BESST: a novel LncRNA knockout strategy with less genome perturbance

Abstract

Long noncoding RNAs (lncRNAs) are >200 nt RNA transcripts without protein-coding potential. LncRNAs can be categorized into intergenic, intronic, bidirectional, sense, and antisense lncRNAs based on the genomic localization to nearby protein-coding genes. The current CRISPR-based lncRNA knockout strategy works efficiently for lncRNAs distant from the protein-coding gene, whereas it causes genomic perturbance inevitably due to technical limitations. In this study, we introduce a novel lncRNA knockout strategy, BESST, by deleting the genomic DNA fragment from the branch point to the 3′ splicing site in the last intron of the target lncRNA. The BESST knockout exhibited comparable or superior repressive efficiency to RNA silencing or conventional promoter-exon1 deletion. Significantly, the BESST knockout strategy minimized the intervention of adjacent/overlap protein-coding genes by removing an average of ∼130 bp from genomic DNA. Our data also found that the BESST knockout strategy causes lncRNA nuclear retention, resulting in decapping and deadenylation of the lncRNA poly(A) tail. Further study revealed that PABPN1 is essential for the BESST-mediated decay and subsequent poly(A) deadenylation and decapping. Together, the BESST knockout strategy provides a versatile tool for investigating gene function by generating knockout cells or animals with high specificity and efficiency.

INTRODUCTION

Long noncoding RNAs (lncRNAs) are RNA transcripts longer than 200 nt with no protein-coding potential that were categorized into long intergenic noncoding RNAs (lincRNAs), antisense, sense, embedded, and divergent lncRNAs according to the relative genomic position to the adjacent protein-coding genes (1). Approximately 40% of human lncRNAs were intergenic lncRNA, whereas the remaining 60% localized overlap or adjacent to protein-coding genes with different patterns (Supplementary Figure S1A) (2). The biological function of most lncRNAs, especially natural antisense lncRNAs (NAT-lncRNAs), was understudied due to technical limitations (3,4). CRISPR-based genome editing effectively knock-out protein-coding genes by introducing nonsense frameshift or deleting functional exons (5,6). However, these strategies do not work efficiently in lncRNAs absent from the open reading frame (7). On the one hand, the deletion of the promoter region and the first exon downregulates the transcription level of lncRNAs. However, removing a big chunk of DNA could abolish the genomic structure of the target gene and causes unexplainable results (8). On the other hand, RNA silencing triggers the effective degradation of cytosolic lncRNAs, whereas it works poorly for the nuclear lncRNAs (9,10). Therefore, it is urgent to establish an efficient and specific knockout strategy to uncover the promiscuous biological function of lncRNAs.

Most lncRNAs are transcribed by RNA polymerase II and processed by the maturation steps, including RNA splicing, 5′ capping, and 3′ polyadenylation (11). Notably, over 90% of lncRNAs are multiple-exon (12). Similar to mRNA splicing, the 5′ splicing site (5′SS), branch point (BP), and 3′ splicing site (3′SS) are indispensable elements for lncRNA splicing lariat formation and intron excision (13). RNA splicing is essential for multi-exon RNA maturation before being exported to the cytoplasm (14). It is reported that introns localized near the 5′ end of the transcript are removed at a higher frequency via co-transcriptional splicing (15). The splicing of the last intron strongly couples with 3′ end processing, including polyadenylation and poly(A) site cleavage (16–18). Thus, the splicing of the last intron is more critical than upstream introns during pre-RNA maturation (18). The normal RNA splicing needs to recognize 5′SS and 3′SS by U1 snRNP and U2 snRNP, respectively (19). The lncRNAs with intact 5′SS and mutation of the 3′SS cause intron-retention and increased lncRNA nuclear retention in a U1 snRNP-dependent manner (20). Mutation in the branch point or polypyrimidine tract or 3′SS of the last intron suppresses the splicing and cleavage of the 3′ end (16,17). The last-intron retained IncRNAs undergo rapid RNA degradation via XRN2 (5′-3′ exonuclease) or exosome (3′-5′ exonuclease) in the nucleus (16,21). Thus, we hypothesized that the lncRNA could be knocked down by triggering lncRNA nuclear retention and RNA degradation via deleting the BP-3′SS in the last intron, which we called the branchpoint to 3′ splicing site targeting (BESST) (Supplementary Figure S1B).

MATERIALS AND METHODS

Cell culture

Human embryonic kidney cells (HEK293), human pancreatic duct epithelial cells (HEPD6-C7), and Hela cells were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), Otwo Biotech Inc. (Shenzhen, China) and the American Type Culture Collection (U.S.), respectively. Cells were cultured in high-glucose Dulbecco modified Eagle's medium (Gibco, #11965092) or Roswell Park Memorial Institute (RPMI) 1640 Medium (Gibco, #C11875500BT) with 10% fetal bovine serum (Yeasen, #40130ES76) and in a 37°C humid incubator with 5% CO₂.

BESST sgRNA designing

All multi-exon lncRNA sequences were fetched from the UCSC genome browser (http://genome.ucsc.edu/index.html). To predict the branch point of multi-exon lncRNAs, all sequences from the 5′ end of the last exons to 70 bp upstream were processed by RNABP (http://nsclbio.jbnu.ac.kr/tools/RNABP/) to pinpoint the putative branch point. Next, sequences from 100 bp upstream of the branch point to 100 bp downstream of the start of the last exon for each multi-exon lncRNA were used to locate suitable sgRNAs by E-CRISP (http://www.e-crisp.org/E-CRISP/index.html). All sgRNAs with the highest on-targeting and lowest off-targeting scores were shown (Supplementary Table S1).

Preparation of sgRNA expression construct

All sgRNA constructs are produced based on the PX459 backbone (Addgene, #62988) (22). Briefly, the PX459 plasmid was linearized by BbsI (New England Biolabs, #R0539) and cleaned using the agarose gel DNA extraction kit (Takara, #9762). Subsequently, primers of each sgRNA were diluted to a final concentration of 10 μM in NEB buffer 2.1 (New England Biolabs, #B7202). The mixture was heated to 100°C and cooled to room temperature for proper annealing. Next, the ligation reaction mixture was prepared as follows.

Component	Volume (μl)
T4 DNA ligase buffer (10x)	1
Linearized PX459 (20 ng/ μl)	1
Annealed sgRNAs	3
T4 DNA ligase (New England Biolabs, M0202)	1
Nuclease-free water	4
Total	10

The reaction mix was incubated at 22°C for 2 h, and 2 μl of the ligation reaction mixture was added to 100 μl competent E. coli (Stbl3) (AlpaLife, #KTSM110L) for transformation. The agar plates were incubated overnight at 37°C. Two or three colonies were picked for genotyping by PCR on the second day. The PCR system for genotyping colony was set up as follows:

Component	Volume (μl)
PCR MasterMix (TIANYA BIO, #M0121-1ml)	10
Forward primer (px459-hU6-F) (10 μM)	1
Reverse primer (each sgRNA reverse) (10 μM)	1
Nuclease-free water	8
Total	20

The thermocycler was set up as follows: 95°C for 5 min; 39 cycles of 95°C for the 30 s and 61°C for the 30 s, and 72°C for the 30 s; and an extra extension at 72°C for 10 min. 1.5% agarose gel electrophoresis was performed in 1× TAE buffer. A successfully sgRNA-ligated plasmid should yield a single band at 266 bp. The positive clones were amplified, and purified plasmid DNA was produced using Plasmid Miniprep Extraction Kit (Tiangen, #DP103-03). All sgRNA plasmids were sequenced and stored at −20°C.

Transfections of plasmids and siRNA duplexes

Plasmids were transfected into HEK293 cells using Lipofectamine 3000 reagent (Thermo Fisher, #L3000015) following the manufacturer's instructions. Briefly, two sgRNAs-embedded plasmids were mixed at 1:1 (500 ng/500 ng). Each sgRNA was cloned into a PX459 plasmid individually (22). Plasmids were transfected to cells cultured in 6-well plates, and fresh culture medium was replaced 6 h post-transfection. The Puromycin (Gibco, #A1113803) was added into the culture supernatant of transfected cells with a final concentration of 1 μg/ml 24 h post-transfection.

All siRNA duplexes (Supplementary Table S2) were synthesized and purchased commercially (GenePharma, Shanghai). siRNAs were transfected into cultured HEK293 cells at a final concentration of 50 nM using siRNA transfection reagent Lipofectamine RNAiMAX (Thermo Fisher, #13778075) following the manufacturer's guide. All cells were collected 72 h post-transfection for further analysis.

CRISPR-Cas9 genome editing and single clone screening

When sgRNA-PX459 transfected cells reached 90% confluency after puromycin selection, the genomic DNA was extracted (Tiangen, #DP304), and PCR genotyping was performed using 200 ng of genomic DNA: 95°C for 5 min; 39 cycles of 95°C for the 30 s and 61°C for 30 s, and 72°C for 1.5 min; and an extra extension at 72°C for 10 min (Supplementary Table S3). If the DNA from pooled cells indicated an intron-exon deletion band, ∼100 cells were sub-cultured in a 10 cm dish and incubated for ∼2 weeks. The single clones were then marked under a microscope, and clones were transferred to 96-well plates. All clones were amplified in serial 96–48–24–12 well plates. Genomic DNAs were extracted when cells reached 90% confluency in a 12-well plate (Transgen, #AD201), and PCR was performed to identify the knockout clones. Next, the positive lncRNA knockout single clones were cultured for further experiments.

RNA extraction, quantitative real-time PCR, and primer walking PCR

The total RNA was lysed in the TRIzol Reagent (Thermo Fisher, #15596026), followed by conventional chloroform extraction. 500 ng of total RNA was reverse transcribed using ReverTra Ace qPCR RT Master Mix with genomic DNA Remover (Toyobo, #FSQ-301). Realtime qPCR was performed using PerfectStart Green qPCR SuperMix (Transgen, #AQ601) in qTOWER 2.0 (Analytik Jena AG, Germany) with thermocycler setup: 95°C for 1 min, 40 cycles of 95°C for 15 s and 60°C for 45 s (plate read), followed by 60–95°C melting curve. Primers of ACTB, MT-CYB, or U6 that were used as internal control and all the lncRNAs and NATs were marked in Supplementary Figure S1C (Supplementary Table S3). All tests were measured triplicated. The 2^−ΔΔCt method was used to analyze and compare the fold change. Primer walking PCR was performed using PCR MasterMix (TIANYA BIO, #M0121-1ml): 95°C for 5 min; 35 cycles of 95°C for the 30 s and 61°C for 30 s, and 72°C for 2.5 min; and extra extension at 72°C for 10 min (Supplementary Table S3).

Cytoplasmic and nuclear fractionation

The subcellular location of all genes was assessed using HEK293 cells. The nuclear and cytoplasmic RNA was isolated following the manufacturer's protocols of the nuclear/cytoplasm fractionation kit (Beyotime, #P0028). Briefly, 1.5 × 10⁶ cells were trypsinized and centrifuged at 1000 × g at 4°C for 3 min. The supernatant was removed, and the cell pellet was resuspended in 200 μl cytoplasmic protein extraction reagent A with 1 U/ml RNase inhibitor (Thermo Fisher, #EO0382). The mixture was incubated in an ice bath for 15 min. Then, 10 μl cytoplasmic protein extraction reagent B was added, followed by a vortex for 5 s, and incubation in an ice bath for another 1 min and vortex for 5 s again. Centrifuge with 16 000 × g for 10 min at 4°C and transfer supernatant to a fresh microfuge tube. The supernatant is cytoplasmic fractionation, and the sediment is nuclear fractionation. For western blotting, the sediment is lysed by protein lysis buffer and denatured in loading buffer by boiling. Cytoplasmic and nuclear fractionation were lysed for RNA purification in TRIzol Reagent (Thermo Fisher, #15596026). ACTB, GAPDH, and MT-CYB were used as the cytoplasmic control, and U3, U6 and NEAT1 were used as nuclear control. The nuclear/cytoplasmic fractions ratio was calculated using the formula below:

Splicing inhibition assay

HEK293 cells were plated in 6-well plates. Cells were treated with 100 μM isoginkgetin (23) (MCE, #HY-N2117) or DMSO (control) for 18 hr. Half cells were collected for total RNA extraction and another half for nuclear/cytoplasmic fractionation. Due to the splicing inhibition, we used primers of GAPDH and U6 without intron-spanning to determine the cytoplasm and nucleus.

RNA degradation velocity assay

HEK293 cells were cultured and administrated with 10 μg/ml actinomycin D (Sigma, #A4262-5MG) for 0, 2, and 4 hr. Total RNAs were extracted for real-time PCR. RNA decay velocities were determined following the previous publication (24).

Western blotting

Cells were lysed in the Protein Lysis Buffer (50 mM Tris–HCl, pH 8.0, 4 M urea, and 1% Triton X-100) with the supplement of protease inhibitor mixture (MCE, #HY-K0011). 20 μg of whole-cell protein lysate were resolved with 12% discontinuous SDS-PAGE gel. The nitrocellulose membranes (PALL, #66485) were blocked with 5% non-fat milk in 1× TBST buffer for 1 h and incubated on an orbital shaker for 1hr at room temperature with primary antibodies: anti-B2M antibody (Cell Signaling Technology, #12851), anti-GAPDH antibody (Proteintech, #10494-1-AP), anti-lamin B1 antibody (Novus, #NBP1-19804), anti-CNOT7 antibody (Proteintech, #14102-1-AP), and anti-CNOT8 antibody (Proteintech, #10752-1-AP). The membranes were washed in 1× TBST three times and incubated for 1hr in second antibodies: anti-rabbit IgG (KPL, #074-1506). Finally, the membrane was incubated in Chemiluminescence ECL reagent (Thermo Fisher, #32106) and visualized by exposure on an X-ray film.

Cap-RIP assay

The mRNA was isolated from 1 × 10⁸ HEK293 cells plated in 15-cm dishes using Dynabead mRNA DIRECT Purification Kit (Thermo Fisher, #61012) following the manufacturer's protocol. The mRNA samples were divided into three groups: (i) 500 ng of the treated RNA (10%) was saved as input; (ii) 5 μg mRNA was diluted to a final volume of 250 μl with reaction buffer (10 mM Tris pH7.4, 150 mM NaCl, 0.1% Igepal) as RIP group; (iii) the rest 5 μg mRNA was de-capped with 3 U of Cap-Clip enzyme (CellScript, #C-CC15011H) in 30 μl reaction volume for an hour at 37°C, purified with the RNA Clean and Concentrator 5 kit (Zymo Research, #R1016) and also diluted to a final volume of 250 μl with reaction buffer as Cap free negative control. The RIP groups and Cap-free negative controls were incubated with 2.5 μg Anti-7-methylguanosine (m7G) antibody (MBL, #RN017M) and 250 units RiboLock RNase inhibitor (Thermo Fisher, #EO0382) in rotator overnight at 4°C. 50 μl of Dynabead Protein G for Immunoprecipitation (Thermo Fisher, #10003D) were then added and incubated at 4°C for 2 h. The magnetic-bead-bound complex is washed twice with 200 μl of reaction buffer at room temperature, twice with 200 μl of low salt buffer (10 mM Tris pH 7.4, 50 mM NaCl, 0.1% Igepal) at room temperature, and twice with 200 μl of high salt buffer (10 mM Tris pH7.4, 500 mM NaCl, 0.1% Igepal). Bound RNA was isolated and purified by phenol-chloroform-isoamyl alcohol for RT-qPCR assay.

Luciferase activity assay

Plasmids containing different lengths of the promoter region of DHRS4-AS1 were cloned and synthesized by Sangon Biotech (Shanghai) Co., Ltd. To determine the luciferase activity in 96-well plates, 100 ng plasmids were transfected into HPDE6-C7 cells and cultured for 48 hr before harvest. Following the manufacturer's instructions, the firefly luciferase activity was measured by Dual-Luciferase Reporter Assay System (Promega, #E1910). The luminescence for each sample was read on Spark® multimode microplate reader (TECAN).

Poly (A) tail length determination

The Poly(A) tail length was determined using the Poly(A) tail length assay kit (Thermo Fisher, #764551KT) following the manufacturer's protocol. The total RNA of HEK293 NC and KO cells was extracted with TRIzol reagent (Thermo Fisher, #15596026), treated with DNase I (Thermo Fisher, #AM2222), and purified with phenol-chloroform again. 1 mg total RNA sample was added with poly(G/I) tails, reverse transcribed the poly(G/I) tailed RNA and the poly(G/I) tailed cDNA was PCR amplified. The size of PCR products can be assessed by running one-half of each PCR reaction (12.5 μl) per lane on a 5% non-denaturing polyacrylamide TBE gel. Stain gels with 50 ml GelRed Nucleic Acid Gel Stain (Sigma Aldrich, #SCT123), diluted in 0.1 M NaCl, for 1 h, and images were captured by a GelDoc Gel Documentation System (Bio-Rad, USA).

High throughput RNA sequencing

The total RNA of cells was isolated and reverse transcribed by qPCR RT Master Mix with gDNA Remover (Toyobo, #FSQ-301). RNA integrity was assessed using Agilent 2100 Bioanalyzer (Agilent Technologies). Then the libraries were constructed using TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, #20020594) according to the manufacturer's instructions. The transcriptome sequencing and analysis were conducted by OE Biotech Co., Ltd (Shanghai, China) and Lianchuan Biotech Co., Ltd (Hangzhou, China). The libraries were sequenced on an Illumina HiSeq X Ten platform, and 150 bp paired-end reads were generated. Raw data (raw reads) of fastq format were firstly processed using Trimmomatic to remove low-quality reads and obtain the clean reads. Then about 48.3 M clean reads for each sample were retained for subsequent analyses. The clean reads were mapped to the human genome (GRCh38) using HISAT2. FPKM of each gene was calculated using Cufflinks, and HTSeq-count obtained the read counts of each gene. Differential expression analysis was performed using the DESeq (2012) R package. P-value <0.05 and fold change >2 or fold change <0.5 was regarded as significantly differential expression. Hierarchical cluster analysis of differentially expressed genes (DEGs) was performed to demonstrate the expression pattern of genes in different groups and samples.

Statistical analysis

All the result was performed as the mean ± SD, and data plotting was presented by GraphPad Prism 8.0. A value of P < 0.05 was considered statistical significance. Unpaired, two-tailed Student's t-test and Mann-Whitney test were used to compare the results between the two groups. One-way ANOVA was applied to compare multiple subjects. Each experiment was repeated with at least three biological replicates.

RESULTS

The BESST strategy efficiently suppresses lncRNA transcripts

Most human lncRNA transcripts are multi-exon lncRNAs and carry 2–7 exons (Figure 1A). To verify the efficacy of the BESST lncRNA knockout strategy, eight candidate lncRNAs comprising 2–7 exons were selected: LOC646762 (2-exon), LOC107985050 (2-exon), DHRS4-AS1 (3-exon), EPB41L4A-AS1 (3-exon), EMX2OS (4-exon), HOXA-AS2 (4-exon), NPSR1-AS1 (5-exon) and HOTAIR (7-exon) (Figure 1B, Supplementary Table S4). The selected lncRNAs are close, overlapping, or embedded in adjacent protein-coding genes (head-to-head, tail-to-tail, and embedded) (Supplementary Figure S1A). To knock out the target lncRNAs, a pair of small guide RNAs (sgRNAs) were designed and synthesized, flanking the branch point and 3′ splicing site in the last intron of each lncRNA (Figure 1C, Supplementary Figure S1B). By co-delivering constructs expressing Cas9 protein and sgRNAs, we created target DNA fragment excision from the genome, and the cell colonies with correct excision were genotyped and amplified (Figure 1C).

Figure 1.

LncRNA knockout strategy and genotyping results of the BESST. (A) The summary for the exon number of human lncRNAs based on the human genome database (GRCh38.104). (B) The schematics for the lncRNAs and overlap/adjacent protein-coding genes were selected in this study. (C) The sgRNA and primer location to perform the BESST knockout and the genotyping agarose gel results for each lncRNA candidate.

To investigate the efficiency of the BESST lncRNA knockout strategy, we generated stable knockout cells of candidate lncRNAs, including LOC646762, LOC107985050, DHRS4-AS1, EPB41L4A-AS1, EMX2OS, HOXA-AS2, NPSR1-AS1 and HOTAIR. In comparison, two sets of siRNAs duplexes were transfected into cultured cells individually to suppress the target lncRNA level. The real-time qPCR was performed to determine the expression level of candidate lncRNAs and adjacent protein-coding genes. Our results showed that the BESST knockout significantly downregulates all candidate lncRNAs at an average of ∼72% repressive efficiency, whereas siRNA silencing exhibits an average of ∼58% repression for lncRNAs (Figure 2A, B, Supplementary Figure S1D). Our data also indicated that the BESST knockout similarly affects adjacent/overlap protein-coding genes as RNA silencing (Figure 2A, B). Next, we determined the subcellular localization of all candidate lncRNAs. Our results showed a wide range of nucleus/cytosol ratio from 0.3 to 0.95 (Figure 2C), suggesting that the BESST strategy effectively knocked out target lncRNAs regardless of the subcellular localization. To understand the correlation between target lncRNA subcellular localization and the expression level, as well as the inhibitory rate upon gene knock-down, Real-time qPCR data were analyzed, and scatterplots showed a positive correlation between validated lncRNA expression and cytoplasmic localization (Supplementary Figure S1E). Notably, target lncRNA’s nuclear/cytosol ratio does not influence the repressive efficiency in BESST-KO cells. However, RNAi repressive efficiency negatively correlated with nuclear localization since the RNAi machinery is only presented in the cytoplasm (Supplementary Figure S1E). To further evaluate the repression of lncRNAs by BESST, we designed a series of primers for detecting different segments of the lncRNA DHRS4-AS1 and EMX2OS, covering intron, exon, intron-exon junction, upstream editing site, and downstream editing site from 5′ end to 3′ end. Our data showed that the BESST knockout reduced the total and spliced RNA transcripts without producing alternative splicing products while leading to the nuclear accumulation of intron-retained transcripts (Supplementary Figure S2A, B). To study whether the BESST works for messenger RNAs (mRNAs) knockout, an abundant expressed protein-coding gene, B2M, was genome engineered by deleting the BP-3′SS in the last intron. Both real-time qPCR and Western blotting results demonstrated a substantial decrease in B2M mRNA level after BESST knockout, suggesting the feasibility of the BESST for protein-coding gene knockout (Figure 2D, Supplementary Figure S3A).

Figure 2.

The validation of efficiency and specificity of lncRNA knockout using the BESST. (A) The real-time qPCR of the RNA expression level of lncRNA candidates and overlap/adjacent protein-coding genes. Two clones for each lncRNA were included. (B) The real-time qPCR of the RNA expression level of lncRNA candidates and overlap/adjacent protein-coding genes. Two siRNAs for each lncRNA were included. (C) The quantitative results of the subcellular localization of each lncRNA candidate and cytosolic marker (ACTB, GAPDH, and MT-CYB), nuclear marker (U3, U6, and Neat1). (D) The strategy for knockout human B2M mRNA and designing primer sets for testing (upper) and the real-time qPCR results for the B2M RNA level with different primer sets after the BESST knockout. At least three biological replicates and three technical replicates for each sample were performed. One-way ANOVA performed the statistical significance, ** indicates P< 0.01, * indicates P< 0.05, and the ‘ns’ indicates no significance.

The BESST knockout exhibits high specificity and low genomic perturbance

The overall lncRNA knockout strategy includes deletion of the promoter and first exon, removal of the whole transcript, and insertion of polyadenylation signal (Supplementary Figure S3B). Deleting promoter and first exon (PE1) is the typical strategy to inactivate lncRNA when generating stable knockout cell lines or producing knockout mice (25,26). However, it is quite challenging for the PE1 strategy to knock out lncRNA expression when protein-coding genes are nearby or overlap. This defect impeded the study of lncRNAs and the high throughput screening of lncRNA function (2).

To understand the pros- and cons- of the BESST lncRNA knockout strategy versus the conventional approach, we generated a lncRNA knockout cell line by deleting the promoter-exon1 (−506 bp to +77 bp of TSS) of DHRS4-AS1 (Figure 3A, Supplementary Figure S3C). Different from the PE1 knockout that removed 583 base pairs from genomic DNA, the BESST strategy deleted 311 base pairs from the genome. Then, RNA-seq analysis and real-time qPCR were performed to determine the RNA level of DHRS4-AS1 and DHRS4 upon lncRNA knockout. Our results indicated that both BESST and PE1 effectively suppressed the RNA level of DHRS4-AS1 (Figure 3B). In contrast, the PE1 knockout significantly influenced the level of DHRS4, which was not seen in BESST knockout cells (Figure 3B). Next, we compared the DEGs of DHRS4-AS1 knockout by BESST and PE1 transcriptome-wise. The RNA-seq data showed that there are 556 differential expressed genes (DEG) (FC > 2, P < 0.05) in BESST knockout cells, whereas 826 DEGs were found in PE1 knockout cells (Figure 3C, Supplementary Figure S3D). The BESST and PE1 knockout share 116 common down-regulated genes, occupying 42.96% and 25.95% of all down-regulated genes in BESST and PE1, respectively. The gene ontology (GO) analysis revealed that the down-regulated genes from both methods shared common biological processes (Figure 3D), as well as showed distinguished biological processes (Figure 3E, F).

Figure 3.

The comparison between the BESST and promoter-exon1 knockout efficiency and specificity. (A) The schematic plot for the BESST and PE1 knockout and sgRNA localization for DHRS4-AS1 in the HPDE6-C7 cell line. (B) The real-time qPCR and RNAseq results of the DHRS4-AS1 and DHRS4 RNA level using PE1 and BESST knockout approach. (C) The DEGs of high throughput RNA-seq data using BESST and PE1 strategy. (D) The gene ontology (biological process) comparison of down-regulated genes between the BESST and PE1 knockout.(E) The gene ontology (biological process) of down-regulated genes of BESST knockout only. (F) The gene ontology (biological process) of down-regulated genes of PE1 knockout only. (G) The working pipeline of designing SgRNAs for BESST lncRNA knockout based on the UCSC database. (H) The DNA length of the experimentally validated promoter was compared to the 3′ end of exon 1 (PE1) and branchpoint to the 3′ splicing site of the last intron (BESST). (I) The comparison of DNA excision length by PE1 and BESST strategy. (J) The comparison of human and mouse lncRNAs results in genomic perturbance in the 5′/3′ UTR or exon of adjacent protein-coding genes using the BESST and PE1 knockout strategy. At least three biological replicates and three technical replicates for each sample were performed. One-way ANOVA performed the statistical significance, ** indicates P< 0.01, * indicates P< 0.05, and the ‘ns’ indicates no significance.

To understand and compare the perturbance on genomic DNA using the BESST versus PE1 strategy genome-wise, the sgRNAs flanking BP-3′SS in the last intron of human and mouse lncRNAs (>2 exons) were designed following the working pipeline (Figure 3G, Supplementary Material S1-S2). To compare with the PE1 strategy, we acquired experimentally validated promoters from Eukaryotic Promoter Database (http://epd.epfl.ch/) (27), and our analysis showed the length of ∼90% (1666/1879) eukaryotic lncRNA promoter is <1000 bp (Figure 3H). Therefore, we excised 1000 bp upstream of TSS (−1 kb) to the 3′ end of exon 1 to deactivate lncRNA in the PE1 strategy. The minimal required length of DNA fragment for lncRNA knockout was compared. Our data showed that an average of ∼130 bp genomic DNA deletion is necessary for knocking out target lncRNA using BESST (Figure 3I, Supplementary Material S3). Next, we analyzed the 5′ untranslated region (5′UTR), 3′UTR, and exon of human and mouse protein-coding genes that the BESST versus PE1 lncRNA knockout could impact. Our data indicated the 5′/3′ UTR and exon of ∼21.5% (2561/11 906) human lncRNAs and 22.6% (961/4255) mouse lncRNAs are potentially influenced by the BESST knockout, whereas more than half (51.6%, 6149/11 906) human lncRNAs and 48.4% (2061/4255) mouse lncRNA knockout by PE1 strategy results in unspecific genomic perturbation (Figure 3J).

The BESST knockout results in nuclear retention of target lncRNAs

Most lncRNAs were transcribed by RNA polymerase II and exported to the cytoplasm following post-transcriptional processing (28). Although abnormal cytosolic RNA effectively degrades via nonsense mRNA decay (NMD) or Staufen-mediated mRNA decay (SMD) (29), the degradation of lncRNAs was still not evident. Our previous data showed that the BESST lncRNA knockout efficiently reduces the target lncRNA level. To unveil the underlying mechanism, actinomycin D was applied to inhibit the activity of RNA polymerase II, and real-time qPCR results indicated the expression of lncRNAs (LOC646762, DHRS4-AS1, EMX2OS and EPB41L4A-AS1) are markedly decreased (Figure 4A, Supplementary Figure S3E). This result demonstrated that BESST-KO enhanced lncRNA degradation.

Figure 4.

The BESST knockout causes nuclear retention of target lncRNAs. (A) The real-time qPCR of LOC646762, DHRS4-AS1, EMX2OS, and EPB41L4A-AS1 levels at 2 and 4 hr post actinomycin D (10 μg/ml) treatment, all data were normalized by the expression level of ACTB. (B) The real-time qPCR of LOC646762, DHRS4-AS1, EMX2OS, and EPB41L4A-AS1 levels in NC and KO cells with or without treatment of isoginkgetin (100 μM), all data were normalized by the expression level of U6. (C) The quantitative assays for the subcellular localization of candidate lncRNAs by the BESST knockout with or without treatment of isoginkgetin (100μM). At least three biological replicates and three technical replicates for each sample were performed. One-way ANOVA performed the statistical significance, ** indicates P< 0.01, * indicates P< 0.05.

To study the subcellular localization of lncRNA transcripts after the deletion of BP-3′SS in the last intron, isoginkgetin, a splicing inhibitor, was used to suppress the RNA splicing in the cultured cells (23). Four lncRNAs were investigated (LOC646762, DHRS4-AS1, EMX2OS and EPB41L4A-AS1), and lncRNA knockout cell lines were established via the BESST. Next, the nuclear-cytosol fractionation was performed, and the subcellular localization of candidate lncRNAs and mRNA control (GAPDH) were quantitatively determined. NC cells with isoginkgetin-treated reduced transcripts of lncRNAs and had a unique nuclear location. This phenomenon was similar to BESST-KO cells. BESST-KO cells with isoginkgetin could enhance the reduction of lncRNAs. Our data demonstrated that the BESST knockout results in nuclear retention and a decrease in target lncRNAs which quickly was the same with results of splicing inhibition of isoginkgetin (Figure 4B, C), implying the BESST-KO affected lncRNAs by interfering splicing.

The BESST knockout causes lncRNA nuclear degradation via PABPN1-mediated CNOT7/8-dependent poly(A) deadenylation and NCBP1-dependent decapping

Our data showed that the BESST knockout results in lncRNA nuclear retention. However, the RNA degradation pathway of BESST knockout was unknown. Generally, lncRNA and mRNA share common nuclear RNA degradation machinery because lncRNA and mRNA transcription and post-transcriptional processes are almost identical (30,31). To uncover the underlying mechanism of BESST-mediated nuclear RNA degradation, we knocked down the critical components of all RNA nuclear degradation pathways (Figure 5A) (31,32).

Figure 5.

The determination of the molecular mechanism of the BESST knockout-induced RNA degradation. (A) Schematic representation of pathways and associated factors for degradation of lncRNAs. (B) The real-time qPCR results of DHRS4-AS1 level of the Cap-RIP samples using DHRS4-AS1 knockout cells versus control. The BESST strategy was applied to produce the knockout cell line. (C) The RNA level of DHRS4-AS1 in BESST-mediated DHRS4-AS1 knockout cells upon knocking down of cap-binding complex (NCBP1 and ARS2), 5′ decapping enzyme (DCP2, DXO, and NUTD16) and 5′-3′ exonuclease (XRN2) by siRNA duplexes. (D) The RNA level of DHRS4-AS1 in BESST-mediated DHRS4-AS1 knockout cells upon knocking down of endonuclease (RNASEH1) by RNA silencing. (E, F) The RNA level of DHRS4-AS1 in BESST-mediated DHRS4-AS1 knockout cells upon knocking down the critical components of PAXT/PPC-NEXT-TRAMP complex (ZCCHC7, ZC3H18, ZCCHC8, and ZFC3H1) and RNA exosomes (EXOSC3, EXOSC10, DIS3, and MTR4) by siRNA duplexes. (G) The poly(A) tail length assay using BESST-mediated DHRS4-AS1 knockout cell line versus wildtype control. (H, I) The RNA level of DHRS4-AS1 in BESST-mediated DHRS4-AS1 knockout cells upon knocking down the essential proteins of 3′-5′ exonuclease and poly(A) deadenylation complex (CNOT7, CNOT8, PAN2, and PARN) and poly(A) binding protein PABPN1 by RNA silencing. (J) A schematic diagram of BESST-KO strategy to reduce lncRNAs through poly(A) deadenylation and decapping. At least three biological replicates and three technical replicates for each sample were performed. One-way ANOVA performed the statistical significance, ** indicates P< 0.01, * indicates P< 0.05, and the ‘ns’ indicates no significance.

The 5′-capping is essential for protecting RNA stability from 5′-3′ exonuclease (33). To investigate whether the BESST knockout influences the 5′-capping, 5′-cap RNA immunoprecipitation (5′Cap-RIP) was performed. Our data indicated less 5′-capped DHRS4-AS1 was detected in the BESST-knockout cells versus wildtype control (Figure 5B). Therefore, the critical components of the cap-binding complex (NCBP1 and ARS2), decapping enzyme (DCP2, DXO, and NUTD16), and 5′-3′ exonuclease (XRN2) were knocked down. The real-time qPCR results revealed that the DHRS4-AS1 and other candidate lncRNA level was elevated upon suppression of caping-binding complex (NCBP1) but not in decapping enzyme complex or 5′-3′ exonuclease or ARS2, consistently (Figure 5C, Supplementary Figure S4A–C) (34–37). To study whether RNA endonuclease involves in the BESST-mediated lncRNA degradation, RNASEH1 was silenced by siRNA (30,38). However, no change in the target lncRNA level was detected (Figure 5D, Supplementary Figure S5A–C). Next, to determine whether 3′-5′ exonucleases dominate the BESST-mediated lncRNA degradation, the RNA exosome accessory complex protein of NEXT (ZC3H18 and ZCCHC8), PAXT/PPC (ZC3H18 and ZFC3H1) and TRAMP (ZCCHC7) complex (39), were knocked down. Real-time qPCR was performed to determine the abundance of DHRS4-AS1 and other tested lncRNAs. Our data demonstrated no consistent impact on the level of lncRNAs (Figure 5E, Supplementary Figure S6A–C). RNA exosomes are the final step of 5′ end nuclear RNA degradation, where RNA fragments are decomposed into nucleotides (40). We suppressed the core proteins of RNA exosome (EXOSC3, EXOSC10, MTR4 and DIS3) by siRNA transfection in BESST-knockout cells, and our results did not show consistent upregulation of lncRNAs (Figure 5F, Supplementary Figure S7A–C) (41).

Poly(A) tail governs RNA transcripts' stability and nuclear export (42). To investigate the impact of BESST knockout on poly(A) tail length, the poly(A) tail length assay was performed, and our results demonstrated that the poly(A) tail of DHRS4-AS1 was shortened in BESST knockout cells (Figure 5G). The CCR4-NOT, PAN2-PAN3 complexes, and PARN protein are the main conserved protein complex that regulates RNA stability by catalyzing the deadenylation of RNA transcripts (43,44). To investigate whether these complexes or protein mediates the RNA degradation upon BESST-knockout, core proteins, including CNOT7, CNOT8, PAN2 and PARN, were inhibited by siRNA duplexes (45). Our real-time qPCR data demonstrated that inhibition of CNOT7/CNOT8, but not PARN and PAN2, results in the elevation of target lncRNA level (Figure 5H, Supplementary Figure S8A–C), suggesting removing of BP-3′SS in the last intron facilitates the CNOT7/8-dependent poly(A) tail deadenylation of RNA transcripts and therefore compromises RNA stability.

Provided that the CNOT7/8 is essential for regulating the mRNA stability with longer poly(A) tails in the cytoplasm (46), we were curious whether CNOT7/8 plays a similar role in the nucleus. Firstly, our Western blotting results using the nuclear and cytosolic fractionation lysates confirmed the presence of CNOT7/8 in the cell nucleus (Supplementary Figure S9A). To uncover whether these lncRNAs were degraded in the nucleus, PCR was performed using primers targeting different regions of DHRS4-AS1 and EMX2OS after CNOT7/8 siRNA transfection in BESST-KO cells (Supplementary Figure S9B, C). Our results indicated that the intact and unspliced RNA transcripts increased following CNOT7 or CNOT8 knockdown. However, no mature product was found, which implied that the RNA degradation upon BESST-KO is mediated by CNOT7/8 in the cell nucleus.

PABPN1 is a poly(A)-binding protein that is necessary for poly(A) synthesis and poly(A) tail length regulation (47). Our data demonstrated that PABPN1 knockdown results in the elevation of target lncRNA level (Figure 5I, Supplementary Figure S10A-C), implying BESST-KO regulates the poly(A) tail length via PABPN1-dependent pathway. Furthermore, we knocked down NCBP1, CNOT7/8, and PABPN1 in unedited HEK293 cells, respectively. Our data revealed a significant increase of target lncRNA level in siNCBP1, siCNOT7, siCNOT8 transfected cells, whereas siPANPN1 failed to restore the lncRNA levels (Supplementary Figure S11A–D). The results above collectively indicated that BESST-KO reduces lncRNA levels in a PABPN1-dependent manner.

DISCUSSION

The loss-of-function study is one of the most critical approaches to understanding the biological function of mRNAs or lncRNAs. Unlike mRNA, which possesses an open reading frame, we cannot use frameshift to produce aberrant transcripts to deactivate the biological function of lncRNA. Likewise, the nonsense mRNA decay initiated by premature stop codon is not applicable for lncRNA knockout. The conventional approach to knockout lncRNA is to remove genomic DNA spanning the promoter and first exon (PE1) of target lncRNA (25). However, the PE1 strategy only works for intergenic lncRNAs and inevitably impacts adjacent gene structure and expression level on the opposite strand. The alternative method is to knock down lncRNA with RNA silencing, whereas siRNA duplexes work efficiently in the cytoplasm but not in the cell nucleus (10). Thus, the biological function of many lncRNAs was understudied due to lack of an efficient knockout tool. Here, we introduced the efficient and particular lncRNA knockout strategy, BESST. Removing as little as 37 bp from the genome, the target lncRNA was retained within the cell nucleus and triggered a deadenylation-dependent surveillance response, leading to RNA degradation by RNA exosome or other unknown pathways.

The last intron-exon is particularly important for the fate of spliced RNA transcripts. Unlike typical non-last intron removed by co-transcriptional splicing, excision of the last intron is inhibited until 3′ end maturation, including recognition of poly(A) sites, cleavage, and polyadenylation (18). Regularly, an intron consists of 5′SS, recognized by U1 snRNP, whereas BP, polypyrimidine tract and 3′SS are associated with U2 snRNP (48,49). It was reported that coupled splicing and 3′ end cleavage were inhibited after depletion of U2 snRNP or CPSF (cleavage and polyadenylation specificity factor) or mutated BP or polypyrimidine tract (17). Furthermore, the physical interaction between U2AF (U2 auxiliary factor, an essential splicing factor) and poly-A polymerase (PAP) is dispensable for the polyadenylation at the 3′ end of pre-RNA (50–52). Notably, U1 snRNP inhibits pre-RNA premature cleavage and polyadenylation, mediating pre-RNA nuclear retention when U2 snRNP is not associated with the same intron (20,53). Thus, removing the BP-3′SS fragment in the last intron encumbered the U2 SNP binds to RNA transcripts, causing RNA nuclear retention and subsequent RNA degradation by the nuclear RNA exosome.

The pathway for nuclear lncRNA degradation was understudied. In this study, we knocked down the key protein components of pathways for degrading nuclear mRNA one by one. We identified that CNOT7/8 plays an essential role in BESST-mediated lncRNA degradation. CNOT7 and CNOT8 are vertebrate homologs of the yeast CCR4-NOT catalytic subunit Caf1, that is critical for deadenylation and 3′-5′ exoribonuclease degradation of target RNA transcripts (54). Our data demonstrated that the knockdown of either CNOT7 or CNOT8 results in the elevation of target lncRNA transcripts in the BESST knockout cell line. Notably, the poly(A) tail length assay indicated the shortened poly(A) tail upon the BESST knockout of DHRS4-AS1. Therefore, we concluded that the BESST knockout targets by deadenylation of RNA transcripts. The underlying mechanism of last-intron BP-3′SS removal initiating deadenylation of target RNA transcripts in a CNOT7/8-dependent manner is yet to be clarified. In the cytoplasm, the 5′-3′ decay is a multistep process triggered by the deadenylation of poly(A) at the 3′ end (35). In contrast, the correlation between shortened poly(A) tail and reduced 5′ capping in the nuclear still needs to be investigated in future work.

The poly(A) tail of RNA transcripts is synthesized ∼250 nt in the cell nucleus (55). A previous study reported that the defined length of the poly(A) tail was regulated by proteins, including PABPN1, PAP, and CPSF (55). Our data indicated that lncRNA levels were replenished in BESST-KO cells, but not in unedited cells, after PABPN1 knockdown, suggesting BESST-KO initiates lncRNA degradation via the PABPN1-dependent decay pathway. Another study reported that mRNAs with retained introns decay through the PABPN1 and PAPα/γ dependent manner (56). The above results demonstrated that the BESST-KO disrupted the last intron splicing and poly(A) tail processing, leading to RNA transcript intron retention and nuclear product retention. Furthermore, the poly(A) synthesis was dysregulated in BESST-KO cells by impeding PAP/CPSF-PABPN1 interaction, influencing poly(A) deadenylation, and 5′ decapping with CNOT7/8-dependent or NCBP1-dependent pathway (Figure 5J).

NAT-lncRNA is an important lncRNA category that globally exists in the mammalian genome and whose functions are barely known. Even though lots of endeavors have been accomplished to reveal the biological process of NAT-lncRNA (57), the adjacent/host gene perturbance prevents NAT-lncRNAs from being investigated with a high throughput approach. Here we introduced a novel lncRNA knockout strategy with less genomic perturbance than conventional promoter and first exon deletion or whole transcript removing and higher efficiency compared to the premature codon insertion, which requires homologous recombination of a repair template. Importantly, we provided an inheritable, specific, and compelling lncRNA and mRNA knockout strategy that can be applied in generating knockout mice or cell lines, which will expand our understanding of the roles of lncRNAs.

DATA AVAILABILITY

The data has been deposited to the GEO database (http://ncbi.nlm.nih.gov/geo) with the accession number GSE216846.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

International Cooperation Fund of Shenzhen [GJHZ20180929162002061 to Y.-O.Z.]; National Natural Science Foundation of China [82070486 to Q.R.L., 31730043 to C.Z.]; National Key R&D Program of China [2018YFC1312700271 to C.Z.]. Funding for open access charge: International Cooperation Fund of Shenzhen [GJHZ20180929162002061 to Y-O.Z.]; National Natural Science Foundation of China [82070486 to Q.R.L., 31730043 to C.Z.]; National Key R&D Program of China [2018YFC1312700271 to C.Z.].

Conflict of interest statement. None declared.