Context-Dependent Function of Long Noncoding RNA PURPL in Transcriptome Regulation during p53 Activation

ABSTRACT

PURPL is a p53-induced lncRNA that suppresses basal p53 levels. Here, we investigated PURPL upon p53 activation in liver cancer cells, where it is expressed at significantly higher levels than other cell types. Using isoform sequencing, we discovered novel PURPL transcripts that have a retained intron and/or previously unannotated exons. To determine PURPL function upon p53 activation, we performed transcriptome sequencing (RNA-Seq) after depleting PURPL using CRISPR interference (CRISPRi), followed by Nutlin treatment to induce p53. Strikingly, although loss of PURPL in untreated cells altered the expression of only 7 genes, loss of PURPL resulted in altered expression of ~800 genes upon p53 activation, revealing a context-dependent function of PURPL. Pathway analysis suggested that PURPL is important for fine-tuning the expression of specific genes required for mitosis. Consistent with these results, we observed a significant decrease in the percentage of mitotic cells upon PURPL depletion. Collectively, these data identify novel transcripts from the PURPL locus and suggest that PURPL delicately moderates the expression of mitotic genes in the context of p53 activation to control cell cycle arrest.

INTRODUCTION

p53 is a master regulatory transcription factor with a diverse array of gene targets (1–4). When a cell is subject to stressors such as hypoxia, oncogene activation, DNA damage, or nutrient deprivation, p53 is activated (2, 5–7). In its transcriptionally active form, p53 binds to the promoter of its targets and activates their expression. In turn, this activates cellular pathways involved in processes that include DNA repair, metabolism, cellular senescence, cell cycle arrest, apoptosis, ferroptosis, and autophagy, among others (5, 7, 8). p53 is considered the “guardian of the genome,” as it is the most frequently mutated gene in human cancers (9–11).

In addition to traditional mRNA targets, we and others have shown that long noncoding RNAs (lncRNAs), a heterogeneous class of noncoding RNAs >200 nucleotides long, are also regulated by p53 (12–17). We and others have previously reported that the basal and DNA damage-induced expression of the lncRNA p53 upregulated regulator of p53 levels (PURPL) is directly activated by p53 (16, 18). When PURPL (also called LINC01021) expression is depleted in HCT116, a colorectal cancer (CRC) cell line, decreased proliferation, clonogenic potential, and the ability to form tumors in mouse xenografts were observed (16). PURPL acts in these cells by interfering with the interaction of p53 with the protein MYBBP1A that binds and stabilizes p53, resulting in its destabilization and a decrease in the expression of its targets, such as p21, MDM2, and NOXA (16). This was later verified in another study where p53 protein levels and activation were inhibited after PUPRL overexpression in melanoma cells (19). This suggests that PURPL acts as a tumor-promoting factor in CRC cells.

The function of a lncRNA can be influenced by a variety of factors, such as its sequence, structure, binding partners, and subcellular localization. When in the nucleus, lncRNAs can aid in chromatin remodeling and the regulation of gene expression (20–22), while in the cytoplasm, they can influence mRNA stability and translation or act as molecular decoys for molecules such as microRNAs (miRNAs) and RNA-binding proteins (RBPs) (20–22). The exonic structure of a lncRNA can also influence its function. Variations in sequence via alternative splicing can include or remove functional domains that facilitate the formation of diverse subcellular environments (23). The inclusion or exclusion of introns can also influence the subcellular localization of a transcript, as intron retention may predispose an RNA to be detained in the nucleus (24–26).

lncRNAs are very diverse molecules that frequently display cell type-specific functions (27). In one cell line, a lncRNA can play an indispensable role in regulating gene expression, while in another, it may be inessential (27). Differences in the ability of a lncRNA to modulate the expression of its target genes also exist across cancer cell types (28), suggesting that lncRNAs perform unique roles in different contexts. A similar pattern is seen for PURPL. In melanoma cells, PURPL was identified to repress autophagic cell death by promoting the phosphorylation of ULK1 via its interaction with mTOR (19).

To further study the context-dependent functions of this p53-regulated lncRNA, we investigated the role of PURPL in liver cancer cells. Here, we show that there are unique classes of transcripts expressed from the PURPL locus with different subcellular localizations. Contrasting with its role in CRC cells, we unexpectedly found that PURPL does not regulate p53 levels or activity in liver cancer cells. Interestingly, transcriptome analysis following depletion of PURPL and induction of p53 identified ~800 differentially expressed genes, including specific genes associated with pathways involving mitosis. Consistent with these data, we found that the loss of PURPL confers a decrease in the percentage of mitotic cells, and PURPL silencing negatively correlates with the formation of colonies in three-dimensional culture. Taken together, our study identified novel transcript variants expressed from the PURPL locus and uncover a context-dependent function of PURPL in transcriptome regulation upon p53 activation.

RESULTS

Identification of diverse novel transcripts expressed from the PURPL locus in liver cancer cells.

lncRNAs have been shown to have context-dependent functions in cancer (29–33). We have previously shown that PURPL transcription is directly induced by p53, and PURPL functions to suppress basal p53 levels by inhibiting the formation of the p53-MYBBP1A complex in CRC cells (16). To investigate a potential role of PURPL in other cancer types, we mined publicly available transcriptome sequencing (RNA-Seq) data from The Cancer Genome Atlas (TCGA) and observed that PURPL has higher expression levels in liver cancer tissues (LIHC) than colorectal tissues (COAD) (Fig. 1A; see Table S1 in the supplemental material). We next screened a panel of cell lines for PURPL expression and measured the absolute number of molecules of PURPL per cell using droplet digital PCR (ddPCR) with primers that recognize all PURPL transcript variants. Briefly, total RNA was extracted from a known number of cells and reverse transcribed, and droplets were generated. The absolute molecules per cell were assessed using QX200 droplet digital PCR system as previously described (34). We observed higher numbers of PURPL molecules per cell in HepG2 and SKHEP1 liver cancer cell lines (52 and 33, respectively) than in CRC cells LS174T and HCT116 (22 and 5, respectively), while glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels were consistent across cell lines (Fig. 1B). Because PURPL is more highly expressed in liver cancer cells and tissues than CRC, we hypothesized that PURPL could have additional functions in liver cancer cells.

FIG 1.

Iso-Seq identifies three groups of transcript variants produced at the PURPL locus. (A) Violin plot of top 15 TCGA cancer types with the normalized expression of PURPL {log₂(transcripts per million [TPM] + 1)} in descending order from left to right. (B) Droplet digital PCR showing the number of molecules per cell for PURPL and GAPDH in HepG2, SKHEP1, LS174T, and HCT116 cells. The graph shows the average of 3 experiments, and the error bars correspond to SD. (C) IGV snapshot of RNA-Seq read coverage at the PURPL locus in SKHEP1 and HepG2. The representation of the locus as annotated by RefSeq is also indicated. (D) IGV snapshot of Iso-Seq read coverage and mapped reads at the PURPL locus in HepG2 and SKHEP1. (E, Top) Schematic of the three PURPL transcript variant groups identified by Iso-Seq, transcripts spliced to include a distal 3′ exon but not intron 2, transcripts with intron 2 retention, and transcripts with both intron 2 retention and inclusion of the distal 3′ exon. (E, Bottom) Full-length read counts of the three transcript variant types in HepG2 and SKHEP1 identified by Iso-Seq.

Although lncRNAs do not encode peptides, they undergo alternative splicing and produce multiple transcript variants, which are often cancer type specific (35, 36). When we performed RNA-Seq from the liver cancer cell lines SKHEP1 and HepG2, we noticed differences in the distribution of reads across the PURPL locus in comparison with the expected expression profile of PURPL (LINC01021) as annotated by RefSeq (Fig. 1C). To assay for the potential expression of PURPL transcript variants, we performed PacBio isoform sequencing (Iso-Seq) from HepG2 and SKHEP1 cells. SQANTI3 analysis identified 3 major groups of transcripts arising from the PURPL locus (Fig. 1D and Table S2). First, there was an enrichment of reads in the locus between exons 3 and 4, indicating the existence of an unannotated exon with various lengths across the variants. We refer to this new exon as the distal 3′ exon. Second, several of the variants were generated with the retention of intron 2, which is ~1.9 kb long. Third, some variants, dual-featured transcripts, were also observed to contain both intron 2 retention and the distal 3′ exon. Of the three aforementioned groups in SKHEP1, transcript variants with the distal 3′ exon were the most abundant, followed by variants with intron 2 retention and dual-featured transcripts (Fig. 1E). The PURPL transcripts expressed in HepG2 followed a similar pattern, but no dual-featured transcripts were identified (Fig. 1E). These features and combinations likely give rise to unique, functional RNA architecture that contributes to the overall diversity of transcripts produced from the PURPL locus.

The nucleocytoplasmic distribution of PURPL is transcript variant specific.

We have previously shown that PURPL predominantly localizes in the nucleus in HCT116 cells (16). To determine the localization of PURPL in liver cancer cells, nuclear and cytoplasmic fractions were prepared from SKHEP1 cells, followed by reverse transcriptase quantitative PCR (RT-qPCR) analysis. Each transcript was assayed with specific primers pairs amplifying specific known or novel exonic and intronic sequences (Fig. 2A). Using primers to capture total PURPL transcript levels, we observed that, overall, PURPL was predominantly cytoplasmic, although it was also present in the nuclear fraction (Fig. 2B). Transcript variants with the distal 3′ exon group were found to be enriched in the cytoplasmic fraction. Conversely, transcript variants with intron 2 retention were highly enriched in the nuclear fraction than their spliced counterparts. The subcellular localization patterns of these transcript variants did not change after treatment with Nutlin-3a (Nutlin) (Fig. 2B), a small molecule that induces p53 by inhibiting MDM2, resulting in upregulation of all PURPL transcript variants (Fig. 2C). GAPDH mRNA and MALAT1 served as cytoplasmic and nuclear controls, respectively. We conclude that the majority of PURPL is cytoplasmic in liver cancer cells and different transcript variants show specific subcellular distribution depending on their exonic structure.

FIG 2.

PURPL is a predominately cytoplasmic transcript but is not associated with polysomes in SKHEP1 cells. (A) Schematic showing the RT-qPCR primer positions (black bars) for detecting the indicated PURPL transcript variant types. Total primer pairs capture all variants of PURPL, distal 3′ exon primers capture variants with the distal 3′ exon, exon 2-exon 3 primers capture variants without retention of intron 2, and intron 2 primers capture variants with intron 2 retention. (B) RT-qPCR analysis for PURPL transcript variant types from nuclear and cytoplasmic fractions of SKHEP1 cells, untreated or treated with Nutlin (16 h). GAPDH and MALAT1 served as cytoplasmic and nuclear controls, respectively. (C) RT-qPCR analysis showing the induction of PURPL transcript variant types in SKHEP1 cells after treatment with Nutlin (16 h). GAPDH mRNA served as a loading control. (D) Representative polysome profiles of cytoplasmic lysates from SKHEP1 cells treated with or without puromycin treatment. Lysates were fractionated via a sucrose gradient, and the absorbance peaks corresponding to the 40S, 60S, and 80S ribosomal subunits, as well as polysomes, are indicated. (E) Percentage fraction distribution of total PURPL, p21 mRNA, and GAPDH mRNA in SKHEP1 cells with or without puromycin treatment. Error bars represent SD of 3 experiments in panels B and C. ****, P < 0.0001.

Many lncRNAs may not be truly noncoding. For example, we and others have shown that some lncRNAs contain short open reading frames (sORFs) that encode functional micropeptides or small proteins (37–41). To examine a possible coding potential of PURPL transcripts, we used SQANTI analysis and identified one putative sORF of 100 amino acids common between predicted protein-coding isoforms of PURPL in both SKHEP1 and HepG2 (data not shown). Because transcripts that are actively translated are associated with polysomes, we next assayed for the association of PURPL with polysomes by performing sucrose density gradient fractionation and RT-qPCR analysis from individual fractions with or without puromycin, a known disruptor of polysomes (42). As expected, the addition of puromycin disrupted heavy polysome formation (Fig. 2D). Interestingly, although we observed that PURPL sedimented in fractions containing polysomes, puromycin treatment did not shift its distribution across fractions as it did for p21 and GAPDH mRNAs (Fig. 2E). These data suggest that PURPL transcripts are not translated, but they may form a distinct, uncharacterized, high-molecular-weight complex in the cytoplasm.

PURPL depletion does not affect the levels and activity of p53 in liver cancer cells.

We have previously reported that PURPL is a p53 target that negatively regulates p53 protein levels in HCT116 (16). We first confirmed that PURPL is regulated by p53 in liver cancer cells by depleting p53 expression with small interfering RNA (siRNAs), and we concurrently observed a decrease in PURPL expression (Fig. 3A). We next sought to investigate whether PURPL regulates p53 and its targets in liver cancer cells. Utilizing the CRISPR/Cas9 system, we depleted PURPL expression in SKHEP1 cells with single guide RNAs (sgRNAs) targeting the p53 response element (p53RE) in the promoter region of PURPL as we successfully did before in HCT116 cells (16). After generating cell lines by clonal expansion, we assayed for PURPL RNA depletion and obtained two PURPL knockout (KO) clones (referred to as KO 1 and KO 2). A wild-type (WT) clone served as a negative control. In the WT clone, PURPL was upregulated ~7-fold after Nutlin treatment (Fig. 3B). Contrarily, in both KO clones, PURPL expression was significantly depleted compared to the WT clone in the presence or absence of Nutlin (Fig. 3B). Subsequent RT-qPCR analysis for the expression of three p53 targets (i.e., p21, MDM2, and GDF15) upon PURPL depletion did not show consistent changes in their expression (Fig. 3C), whereas the induction of these targets after Nutlin treatment was comparable between WT and KO clones (Fig. 3D). To examine if PURPL regulates p53 protein levels in liver cancer cells, we assayed for p53 protein expression. Surprisingly, we found that p53 protein levels were equivalent in the PURPL WT and KO clones with or without Nutlin, suggesting a context-dependent function of PURPL on the p53 protein and its targets in CRC cells (Fig. 3E). These findings suggest that PURPL does not regulate p53 protein levels or its transcriptional targets after p53 induction in liver cancer cells and indicate that PURPL could have a context-dependent function.

FIG 3.

PURPL does not regulate the levels or activity of p53 in SKHEP1 cells. (A) RT-qPCR for p53 and PURPL after knockdown of p53 with siRNAs. (B) RT-qPCR for PURPL in SKHEP1 PURPL WT and KO clones with or without Nutlin treatment (16 h). (C) RT-qPCR for p53 targets p21 mRNA (left), MDM2 mRNA (middle), and GDF15 mRNA (right) in SKHEP1 PURPL WT and KO clones. (D) RT-qPCR analysis for p53 targets p21 mRNA (left), MDM2 mRNA (middle), and GDF15 mRNA (right) in SKHEP1 PURPL WT and KO clones with or without Nutlin treatment (16 h). (E) Immunoblot for p53 in SKHEP1 PURPL WT and KO clones with or without Nutlin treatment (16 h). GAPDH served as the loading control. GAPDH mRNA served as a loading control in panels A through D. Error bars represent SD of 2 (A) and 3 (B to D) experiments. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

When generating knockout clones, it is often observed that clones have unique features compared to the cell line from which they were derived (43, 44). To ensure that the phenotypes observed were not due to clonal differences, we therefore utilized CRISPR interference (CRISPRi) to determine the effects of PURPL knockdown on a pool of SKHEP1 cells. To do this, we utilized publicly available HepG2 p53 ChIP-Seq data from ENCODE (45) to design six sgRNAs for dCas9-KRAB targeting at or near the p53RE in the promoter of PURPL in SKHEP1 cells (Fig. 4A). Five of the six sgRNAs were designed near the transcription start site (TSS), and one was designed upstream of the p53 chromatin immunoprecipitation sequencing (ChIP-Seq) peak. The sgRNA that was farthest upstream of the TSS (sg1) did not affect PURPL expression, whereas four of the five remaining sgRNAs substantially depleted PURPL expression (sgPURPL) compared to two nontargeting control sgRNAs (sgNTC) (Fig. 4B). This confirms that targeting the p53RE is necessary and sufficient to deplete PURPL expression. We next assayed for the expression of PURPL and p21 in the CRISPRi cells with or without Nutlin treatment. PURPL expression was not induced in either treatment condition, while p21 mRNA was upregulated after Nutlin treatment consistently between the sgNTC and sgPURPL CRISPRi cells (Fig. 4C). There was also no observable change in p53 protein levels after the loss of PURPL expression compared to the control cells with or without Nutlin treatment (Fig. 4D). These data confirm that PURPL depletion does not affect p53 protein levels or its target p21 mRNA in SKHEP1 cells.

FIG 4.

PURPL promotes clonogenicity of SKHEP1 cells. (A) IGV snapshot of CRISPRi sgRNA sequences used in SKHEP1 cells tiling the PURPL promoter aligned with publicly available ENCODE HepG2 p53 ChIP-Seq data. (B) RT-qPCR for PURPL in SKHEP1 CRISPRi cell lines. ACTB served as a housekeeping gene control. (C) RT-qPCR for PURPL (left) and p21 mRNA (right) in SKHEP1 CRISPRi cell lines with or without Nutlin treatment (6 h). (D) Immunoblot for p53 in SKHEP1 CRISPRi cell lines with or without Nutlin treatment (16 h). GAPDH served as the loading control. (E) Live-cell proliferation assay assessed by Incucyte in SKHEP1 CRISPRi cells in the presence or absence of Nutlin. (F) Average number of colonies per field after soft agar colony formation assay for SKHEP1 CRISPRi cell lines (left) and representative images (right). Scale bar represents 300 μm. GAPDH mRNA served as a loading control in panels B and C. Error bars represent SD of 2 (C, E) and 3 (F) experiments. *, P < 0.05; **, P < 0.01.

PURPL has been shown to promote proliferation and clonogenicity in CRC cells (16). To investigate if PURPL depletion influences proliferation in liver cancer cells, we performed Incucyte live-cell proliferation assays. No difference in cell proliferation was observed after depletion of PURPL either in untreated or in Nutlin-treated cells (Fig. 4E). However, soft agar colony formation assays for anchorage-independent growth revealed a modest but significant decrease in colonies in three out of four sgPURPL CRISPRi cells compared to the sgNTC cells (Fig. 4F), indicating that PURPL could act as a growth-promoting factor in liver cancer cells in three-dimensional culture.

PURPL regulates mitosis in the context of p53 activation.

lncRNAs have been shown to play regulatory roles in multiple aspects of cellular functions, such as gene regulation (20, 21). To gain further insight into the potential impact of PURPL on the transcriptome, we performed RNA-Seq from biological duplicates from two sgNTC and four sgPURPL CRISPRi cells with or without Nutlin treatment (Fig. 5A). PURPL levels were not induced in the presence of Nutlin in sgPURPL CRISPRi cells, but its expression increased significantly in sgNTC cells, as shown from the normalized read measurements (Fig. 5B) and depicted in the IGV snapshot (Fig. 5C) confirming PURPL knockdown. Similar to previous results (Fig. 4C), there was no difference in p21 mRNA levels in sgPURPL cells from the control cells in either treatment group, whereas p21 mRNA levels increased substantially in the presence of Nutlin (Fig. 5B and C). Utilizing DESeq2, we identified differentially expressed genes after depletion of PURPL expression and/or treatment with Nutlin. Although there were only 7 differentially expressed genes (all upregulated) resulting from the loss of PURPL expression in dimethyl sulfoxide (DMSO)-treated cells, there were robust changes in gene expression after Nutlin treatment compared to DMSO treatment in both the sgNTC and sgPURPL CRISPRi cells (Fig. 5D and Table S3). After overlapping the genes that were significantly upregulated or downregulated in the sgNTC or sgPURPL CRISPRi cells after Nutlin treatment, there were 803 differentially expressed genes (adjusted P value [P_adj] < 0.05, |log₂FoldChange| > 0.6) only in the sgPURPL CRISPRi cells (Fig. 5E and F and Table S3). These genes did not show a significant change in sgNTC cells. Out of these 803 genes, 251 were downregulated, and 552 genes were upregulated. Because lncRNAs can act in cis or in trans, we assayed for the enrichment of differentially expressed genes across the genome using chromPlot. There was no observable enrichment of the 803 genes on any one chromosome (Fig. 6A). This was further confirmed by gene set enrichment analysis (GSEA) of these genes against the MSigDB positional gene set (46) (Fig. 6B).

FIG 5.

RNA-Seq identifies PURPL-regulated transcriptome upon p53 activation. (A) Experimental design of SKHEP1 CRISPRi RNA samples for RNA-Seq. (B) Expression of PURPL and p21 mRNA measured as transcripts per million (TPM) in SKHEP1 CRISPRi cells with or without Nutlin treatment (6 h). (C) IGV snapshot of representative SKHEP1 CRISPRi samples treated with DMSO (blue tracks) or Nutlin (red tracks) for 6 h aligned to the PURPL (left) or p21 mRNA (right) locus. (D) Volcano plots of logarithmic fold change (log₂FC) in expression and adjusted P value (P_adj) of differentially expressed genes after RNA-Seq analysis for the indicated comparisons. A significant increase or decrease in gene expression is indicated in red or blue, respectively. Black color represents any gene expression changes with P_adj values adjusted to the smallest value in a 64-bit system. (E) Venn diagram of indicated comparisons after RNA-Seq analysis. Eight hundred three significant gene expression changes were unique to the SKHEP1 sgPURPL CRISPRi cells following Nutlin treatment. (F) Volcano plots of RNA-Seq analysis with calculated log₂FC and P_adj values present in both the sgNTC and sgPURPL CRISPRi cell lines. Gray and blue/red coloring represent gene expression changes in the sgNTC and sgPURPL CRISPRi cell lines, respectively. Error bars in panel B represent SD of 2 independent samples.

FIG 6.

PURPL target genes are not regulated in cis but are enriched for regulation of mitosis. (A and B) Chromosomal mapping with chromPlot (A) and gene set enrichment analysis against MSigDB positional gene sets (B) of 803 differentially expressed genes unique to the SKHEP1 sgPURPL CRISPRi cells with or without Nutlin treatment. (C) Most significantly enriched pathways in the gene set enrichment analysis of upregulated (left) or downregulated (right) genes in sgPURPL CRISPRi cells after Nutlin treatment. Genes without an annotated biotype were filtered out prior to analysis. (D) Heatmap of downregulated genes associated with mitotic spindle and G₂/M checkpoint pathways identified from gene set enrichment analysis in panel C. Gray and blue coloring indicate non- and differentially expressed changes, respectively.

To identify any cellular pathways regulated by PURPL, we performed GSEA analysis for the down- and upregulated genes against the MSigDB hallmark gene sets (46, 47). Expectedly, the upregulated genes showed significant enrichment of genes involved in the p53 pathway, among others (Fig. 6C). On the other hand, the downregulated genes were most significantly enriched in pathways associated with mitosis, such as the assembly of mitotic spindle and the G₂/M checkpoint (Fig. 6C). When comparing the logarithmic fold change of the downregulated genes associated with these two pathways, the decrease in expression of these genes after Nutlin treatment was more robust in the sgPURPL cells than the sgNTC cells (Fig. 6D).

We then investigated the role of PURPL in cell cycle progression and mitosis by examining nuclear morphology for different stages of the cell cycle in sgNTC and sgPURPL CRISPRi cells with or without Nutlin treatment (see Materials and Methods). Our results showed a significant reduction in mitotic population in PURPL-depleted cells compared to control cells (Fig. 7A). As expected, a reduced proportion of mitotic cells was observed in Nutlin-treated cells due to activation of p53 compared to control DMSO-treated cells (Fig. 7A) (48), and this effect was further exacerbated upon depletion of PURPL (Fig. 7A). We next examined which specific stage of the cell cycle was affected upon depletion of PURPL by quantification of cells in interphase, prometaphase, metaphase, anaphase, and cytokinesis with or without Nutlin treatment (Fig. 7B to F). Our results showed that the reduced proportion of mitotic cells (Fig. 7A) correlates with an increase in interphase cells upon treatment with Nutlin in sgPURPL CRISPRi cells (Fig. 7B). Furthermore, the proportions of cells in all stages of mitosis were lower in sgPURPL CRISPRi cells upon Nutlin treatment, indicating a role of PURPL in cell cycle progression (Fig. 7C to F). We also assayed for the levels of phosphorylation of histone H3, known to correlate with mitosis (49). In the DMSO-treated cells, there was no change in the phosphorylation status of histone H3 (Fig. 7G). However, following Nutlin treatment, a substantial decrease in the phosphorylation of histone H3 was observed after PURPL depletion using two independent NTC and two independent PURPL sgRNAs (Fig. 7G). Taken together, our results show that depletion of PURPL in the context of p53 activation affects cell cycle progression, and this contributes to a reduced mitotic population.

FIG 7.

PURPL loss causes changes in mitosis phases. (A to F) sgNTC and sgPURPL SKHEP1 cells were treated with DMSO or Nutlin (16 h) and stained with DAPI to visualize nuclear morphology. Cells were quantified for different stages of the cell cycle as indicated. At least 1,000 cells were counted for each condition. Error bars represent the SD of 4 independent experiments. (G) Immunoblot for phosphorylated histone H3 (S10) in SKHEP1 CRISPRi cells with or without Nutlin treatment (16 h). GAPDH was used as a loading control. *, P < 0.05; **, P < 0.01.

Overall, we observed a subset of genes that are not differentially expressed in the presence of Nutlin but become downregulated when PURPL expression is depleted. This suggests that upon p53 induction following Nutlin treatment, the expression of genes involved in mitosis-related pathways is decreased. At the same time, PURPL levels are elevated to sustain the expression of the aforementioned genes.

DISCUSSION

p53 has numerous targets that mediate its functions, depending on the context. While the protein-coding gene targets of p53 have been extensively studied, the mechanisms by which p53 confers its tumor-suppressive activity through noncoding RNAs have only begun to emerge. We previously identified PURPL as a growth-promoting lncRNA transcriptionally activated by p53 in CRC cells (16). We found that it promotes cell proliferation by reducing the stability of the p53 protein. Because lncRNAs have been shown to have higher variability in expression and function across cell types and tissues than protein-coding genes, we sought to study the function of PURPL in a different context. We chose to study PURPL in a liver cancer model due to its higher expression in these cells than in CRC. We indeed observed a liver cancer-specific function of PURPL. Most unexpectedly, we did not see any effect on p53 protein levels when PURPL was knocked out using a CRISPR/Cas9 approach or after knockdown using the CRISPRi system. Although we were not able to see the same effect on cell proliferation and cell growth as we saw in CRC, in three-dimensional culture, the loss of PURPL correlated with a decrease in the formation of colonies, suggesting that PURPL is a growth-promoting factor in liver cancer cells. This phenomenon has been observed before in the case of the tumor-suppressive lncRNA DRAIC in prostate cancer and glioblastoma cell lines. In both cases, DRAIC overexpression was not able to affect the proliferation and growth of cells attached to plastic, but it did so in soft agar colony formation assays, implying that these two processes may be regulated independently (50, 51).

lncRNAs do not follow a one-size-fits-all model. The cell type in which they are expressed plays a large role in determining their functions (52, 53). This is exemplified by a study where the same CRISPRi system was used across multiple cell lines (27). The vast majority of lncRNAs proffered a growth advantage in only one cell line (27). In contrast, previous studies have shown that most protein-coding genes have similar functions in two or more cell lines, even in cell lines arising from different tissues (54, 55). lncRNAs may also have opposing effects on the proliferation of cells, depending on the type of cancer. For example, the lncRNA MALAT1 was originally identified as a tumor-promoting factor in lung cancer (56), but, surprisingly, it was found to be a tumor suppressor in breast cancer cells (57). Another characteristic example is H19, a lncRNA that has undergone much scrutiny due to ongoing debates as to whether it has a tumorigenic or tumor-suppressive function (58–61). Previous studies have shown that H19 has different mechanisms of action to promote invasion and metastasis, depending on the tissue of origin. In CRC cells, it functions as a competing endogenous RNA (ceRNA) that sponges microRNAs (miRNAs) involved in epithelial-to-mesenchymal transition (EMT) progression (32), while in bladder cancer, H19 interacts with enhancer of zeste homolog 2 (EZH2) to ultimately downregulate E-cadherin, a marker of epithelial cells (33). These studies and our work on the role of PURPL in different cancer types highlight the importance of studying lncRNAs across multiple tissues.

Alternative splicing leads to the formation of a diverse array of RNA products arising from a single locus. Splicing regulation can be cell type-specific and can lead to the generation of lncRNA transcript variants with distinct roles. This is the case for the lncRNA MUNC, where inclusion or exclusion of its intron confers different functions in myogenesis (62). By conducting PacBio Iso-Seq, we found novel classes of transcripts transcribed from the PURPL locus in liver cancer cells, some of which result from alternative splicing of intron 2, an observation which has been reported previously (16, 63). Alternative splicing may occur in cases where splice sites have weak context sequences. Regulation is achieved by cis-acting regulatory elements on the RNA, and by trans-acting factors, such as RBPs, which bind to these elements to block or promote the assembly of the spliceosome (64–66). Further studies are required to parse out the driver(s) of alternative splicing in the PURPL locus. Additionally, we found that these transcripts localize to different compartments of the cell. Retention of intron 2 results in nuclear sequestration, and spliced transcripts with the distal 3′ exon localize to the cytoplasm. Intron retention can be a regulator of the subcellular distribution of many coding or noncoding RNAs, and lncRNA transcripts with long intron retention have been shown to be retained in the nucleus (24, 25). This process may serve as a defense mechanism against inefficiently spliced RNAs that would lead to deleterious peptide products. lncRNAs undergo posttranscriptional processing similarly to protein-coding genes but with inefficient splicing and, thus, higher rates of intron retention (67, 68). Like PURPL, the lncRNA TUG1 has been shown to display a nuclear presence after intron retention, while spliced transcripts localize to the cytoplasm (25).

The spatial regulation and distribution of RNAs are tightly linked to their function. While intron-retaining transcripts of PURPL may perform functions in the nucleus, the transcript variants of PURPL that localize to the cytoplasm are exposed to a different cellular environment, opening the possibility for PURPL to play diverse roles. A recent study showed that PURPL has the ability to affect cell proliferation and the cell cycle as well as suppress p53 protein levels in liver cancer cells, which is contradictory to what we observed in liver cancer cells (69). This discrepancy could be due to the significantly higher efficiency of PURPL knockdown (>90% compared to 70%) that we observed in our experiments using 4 PURPL-independent sgRNAs. Consequently, they could be only knocking down cytoplasmic PURPL transcripts with the siRNA used in their study since siRNAs are known to work much more efficiently in the cytoplasm.

Emerging evidence has revealed that many cytoplasmic lncRNAs containing sORFs are associated with ribosomes and translated to produce micropeptides, or small proteins less than 100 amino acids in length, many of which have important biological functions (37–41). However, several of them are not translated despite their association with polysomes, such as lincRNA-p21 (42) and H19 (70, 71). Because the cytoplasmic variants of PURPL do have a predicted sORF, we assayed for their association with polysomes. We found that while PURPL cosediments with polysomes, we did not see a shift in its distribution profile after puromycin treatment, a known disruptor of translation (42). We thus hypothesize that PURPL may assemble a high-molecular-weight ribonucleoprotein (RNP) complex in the cytoplasm but not undergo translation. One possibility is that this RNP may be the RISC complex, an RNP involved in gene silencing in the cytoplasm. Future studies should investigate potential interactors of PURPL in both nuclear and cytoplasmic contexts.

As mentioned above, lncRNAs show functional specialization connected to cell type specificity (51, 52). In the case of PURPL, we saw several striking differences between the CRC cell line HCT116, and SKHEP1 and HepG2 liver cancer cell lines. We previously found that MYBBP1A and HuR mediate the effect of PURPL on p53 protein levels (16). Based on our RNA-Seq data, expression of the genes encoding MYBBP1A and HuR, which is lower in liver cancer cells (data not shown), may partially explain why there was no difference in p53 upon PURPL knockdown. Moreover, we saw a large array of transcript variants being expressed from the PURPL locus in SKHEP1 and HepG2 cells (Fig. 1D), while there is almost no such diversity observed in HCT116 (our unpublished data). We propose that the observed PURPL transcript heterogeneity in SKHEP1 and HepG2, but not HCT116, could be one of the major reasons why PURPL does not regulate basal p53 levels in liver cancer cells and that it regulates the expression of ~800 genes in the context of p53 activation. PURPL may exert a variety of functions in liver cancer cells depending on the expression and subcellular localization of each transcript variant, while in CRC cells, there is little or no transcript variant diversity. Future studies will determine the mechanism(s) by which these novel PURPL transcripts mediate the effects of PURPL in unstressed cells and in the context of p53 activation.

lncRNAs are known to play fundamental roles in gene regulation (22). Our study provides novel insight into the role of PURPL in fine-tuning gene expression. When PURPL is induced in the presence of Nutlin, a small molecule that upregulates p53, a subset of differentially expressed genes is enriched in pathways associated with mitosis. These genes are known to be involved in multiple processes of mitosis, including the regulation of the stability, shape, and orientation of the mitotic spindle, and include centriolin (CNTRL), centromere protein J (CENPJ), aniline (ANLN), epithelial cell transforming 2 (ECT2), and Rac GTPase-activating protein 1 (RACGAP1) (72–76). Other genes control the progression of the cell cycle by regulating the G₂/M checkpoint, and they include cell division cycle 7 (CDC7), NimA-related kinase 2 (NEK2), and thymopoietin (TMPO) (77–79). These genes, in combination with the rest of the total 803 genes changing in sgPURPL CRISPRi cells after Nutlin treatment, sustain their expression in sgNTC cells but become differentially expressed when PURPL is depleted. An attractive hypothesis for this is that p53 activation leads to inhibition of the expression of these genes, and PURPL, as a direct transcriptional target of p53, acts to resist this inhibition as part of a negative feedback loop. Other examples of negative feedback loops of p53 exist, with the most characteristic negative regulator being MDM2 (80–82). lncRNAs can also regulate p53, as we and others have seen in the case of PURPL in CRC (16) and LINC-ROR in breast cancer cells (83). Mechanistically, it remains to be determined how PURPL regulates these genes, but given its cytoplasmic localization, we speculate that it is happening indirectly through a yet unknown factor. We speculate that PURPL could be interacting with miRNAs destined to target these genes in order to modulate and adjust their availability. Because miRNAs work to fine-tune gene regulation (84), this would explain why the change in expression of these genes is modest but more substantial after knockdown of PURPL, specifically in the context of p53 activation. Another possibility may be that PURPL physically or functionally interacts with an RBP(s) that functions to destabilize these genes. Further studies on the role of PURPL in mRNA decay pathways may elucidate how it could regulate posttranscriptional gene expression in the cytoplasm.

In summary, our discovery of a context-dependent function of PURPL emphasizes the importance of studying lncRNAs across cancer tissue types. Within each cancer type, the transcript variants produced at the locus should be characterized, along with their abundance and subcellular localization. Utilizing pooled populations of cells in addition to single-cell clones is also an important step in improving reproducibility. Overall, we report that in a liver cancer context, PURPL takes a step back from regulating p53 and, instead, plays a role focused on fine-tuning gene expression during p53 activation.

MATERIALS AND METHODS

Cell culture and treatment.

SKHEP1, HepG2, LS174T, HCT116, and 293T cells were maintained in Dulbecco’s modified Eagle medium (DMEM) (Gibco) supplemented with 10% (vol/vol) fetal bovine serum (Gibco) and 1% (vol/vol) Pen-Strep (Gibco) in a 5% CO₂ atmosphere at 37°C. All cell lines were routinely confirmed to be free of mycoplasma by using the Venor GeM mycoplasma detection kit (Millipore Sigma-Aldrich). Cells were treated with Nutlin-3a (Nutlin) (Sigma-Aldrich) at a final concentration of 10 μM or an equal volume of DMSO for the indicated times.

RNA extraction and RT-qPCR.

Total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s protocol and then used for cDNA synthesis using iScript reverse transcription supermix (Bio-Rad). For real-time qPCR, all reactions were carried out on the StepOnePlus real-time PCR system (Applied Biosystems) using FastStart SYBR green master mix (MilliporeSigma). GAPDH mRNA was used to normalize expression, and the relative expression of RNA was calculated using the threshold cycle (2^−ΔΔCT) method. RT-qPCR primer sequences for each gene are indicated as follows: ACTB, TGACCCAGATCATGTTTGAGA and AGGGCATACCCCTCGTAGAT; GAPDH, TGCACCACCAACTGCTTAGC and GGCATGGACTGTGGTCATGAG; MALAT1, GACGGAGGTTGAGATGAAGC and ATTCGGGGCTCTGTAGTCCT; MDM2, TGGTGCTGTAACCACCTCAC and TTTTTGTGCACCAACAGACTT; GDF15, GAGCTGGGAAGATTCGAACA and AGAGATACGCAGGTGCAGGT; p21, GGATTAGGGCTTCCTCTTGG and GACTCTCAGGGTCGAAAACG; p53, GAGCGTGCTTTCCACGAC and GCTCGACGCTAGGATCTGAC; PURPL total, CGTGTGAAAAGAACCCAGGTA and CGCCTGGTAAAACAACCAGT; PURPL distal 3′ exon, GAGCCTGCTCTCTTCAGCTC and TGCTTTTCCTCATCCATGTTT; PURPL Ex2-Ex3, GGTTGTTTTACCAGGCGTTG and TCGATTCCCACATTACCACA; and PURPL intron 2, TGCTTCCCTTTAAGCAGTCAA and GGCCTGTTGGAAAGTGACAT.

Droplet digital PCR analysis.

Total RNA was extracted as described above from a known number of cells counted using the TC10 automated cell counter device (Bio-Rad) with trypan blue (Gibco) staining. The copy numbers of PURPL and GAPDH were calculated as previously described (85). Briefly, we reverse transcribed 1 μg of total RNA, and droplets were generated using EvaGreen supermix (Bio-Rad) on the QX200 AutoDG droplet digital PCR system (Bio-Rad Laboratories); for GAPDH, we used 1 μL of 1.25% (vol/vol) cDNA, whereas for PURPL, we used 3 μL of 12.5% (vol/vol) cDNA and 250 nM primers. The PCR amplification setup was 5 min at 95°C and 30 s at 95°C followed by 60 s at 60°C for 40 cycles, 5 min at 95°C, and then a hold at 4°C. The absolute copy numbers were assessed using QX200 droplet digital PCR system and calculated as previously described (34). ddPCR genes and corresponding primer sequences are indicated as follows: PURPL, CGTGTGAAAAGAACCCAGGTA and CGCCTGGTAAAACAACCAGT; and GAPDH, ACAGTCAGCGCGATCTTCTT and TGGAAGATGGTGATGGGATT. The primers for PURPL target exons 1 and 2, which are common for all the transcripts.

siRNA transfections.

AllStars negative-control siRNA was purchased from Qiagen, and TP53 SMARTpool siRNAs were purchased from Dharmacon (catalog no. L-003329-00-0005). Cells were forward transfected with siRNAs at a final concentration of 20 nM using Lipofectamine RNAiMAX (Invitrogen) in Opti-MEM I reduced-serum medium (Gibco) according to the manufacturer’s protocol. After 72 h, cells were reverse transfected and harvested for RNA extraction.

Generation of PURPL KO and CRISPRi stable cells.

For SKHEP1 PURPL KO, sgRNAs targeting the p53RE were previously designed (16). sgRNAs were cloned into U6-gRNA vector (86) having the BsmB1 restriction enzyme site. CRISPR-mediated PURPL KO SKHEP1 cells were generated using piggyBac cotransposition as previously described (86). Briefly, SKHEP1 cells were cotransfected with 5 plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol as follows: 2 μg of hpT3.5Cagg5-FLAG-hCas9, 2 μg each of the 5′ and 3′ PURPL gRNAs cloned in U6-gRNA vector, 500 ng of pcDNA-pPB7 transposase, and 500 ng of pPBSB-CG-LUC-GFP (Puro) (+CRE) transposon vector. After 48 h, transfected cells were selected with 2 μg/mL puromycin for 72 h. Cells were then reseeded at one cell per well in 96-well plates to obtain clones. PURPL expression was measured by RT-qPCR in individual clones, and deletion of the region around the p53RE of PURPL was confirmed by genomic DNA extraction and PCR amplification subjected to Sanger sequencing.

For CRISPRi, sgRNAs were designed using the Broad Institute CRISPick tool (https://portals.broadinstitute.org/gppx/crispick/public), and specificity was checked with GGGenome Ultrafast sequence search (https://gggenome.dbcls.jp/hg38/). For the generation of stable cells, each sgRNA was cloned into pLV hU6-sgRNA hUbC-dCas9-KRAB-T2a-Puro (Addgene; catalog no. 71236) with BsmBI-v2 sites. Oligonucleotides that were annealed and used for cloning are as follows: PURPL sg1, CACCGCTAACTTCTTATTGCCTACAAGG and AAACCCTTGTAGGCAATAAGAAGTTAGC; PURPL sg2, CACCGGGGCATGCCCAGACAAGCCC and AAACGGGCTTGTCTGGGCATGCCCC; PURPL sg3, CACCGCTGGAGCCTGACATGCTCACTGG and AAACCCAGTGAGCATGTCAGGCTCCAGC; PURPL sg4, CACCGGGAATTCATGCCTTGCAAAG and AAACCTTTGCAAGGCATGAATTCCC; PURPL sg5, CACCGTTGAGCTCTTGTGGTGACCT and AAACAGGTCACCACAAGAGCTCAAC; and PURPL sg6, CACCGGACCTGGGAATCAATGTGTG and AAACCACACATTGATTCCCAGGTCC. Successfully cloning was confirmed via Sanger sequencing using the hU6-F primer. Vectors containing the cloned NTC sg1 and sg2 sgRNAs (CCGGCGCCGAGCCGGACTTCG and AGTCGCTTCTCGATTATGGG, respectively) were a generous gift from Javed Khan, Genetics Branch, CCR, NCI, NIH (Bethesda, MD, USA).

For lentivirus particle packaging, 293T cells were seeded at 2.0 × 10⁵ cells/well onto 6-well plates, and 1,200 ng of pLV hU6-sgRNA hUbC-dCas9-KRAB-T2a-Puro was transfected with lentiviral package vectors using Lipofectamine 2000 (Life Technologies, Invitrogen) according to the manufacturer’s protocol. The medium containing packaged viral particles was collected at 48, 56, and 72 h posttransfection. A viral titer was determined using the serial dilution method. A multiplicity of infection (MOI) equal to 1 was used for transducing SKHEP1 cells, and stably transduced cells were generated via 2 μg/mL puromycin selection.

Immunoblotting.

The total cell lysate was prepared using radioimmunoprecipitation assay (RIPA) buffer (Thermo Scientific) containing protease inhibitor cocktail (Roche). Protein concentration was determined using the bicinchoninic acid (BCA) protein quantitation kit (Thermo Scientific). Ten micrograms of whole-cell lysate was loaded per lane onto a 10% or 12% SDS-PAGE gel and transferred to either a nitrocellulose (Thermo Scientific) or polyvinylidene difluoride (PVDF) membrane (Thermo Scientific) using a semidry apparatus (Bio-Rad). The membrane was blocked in 5% skim milk (Oxoid) in Tris-buffered saline (TBS) containing 0.05% Tween. The following antibodies were used: anti-p53 (DO-1) (1:3,000; mouse) (Santa Cruz), anti-GAPDH (1:10,000; rabbit) (CST), anti-phospho-histone H3 (S10) (1:3,000; rabbit) (CST), and secondary anti-rabbit (1:5,000) (CST) and anti-mouse (1:5,000) (CST).

Polysome fractionation.

SKHEP1 cells were grown in a 10-cm dish to confluence and then incubated for 10 min with 100 μg/mL cycloheximide (Sigma). For puromycin treatment, 0.5 mM puromycin was added to the media 1 h before harvest. Cytoplasmic lysates were prepared using polysome extraction buffer (100 mM KCl, 5 mM MgCL₂, 0.5% NP-40, 100 μg/mL cycloheximide, 2 mM dichlorodiphenyltrichloroethane [DDT], 40 U/mL RNase Out [Invitrogen], and 1× complete protease inhibitor cocktail) and then fractionated by ultracentrifugation through a 10% to 50% linear sucrose gradient in an SW 40 Ti swinging bucket rotor (Beckman Coulter) at 37,000 rpm at 4°C for 2 h. Twelve fractions were obtained via a fraction collector (BioComp Instruments) while being monitored by optical density measurement at 260 nm. Following fractionation, 250 μL from each fraction was used to extract RNA using TRIzol LS (Thermo Fisher), and RNA was used for RT-qPCR analysis as described above.

Soft agar colony formation and microscopy.

Six-well plates were prepared with a layer of 0.5% low-melt agarose (FMC BioProducts) in prewarmed full medium diluted from 2% low-melt agarose in 1× phosphate-buffered saline (PBS). Cells were seeded at a density of 5.0 × 10⁴ cells per well in 0.37% low-melt agarose in prewarmed full medium on top of the 0.5% solidified layer. Plates were chilled at 4°C for 20 min until they were solidified and then grown for at least a week until they were imaged. Phase images were collected with an Invitrogen EVOS M5000, and the average number of colonies from 9 randomized field images was counted from 3 independent experiments.

Proliferation assay.

Cells were plated at low confluence in 96-well plates. After the cells had settled overnight, Nutlin was added at a final concentration of 10 μM, live-cell proliferation assays were performed using the Incucyte system (Essen BioScience) system, and images were captured at the indicated times. Confluence was measured using the Incucyte base analysis software.

PacBio isoform sequencing.

Total RNA from SKHEP1 and HepG2 cells was isolated using the RNeasy Plus kit (Qiagen). We used 300 ng of RNA with RNA integrity number of 10 to prepare libraries for sequencing following the Iso-Seq Express template preparation protocol (Pacific Biosciences, CA, USA) selecting for transcripts <2 kb. Sequencing primer v4 was annealed, and Sequel II polymerase 2.1 was bound to libraries prior to loading each on one 8M SMRT cell on the Sequel II system by using diffusion loading. Sequencing was performed with a 2-h preextension and a 24-h movie. PacBio SMRTlink (smrtlink-release_9.0.0.92188)-generated raw subreads were converted into HiFi circular consensus sequences (CCSs) with a minimum of 3 passes. The PacBio IsoSeq v3 pipeline was used to demultiplex the barcodes and remove primers. Additional refining steps included trimming poly(A) tails and removing concatemers to generate full-length nonconcatemer (FLNC) reads. The FLNC reads were used to map to human reference (hg38) using minimap2 software to generate alignment .bam files and coverage tracks. In addition, hierarchical clustering, iterative clustering, and merging steps were performed to obtain consensus isoforms and full-length (FL) consensus sequences. The high-quality FL transcripts were mapped to the reference genome by using the minimap2 software. Isoform classification and quality control were done using SQANTI3, and Illumina short-read gene expression data were combined with PacBio Iso-Seq for transcript quantification.

RNA-Seq.

The same SKHEP1 and HepG2 RNA samples that underwent Iso-Seq were also subject to RNA-Seq. The samples were sequenced on NextSeq using Illumina TruSeq stranded mRNA library prep and paired-end sequencing. Sample reads were trimmed for adapters and low-quality bases using Cutadapt (87) (v1.18) before alignment with the hg19 reference genome and annotated transcripts using the STAR aligner (88) (v2.7.0f) using the Gencode v19 gene annotation.

For the SKHEP1 CRISPRi cell lines, RNA-Seq samples were performed in biological duplicate using two independent NTCs and four independent PURPL sgRNAs treated with Nutlin (10 μM final concentration) or DMSO for 6 h. Total RNA was isolated using the RNeasy Plus kit (Qiagen). Samples were sequenced on NovaSeq 6000 S1 using TruSeq stranded mRNA prep and paired-end sequencing. Raw reads were first trimmed with Cutadapt (87) (v1.18) followed by alignment to the hg38 human genome using the STAR aligner (88) (v2.7.0f) using the Gencode v30 gene annotation. Raw reads were quantified using the RSEM (89) package (v1.3.1). Differential gene expression analysis was calculated using DESeq2 (90) (v1.32.0) in R (v4.1.1) using the default settings, including the results from the 2 sgNTC and the 4 sgPURPL CRISPRi cell lines and from both biological replicates. Normalized RNA-seq coverage tracks were created with BAMscale (91) (v0.2) using the scale factors obtained from DESeq2 sizeFactors function.

TCGA gene expression data.

TCGA cancer gene expression data were accessed via the publicly available MiPanda resource (92) querying for LINC01021.

Quantification and statistical analysis.

Data were plotted in GraphPad Prism (v8) unless otherwise indicated. In all figures, error bars represent standard deviation, and statistical analysis was performed using Student’s t test. Using R software (v4.0.3), the Venn diagram was plotted using the VennDiagram library (93), the chromPlot was plotted using the chromPlot library (94), and volcano plots were graphed using the ggplot2 library. When plotting the volcano plots, the DESeq2-adjusted P values with 0 value were converted to the smallest value in a 64-bit system for plotting. This paper does not report original code.

Gene set enrichment analysis.

GSEA using MSigDB hallmark gene sets was performed on the website of the Broad Institute (http://www.gsea-msigdb.org/gsea/msigdb/annotate.jsp). The genes that were used as input were filtered for unannotated genes. For MSigDB positional gene sets, 803 genes were used as input in the GSEA software (v4.2.3) from the Broad Institute (46, 47).

Nucleocytoplasmic fractionation.

SKHEP1 cells were grown to 80% confluence. Cells were then washed twice with 1× PBS and harvested via trypsinization. Cell viability was checked with trypan blue. Cell pellets were resuspended in 500 μL RSB buffer (10 mM Tris-HCl [pH 7.4], 100 mM NaCl, and 2.5 mM MgCl₂) with 40 μg/mL digitonin (Invitrogen), 80 U of RNase Out, and 1× protease inhibitor cocktail, mixed by inversion, and kept on ice for 10 min before confirming cell viability (<10%) with trypan blue. Cells were centrifuged at 1,500 × g for 5 min at 4°C. The cytoplasmic supernatant was collected, and one-fifth of the supernatant was separated for RNA isolation as described above. The nuclear pellet was resuspended in RSB with the same volume as one-fifth of the supernatant before RNA isolation as described above. The RNA pellets were resuspended in equivalent volumes of water, and the same volume of RNA was used for cDNA synthesis. The fact that one-fifth of the cytoplasmic fraction was used in RNA extraction was eventually considered when the relative expression of RNA was calculated using the 2^−ΔΔCT method.

Cell cycle analysis.

SKHEP1 cells were grown on glass coverslips to 40 to 50% confluence and treated overnight with10 μM Nutlin-3a before fixing with ice-cold methanol for 1 min followed by three washes, 5 min each with PBS. The cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) for 45 min at room temperature, followed by washes, and mounted on slides by using Prolong Gold antifade mounting medium (Thermo Fisher Scientific). For microscopy and image analysis, cells were imaged on Delta Vision Core system (Applied Precision/GE Healthcare) with an Olympus IX70 inverted microscope (Olympus America) using 100× (numerical aperture [NA], 1.4) oil immersion objective and a CoolSnap HQ 12-bit camera (Photometrics) controlled by Softworx software. The filter used for imaging was DAPI (excitation, 360/40 nm; emission, 457/50 nm) with exposure of 100% transmission for 0.1 s. At least 1,000 cells were counted for each experiment manually for different stages of the cell cycle. Cells with intact nuclei were considered interphase; cells with condensed but uncongressed chromosomes were considered prometaphase. Cells that had all chromosomes congressed on the equatorial plate were considered metaphase. Cells that displayed chromosome segregation were considered anaphase, and cells with segregated chromosomes with cytokinetic furrows were considered cytokinesis.

Data availability.

The Iso-Seq and RNA-Seq data have been deposited at GEO under accession no. GSE206702.