Long Noncoding RNA DLGAP1-AS2 Promotes Tumorigenesis and Metastasis by Regulating the Trim21/ELOA/LHPP Signaling Axis in Colorectal Cancer

Background: Long noncoding RNAs (lncRNAs) have driven research focused on their effects as oncogenes or tumor suppressors involved in carcinogenesis. However, the functions and mechanisms of most lncRNAs in colorectal cancer (CRC) remain unclear. Methods: The expression of DLGAP1-AS2 was assessed by quantitative RT-PCR in multiple CRC cohorts. The impacts of DLGAP1-AS2 on CRC growth and metastasis were evaluated by a series of in vitro and in vivo assays. Furthermore, the underlying mechanism of DLGAP1-AS2 in CRC was revealed by RNA pull down, RNA immunoprecipitation, RNA sequencing, luciferase assays, chromatin immunoprecipitation, and rescue experiments. Results: We discovered that DLGAP1-AS2promoted CRC tumorigenesis and metastasis by physically interacting with Elongin A (ELOA) and inhibiting its protein stability by promoting tripartite motif containing 21(Trim21)-mediated ubiquitination modi�cation and degradation of ELOA. In particular, we revealed that DLGAP1-AS2decreases phospholysine phosphohistidine inorganic pyrophosphate phosphatase (LHPP) expression by inhibiting ELOA-mediated transcriptional activating of LHPP and thus blocking LHPP-dependent suppression of the AKT signaling pathway. In addition, we also demonstrated that DLGAP1-AS2 was bound and stabilized by cleavage and polyadenylation speci�city factor (CPSF2) and cleavage stimulation factor (CSTF3). Conclusions: The discovery of DLGAP1-AS2, a promising prognostic biomarker, reveals a new dimension into the molecular pathogenesis of CRC and provides a prospective treatment target for this disease.


Background
Colorectal cancer (CRC) is the third most common malignant carcinoma and the second leading cause of cancer-related death worldwide. The incidence of CRC is continually increasing, and it is estimated that approximately 1.9 million new CRC cases emerged and 935,000 deaths occurred in 2020 [1].The molecular pathogenesis of CRC has not been fully elucidated, and limited success has been achieved in improving the survival of CRC patients. Thus, constant efforts are required to elucidate the underlying molecular mechanisms and identify novel therapeutic targets.
Current progress in cancer transcriptomics has demonstrated that many cancer-related genes are noncoding RNAs (ncRNAs).As a major member of the ncRNA family, long ncRNAs (lncRNAs) have gained widespread attention. Gradually accumulating evidence has shown that lncRNAs participate in various physiological and pathological processes by regulating protein-protein, RNA-protein or protein-DNA interactions, as well as by sponging miRNAs [2].Some lncRNAs have been shown to contribute to CRC development and could be used as biomarkers for cancer diagnostics and therapy [3][4][5][6][7][8][9].

Vector constructs and siRNA
The DLGAP1-AS2 sequence was cloned into the lentiviral expression vector pLenti-EF1a-EGFP-F2A-Puro-CMV-MCS. The CPSF2 and ELOA sequences were cloned into the expression vector pcDNA3.1-Flag. The CSTF3 sequence or Trim21 sequence was cloned into the expression vector PCMV5-HA or pcDNA3.1-Myc, respectively. SiRNAs targeting DLGAP1-AS2, CPSF2, CSTF3, ELOA and Trim21 were purchased from GenePharma (China). The validated shRNA sequence of DLGAP1-AS2 or ELOA was synthesized and cloned into the pLKO.1 lentiviral expression vector. The promoter of LHPP was ampli ed from human genomic DNA by PCR and cloned into the pGL3-Basic vector. The related sequences are listed in Table  S2-3.

Cell proliferation and colony formation assays
Cell viability was measured with the Cell Counting Kit 8 (CCK8, Beyotime, China) according to the manufacturer's instructions. For the colony formation assay, 800 to 1,500 CRC cells were seeded into each well of a 6-well plate and maintained in medium containing 10% FBS for 10-15 days. The colonies were xed with methanol, stained with 0.1% crystal violet and counted using an inverted microscope.

Invasion and migration assays
Migration and invasion assays were performed in Transwell chambers (Corning, USA) as we previously described [9].

In vivo assays
Male athymic BALB/c nude mice, aged 5 weeks old, were injected subcutaneously with 0.2 ml of a cell suspension containing 2×10 6 CRC cells. The tumor size was measured, and the tumor volume was calculated according to the formula volume=length×width 2 ×0.5. For the in vivo metastasis model, 2×10 6 CRC cells were injected into 7-week-old male BALB/c nude mice via the tail vein. Five weeks after injection, the lung nodules in the mice were observed to measure the capability of the cells to form metastatic tumors.
RNA pull-down assays and mass spectrometry analyses RNA pull-down assays were performed using the Pierce™ Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher, USA) according to the manufacturer's instructions. The RNA pull-down samples were separated by gel electrophoresis and visualized with silver staining. Speci c bands were excised for proteomics screening by mass spectrometry analyses and retrieved from theHuman Protein Reference Database(http://www.hprd.org/). The primers for DLAGP1-AS2 and its deletion fragments for in vitro transcription are provided in Table S4.
RNA immunoprecipitation (RIP) assays RIP assays were performed using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, USA)according to the manufacturer's instructions. Cell lysates were incubated overnight at 4°C with magnetic beads conjugated to anti-CPSF2, anti-CSTF3, anti-ELOA or anti-IgG. The RIP samples were subjected to RNA extraction and subsequent RT-PCR analyses to detect the abundance of DLGAP1-AS2.
Information about the antibodies is listed in Table S5.

Immunoblotting analyses
Cells were lysed in lysis buffer (Beyotime) with protease inhibitor cocktail treatment (Roche, USA). The protein extracts were then separated by SDS-PAGE and transferred to PVDF membranes (Millipore,USA). After blocking, the membranes were incubated with primary antibodies at 4°C overnight and were then incubated with a peroxidase conjugated secondary antibody (1:5000, Thermo Fisher) for 1 h at room temperature. Finally, the membranes were visualized with ECL substrate (Vazyme).

Immunoprecipitation (IP) assay
The indicated plasmids were transfected into HCT116 cells and lysed in IP buffer (Beyotime) containing protease inhibitors. The supernatants were incubated with the beads overnight at 4°C with gentle rotation. The beads with the attached immune complexes were washed three times and analyzed by immunoblotting with the indicated antibodies.
Immunohistochemistry(IHC)staining ELOA protein levels in CRC tissues were determined by IHC. IHC staining was performed on 4-mm sections of para n-embedded tissue samples. Brie y, the slides were incubated with an anti-ELOA antibody (Santa, 1:200) at 4 °C overnight. The subsequent steps were performed using the GT Vision III Detection System/Mo&Rb (GeneTech, China).

ChIP-on-chip arrays and data analyses
The protein-DNA complexes were precipitated using an antibody against ELOA or IgG. The ChIP samples were then ampli ed, labeled and hybridized to Nimblegen human 720K RefSeq promoter arrays (Roche Nimblegen, USA). The signi cant peak regions with false discovery rates≤0.05 were mapped to the nearest genes. The fasta sequences were extracted from the promoters of the differentially expressed genes (DEGs) from ChIP data. The MEME-ChIP database and Markov model were used to search the candidate motif sites of ELOA in the promoters of the potential ELOA target genes.

Chromatin immunoprecipitation(ChIP)
ChIP was performed using the ChIP-IT Express Kit (Active Motif, USA). In brief, cells were cross-linked with 1% formaldehyde, lysed and sonicated. The sheared chromatin was immunoprecipitated with anti-ELOA or IgG followed by RT-PCR with primers designed to amplify speci c promoters.

Dual luciferase reporter assays
Luciferase reporter plasmids were co-transfected with ELOA expression plasmids into 293T cells using Lipofectamine 293(SignaGen, China). The luciferase activities of these cells were detected at 48h after transfection using the Dual-Luciferase® Reporter Assay System (Beyotime).

Statistical analyses
The data are presented as the mean±standard deviation. The statistical analyses were conducted using GraphPad Prism 8.0 (GraphPad Software, USA) and SPSS 20.0 (SPSS Inc., USA), andp<0.05 was considered to be statistically signi cant.

Results
Increased DLGAP1-AS2 expression is associated with poor clinical outcomes in CRC patients To identify potential CRC-related lncRNAs, we performed transcriptional pro ling analyses using nextgeneration sequencing in nine paired CRC and NCTs. The top 50 differentially expressed lncRNAs were veri ed with the data from the TCGA CRC cohort, and DLGAP1-AS2 showed signi cant upregulation in both CRC cohorts (Fig.1A). The aberrant upregulation of DLGAP1-AS2 was further validated in four additional CRC cohorts(GSE8671, GSE32323, GSE18105 and GSE22598, Fig.1B-C, Fig.S1). Moreover, pancancer analyses revealed that DLGAP1-AS2 was highly expressed in other types of cancer, including stomach adenocarcinoma, esophageal carcinoma, pancreatic adenocarcinoma, cholangiocarcinoma and kidney renal papillary cell carcinoma, suggesting that DLGAP1-AS2 is a key cancer-related lncRNA (Fig.1D). Thus, we focused on DLGAP1-AS2 for subsequent study.
Further experimental validation using an independent CRC cohort we collected showed that 67% (67 of 101) of CRCs showed more than 1.5-fold upregulation of DLGAP1-AS2 in CRCs compared with the adjacent NCTs ( Fig.1E-F). Kaplan-Meier survival analyses showed that high DLGAP1-AS2 expression was signi cantly correlated with poor overall survival (Fig.1G) and disease-free survival (Fig.  1H).Correlation analyses showed that DLGAP1-AS2 expression levels were correlated with tumor differentiation, lymph node metastasis and tumor stage (Table S6). Furthermore, univariate and multivariate Cox proportional hazard analyses identi ed DLGAP1-AS2 as an independent prognostic factor for CRC ( Fig.1I-J).
Identi cation of a novel transcript of DLGAP1-AS2 in CRC ThreeDLGAP1-AS2transcripts,923 bp, 1138 bp and 2261 bp in length, were listed in GENECODE or GenBank ( Fig. 2A). Interestingly, when we cloned DLGAP1-AS2 based on the only transcript provided by GenBank (NR_119377), a novel transcript with an additional 58 bp in the second exon was identi ed, which we submitted to GenBank(MK336171, Fig.2B).Further analyses using qRT-PCR and semi quantitative RT-PCR revealed that the novel transcript was the predominant transcript in CRC cells ( Fig.  2C-D).In addition, further analyses using the PhyloCSF and Coding Potential Assessment Tool (CPAT) indicated that DLGAP1-AS2 lacks protein-coding potential (Fig. S2A-B). Consequently, we focused on this new and predominant transcript for subsequent studies in CRC. DLGAP1-AS2 is highly expressed in different CRC cell lines (Fig.S2C) and distributed in both the cytoplasm and nucleus of CRC cells (Fig.  S2D).

DLGAP1-AS2 promotes CRC growth and metastasis.
CRC cells with relatively higher(HCT116 and SW480) or lower (LoVo and DLD1) DLGAP1-AS2 expression were selected for gene knockdown or overexpression and subsequent functional assays, respectively.CCK-8 and colony formation assays demonstrated that DLGAP1-AS2 knockdown signi cantly inhibited, whereas ectopic DLGAP1-AS2 expression promoted the proliferation and colony formation abilities of CRC cells ( Fig.3A-C).Transwell assays showed that DLGAP1-AS2 induction enhanced the migration and invasion of LoVo and DLD1 cells. In contrast, DLGAP1-AS2 knockdown drastically inhibited the migration and invasion activities of HCT116 and SW480 cells ( Fig.3D-E).
To further explore the growth-promoting effects of DLGAP1-AS2 on CRC in vivo, we subcutaneously injected CRC cells with stable knockdown or overexpression of DLGAP1-AS2 into nude mice. Both the volumes and weights of the xenograft tumors in the knockdown group were markedly lower than those in the control group. In contrast, ectopic DLGAP1-AS2 expression signi cantly promoted CRC tumorigenesis (Fig. 3F).
We also used a mouse lung metastasis model to evaluate the effect of DLGAP1-AS2 on CRC metastasis. The results showed that DLGPA1-AS2 knockdown drastically inhibited, whereas DLGAP1-AS2 overexpression promoted CRC metastasis (Fig.3G, Fig. S3A-B).Taken together, these data demonstrate that DLGPA1-AS2 promotes CRC growth and metastasis.

DLGAP1-AS2 interacts with CPSF2, CSTF3 and ELOA in CRC cells
To explore the molecular mechanism underlying the oncogenic role of DLGAP1-AS2 in colorectal carcinogenesis, we performed RNA pull-down assays to identify the proteins associated with DLGAP1-AS2 in CRC cells. The retrieved proteins were subjected to SDS-PAGE electrophoresis, mass spectrum and subsequent western blotting analyses. The results showed that CPSF2, CSTF3 and ELOA were potential DLGAP1-AS2-associated proteins( Fig. 4A-B, Fig.S4A-C). Moreover, RIP assays further con rmed the associations between these three proteins and DLGAP1-AS2 (Fig. 4C,Fig.S5A).
We then constructed several deletion mutants of these three proteins for RIP assays. The results showed that the deletion of 376-728 aa of CPSF2 signi cantly abolished the association between CPSF2 and DLGAP1-AS2 (Fig. 4E). Additionally, the 1-374 aa domain of CSTF3 mediates its association with DLGAP1-AS2 (Fig. 4F), and the 251-500 aa domain of ELOA physically associates with DLGAP1-AS2 in CRC cells (Fig.4G). Together, these data indicate that DLGAP1-AS2 speci cally binds to CPSF2, CSTF3 and ELOA in CRC cells.
Although we revealed that DLGAP1-AS2 interacted withCPSF2, CSTF3 and ELOA in CRC cells, their underlying functional and mechanistic effects were unclear. We evaluated the effects of DLGAP1-AS2 on the expression of these targets, and no obvious changes were observed at either the protein or mRNA levels of CPSF2 and CSTF3. The mRNA levels of ELOA also did not change inDLGAP1-AS2-depleted or DLGAP1-AS2-overexpressing CRC cells (Fig.S7A-C). However, the protein levels of ELOA were dramatically increased in DLGAP1-AS2-depleted CRC cells and were notably reduced with DLGAP1-AS2 overexpression (Fig.5A).Moreover, ectopic DLGAP1-AS2 expression decreased the half-life of the ELOA protein in CRC cells treated with the protein synthesis inhibitor cycloheximide (CHX) (Fig.5B,Fig.S8A).DLGAP1-AS2-induceddownregulation of ELOA protein was blocked in CRC cells treated with the proteasome inhibitor MG132 (Fig.5C).Furthermore, we found that the ubiquitination levels of ELOA were signi cantly increased inDLGAP1-AS2-overexpressing CRC cells and were signi cantly decreased in DLGAP1-AS2-depleted CRC cells (Fig.5D).Taken together, these data suggest that DLGAP1-AS2 promotes the proteasome-dependent degradation of ELOA in CRC cells.
To investigate howDLGAP1-AS2 accelerates the ubiquitin-mediated proteasome degradation of ELOA, we screened out six E3 ligases using ELOA IP and subsequent mass spectrometry analyses (Fig.5E). Of these E3 ligases, Trim21 showed the highest abundance, and co-IP assays further con rmed the association between Trim21 and ELOA (Fig.5F). Immuno uorescence assays also suggested that ELOA and Trim21 colocalized with each other in CRC cells (Fig.S8B). The protein levels of ELOA were dramatically increased in Trim21-depleted CRC cells and were signi cantly reduced in Trim21-overexpressing CRC cells (Fig.5G).MG132 treatment rescued the Trim21-induced downregulation of ELOA (Fig.5H), suggesting that Trim21 promotes the proteasome-dependent degradation of ELOA in CRC cells. Moreover, Trim21 knockdown increased the half-life of ELOA in CRC cells treated with CHX (Fig.5I). Furthermore, we demonstrated that the ubiquitination levels of ELOA were signi cantly decreased in Trim21-depleted CRC cells and increased in Trim21-overexpressing CRC cells (Fig.5J).
ELOA inhibits CRC growth and metastasis ELOA, a subunit of the transcription factor B (SIII) complex, is relatively understudied and its role in tumorigenesis and progression is unclear. Therefore, we studied the functional role of ELOA in CRC by using a series of in vitro and in vivo assays. We demonstrated that ELOA knockdown signi cantly promoted, whereas ectopic ELOA expression inhibited, CRC cell proliferation and colony formation ( Fig.6A-C,Fig.S8C-D). Transwell assays demonstrated the inhibitory functions of ELOA on the migratory and invasive abilities of CRC cells (Fig. 6D-E).
We further con rmed the growth-suppressive effects of ELOA on CRC in vivo by using a xenograft nude mouse model (Fig. 6F).In addition, a mouse lung metastasis model was applied to evaluate the effect of ELOA on CRC metastasis. The results showed that the number of lung metastatic nodules was decreased in the ELOA-overexpressing group compared with the control group (Fig.6G). Collectively, the above results reveal that ELOA inhibits the growth and metastasis of CRC.
ELOA protein expression negatively correlates with DLGAP1-AS2 and is associated with good prognosis in CRC To further evaluate the role of ELOA in CRC, we detected its expression in clinical CRC tissues using IHC (Fig. 7A). Kaplan-Meier survival analyses showed that low ELOA expression was signi cantly correlated with poor overall survival (Fig. 7B).Correlation analyses showed that ELOA expression levels were correlated with tumor differentiation, lymph node metastasis and tumor stage (Fig. 7C, Table  S7).Furthermore, univariate and multivariate Cox proportional hazard analyses identi ed ELOA as a dependent prognostic factor for CRC (Fig. 7D).Importantly, a signi cant negative correlation was observed between the protein expression levels of DLGAP1-AS2 and ELOA (Fig. 7E).These results suggest that ELOA is involved in CRC tumorigenesis and is regulated by DLGAP1-AS2.

ELOA is a downstream functional target of DLGAP1-AS2
To verify whether ELOA is a functional target of DLGAP1-AS2, we performed a rescue assay. The results revealed that silencing ELOA expression restored the impaired proliferation and colony formation abilities induced by DLGAP1-AS2 knockdown, whereas ectopic expression of ELOA remarkably impaired the proliferation-promoting effects of DLGAP1-AS2 overexpression in HCT116cells.Transwell assays demonstrated the ELOA rescue the promoting effect of DLGAP1-AS2 on the invasive abilities of CRC cells (Fig.7F-G, Fig.S8E-F).These data suggest that ELOA is adownstream functional target of DLGAP1-AS2.
To further investigate the mechanistic association between DLGAP1-AS2 and ELOA, we compared the transcriptome pro les in HCT116 cells transfected with si-DLGAP1-AS2, ELOA plasmid or their corresponding control. A total of 682 DEGs, including 478upregulated genes and 204 downregulated genes, were identi ed in DLGAP1-AS2-silenced HCT116 cells compared with control cells. In contrast, a total of 1420 DEGs, including 768 downregulated genes and 652 upregulated genes, were observed in ELOA-overexpressing HCT116 cells. We performed GSEA pathway enrichment analyses using these 682 and 1420DEGs and found that the pathways potentially affected byDLGAP1-AS2 or ELOA highly overlapped, including the in ammatory response, mitotic spindle, PI3K/AKT, interferon and UV response pathways (Fig. 7H). Furthermore, some genes in these pathways potentially regulated by both DLGAP1-AS2 and ELOA were veri ed using qRT-PCR. We observed that ELOA overexpression signi cantly rescued the effects of DLGAP1-AS2 knockdown on the expression of several genes in the in ammatory response, AKT and interferon pathways (Fig. 7I), suggesting that DLGAP1-AS2 regulates the expression of its downstream genes through ELOA.

ELOA transcriptionally regulates LHPP expression by speci cally binding to its promoter
To screen gene harboring the speci c binding site of ELOA in their promoters, a ChIP-on-chip assay was employed in HCT116 cells using Nimblegen human 720K RefSeq promoter arrays. A total of 836 promoters were enriched by the ChIP-on-chip assay using an ELOA antibody (false discovery rate<0.05). By combining the results of the ChIP-on-chip assay with those of the above-mentioned transcriptome pro le data, three genes (LHPP, IL7 and CMPK2) potentially regulated by ELOA were screened out. Of them, only LHPP, a tumor suppressor [16], appeared to be upregulated by ELOA and was selected for further validations. As expected, LHPP mRNA expression was signi cantly increased in ELOAoverexpressing CRC cells (Fig. 7J).
To validate the potential ELOA-binding region(-1655/-1134) revealed by the ChIP-on-chip assay, we constructed two mutants with deletion of -1034/-513 or-1655/-1134 (as a negative control) of the LHPP promoter. Luciferase assays showed that ELOA failed to stimulate the reporter expression of the constructs containing the -1034/-513 deletion of the LHPP promoter, suggesting that ELOA transcriptionally regulated LHPP expression by speci cally binding to its promoter (Fig.7K).To validate it, we performed ChIP-qPCR assays using primers anking the promoter segments of LHPP(-1034/-513). We observed a signi cant amount of ELOA bound to the -684/-513 region of the LHPP promoter (Fig.7L), con rming that ELOA enhances the transcription of LHPP by directly binding to its promoter.
Using MEME-ChIP database and Markov model, we searched the conserved sequences in the promoters of 836 genes screened by the ChIP-on-chip assay, and revealed two possible motif binding sequences of ELOA(GCTGGGATTACAGGC and CCAGCCTGGGCAACA).One of them existed in the -579/-565 region of the LHPP promoter (TGTCTGTCAGGGTGT, partially reverse complement to the sequence of CCAGCCTGGGCAACA).When this region was mutated, ELOA failed to induce luciferase expression of the recombinant reporter plasmids (Fig. 7M).Based on the above results, we conclude that ELOA promotes the transcription of LHPP by binding to the-579/-565 region of the LHPP promoter.

DLGAP1-AS2 activates the AKT pathway by regulating the ELOA/LHPP axis
We revealed thatDLGAP1-AS1 regulated several key cancer-related pathways, including PI3K/AKT (Fig.   7H-I). Interestingly, recent studies have reported that LHPP inhibits the PI3K/AKT signaling pathway in CRC [17,18]. We found that LHPP was expressed at lower levels in CRC tissues compared with NCTs based on multiple public CRC cohorts (Fig. S9A). We also showed that LHPP expression was negatively correlated with DLGPA1-AS2, and was positively correlated with ELOA in CRC (Fig. 7N,Fig. S9B).To determine whether LHPP mediated the regulation of the PI3K/AKT pathway by ELOA in CRC, we measured the AKT activity(phosphorylated AKT at Ser473) in ELOA-overexpressing CRC cells and con rmed that ELOA increased LHPP expression and suppressed AKT pathway activity (Fig. S9C). What is more, the rescue assays revealed that ELOA overexpression inhibitedDLGAP1-AS2-inducedAKT activation (Fig. 7O).Taken together, these data demonstrate that DLGPA1-AS2 promotes CRC development and progression by regulating the ELOA/LHPP/AKT signaling axis in CRC.
CPSF2 and CSTF3bind to DLGAP1-AS2and increaseits stability CPSF2 and CSTF3, which are part of the C/P machinery family, function as multi protein complexes to regulate RNA processing and stability. We have shown that both CPSF2 and CSTF3 are DLGAP1-AS2associated proteins. Analyses of public CRC databases revealed that both CPSF2 and CSTF3 were upregulated in tumor tissues compared with paired NCTs, and higher CPSF2 and CSTF3 expression was associated with worse survival (Fig. S10A-D).We observed that DLGAP1-AS2 did not affect the expression of CPSF2 and CSTF3 in CRC cells, whereas CPSF2 or CSTF3 positively regulated the expression of DLGAP1-AS2 in CRC cells (Fig.8A).Moreover, actinomycin D (2ug/ml) treatment showed that both CPSF2 and CSTF3 increased the stability of DLGAP1-AS2 (Fig.8B). Functionally, the knockdown of CPSF2 or CSTF3 remarkably reduced, whereas the overexpression of CPSF2 or CSTF3 promoted the proliferation of CRC cells (Fig. S11A-D). Furthermore, DLGPA1-AS2 overexpression rescued the decreased cell proliferation induced by the knockdown of CPSF2 or CSTF3 in CRC cells (Fig. S12A-D).As expected, CPSF2 or CSTF3 negatively regulated the expression of ELOA through DLGAP1-AS2 in CRC cells (Fig. 8C).
We further revealed that CPSF2 was able to bind to CSTF3 (Fig.8D), and their interaction was enhanced by DLGAP1-AS2 in CRC cells (Fig.8E), suggesting that DLGAP1-AS2 functions as a molecular scaffold to enhance the association between them. In addition, we showed that CPSF2 and CSTF3 might work together to increase the levels of DLGA1-AS2 (Fig.8F).Interestingly, we found a CPSF2-speci c binding motif (AAUAAA) in the 89-94 nt region of DLGAP1-AS2. When the motif was deleted (ΔAATAAA), the promoting effect of CPSF2 and CSTF3 on DLGAP1-AS2 was signi cantly suppressed (Fig. 8G). Taken together, these data suggest that CPSF2 and CSTF3 may work together to stabilizeDLGAP1-AS2 in CRC cells.

Discussion
The identi ed number of lncRNA increased rapidly with the progress of RNA sequencing technology. In our previous studies, we screened and identi ed several tumor-related lncRNAs in CRC using microarrays [7][8][9][10][11][12][13][14][15].In this study, we further screened CRC-related lncRNAs using next-generation sequencing and identi ed a cancer-promoting lncRNA, DLGAP1-AS2, is upregulated in many types of cancers, including CRC. DLGAP1-AS2 is located at chromosome 18q11.31, and a total of three transcripts of DLGAP1-AS2have been included in GENECODE or GenBank. In this study, we identi ed a novel DLGAP1-AS2 transcript that was predominantly expressed in CRC cells compared with the three known transcripts. Based on multiple CRC cohorts, we showed that DLGAP1-AS2 is a promising prognostic biomarker for CRC. Functionally, we demonstrated that DLGAP1-AS2 has strong oncogenic activity by promoting CRC tumorigenicity and progression. Recent studies also reported the upregulation and prognostic value of DLGAP1-AS2 in several cancer types, including glioma [19], gastric cancer [20], cholangiocarcinoma [21], and non-small cell lung cancer [22].Similar tumor-promoting functions have also been observed in these cancer types [19][20][21][22][23]. These ndings indicate that DLGAP1-AS2 is a potential oncogene, and may act as a pancancer therapeutic target and biomarker.
LncRNAs typically exert their biological functions through physical interactions with DNA, RNA, or proteins [24].Little was known about the molecular mechanisms of DLGAP1-AS2 in cancers. Limited studies reported that DLGAP1-AS2sponges and inhibits the activities of miR-503 [22]or miR-505 [21].In addition, a recent study reported that DLGAP1-AS2 binds to the transcription factorSix3and inhibits its occupancy in Wnt1 promoter, resulting in the activation of Wnt/β-catenin signaling in gastric cancer cells [20].In view of relative low abundance of miR-503 and miR-505 and mainly cytoplasmlocation of DLGAP1-AS2 in CRC cells, we speculated that there are other mechanisms mediating the cancerpromoting functions of DLGAP1-AS2 in CRC.
In this study, we identi ed CPSF2, CSTF3 and ELOA as bona de interacting partners of DLGAP1-AS2. By a series of experimental validations, we identi ed ELOA as a key downstream target of DLGAP1-AS2.
ELOA was originally identi ed as a transcriptionally active subunit of RNA polymerase II (Pol II) transcription factor Elongin (SIII), which stimulates the overall rate of Pol II elongation through direct interactions with the enzyme [25,26]. Elongin is composed of ELOA and a heterodimeric submodule comprised of Elongin B and C proteins, which bind to a short ELOA sequence motif referred to as the BC box that has potent transcriptional regulation activity [27]. Inaddition to being a part of SIII, ELOA also functions as the substrate recognition subunit of a Cullin-RING E3 ubiquitin ligase that ubiquitinates theRPB1 subunit of Pol II in response to DNA damage. Although Elongins B and C have been well characterized in tumorigenesis, little is known about the role of ELOA in tumorigenesis and how its activity is regulated. Here, we revealed that DLGAP1-AS2 enhances ELOA ubiquitination and degradation by promoting the binding of the E3 ligase Trim21 to ELOA, revealing a novel regulatory mechanism for ELOA at the posttranslational level. Using a series of in vitro and in vivo assays, we revealed, for the rst time, the inhibitory functions of ELOA in CRC growth and metastasis, discovering a novel role of ELOA in tumorigenesis and progression.
By expression pro le analyses, we demonstrated that DLGAP1-AS2 and ELOA share highly similar regulatory effects on the patterns of signaling pathways. Moreover, rescue assays also con rmed that ELOA is a downstream target of DLGAP1-AS2.Although ELOA was reported to be a transcription factor that induces the transcription of ATF3 and p21 under stress conditions [28], little is known about the downstream targets of ELOA in cancer cells. LHPP, a histidine phosphatase, has been suggested to be a novel tumor suppressor [29]. LHPP impedes tumor proliferation and metastasis in multiple cancer types, including CRC[18, [30][31][32][33][34][35]. For example, LHPP inhibits CRC cell proliferation through the PI3K/AKT pathway [17,18].In this study, for the rst time, we provided evidence that ELOA directly binds to the speci c promoter region of LHPP and promotes its transcription, resulting in increased LHPP expression and decreased AKT activity.
Although several studies have reported aberrant upregulation of DLAGP1-AS2 in human cancers, its underlying mechanism is unclear. The RNA-binding proteins CPSF2 and CSTF3 are parts of a multi protein complex essential for the formation of the mRNA 3' end-processing machinery [36]. CPSF2 recognizes the AAUAAA polyadenylation signal(PAS), which determines the site of 3'-end cleavage interactions within premRNAs [37], whereas CSTF3 binds GU-rich sequences located downstream from the cleavage site AAUAAA and adds a poly(A) tail for polyadenylation, resulting in increased RNA stability [38].CPSF2 and CSTF3 have been reported to be members of the cytoplasmic polyadenylation element(CPF) family as regulators of mRNA 3' end-processing machinery [39]. Here, we proved that CPSF2 and CSTF3 bind to DLGAP1-AS2 and enhance its stability, thereby inhibiting the ELOA/LHPP axis in CRC. Our data reveal a posttranscriptional regulatory mechanism that, at least partially, mediates the upregulation of DLGAP1-AS2 in CRC. However, how CPSF2 and CSTF3 regulate the stability of DLGAP1-AS2 is not clear, and whether the regulatory mechanism could be validated in other cancer types also remains to be elucidated.

Conclusion
In conclusion, we demonstrated that DLGAP1-AS2 functions as an oncogenic lncRNA in CRC by promoting tumor growth and metastasis. Mechanistically, DLGAP1-AS2 inhibits ELOA protein stability by promoting Trim21-mediated ubiquitination and degradation of ELOA. We also demonstrated that ELOA enhances the transcription of LHPP by binding to its promoter and thereby revealed uncovered a novel regulatory axis ofDLGAP1-AS2/Trim21/ELOA/LHPP in CRC. In addition, we uncovered that CPSF2 and CSTF3bind to DLGAP1-AS2 and increase its stability. These data provide insight into colorectal carcinogenesis and suggest DLGAP1-AS2 as a promising prognostic biomarker and therapeutic target for CRC (Fig.9).