lncRNA RMST Enhances DNMT3 Expression through Interaction with HuR.

Large bodies of studies have shown that the CRISPR/Cas9-based library screening is a very powerful tool for the identification of gene functions. However, most of these studies have focused on protein-coding genes, and, furthermore, very few studies have used gene reporters for screening. In the present study, we generated DNA methyltransferase 3B (DNMT3B) reporter and screened a CRISPR/Cas9 synergistic activation mediator (SAM) library against a focused group of lncRNAs. With this screening approach, we identified Rhabdomyosarcoma 2-Associated Transcript (RMST) as a positive regulator for DNMT3B. This was confirmed by activation of the endogenous RMST by SAM or ectopic expression of RMST. Moreover, RMST knockout (KO) suppresses DNMT3, while rescue with RMST in the KO cells restores the DNMT3 level. Finally, RMST KO suppresses global DNA methylation, leading to the upregulation of methylation-regulated genes. Mechanistically, RMST promotes the interaction between the RNA-binding protein HuR and DNMT3B 3' UTR, increasing the DNMT3B stability. Together, these results not only provide the feasibility of a reporter system for CRISPR library screening but also demonstrate the previously uncharacterized factor RMST as an important player in the modulation of DNA methylation.

Introduction

The development of cancer often involves the aberrant expression of multiple genes due to genetic or epigenetic alterations. In this regard, DNA methylation plays an important role in gene expression. In mammals, DNA methylation patterns are written and regulated mainly by canonical DNA methyltransferases (DNMTs), including DNMT1 and DNMT3, including DNMT3A and DNMT3B. DNMT1 is required for maintaining patterns of DNA methylation, whereas both DNMT3A and DNMT3B serve as de novo methyltransferases, and they are often required to establish genomic methylation during embryogenesis and development.2, 3, 4

DNMTs are often dysregulated in cancer. For instance, DNMT3B is upregulated in colon, prostate, and breast cancers.5, 6, 7 Several factors have been shown to be important to the regulation of DNMT3B, and among them is a dynamic interplay through the 3′ UTR. For instance, HuR is a well-known stabilizing factor that can interact with its target mRNAs through an RNA recognition motif at the 3′ UTR, and modify their expression by altering their stability. In this case, the binding of HuR to DNMT3B mRNA increases the expression of DNMT3B in cancer cells by enhancing its stability, which in turn affects both global DNA methylation and DNMT3B-specific target DNA methylation levels. Similarly, a large number of microRNAs can also interact with DNMT3 at the 3′ UTR and silence its expression. However, little is known about whether long non-coding RNAs (lncRNAs) are involved in the regulation of DNMT3, leading to alterations in global methylation.

Therefore, we asked whether lncRNAs can regulate DNMT3 expression through the 3′ UTR. Toward this direction, we developed a DNMT3B reporter system for CRISPR/Cas9 synergistic activation mediator (SAM) guide RNA (gRNA) library against a focus group of lncRNAs. We showed that lncRNA Rhabdomyosarcoma 2-Associated Transcript (RMST) is capable of upregulating DNMT3B expression by interaction with the stabilizing factor HuR.

Results

Identification of RMST as a Positive Regulator for DNMT3B

Overwhelming numbers of lncRNAs have been identified, however, little is known about their function for the vast majority of them. To determine whether lncRNAs are involved in the regulation of DNMT3B expression, we adopted a CRISPR-based library screening that has been shown to be a powerful research tool for the identification of novel gene function.14, 15 It is well known that, in addition to gene knockout (KO), CRISPR technology can also be used as a tool for gene activation or knockdown. We used a SAM-based lncRNA library screen system for the identification of lncRNAs involved in the regulation of DNMT3B expression. The SAM system is a gene activation tool for the transcriptional activation of endogenous genes. It is capable of robust activation of coding genes as well as lncRNAs by a combination with gRNA library for gain-of-function screening.

It is well known that lncRNAs can regulate gene expression through various mechanisms at the transcriptional and posttranscriptional levels. In particular, emerging evidence indicates that lncRNAs can regulate mRNA stability through interaction with various RNA-binding proteins. For example, lncRNA OCC-1 can interact with HuR and recruit ubiquitin E3 ligase β-TrCP1 to HuR, such that HuR is downregulated by destabilization. Linc-RoR interacts with heterogeneous nuclear ribonucleoprotein (hnRNP) I (stabilizing factor) and AUF1 (destabilizing factor), respectively, with an opposite consequence to their interaction with c-Myc mRNA. FIRRE physically interacts with hnRNP U, regulating the stability of mRNAs of selected inflammatory genes through targeting the adenine and uridine (AU)-rich elements of their mRNAs in cells following lipopolysaccharide (LPS) stimulation.

Therefore, we asked whether we can identify this type of lncRNAs through a 3′ UTR reporter from a CRISPR-based SAM gRNA library. Thus, we generated a DNMT3B 3′ UTR reporter based on pGL3-control vector (Promega). In this case, the luciferase sequence in pGL3-control vector was replaced by puromycin gene (Pu) and then DNMT3B 3′ UTR was cloned into the downstream of Pu. In this way, DNMT3B 3′ UTR stability can control Pu expression. Cells carrying this reporter became resistant to puromycin when DNMT3B 3′ UTR level (i.e., Pu level) was increased due to the presence of candidate gRNAs (Figure 1A). On the other hand, cells carrying vector control (i.e., no gRNA) or non-related gRNAs were sensitive to puromycin.

Figure 1

Identification of RMST as a Candidate lncRNA Capable of Promoting DNMT3B mRNA Stability from the SAM Library Screen

(A) Schematic description of principle for the DNMT3B 3′ UTR reporter used to screen the SAM library. SAM gRNAs that can stabilize DNMT3B mRNA stability are enriched by puromycin selection. (B) Detection of DNMT1 and DNMT3A/3B mRNA levels for cells before selection (BS) and cells after selection (AS) by qRT-PCR. (C) Detection of DNMT1 and DNMT3A/3B protein levels for BS and AS groups by western blot. (D) SAM gRNAs activate the endogenous RSMT. MCF-7 cells were first infected with lenti MS2-P65-HSF1_Hygro and lenti dCas9-VP64_Blast, and then a pool of 5 RMST-specific gRNAs was introduced into these cells by infection. The cells were harvested 4 days after infection and RMST RNA level was detected by qRT-PCR. (E) RMST SAM gRNA (top) or ectopic expression of RMST (bottom) confers resistance to puromycin. For RMST SAM gRNA experiments, the test MCF-7 cells carried all components plus the reporter; for ectopic expression of RMST, the test MCF-7 cells carried RMST expression vector plus the reporter. Data are presented as mean ± SEM (n = 3). **p < 0.01.

The overall procedure for this screening was as follows. First, we introduced dCas9-VP64 and pMS2-p65-HSF1 into MCF-7 cells sequentially by infection and established stable transductants. Next, we introduced the SAM library into the transductants by infection. Finally, we introduced the DNMT3B 3′ UTR reporter into the cells carrying all these components by transfection, and then we treated them with puromycin for 4 days, followed by further culture for 10 days with puromycin. As a control, we also introduced the reporter into the host cells carrying the vector control instead of the library.

The 1.5-μg/mL concentration was determined to be optimal to see the difference between vector control and reporter group in terms of cell survival. For example, the vector control cells barely survived under 1.5 μg/mL puromycin. By contrast, a number of colonies were formed in the reporter group, suggesting that the selection procedure is successful (data not shown). Therefore, we pooled the cells from the reporter group before selection (BS) and after selection (AS), respectively, for RNA and protein extraction. We detected an increase in DNMT3B expression at both the mRNA and protein levels in the AS group cells as compared with the BS group (Figures 1B and 1C). Of interest, in addition to DNMT3B, we also detected the upregulation of DNMT3A (both mRNA and protein), but no difference was seen for DNMT1 (Figures 1B and 1C).

To identify which lncRNAs are upregulated by SAM gRNA, we performed qRT-PCR profiling against these 421 lncRNAs, and we found that this selection led to over a 4-fold enrichment of 3 lncRNAs (H19, NPPA-AS1, and RMST). However, only RMST gRNAs were able to consistently enhance the expression of DNMT3B (Figure 1D). Moreover, this SAM-induced upregulation of RMST increased the resistance to puromycin in the testing cells that carried the DNMT3B 3′ UTR reporter as compared to the vector control (Figure 1E, upper panel). Similarly, ectopic expression of RMST in the parental MCF-7 cells also resulted in a high number of colonies as compared to the vector control (Figure 1E, bottom panel). Therefore, we focused on RMST for further characterization.

RMST Is a Positive Regulator for DNMT3 Expression

To further define the role of RMST in the regulation of DNMT3 (3A and 3B), we established stable RMST KO cells by the CRISPR/Cas9 approach in three different cancer cell lines, i.e., breast cancer MCF-7, pancreatic cancer MIA PaCa-2, and colon cancer HCT-116 (Figures 2A, 2D, and 2G). It was evident that both DNMT3A and DNMT3B mRNA levels were decreased in all three cell lines after RMST KO (Figures 2B, 2E, and 2H). Consistent with the mRNA results, DNMT3A and DNMT3B protein levels were also decreased in these KO cells (Figures 2C, 2F, and 2I).

Figure 2

RMST KO Suppresses DNMT3 Expression

(A, D, and G) Detection of RMST expression in KO clones by qRT-PCR. RMST knockout (KO) was generated by the CRISPR/Cas9 approach in MCF-7 (A–C), MIA PaCa-2 (D–F), and HCT-116 (G–I) cells; two KO clones of each cell line were identified by qRT-PCR. (B, E, and H) Detection of DNMT3A and DNMT 3B mRNA in RMST KO of MCF-7, MIA PaCa-2, and HCT-116 cells, respectively, as determined by qRT-PCR. (C, F, and I) RMST KO causes a decrease in DNMT3A and DNMT 3B protein levels, as determined by western blot. Data are presented as mean ± SEM (n = 3). **p < 0.01.

To better characterize the role of RMST in the regulation of DNMT3, we performed rescue experiments, i.e., re-expression of RMST in the KO cells (Figures 3A, 3D, and 3G). In all KO clones, re-expression of RMST was able to enhance the DNMT3A and DNMT3B levels at the mRNA (Figures 3B, 3E, and 3H) and protein levels (Figures 3C, 3F, and 3I), providing further evidence that RMST is able to enhance DNMT3 expression.

Figure 3

Re-expression of RMST Upregulates DNMT3

(A, D, and G) Detection of RMST in the rescued KO cells by qRT-PCR. Lentiviral expression vector carrying RMST and empty vector was used to infect RMST KO cells: at 2 days after infection, the cells were harvested for RNA extraction. The following clones were used in this experiment: MCF-7 KO 8 (A–C), PaCa-2 KO 1 (D–F), and HCT-116 KO 8 (G–H). (B, E, and H) Detection of DNMT3A and DNMT3B mRNA in the rescued cells by qRT-PCR. (C, F, and I) Detection of protein levels in the rescued cells by western blot. Data are presented as mean ± SEM (n = 3). **p < 0.01.

RMST KO Causes a Decrease in Global DNA Methylation and Reactivation of Tumor Suppressor Genes

It is known that a reduction of DNMT activity induces hypomethylation of the global DNA. Thus, we next asked whether the loss of RMST can also functionally result in DNA hypomethylation. Genomic DNA was extracted from the RMST KO cells, and global DNA methylation was measured using a methylated DNA 5-mC quantification kit. We observed an ∼40% reduction in global DNA methylation for KO cells lines compared with the vector controls (Figures 4A–4C). It is known that the hypermethylation can silence gene expression whereas the silenced genes can be reactivated by demethylating agents. To further confirm that RMST KO can reactivate these genes, we chose three tumor suppressor genes (XEDAR, CDH13,26, 27 and miR-34a) because they have been shown to be silenced by hypermethylation. We tested XEDAR in MCF-7 RMST KO cells (Figure 4D), and we tested CDH13 in HCT116 RMST KO cells (Figure 4E). Both XEDAR and CDH13 were increased in these RMST KO cells compared with the vector control. For miR-34a, we tested all of these three cell lines, and we found that miR-34a was increased by RMST KO in all the cases (Figure 4F), suggesting that methylation-regulated genes can be activated by RSMT KO.

Figure 4

RMST Is Required for Global DNA Methylation and Methylation-Mediated Gene Expression

(A–C) Knockout of RMST results in a decrease in global DNA methylation. DNA methylation status of vector control and RMST KO for MCF-7 (A), PaCa-2 (B), and HCT-116 (C) cell lines were measured by colorimetric MethylFlash DNA Methylation Quantification Kit per the manufacturer’s instruction. (D–F) RMST KO upregulates the expression of XEADR (D), CDH13 (E) and miR-34a (F), as determined by qRT-PCR. Note that the relative expression of miR-34a in the RMST KO group represents the mean level of miR-34a in two different KO clones. Data are presented as mean ± SEM (n = 3). *p < 0.05; **p < 0.01.

RMST Upregulates DNMT3 by Increasing mRNA Stability

Since DNMT3 protein is subject to lysosomal degradation, we examined the protein stability of DNMT3A and DNMT3B for vector control and RMST overexpression. After treatment with protein synthesis inhibitor cycloheximide (CHX) at 20 μg/mL, protein extracts were prepared at various time points. No difference was seen for the protein half-lives of DNMT3A and DNMT3B between vector control and RMST overexpression (Figures 5A and 5B).

Figure 5

RMST Increases DNMT3 mRNA Stability

(A and B) RMST has no effect on DNMT3 protein stability. MCF-7 cells were infected with vector control and RMST expression vector, and then treated with CHX at 20 μg/mL for the indicated time points. The cells were harvested for protein extraction and western blot (A) and quantitive analysis (B). (C and D) RMST stabilizes DNMT3 mRNA. MCF-7 cells were infected with vector control and RMST expression vector as above. The infected cells were treated with actinomycin D (2 μg/mL) for 0, 2, 4, 6, and 8 h. RNA was extracted from these cells to determine the level of DNMT3A (C) or DNMT3B mRNA (D). Data are presented as mean ± SEM (n = 3). *p < 0.05.

Next, we determined whether RMST affects the stability of DNMT3A or DNMT3B mRNA. Using the RNA synthesis inhibitor actinomycin D, we were able to show that RMST increased the stability for both DNMT3A and DNMT3B mRNA. For example, the DNMT3A mRNA half-life was ∼8 h for RMST overexpression cells, as compared to ∼4.8 h for vector control (Figure 5C); similarly, the half-life of DNMT3B mRNA for RMST overexpression cells was ∼6.4 h, as compared to ∼3.8 h for vector control (Figure 5D).

RMST Interacts with HuR to Regulate the mRNA Stability of DNMT3

To determine the possible mechanism of RMST-mediated upregulation of DNMT3, we focused on RNA-binding proteins (RBPs), such as HuR and AUF1, because these factors are known to interact with many AU-rich target mRNAs and affect target mRNA stability. In particular, it has been shown that HuR binding can stabilize the DNMT3B mRNA. As a destabilizing factor, AUF1 can also interact with DNMT3B mRNA. Therefore, we asked whether RMST can regulate the RBP-mediated DNMT3B stability through these two factors.

We first determined the effect of RMST on the level of HuR or AUF1, and we found that there was no difference in HuR or AUF1 between vector control and RMST KO (Figure 6A). We then performed RNA precipitation with the biotin-labeled DNMT3B 3′ UTR probe, and we confirmed that DNMT3B mRNA can interact with HuR and AUF1 in MCF-7 cells (Figure 6B, top). By contrast, the DNMT3B 5′ UTR probe was not able to interact with HuR or AUF1 (Figure 6B, top). Of interest, RNA precipitation using biotin-labeled RMST as a probe showed that both HuR and AUF1 interacted with RMST (Figure 6B, bottom).

Figure 6

RMST Stabilizes DNMT3 mRNA by Interaction with HuR

(A) RMST KO has no effect on the protein levels of HuR and AUF1. The HuR and AUF1 protein levels after RMST KO compared to vector control were detected by western blot. (B) HuR and AUF1 interact with DNMT3B RNA and RMST. RNA pull-down assay was performed in HCT-116 cells using biotin-labeled DNMT3B RNA as a probe (top) or RMST RNA probe (bottom). DNMT3B 5′ UTR RNA probe was used as a negative control. HuR or AUF1 protein level was detected by western blots. (C) UV cross-linked RIP assay with anti-HuR antibody indicates that HuR interacts with RMST, DNMT3A, and DNMT3B RNA. The levels of RMST, DNMT3A, and DNMT3B mRNA were determined by qRT-PCR. (D) AUF1 also interacts with RMST, DNMT3A, and DNMT3B mRNA. The same procedure was used as in (C) except for AUF1 antibody. (E) RMST KO impairs the binding of HuR to DNMT3B 3′ UTR. RNA precipitation assay was performed in vector control (LCV2-m) or RMST KO cells, using the DNMT3B RNA probe followed by western blot with HuR antibody. (F) RMST KO has no effect on AUF1 binding to DNMT3B RNA. Detection of AUF1 after RNA precipitation using the DNMT3B RNA probe in vector control (LCV2-m) or RMST KO cells. (G) Re-expression of RMST in KO cells increases the binding of HuR to DNMT3B 3′ UTR. HCT-116 KO8 cells were infected with lentiviral expression vector carrying RMST or empty vector. RNA precipitation assay was performed as in (E).

Next, RNA immunoprecipitation (RIP) with HuR antibody revealed that HuR antibody was able to enrich RMST over 10-fold as compared to immunoglobulin G (IgG) control (Figure 6C). AUF1 antibody was also able to pull down RMST, but to a slightly less extent (Figure 6D). In addition, HuR and AUF1 interacted with DNMT3A mRNA (Figures 6B and 6C). No significant enrichment was detected for the interaction of DNMT1 with HuR or AUF1 (data not shown). Importantly, we were able to show that RMST KO reduced the interaction between DNMT3B mRNA and HuR (Figure 6E, red arrows). However, we detected no difference between vector control and RMST KO for AUF1 (Figure 6F). Finally, re-expression of RMST in the KO cells enhanced the interaction between DNMT3B mRNA and HuR (Figure 6G, purple arrows). These results suggest that RMST promotes the mRNA stability of DNMT3 by enhancing the interaction between DNMT3 and HuR.

Discussion

DNA methylation is an important part of the epigenetic regulation system for gene expression. A well-known example is CpG island methylation, which is primarily attributed to a group of DNMTs, including DNMT1 and DNMT3A/B. Since DNA methylation in the promoter region often suppresses gene expression, it controls large numbers of genes. For example, DNA methylation in cancers may reveal a very different pattern from that of normal tissue, presumably due to the dysregulation of these DNMTs. However, it is largely unknown as to how these DNMT genes are regulated in cancer. The present study suggests that RMST enhances DNMT3 expression by stabilizing its mRNA through interaction with HuR.

The supporting evidence comes from the following: first, RMST-specific SAM gRNAs are enriched after selection against the library using the DNMT3B 3′ UTR reporter; second, activation of the endogenous RMST by SAM gRNAs or ectopic expression of RMST enhances the expression of DNMT3B; in addition, we found that it is also capable of enhancing DNMT3A; third, in multiple cell lines, RMST KO reduces DNMT3 at the mRNA and protein levels, whereas re-expression of RMST is able to increase the DNMT3 level; lastly, RIP assays using HuR antibody as well as RNA precipitation assays using RMST RNA probe demonstrate that RMST interacts with HuR, through which the interaction between HuR and the DNMT3 3′ UTR is enhanced.

The 3′ UTR is an important regulatory element that can influence polyadenylation, efficiency, localization, and stability of the mRNA. In this regard, several factors have been shown to interact with the 3′ UTR to regulate mRNA levels, because it carries binding sites for regulatory proteins and microRNAs. For example, by binding to specific sites within the 3′ UTR, microRNAs can silence various mRNAs by either inhibiting translation or directly causing degradation of the transcript. Furthermore, many 3′ UTRs also contain AU-rich elements (AREs). Proteins that can bind AREs often affect the stability or decay rate of transcripts in a localized manner or affect translation initiation. Therefore, the 3′ UTR reporters, such as those luciferase-3′ UTR reporters, can serve as an important research tool. They have been extensively used in microRNA studies, because the interaction between microRNA and 3′ UTR of a particular gene can often lead to downregulation of this gene. However, very few studies have used this type of reporter as we reported here to identify gene function, in particular for lncRNAs involved in RBPs and the 3′ UTR. Thus, our study provides a new utility of the 3′ UTR reporters.

Regulation of DNMT3 is complex and can occur at transcriptional and posttranscriptional levels. At the mRNA level, DNMT3 is subject to regulation by RBPs and microRNAs. HuR is a member of the RBPs that bind AREs commonly found in the 3′ UTR. Many mRNA species carry AREs that can be targeted by HuR. However, HuR-mediated regulation may also involve other factors that may play a fine-tuning role. Our study suggests that RMST is one of such factors that promote their interaction. On the other hand, competition for RBPs between lncRNAs and target genes could serve as an important mechanism for lncRNA-mediated mRNA stability. For instance, Linc-RoR interacts with hnRNP I (a stabilizing factor) and AUF1 (destabilizing factor), respectively, with an opposite consequence to their interaction with c-Myc mRNA. Interaction of Linc-RoR with hnRNP I inhibits AUF1 to bind to c-Myc mRNA such that the c-Myc mRNA half-life is increased. With regard to the regulation of DNMT3, although RMST interacts with both HuR and AUF1, impacting the binding of HuR to DNMT3B 3′ UTR, there is no effect of RMST KO on the binding of AUF1 to the DNMT3B 3′ UTR. However, it remains to be determined yet as to whether this might occur under different conditions.

Regarding microRNA regulation, DNMT3B has been shown to be a target gene directly regulated by miR-29c-3p, as shown by luciferase reporter and western blotting assays in hepatocellular carcinoma. On the other hand, miR-29a and miR-29b are capable of suppressing both DNMT3A and DNMT3B levels in an HNF4α-dependent manner in hepatocytes. Similarly, both miR-200b and miR-200c can directly target DNMT3A and DNMT3B in gastric cancer. Thus, it would be interesting to determine whether our reporter can be used for the identification of additional microRNAs involved in the regulation of DNMT3 through its 3′ UTR.

RMST was initially identified as a potential functional non-coding RNA in rhabdomyosarcoma, and its expression is higher in the alveolar subtype compared to the embryonal subtype. Two subsequent reports demonstrated that the higher expression level of RMST is critical to mouse brain development.37, 38 More recently, a study suggests that RMST functionally interacts with SOX2 to regulate human neurogenesis, suggesting a vital role of RMST in both mouse and human brain development. Thus, the identification of RMST as an important factor for DNA methylation through the regulation of DNMT3 expands the repertoire of RMST targets.

In summary, we adopted the DNMT3B 3′ UTR reporter combined with the CRISPR library, and we identified RMST as an important player in the regulation of DNMT3. Further characterization of this RMST/DNMT3 axis will provide new insight into lncRNA-mediated regulation of DNA methylation pathways. Given the importance of DNA methylation in gene expression, RMST may serve as a potential target for cancer therapy.

Materials and Methods

Reagents

Sources of primary antibodies were as follows: DNMT3A (2160), DNMT3B (67259), and DNMT1 (5032) were from Cell Signaling Technology (Danvers, MA); AUF1 (sc-166577) and HuR (sc-5261) were from Santa Cruz Biotechnology (Dallas, TX); and GAPDH was from Protein Tech (Chicago, IL, USA). Secondary antibodies conjugated with IRDye 800CW or IRDye 680 were purchased from LI-COR Biosciences (Lincoln, NE). PCR primers (Table S1) were purchased from Integrated DNA Technologies (Coralville, IA).

Cell Culture

MCF-7, MIA PaCa-2, HCT-116, SJCRH30, and 293T cells were obtained from ATCC (Manassas, VA). MCF-7, HCT-116, and SJCRH30 cells were grown in RPMI 1640 from Sigma (St. Louis, MO), supplemented with 10% fetal bovine serum (FBS). MIA PaCa-2 cells were grown in DMEM with 10% FBS and 2.5% horse serum (Gibco). 293T cells were cultured with DMEM with 10% FBS. All media were supplemented with 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (Lonza, Walkersville, MD). Cells were incubated at 37°C and supplemented with 5% CO2 in the humidified chamber.

SAM Library

Five gRNAs were designed for each lncRNA against 421 lncRNAs, as shown in Table S2, and, thus, a total of 1,205 gRNAs, plus 10 gRNA controls, was designed and synthesized as a mixed pool (CustomArray), as described previously. These gRNA sequences were derived from ∼1 kb upstream of the first exon of lncRNAs. Oligonucleotides carrying each gRNA were synthesized as a mixed pool; they were amplified by PCR and then cloned into lenti single guide RNA (sgRNA) (MS2)_zeo backbone (Lenti-MS2) (Addgene, plasmid 61427), as described previously.

Library Screening

For library screening, MCF-7 cells were first infected with lenti MS2-P65-HSF1_Hygro (Addgene, plasmid 61426) and lenti dCas9-VP64_Blast (Addgene, plasmid 61425) and then selected with hygromycin and blasticidin. Next, the SAM library was introduced into these cells, followed by zeocin selection. Finally, the DNMT3B 3′ UTR reporter was introduced into these cells by transfection and then cultured in the presence of puromycin (1.5 μg/mL) for a period of 4 days, during which the medium was changed every other day and then further cultured for 10 days without puromycin. The BS group was the cells that received the reporter but without exposure to puromycin.

qRT-PCR

Total RNA was extracted using Direct-zol RNA MiniPrep (Zymo Research, Irvine, CA), and cDNA synthesis was carried out using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) with random primers. PCR was performed using a standard SYBR Green method. Delta delta Ct values were used to determine their relative expression as fold changes, as previously described.

lncRNA Profiling

Since the number of lncRNAs in this SAM gRNA library was relatively small, we profiled lncRNA expression using PCR arrays consisting of primer sets derived from each of these lncRNAs included in the library. Total RNA from cells was isolated using Direct-zol RNA MiniPrep kit (Zymo Research, Irvine, CA, USA). Reverse transcription was carried out by using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) and random primer mix (New England Biolabs). The expression of lncRNAs was detected by qRT-PCR using SYBR Green method. Analysis of qRT-PCR was performed as described previously. For comparison, we included the cells carrying the same SAM library BS as a control.

Transfection

Cells were transfected with plasmid DNA using DNAfectin (Applied Biological Materials), following the manufacturer’s protocol.

Plasmid Construction

The DNMT3B-3′ UTR reporter was constructed as follows. First, ∼1.4 kb DNMT3B-3′ UTR fragment was amplified from MCF-7 cells by PCR using primers Pu-DNMT3B-3 UTR-5.1 and DNMT3B-3′ UTR-Xba1-3.1. Next, the puromycin gene was amplified using primers SV40-Pu-Nco1-5.1 and Pu-DNMT3B-3 UTR-3.1. These two overlapped PCR products were cloned into pGL3-control vector at Nco1 and Xba1 sites using NEBuilder HiFi DNA Assembly kit (New England Biolabs, Ipswich, MA). RSMT expression vector was constructed in pCDH-MSCV-copGFP-T2A-Pu (System Biosciences, Mountain View, CA). We first amplified RSMT by PCR using primers RMST-EcoRI-5.1 and RMST-NotI-3.1, and then we cloned into pCDH-MSCV-coPFP-T2A-Pu at EcoR I and Not I sites using the NEBuilder HiFi DNA Assembly kit. For cloning dual gRNAs, we modified LentiCRISPR version (v.)2 by using the optimized scaffold to generate LCV2-m. Two gRNAs and mouse U6 were introduced by PCR using primers RMST-T3B-5.1 and RMST-T3B-3.1, and mouse U6 as a template. Finally, this PCR product was cloned into LCV2-m at the BsmB1 site using the NEBuilder HiFi DNA Assembly kit. PCR reactions for cloning purpose used Phusion enzyme from Thermo Fisher Scientific (Pittsburgh, PA, USA). All clones were verified by DNA sequencing.

KO of RMST by the CRISPR/Cas9 Approach

The lentiviral vector carrying Cas9 and dual gRNAs, driven by human U6 and mouse 6U promoter, respectively, were introduced into MCF-7, MIA PaCa-2, and HCT-116 cells by infection. At 3 days after infection, cells were subject to puromycin selection. Then 2 weeks later, individual colonies were manually collected and expanded in 24-well plates. KO clones were characterized by genomic PCR using outside primers RMST-OS-5.1 and RMST-OS-3.1 (Table S1) and then verified by inside primers RMST-IS-5.1 and RMST-IS-3.1. Finally, it was further confirmed by qRT-PCR using primers RMST-RT-5.3 and RMST-RT-3.3.

Western Blot

Cells were harvested and proteins were extracted and quantified as previously described. Samples were separated in a polyacrylamide SDS gel before transferring to polyvinylidene difluoride (PVDF) membrane. Signals were detected using Odyssey systems (LI-COR Biosciences).

RNA Stability Assay

For the analysis of RNA stability, cells were treated with actinomycin D (2 μg/mL) and collected at 0, 2, 4, 6, and 8 h for RNA extraction. Reverse transcription was carried out by using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) and random primer mix (New England Biolabs) and DNMT3 mRNA levels were measured by qRT-PCR.

Global DNA Methylation Analysis

Genomic DNA was extracted by using Quick-DNA Microprep kit (Zymo Research, Irvine, CA). MethylFlash DNA Methylation Quantification Kit (Epigentek, Farmingdale, NY) was used to determine the global DNA methylation levels, following the manufacturer’s instructions. Relative 5-mC levels were calculated using the formula provided in the kit.

RNA Precipitation

We performed the RNA precipitation assay using a biotin-labeled RSMT RNA probe. We first amplified a DNA fragment covering the entire RMST by PCR using a T7 promoter-containing forward primer T7-RMST-5.1 and T7-RMST-NotI-3.1 (Table S1) and then cloned into pCR8 (Invitrogen, Grand Island, NY). The resultant plasmid DNA was linearized with restriction enzyme Not I, which was introduced from the reverse PCR primer, and then used to synthesize RNA in vitro by T7 polymerase in the presence of biotin-labeled UTP (PerkinElmer, Boston, MA). For RNA-bound protein detection, 2 μg biotin-16-UTP-labeled RNA was incubated with cellular lysate for 2 h at 4°C. After the addition of 20 μL streptavidin beads (Pierce, Rockford, IL), the mixture was incubated for another 30 min and washed 5 times. After the final wash, agarose beads were resuspended in 20 μL protein sample buffer, and RNA-bound protein was separated by SDS-PAGE and detected with anti-HuR or anti-AUF1 antibody. The same procedure was used for DNMT3B 3′ UTR and DNMT3B 5′ UTR probes to determine the interaction of DNMT3B 3′ UTR with HuR or AUF1.

UV Crosslinking RNA Immunoprecipitation

Approximately 7 × 106 cells per UV-RIP were UV crosslinked at 254 nm with 150 mJ/cm2 in Stratalinker (Stratagene, San Diego, CA). To determine the interaction of HuR or AUF1 with RMST, DNMT3A, or DNMT3B mRNA, we used HuR or AUF1 antibody for pull-down from cell lysate. Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Burlington, MA) was used for RIP procedures, according to the manufacturer’s protocol. After the antibody was recovered by protein A/G beads, standard qRT-PCR was performed to detect RMST, DNMT1, DNMT3A, or DNMT3B mRNA in the precipitates.

Statistical Analysis

Each experiment was repeated at least three times. The continuous variables are summarized as mean and SEM unless stated. The two sample t test or two-way ANOVA was used to compare the mean of a continuous variable between two samples. Association between two categorical variables was evaluated by using the Fisher’s exact test. All p values were two sided and p values < 0.05 were considered as significant.

Author Contributions

W.-X.P. conceived the study, performed experiments, analyzed the data, and wrote the manuscript. P.K., W.Z., and C.N. performed the experiments and analyzed the data. L.Y. and Y.-Y.M. conceived and designed this research. Y.-Y.M. supervised the study and wrote the manuscript. All authors read and approved the final manuscript.

Conflicts of Interest

The authors declare no competing interests.