) or mCherry (dCas9) expression by FACS analysis, and the positive cells were sorted by a FACSAria machine (BD Biosciences).

5-Aza-2′-deoxycytidine(decitabine) treatment

SNU387 cells were treated with 1.25 μM of 5-aza-2′-deoxycytidine (Sigma-Aldrich) according to the manufacturer’s instructions. Culture medium and drug were refreshed every 24 hours because of the drug being light sensitive. RNA (for RT-PCR) and genomic DNA (for bisulfite sequencing) were isolated after 5 days of treatment.

Digital droplet PCR

Reactions for the ddPCR were prepared by harvesting 100,000 cells on each day for RNA extraction and cDNA preparation. The reaction mixture was prepared with the 2× ddPCR supermix for probes (Bio-Rad, catalog no. 186-3026), 10-fold diluted cDNA, nuclease-free water, and forward and reverse primers. Once the reaction mixture was ready, it was loaded into a DG8 cartridge for the QX200 Automated Droplet Generator (Bio-Rad, catalog no. 186-4003). We then proceeded to the thermal cycling with a Bio-Rad C1000 Touch Thermal Cycler with the following cycle conditions: 95°C for 10 min, 94°C for 30 s (40 cycles), 60°C for 2 min (40 cycles), 98°C for 10 min, and 4°C hold. The reaction plate was loaded into the QX200 Droplet Reader (Bio-Rad, catalog no. 186-4003) for gene expression analysis.

SALL4 methylation and expression correlation analysis

Pearson correlations between SALL4 expression and methylation levels were performed for the sites within the 5′UTR–exon 1–intron 1 region and distant sites in the intron 1 (fig. S2). To show that a negative correlation is specific to primary patients with HBV⁺ HCC, adjacent normal samples were used as negative controls. The data used for the correlation were taken from Yang et al. (23), containing 19 pairs of primary patients with HBV⁺ HCC and their matched adjacent normal tissues.

Statistical analysis

All in vitro cell line experiments were repeated three times (n = 3), and statistical significance is represented as *P < 0.05, **P < 0.01, and ***P < 0.001 for Student’s t test.

Supplementary Materials

This PDF file includes:

Figs. S1 to S5

Tables S1 to S5

View/request a protocol for this paper from Bio-protocol.