L3MBTL2-mediated CGA transcriptional suppression promotes pancreatic cancer progression through modulating autophagy

Summary L3MBTL2 is a crucial component of ncPRC1.6 and has been implicated in transcriptional repression and chromatin compaction. However, the repression mechanism of L3MBTL2 and its biological functions are largely undefined. Here, we found that L3MBTL2 plays a distinct oncogenic role in tumor development. We demonstrated that L3MBTL2 repressed downstream CGA through an H2AK119ub1-dependent mechanism. Importantly, the binding of the MGA/MAX heterodimer to the E-box on the CGA promoter enhanced the specific selective repression of CGA by L3MBTL2. CGA encodes the alpha subunit of glycoprotein hormones; however, we showed that CGA plays an individual tumor suppressor role in PDAC. Moreover, CGA-transcript1 (T1) was identified as the major transcript, and the tumor suppression function of CGA-T1 depends on its own glycosylation. Furthermore, glycosylated CGA-T1 inhibited PDAC, partly by repression of autophagy through multiple pathways, including PI3K/Akt/mTOR and TP53INP2 pathways. These findings reveal the important roles of L3MBTL2 and CGA in tumor development.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDAC), among all other malignancies, is one of the primary causes of cancer-related death globally because of its characteristically deadly aftermath (Gomez-Rubio et al., 2017). The frequency of PDAC is substantially increasing worldwide, and it accounts for the worst prognosis among all digestive cancers because of its masked anatomical position, lack of effective treatment, and late clinical diagnosis. The average survival of a patient with PDAC is 2-8 months, and they only have a 6% chance of survival five years after diagnosis (Jemal et al., 2011). A deep understanding is a prerequisite to unveil the molecular mechanism, protein interactions, and genomic mutability to grasp how these factors are involved in the development and progression of pancreatic cancer, and this might be helpful for therapeutic developments.
Lethal-3-malignant brain tumor-like 2 (L3MBTL2) is an essential member of the Polycomb group proteins (PcG), and it belongs to the family of Malignant brain tumor (MBT) proteins that are associated with modification of the chromatin structure by binding to histones, which results in compaction of the chromatin (Bonasio et al., 2010;Qin et al., 2012;Trojer et al., 2007;Ogawa et al., 2002). L3MBTL2 possesses a zinc finger domain at the N-terminal end and four MBT domains, which facilitate the recognition of methylated histones (Guo et al., 2009). Notably, in contrast to L3MBTL1 that compacts nucleosomes only in the presence of methyl marks, L3MBTL2 is involved in chromatin compaction independent of histone modification (Frankenberg et al., 2007;Trojer et al., 2007Trojer et al., , 2011. In addition, L3MBTL2 functions as a transcription suppressor because it is an important component of the Polycomb suppression Complex-1 family (PRC1), which is one of the two major PcG complexes (the other is PRC2) (Qin et al., 2012;Montellier et al., 2013;Yoo et al., 2010;Trojer et al., 2011). PRC1 and PRC2 catalyze two repressive histone modifications: monoubiquitination of histone H2A at lysine-119 (H2AK119ub1) and trimethylation of histone H3K27, respectively (Tavares et al., 2012;Muller and Verrijzer, 2009). The combined activity of PRC1 and PRC2 is considered to be critical for normal Polycomb-mediated transcriptional repression. In mammals, six independent groups of PRC1 complexes (PRC1.1-PRC1.6) have been identified based on the availability of the PCGF subunit, and L3MBTL2 is a component of the noncanonical PRC1.6 (ncPRC1.6) complex (Gao et al., iScience Article 2012). Although it remains elusive how L3MBTL2 transcription represses its target genes, several recent studies have provided a mechanism by which L3MBTL2 interacts with the core constituents of PRC1.6 and facilitates recruitment of the PRC1.6 complex to the promoters of specific target genes (Huang et al., 2018b;Stielow et al., 2018).
The revealed biological functions of L3MBTL2 are still limited. L3MBTL2 has been shown to have an essential role in early embryonic development and pluripotent stem cells (Qin et al., 2012;Eid and Torres-Padilla, 2016). L3MBTL2 is highly expressed in testicular tissues and is involved in chromatin remodeling throughout meiosis and spermatogenesis (Meng et al., 2019). L3MBTL2 inhibits the DNA damage-p53apoptosis pathway and therefore has been reported to play a protective role in kidney injury and silica nanoparticle-induced spermatogenesis disorders (Huang et al., 2018a;Liu et al., 2020;Nowsheen et al., 2018). However, the other biological functions of L3MBTL2 are largely undefined, and there is no direct evidence that L3MBTL2 plays a role in cancer (Khan et al., 2015).
Glycoprotein hormone alpha subunit (CGA) encodes the alpha subunit of the glycoprotein hormone family, which includes human chorionic gonadotropin (hCG), luteinizing hormone (LH), follicle-stimulating hormone (FSH), and thyroid-stimulating hormone (TSH) (Chin et al., 1981). These four glycoprotein hormones are composed of two noncovalently bound subunits labeled a and b. In contrast to the shared same a subunit, the b-subunit confers both receptor and biological specificity that are distinct for each glycoprotein hormone (Pierce and Parsons, 1981). Therefore, the majority of studies concerning their role and mechanism in cancer have focused on the b-subunits, such as b-hCG, which was shown to promote metastasis through the ERK/MMP2 pathway in ovarian cancer (Wu et al., 2019). Aberrant expression of CGA has been detected in gastric cancer, ER-positive breast cancer, and prostate cancer (Li et al., 2014;Bieche et al., 2001Bieche et al., , 2002. However, the reasons for the aberrant CGA expression and whether CGA has an individual role in cancer are still elusive. Autophagy is an evolutionarily conserved catabolic process that targets cellular organelles and macromolecules to lysosomal compartment for degradation, which plays a major role in cellular homeostasis and participates in a variety of biological activities (Kundu and Thompson, 2008). Dysfunctional autophagy is implicated in various forms of human disease. In cancer, the role of autophagy seems to be complex and biphasic, i.e., either tumor-suppressive or tumor-promoting role depending on tumor type or contexts (White and DiPaola, 2009). Accumulating evidence demonstrates that autophagy activation is prevalent in PDAC and meditated pancreatic cancer development (Tsang et al., 2020;Zhou et al., 2020). However, different from other cancers, pancreatic cancers have a distinct dependence on autophagy. A study by Yang et al. demonstrated that autophagy is required for tumorigenic growth of pancreatic cancers de novo and the continued malignant growth during the late stages of PDAC transformation (Yang et al., 2011). Therefore, the exploration of genes regulating autophagy in PDAC may provide new strategies for treatment.
In this study, we examined the contribution of L3MBTL2 to tumor progression. We demonstrated that L3MBTL2 mediates the repression of CGA by interacting with ncPRC1.6 complexes and regulating histone modifications at the CGA promoter. Importantly, the binding of the MGA/MAX heterodimer to the E-box on the CGA promoter enhanced the specific selective repression of CGA by L3MBTL2. We further demonstrated that CGA plays an individual tumor suppressor role in PDAC, which is independent of glycoprotein hormones. We also investigated the underlying mechanism of the individual tumor suppressor role of CGA in PDAC and indicated that the function of CGA depends on its alternative pre-mRNA splicing and protein glycosylation. Collectively, we proposed that L3MBTL2-mediated CGA repression facilitates tumorigenesis and it represents a potential therapeutic strategy. (F and G), wound healing assays (F), Transwell assays (G) of PANC-1 cells expressing exogenous L3MBTL2, vector, shL3MBTL2(1,2), and shCtrl. Scale bars, 100 mm (H and I), the volumes and weight of subcutaneous tumors from the indicated groups. Data are presented as means G SD (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001, t-tests.

L3MBTL2 promotes tumor growth and metastasis in pancreatic cancer
The role of L3MBTL2 in cancer is still unclear. We first investigated the role of L3MBTL2 in PDAC, because the expression level of L3MBTL2 shows a greater discrepancy between tumor and adjacent tissues in PDAC than in other cancers ( Figure S1). According to the TCGA and GTEx database, L3MBTL2 expression was preferentially higher in pancreatic adenocarcinoma (PAAD) tissues (n = 179) than in nontumor tissues (n = 171) ( Figure 1A). To further verify the L3MBTL2 expression level in PDAC, we selected three pancreatic cancer cell lines (ASPC-1, BXPC-3, and PANC-1) and one pancreatic ductal epithelial cell line (HPDE6-C7). Quantitative PCR and western-blot detected increased L3MBTL2 mRNA and protein in PDAC cells compared to HPDE6-C7 cells, shedding light on the oncogenic role of L3MBTL2 in PDAC ( Figure 1B.).
To comprehensively investigate the function of L3MBTL2 in PDAC, we established L3MBTL2 overexpression and knockdown stably transfected PANC-1 and ASPC-1 cell lines. Transfection was confirmed by analyzing the mRNA and protein levels of L3MBTL2 ( Figure 1C). Upregulation of L3MBTL2 significantly promoted the proliferation, colony formation, cell migration and invasion capacity, and vice versa for its knockdown . In addition, we also established a subcutaneous xenograft mouse model to evaluate the role of L3MBTL2 in tumorigenesis in vivo (n = 3/group). The tumor weight and volume of the L3MBTL2 overexpression group were significantly increased, whereas L3MBTL2 knockdown had the opposite effect ( Figures 1H and 1I). Collectively, these results demonstrated that L3MBTL2 promotes cell proliferation, migration, invasion, and tumorigenesis in both PANC-1 and ASPC-1 cells.

L3MBTL2 transcriptional suppresses the expression of CGA
L3MBTL2 has been suggested to function as a transcriptional repressor. We thus hypothesized that the oncogenic function of L3MBTL2 in PDAC may be closely related to its suppressive effect on key downstream target genes. To explore its downstream pathway, we performed RNA sequencing (RNA-seq) using PANC-1-shL3MBTL2 and the corresponding control cells (Figure 2A). We identified 650 differentially expressed genes (DEGs) in shL3MBTL2 cells relative to the control group (fold changeR2, p < 0.05). Many of these DEGs, including CGA, CDKN1A, FN1, CLIC3, and S100P, have been reported to play essential roles in cancer progression ( Figure 2B). Subsequently, KEGG pathway enrichment analysis revealed that the Pathway in cancer was the top most enriched signaling pathway, and there were also other pathways closely related to tumor development, such as the PI3K-Akt signaling pathway, among the top 10 most enriched pathways ( Figure 2C). This suggests that L3MBTL2 plays a critical role in cancer, and the abnormal expression of L3MBTL2 may be involved in multiple mechanisms of tumor development. Therefore, we assumed that the target gene of L3MBTL2 might be a tumor suppressor gene with a certain role in cancer. However, we found that among the DEGs, one of the genes most strongly upregulated in response to L3MBTL2 knockdown was a hormone-related gene (CGA), whose role in cancer is still unclear. This aroused our great interest. Subsequently, we confirmed that L3MBTL2 remarkably negatively regulated CGA expression at the mRNA and protein levels ( Figure 2D). Notably, two CGA bands appeared in shL3MBTL2 cells, whereas there is only single band in other cells ( Figure 2D). According to the NCBI database, CGA has two transcripts. Transcript 1 is a full-length transcript of 147 amino acids, and transcript 2 lacks 32 amino acids (30E$61D). Besides, CGA encodes the alpha subunit of glycoprotein hormones, which is a glycosylated protein containing 92 amino acids with two N-glycosylation sites (107N and 133N). We hypothesized that L3MBTL2 may also inhibit the glycosylation of CGA. To test this, we established four stable overexpressed cell lines that overexpressed CGA transcript 1(CGA-T1), CGA transcript 2 (CGA-T2), CGA-Transcript1-N107Q-N133Q (CGA-T1-NQ), and CGA-Transcript2-N76Q-N102Q (CGA-T2-NQ), respectively ( Figure 2E). CGA-T1-NQ and CGA-T2-NQ are mutations at two same N glycosylation sites. We found that CGA-T1 is the primary transcript in PANC-1 cells according to the protein weight ( Figure 2E). Furthermore, the shL3MBTL2 promoted the expression of total CGA while also possibly promoting the glycosylation of CGA through other mechanisms. This explains why two CGA bands appeared in shL3MBTL2 cells. In addition, to evaluate the transcriptional regulatory role of L3MBTL2 on CGA, a luciferase promoter-reporter assay was performed. L3MBTL2 overexpression significantly decreased the activity of the CGA promoter, whereas L3MBTL2 knockdown had the opposite effect ( Figure 2F). These results demonstrate that CGA is a direct target of L3MBTL2, which transcriptionally suppresses the CGA in pancreatic cancer.

L3MBTL2 represses CGA by mediating ubiquitination of H2A
L3MBTL2 functions as a transcriptional repressor as L3MBTL2 has been reported to interact with the core components of ncPRC1.6 and contribute to the recruitment of this complex to the promoter of target iScience Article genes. To investigate the local impact of L3MBTL2 on chromatin modifications at the CGA promoter, we performed ChIP analysis. All PRC1 complexes, including PRC1.6, contained RING1B, which catalyzes H2AK119ub1 and thereby contributes to chromatin compaction and transcriptional silencing. Therefore, we investigated whether L3MBTL2 represses CGA by mediating H2AK119ub1. As expected, ChIP-qPCR analysis revealed that H2AK119ub1 at the CGA promoter was markedly increased after overexpression of L3MBTL2, revealing a role for L3MBTL2 in CGA repression by mediating H2AK119ub1 deposition (Figure 3A). In addition, histone acetylation has been suggested to be involved in transcriptional gene regulation, and HDACs have been identified to interact with L3MBTL2 (Qin et al., 2012;Yoo et al., 2010). We therefore investigated whether L3MBTL2-mediated CGA inhibition was related to histone acetylation. The expression of HDAC1, a histone deacetylase, was increased after L3MBTL2 overexpression, indicating that histone acetylation (H3Ac and H4Ac) at the CGA promoter was decreased. Indeed, we confirmed the decreased acetylation changes by detecting H3Ac on CGA promoter. Moreover, H3K27me3 was iScience Article increased at the CGA promoter. The increase of H3K27me3 may be the result of an increase in H2AK119ub1, which has been indicated to recruit PRC2 and downstream H3K27me3 deposition to effectively initiate a Polycomb domain (Blackledge et al., 2014). Other histone modifications, such as H3K9me3 and H3K36me3, were virtually unchanged at the CGA promoter. Collectively, the upregulation of L3MBTL2 resulted in a significant increase in H2AK119ub1 on the CGA promoter, whereas the changes in histone acetylation and H3K27me3 were relatively minor.
To gain insight into the functional involvement of L3MBTL2-mediated H2AK119ub1, acetylation, and H3K27me3 deposition in CGA transcription, we further rescued these histone modifications upstream of the TSS of the CGA promoter and evaluated the effects of their alterations on the CGA transcription level. A CRISPR-Cas9-based fusion system was applied to modify histone modifications of the CGA promoter. USP16, a deubiquitinating enzyme, was fused with the dCas9 protein by a GSG flexible connection linker (dCas9-USP16), which decreased H2AK119ub1 at the target loci. In addition, 3 independent gRNAs were designed to lead dCas9-USP16 to upstream of the TATA box of the CGA promoter. Furthermore, we combined three gRNA transcription frames into a single vector, named Rainbow gRNA, to shield the problem of uneven efficiency in co-transfection ( Figures 3B and S3). The specificity and effectiveness of gRNAs were verified by co-transfected dCas9-USP16 and gRNAs into 293T RING1BÀOE cells and analyzed the H2AK119ub1 levels at the CGA promoter by ChIP-qPCR ( Figure S4).
L3MBTL2-overexpressing PANC-1 cells transfected with dCas9-USP16 alone or co-transfected with a scrambled gRNA barely affected L3MBTL2-mediated CGA repression. Notably, L3MBTL2-overexpressing cells co-transfected with the dCas9-USP16 fusion construct. and each of the 4 gRNAs failed to maintain transcriptional repression, indicating that L3MBTL2-mediated repression of CGA was significantly abolished by the loss of H2AK119ub1 deposition at the CGA promoter ( Figure 3C). Subsequently, the fusion of dCas9 with histone demethylase KDM6B, histone deacetylase HDAC3, and acetyltransferase EP300 was constructed in a similar way to decrease H3K27me3 and to reduce and increase histone acetylation (H3Ac and H4Ac) at the CGA promoter, respectively. The results showed that the CGA repressed by L3MBTL2-OE (L3MBTL2-overexpressing cells) was partially rescued by the increase in histone acetylation at the CGA promoter ( Figure 3D). In contrast, the reduced histone acetylation at the CGA promoter inhibited CGA transcription ( Figure 3E). However, compared with the fold change of CGA expression affected by histone ubiquitination, the effect of histone acetylation on CGA was obviously weaker. In addition, decreased H3K27me3 at the CGA promoter induced by dCas9-KDM6B transfection of L3MBTL2-OE cells did not impact CGA transcription, suggesting that L3MBTL2-mediated CGA repression is H3K27me3 independent ( Figure 3F). Collectively, L3MBTL2-mediated CGA repression was significantly abolished by a loss of H2AK119ub1 deposition at the CGA promoter but was partially rescued by increased histone acetylation independent of H3K27me3. Thus, we suggested that L3MBTL2 represses CGA mainly by regulating H2AK119ub1, whereas changes in acetylation (H3Ac and H4Ac) and H3K27me3 may contribute to optimal L3MBTL2-mediated repression of CGA.
The E-boxes on the CGA promoter are essential for specific repression of CGA by L3MBTL2 We proposed that L3MBTL2 represses CGA by mediating histone modifications; however, the mechanism by which L3MBTL2 targets specific loci is not clear. L3MBTL2 selectively represses specific genes by recruiting the ncPRC1.6 complex to the promoters of its specific target genes. L3MBTL2-dependent ncPRC1.6 binding sites have been reported to be enriched for bHLH E-box motifs (Stielow et al., 2018). L3MBTL2 facilitates PRC1.6 binding site selection by promoting binding of the MGA/MAX heterodimer to the E-box or T-box on the target promoter. To examine this mechanism in L3MBTL2-mediated CGA repression, we mutated three E-boxes of CGA ( Figure S3). We compared the ability of L3MBTL2 to repress different reporter constructs carrying different combinations of mutated E-boxes ( Figure 4A). The results showed Overexpression of L3MBTL2 correlates with increased H2AK119ub1, HDAC1, and H3K27me3 at the CGA promoter but does not affect H3K9me3 and H3K36me3. ChIP of IgG control antiserum, HDAC1, H3Ac, H3K27me3, H3K9me3, and H3K36me3 (left to right) followed by qPCR analysis of CGA in PANC-1 cells.

OPEN ACCESS
iScience 25, 104249, May 20, 2022 7 iScience Article that the three E-boxes cooperate in L3MBTL2-mediated CGA repression and that triple-mutant constructs significantly blocked L3MBTL2-mediated CGA repression, indicating that the E-boxes on the CGA promoter are essential for specific selective repression of CGA by L3MBTL2.
Collectively, L3MBTL2 represses CGA via an H2AK119ub1-dependent mechanism, and the E-box on the CGA promoter determined it to be a specific target of L3MBTL2. Thus, the interaction of L3MBTL2, RING1B, and MAX should play a crucial role in suppressing CGA. To further verify our statement, we applied the PUP-IT (pupylation-based interaction tagging) proximity-tagging system  to study the interactome of the CGA promoter. PafA was fused with the CRISPR-dxCas3.7 protein (dxCas3.7-PafA), and gRNAs were designed to lead dxCas3.7-PafA to the position near the TSS sequence of the CGA promoter. In addition, we established a Dox inducible cell line based on L3MBTL2 overexpression cells (PANC1-L3MBTL2-OE -Pup(E) TET-ON ) for controlling the expression of Pup(E). The PANC1-L3MBTL2-OE-Pup(E) TET-ON cells were co-transfected with dxCas3.7-PafA and gRNAs, followed by dox induction. In this case, the Pup(E)-modified proteins are the ones that interact with the CGA promoter. We observed a high-molecular-weight band in the co-transfection of dxCas3.7-PafA and gRNAs groups, suggesting Pup(E)-modified L3MBTL2, MAX, and RING1B ( Figure 4B). By contrast, cells transfected with dxCas3.7-PafA and gRNAs alone lacked the Pup(E)-modified proteins. Therefore, these results reveal that L3MBTL2, MAX, and RING1B are colocalized to the CGA promoter, suggesting that these three proteins may form protein complexes and synergistically inhibit CGA expression. In summary, we propose a mechanism that MGA/MAX assists L3MBTL2 in targeting specific CGA promoters, and L3MBTL2 interacts with ncPRC1.6 core components such as RING1B and HDACs, thereby inhibiting CGA transcription by regulating the histone H2AK119ub1, acetylation (H3Ac and H4Ac), and H3K27me3.   Figure 5A). To further verify the decreased CGA expression level in PDAC, we analyzed the protein levels of CGA in selected three PDAC cell lines (ASPC-1, PANC-1, and BXPC-3) and one pancreatic ductal epithelial cell line (HPDE6-C7). The western blot detected a significantly decreased CGA expression level in PDAC cells compared to HPDE6-C7 cells, suggesting the tumor suppressor role of CGA in PDAC ( Figure 5B). To determine the role of CGA in PDAC and further validate its association with L3MBTL2, we stably overexpressed CGA into PANC-1 cells and into PANC-1 cells overexpressing L3MBTL2. CGA overexpression significantly attenuated cell proliferation, colony formation, migration, and invasive abilities ( Figure 5C and 5E-G) compared with the corresponding control. In addition, elevated CGA expression inhibited tumor growth in subcutaneous transplantation models in nude mice ( Figure 5D). Simultaneously, we found that the tumor-promoting capacity of L3MBTL2 could be markedly abolished by CGA overexpression (Figures 5H-5K), whereas the loss of CGA could also rescued the antitumor effect of L3MBTL2 knockdown ( Figure S5). In summary, these results reveal that the oncogenic function of L3MBTL2 in pancreatic cancer is closely related to its negative regulatory effect on CGA, which plays a tumor suppressor role in PDAC.
Many studies have presented the critical roles of glycoprotein hormones in the diagnosis, monitoring, and treatment of different types of cancer, including breast, gastric, prostate, thyroid, and many other cancers (Kolbl et al., 2018;Crawford and Schally, 2020;Dizeyi et al., 2019;Song et al., 2019;Zhao et al., 2018;Wu et al., 2019). CGA encodes the alpha subunit of glycoprotein hormones, including hCG, LH, FSH, and TSH. Therefore, we wondered whether the role of CGA in PDAC is associated with these four glycoprotein hormones. However, according to the qPCR results, CGA overexpression was not accompanied by overexpression of CGB, LHB, TSHB, or FSHB to produce ectopic hormones in PDAC ( Figure 5L). In addition, loss of CGB, LHB, TSHB, and FSHB barely affected the antitumor effect of CGA overexpressing in PANC-1 cells ( Figure S6). Therefore, CGA might have an individual role in PDAC and is independent of ectopic glycoprotein hormone production.

Glycosylated CGA isoform 1 plays a major role in PDAC
Having established the function of CGA in inhibiting tumor progression, we next sought to determine its underlying mechanisms. CGA has two transcripts, and CGA is a glycosylated protein containing two N-glycosylation sites. We wonder about the function of the two transcripts of CGA in PDAC, and whether the tumor suppressive function of CGA is related to protein glycosylation. To test this, we compared the function of four established stable overexpressed cell lines (CGA-T1, CGA-T2, CGA-T1-NQ, and CGA-T2-NQ) in PDAC progression. CGA-T1 was identified as the major transcript that functions in PDAC, which promoted the proliferation and migration of PDAC cells, whereas CGA-T2 barely affected PDAC progression ( Figures 6A and 6B). In addition, mutation of N glycosylation sites of CGA-T1 abolished the suppressor function of CGA-T1 ( Figures 6C and 6D), indicating that the function of CGA-T1 depends on its own glycosylation. Meanwhile, the CGA-T2 and CGA-T2-NQ cells did not present any function in PDAC ( Figure S7). Collectively, the glycosylated CGA isoform 1 plays a major role in PDAC.

Glycosylated CGA-T1 inhibits autophagy through multiple pathways
To further investigate the mechanism of the individual tumor suppressor role of CGA-T1 in PDAC, we performed RNA-Seq to analyze the gene expression profiles of PANC-1-CGA-T1-OE compared with control cells. We identified 257 upregulated genes and 174 downregulated genes after CGA overexpression iScience Article ( Figure 7A). Among the top 15 enriched KEGG pathways, there were several pathways closely related to tumor development, including focal adhesion, Hippo signaling, PI3K-Akt signaling, and the TGF-beta signaling pathway ( Figure S8). These results suggest that CGA has individual biological functions and may therefore be involved in multiple tumorigenesis mechanisms.
Among the DEGs, TP53INP2/DOR (tumor protein p53 inducible nuclear protein 2) is one of the most strongly downregulated genes in response to CGA-T1 overexpression. TP53INP2 has been shown to participate in many cancers and is involved in multiple stages of autophagy. Previous studies show that upon starvation, TP531NP2 interacts with nuclear LC3 and redistributes LC3 by shifting from the nucleus to the cytoplasm, further facilitating autophagosome formation by enhancing the LC3-ATG7 interaction (Huang et al., 2015;You et al., 2019). In addition, the autophagy activation is prevalent in PDAC and promotes the development of pancreatic cancer (Tsang et al., 2020;Yang et al., 2011;Li et al., 2016). We speculated that the role of CGA-T1 in PDAC might be associated with the TP53INP2-mediated autophagy process. To test our hypothesis, we confirmed the significant negative regulation of TP53INP2 by CGA-T1 ( Figure 7B) and subsequently investigated the effect of CGA-T1 on autophagy. According to western blot assays, the expression levels of LC3II decreased markedly in CGA-T1 overexpressed PANC-1 cells, whereas p62 protein expression was significantly increased compared with the control cells ( Figure 7C).
To confirm the effect of CGA-T1 on autophagy, an mRFP-GFP-LC3B construct was used to detect the autophagic flux. As shown in Figure 7H, immunofluorescence assay revealed a significant decrease in both autophagosome and autolysosome in CGA-T1-overexpressed cells ( Figure 7H). These results indicated that CGA-T1 inhibited the autophagy process. Considering the L3MBTL2 strongly suppresses transcription of CGA, we also verified the inhibitory effect of shL3MBTL2 on autophagy ( Figures 7D and S11). Notably, the reverse trend was observed in CGA-T1-NQ cells, where overexpression of CGA-T1-NQ increased LC3II expression and autophagosomes, which promoted autophagy ( Figure 7C). The results suggested that CGA-T1 prevents the autophagy depending on its own glycosylation. These results are consistent with our previous conclusion that glycosylated CGA-T1 plays a major role in PDAC. iScience Article Furthermore, the immunofluorescence assay showed that LC3 distribution in CGA-T1-overexpressed cells was largely enriched in the nucleus ( Figure 7H), suggesting that CGA-T1 might influenced the translocation efficiency of LC3 from the nucleus to the cytoplasm. This might be because of CGA-T1 preventing TP53INP2 from assisting LC3 movement from the nucleus to the cytoplasm. Subsequently, we confirmed the LC3-TP53INP2 and LC3-ATG7 interaction in PANC-1 cells. In addition, the decreased PupE-modified bands indicated that CGA-T1 overexpression reduced the LC3-TP53INP2 and LC3-ATG7 interaction ( Figures 7E and 7F). The decreased interaction can be explained as CGA overexpression leads to reduced TP53INP2 expression, decreasing participation in LC3-TP53INP2 interactions, and diminishing the promotion effect of TP53INP2 on LC3-ATG7 interaction. These results support our hypothesis that glycosylated CGA-T1 inhibits autophagy partially through the TP53INP2-mediated pathway.
Although TP53INP2 plays an important role in the early stage of autophagy, TP53INP2 alone is not enough to support the significant inhibition of CGA on autophagy. Therefore, we suspect that the effect of CGA on autophagy is not limited to the regulation of TP53INP2. According to the RNA-seq results, we found that among the enriched KEGG pathways ( Figure S6), PI3K-Akt-mTOR signaling pathway is frequently activated and plays an important role in the pathogenesis of PDAC (Ebrahimi et al., 2017). As a critical modulator of autophagy, PI3K/AKT/mTOR-mediated autophagy is involved in many cancers, including PDAC (Ebrahimi et al., 2017;Liu et al., 2018a;Xu et al., 2020;Zhang et al., 2019). Therefore, we assumed that CGA-T1 might influence autophagy partially by regulating the PI3K/AKT/mTOR-mediated autophagy process. Afterward, we investigated the effect of CGA-T1 on the central checkpoints of PI3K-Akt-mTOR-mediated autophagy, mTOR. According to western blot assays, CGA-T1 overexpression upregulated the expression levels of phospho-Akt (p-Akt) and phospho-mTOR (p-mTOR), whereas CGA-T1-NQ had no effect ( Figure 7G). Furthermore, we used the mTOR inhibitor Rapamycin (Rap), confirmed the autophagy promoted by Rap alone in PDAC, and verified the relationship between CGA and the mTOR pathway in autophagy (Figure S9). These results indicated that glycosylated CGA-T1 inhibits the initiation of autophagy partially through the PI3K/Akt/mTOR pathway. Besides, we do not exclude CGA-T1 inhibiting autophagy through other mechanisms, such as modulates other key proteins, and further studies are needed. Collectively, we demonstrated that the tumor suppressor role of glycosylated CGA-T1 in PDAC is associated with its negative regulation of autophagy through multiple pathways, including PI3K/Akt/mTOR and TP53INP2 pathways.

L3MBTL2 plays oncogenic roles in lung and hepatology carcinoma
Data from the GEPIA database indicated that L3MBTL2 mRNA levels are significantly upregulated in various types of cancers, including lung cancer, liver cancer, breast cancer, skin cutaneous melanoma, and stomach adenocarcinoma ( Figure S1). To explore the function of L3MBTL2 in other cancers, we established L3MBTL2 knockdown stably transfected H1299 and HepG2 cell lines. L3MBTL2 knockdown significantly decreased cell proliferation, migration, and invasion in H1299 (Figures S10A-S10C) and HepG2 iScience Article (Figures S10D-S10F) cells. These data suggest that L3MBTL2 plays an oncogenic role in lung and hepatology carcinoma and it may also be critical for other cancers.

DISCUSSION
L3MBTL2 is a crucial component of a noncanonical PRC1 subgroup termed polycomb repressive complex 1.6 (PRC1.6), and it has been implicated in transcriptional repression and chromatin compaction (Gao et al., 2012). However, the mechanism by which L3MBTL2 represses its target genes is still unclear, and its biological functions are largely undefined. Our study shows for the first time that L3MBTL2 functions as an oncogene, significantly promoting cell proliferation, migration, invasion, and tumorigenesis in pancreatic cancer. We also investigated the effects of L3MBTL2 in liver (HepG2) and lung (H1299) cancer cells, indicating a critical oncogenic role for L3MBTL2 in tumor initiation and progression. The effect of L3MBTL2 may involve transcriptional repression of the expression of the downstream target gene CGA, and the E-box of CGA plays a critical role in specific selective repression of CGA by L3MBTL2. Mechanistically, we suggested that L3MBTL2 represses CGA mainly by regulating H2AK119ub1, whereas changes in acetylation (H3Ac and H4Ac) and H3K27me3 may contribute to optimal L3MBTL2-mediated repression of CGA. In addition, we demonstrated CGA acts as a tumor suppressor gene in PDAC and verified that the oncogenic function of L3MBTL2 is closely related to its regulatory effect on CGA. Furthermore, we also explored the mechanism of CGA in PDAC, which is glycosylated CGA-T1 inhibits multiple stages of autophagy through multiple pathways. Thus, we propose the L3MBTL2/CGA pathway in PDAC, which may represent a potential therapeutic strategy for treatment.
Polycomb group (PcG) proteins assemble into two major repressive complexes-PRC1 and PRC2-that catalyze two repressive histone modifications: H2AK119ub1 and H3K27me3. The canonical view suggests that H3K27me3 is required for the recruitment of PRC1 to genomic sites (Simon and Kingston, 2009). However, recent studies have suggested that H2AK119ub1 deposition by noncanonical PRC1 facilitates the recruitment of PRC2 and downstream H3K27me3 (Blackledge et al., 2014;Cooper et al., 2014). L3MBTL2 is a component of the ncPRC1.6 complex, which also contains other core components, such as MGA, RING1B, HDAC1/2, PCGF6, and E2F6. In this study, ChIP-qPCR analysis revealed that chromatin modifications, including H2AK119ub1, acetylation, and H3K27me3, at the CGA promoter were changed after overexpression of L3MBTL2, which might be associated with the interaction between L3MBTL2 and the associated enzymes (RING1B and HDAC1). Moreover, our data indicated that L3MBTL2-mediated CGA repression was significantly abolished by a loss of H2AK119ub1 deposition at the CGA promoter, but it was partially rescued by increased histone acetylation. In addition, the decreased H3K27me3 at the CGA promoter did not impact CGA transcription, suggesting that L3MBTL2-mediated CGA repression is H3K27me3 independent. Our study is consistent with a previous study claiming that ncPRCs target chromatin by H3K27me3-independent mechanisms (Blackledge et al., 2014). Collectively, we demonstrated that L3MBTL2 represses CGA transcription mainly by regulating H2AK119ub1, whereas changes in acetylation (H3Ac and H4Ac) and H3K27me3 may contribute to optimal L3MBTL2-mediated repression of CGA. It will be of interest to determine whether this represents a general mechanism by which L3MBTL2 represses target genes.
L3MBTL2 mediates transcriptional repression of specific target genes. However, the mechanism by which L3MBTL2 targets specific loci remains unclear. A recent study indicated that L3MBTL2-mediated genomic targeting is closely related to its interaction with the PRC1.6 complex. L3MBTL2 facilitates PRC1.6 binding site selection by promoting and stabilizing the binding of the MGA/MAX heterodimer to the E-box or T-box in the target sequence (Stielow et al., 2018). In our study, we confirmed the critical role of the E-box in the CGA promoter in specific selective repression by L3MBTL2. Cells transfected with the reporter constructs carrying triple-mutant E-boxes significantly blocked L3MBTL2-mediated CGA repression. L3MBTL2 interacts with the core components of the PRC1.6 complex, making it critical for transcriptional repression and genomic targeting of PRC1.6. Indeed, we applied the PUP-IT proximity-tagging system to identify the proteins that interacted with the CGA promoter. The results showed that L3MBTL2, MAX, and RING1B are all localized to the CGA promoter. However, other mechanisms that may determine genomic targeting and transcriptional repression still need to be further investigated.
CGA encodes the alpha subunit of glycoprotein hormones that includes hCG, LH, FSH, and TSH. However, in this study, we demonstrated that CGA plays a role as a tumor repressor that is independent of the hormones it is typically associated with in PDAC. CGA overexpression significantly inhibited the proliferation ll OPEN ACCESS iScience 25, 104249, May 20, 2022 iScience Article rate and attenuated the migration and invasive abilities of PDAC cells. In addition, we found no correlation between CGB, LHB, TSHB, or FSHB mRNA levels and CGA mRNA overexpression. Thus, we propose that CGA plays an individual role and is independent of ectopic glycoprotein hormone production. However, the underlying mechanism is still unclear. Interestingly, we found the two transcripts of CGA have different functions, and the tumor suppressive function of CGA is related to the protein glycosylation. In two of the CGA transcripts, CGA-T1 was identified as the primary transcript that functions in PDAC, whereas mutation of N glycosylation sites of CGA-T1 abolished the suppressor function of CGA-T1, indicating that the tumor suppressor function of CGA-T1 depends on its own glycosylation. Collectively, the function of CGA in PDAC depends on its alternative pre-mRNA splicing and protein glycosylation. In addition, we found that the shL3MBTL2 upregulated the transcription of total CGA and also promoted the glycosylation of CGA. We speculated that L3MBTL2 might transcriptionally repress glycotransferase or have other complex biological functions that influence the CGA glycosylation process, which need to be studied further.
In the further mechanism study, we found that CGA-T1 inhibited autophagy, characterized by decreased LC3-II, upregulated p62 expressions, and decreased autophagosomes in the cytoplasm. In addition, glycosylated CGA-T1 negative regulated the TP53INP2 expression and reduced the LC3-TP53INP2 and LC3-ATG7 interaction, suggesting that CGA-T1 inhibits the autophagy through TP53INP2 pathway. Meanwhile, we found that CGA-T1-NQ promoted autophagy, while barley affecting TP53INP2 expression level. Therefore, we suspect that the effect of CGA on autophagy is not limited to a single stage. Indeed, we observed that glycosylated CGA-T1 enhanced the p-AKT and p-mTOR expression, indicating that CGA-T1 may inhibit the initial stage of autophagy. Therefore, the inhibition effect of CGA-T1 on autophagy might involve multiple stages; however, the specific molecular mechanism needs to be further studied. Moreover, our study demonstrated that the glycosylated CGA-T1 plays a major role in PDAC. Thus, the correlation between CGA glycosylation and autophagy also needs to be studied in the future. In addition, glycoprotein hormones have been demonstrated to play a role in many types of cancers. It is worth knowing whether CGA exhibits an independent tumor suppressor effect in other cancers. If CGA plays a universal role in tumor suppression, it will provide a new strategy and therapeutic targets for cancer treatment.
In summary, our study revealed a distinct oncogenic role of L3MBTL2 in tumor development and an L3MBTL2-mediated transcriptional repression mechanism. We demonstrated that L3MBTL2 interacted with ncPRC1.6 core components to repress downstream CGA via an H2AK119ub1-dependent mechanism, and the E-box on the CGA promoter determined it to be a specific target of L3MBTL2. We further validated that CGA plays an individual tumor suppressor role in pancreatic cancer. Indeed, CGA-overexpressing cells showed attenuated proliferation, migration, and invasive abilities, which were independent of glycoprotein hormones. Moreover, CGA-T1 was identified as the major transcript in PDAC, and the tumor suppressor function of CGA-T1 depends on its own glycosylation. Furthermore, glycosylated CGA-T1 inhibited PDAC might be partly through negative regulated autophagy through multiple pathways, including PI3K/Akt/mTOR and TP53INP2 pathways ( Figure 7I). The discovery of the L3MBTL2/CGA axis and its impact on autophagy and PDAC progression provides a way to explore more efficient therapeutic strategies.

Limitations of the study
Despite detecting a significant transcriptional suppression of CGA by L3MBTL2, we proposed the mechanism by mediating histone modifications, which is closely related to the ncPRC1.6 complex in which L3MBTL2 embedded. However, the mechanism by which L3MBTL2 collaborates with PRC1.6 members to rapidly and selectively target specific genes needs to be further explored. In addition, we found that the shL3MBTL2 upregulated the transcription of total CGA and also promoted the glycosylation of CGA. We speculated that L3MBTL2 might transcriptionally repress glycotransferase or have other complex biological functions that influence the CGA glycosylation process, which need to be studied further. Moreover, we showed the antitumor function and regulatory mechanism of CGA mainly in PDAC cell lines; however, CGA is a secreted protein. Thus, the function of secreted CGA and its influence on function of cellular CGA are still unclear, and more in vivo experiments are needed. Furthermore, we observed that glycosylated CGA-T1 showed opposite effects on the autophagy process compared with CGA-T1-NQ cells. Therefore, the correlation between CGA glycosylation and autophagy might be a critical element and needs to be studied in the future.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
iScience Article RESOURCE AVAILABILITY

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Hua Huang (huangh@bjut.edu.cn).

Materials availability
This study did not generate new unique reagents.

Data and code availability
The RNA-seq data generated during this study are available at Gene Expression Omnibus (GEO): GSE184834. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Animal studies
Six-week-old female Balb/c nude mice were purchased from Charles River (Beijing, China) and were housed under a 12-h dark, and 12-h light conditions and a standard temperature of 18-23 C with humidity of 40-60%. For in vivo tumorigenicity assays, PANC-1 cells (L3MBTL2-OE, sh L3MBTL2, CGA-OE and the corresponding control cells) in 200 mL PBS were subcutaneously injected into the right and left flanks of mice with cell density of 2310 6 cells per site. Following 5-6 weeks growth, the mice were anesthetized with 1% pentobarbital sodium (45 mg/kg, intraperitoneal) and subsequently euthanized by cervical dislocation. The tumor weight was measured and the volume was calculated at the ending point. All animal experiments in this study were approved by the Ethics Committee of Beijing University of Technology (protocol number HS202105004).

METHOD DETAILS
Quantitative real time-PCR Cells were harvested, and TRIzolâ Reagent (Invitrogen, USA) was employed to extract total RNA following manufacturer's instructions. Reverse transcription was performed using the HiScript III 1st Strand cDNA Synthesis Kit (Vazyme Biotech Co., Nanjing, China) according to the manufacturer's protocol. Gene expression was evaluated by qPCR using the ChamQ SYBR Color qPCR Master Mix (Vazyme Biotech Co., Nanjing, China) and run on the ABI STEPONE system (Applied Biosystems, Foster City, CA, USA). The thermocycling protocol included an initial denaturation step at 95 C for 3 min, followed by 40 cycles of denaturation at 95 C for 15 sec, annealing at 60 C for 15 sec and extension at 72 C for 30 sec. The 2-DDCq method was calculated to analyze the relative changes in gene expression according to the previous reported method (Livak and Schmittgen, 2001). b-Actin was used as an internal control. Primers used in this study are listed in Table S1.

Western blotting
Proteins were extracted using cell lysis buffer RIPA (Beyotime, Shanghai, China) and a protease and phosphatase inhibitor cocktail (Beyotime, Shanghai, China). A total of 20 mg protein/per lane was separated by 4-20% SDS-PAGE gels and subsequently transferred onto a membrane and blocked with 5% skim milk (Difco; BD Biosciences) at room temperature for 2 h. The membranes were then subjected to immunoblotting with appropriate primary at 4 C overnight and secondary antibodies for 2 h at room temperature. Luminescence was visualized using a BeyoECL Moon super sensitivity detection kit (Beyotime, Shanghai, China) according to the manufacturer's protocol.

Lentivirus transfection
The the coding DNA sequences (CDSs) of human L3MBTL2 and CGA were subcloned into the pCDH-EF1a-MCS-T2A-Puro lentivirus vector (System Biosciences: CD527A-1)). The shRNA sequences of L3MBTL2, CGA and their negative control were cloned into the pLVX-shRNA2-BSD lentivirus vector upgraded by pLVX-shRNA2 (Clonetech: 632179) (in which the ZsGreen gene sequence is replaced with a BSD resistance gene sequence). Each of the purified plasmids was packaged into a lentivirus and then transfected into PANC-1, ASPC-1, HepG2 and H1299 cells. The lentivirus packing and transfection methods are same as previously reported . Then, positive overexpressed and knockdown cells were selected using 2 mg/mL puromycin or 5 mg/mL blasticidin S for 10 days.

Promoter luciferase reporter assay
A promoter sequence 500 bp in length, ranging from À406 base site to the +94 base site of the CGA gene, were copied into the modified pGL4.

Mutant CGA reporter constructs
Three 5'-CANNTG (one 5'-CACCTG and two 5'-CAGGTG) sequences of E-box were mutated to 5'-AANNTA respectively. The mutant reporter constructs used carry the CGA promoters with combinations of the three mutated core sequences. Each combination of mutations was confirmed by sequencing.

Detection of protein complexes on CGA promoters
The Biotin-Pup(E) was subcloned into the Tet-On lentivirus vector which contains blasticidin S resistance. The plasmid was packaged into a lentivirus and then transfected into L3MBTL2-OE PANC-1 cells for 48 ll OPEN ACCESS iScience 25, 104249, May 20, 2022 iScience Article hours. Then, positive PANC1-L3MBTL2-OE -Pup(E)TET-ON cells were selected using 5 mg/mL blasticidin S for 7 days. Afterwards, a CRISPR-dxCas3.7-based fusion system was established by fusion PafA to the C-terminal tail of full-length dxCas3.7 by a flexible GSG connection linker (dxCas3.7-PafA). The PANC1-L3MBTL2-OE -Pup(E)TET-ON cells were applied to nucleofection with 4 mg dxCas3.7-PafA plasmids and 2 mg gRNA donor plasmid according to the nucleofection method mentioned previously. After 24 hours, 2 mg/ml doxycycline (Dox) was used for another 24 h for Pup(E) induction. The gRNA sequences for targeting CGA promoter are listed in the key resources table.
Detection of interaction of LC3 with ATG7 and TP53INP2 LC3B coding sequences was subcloned to the transient vector fused with PafA coding sequence, and the downstream PupE sequence was connected to PafA via an IRES element. PANC1-CGA-T1-OE, PANC1-CGA-T1-NQ-OE and the corresponding control cell line were applied to nucleofection with 4 mg LC3B-PafA-IRES-PupE plasmids according to the nucleofection method mentioned previously. After 36 hours, cells were collected for protein purification and western-blot detection with corresponding antibodies.

Chromatin Immunoprecipitation (ChIP)
To perform ChIP experiments, we utilized PANC-1 cells transfected with L3MBTL2 or control vector. 2310 5 cells were harvested for each antibody reaction and additional same amount cells for input samples and negative control Rabbit (DA1E) mAb IgG. The CUT&RUN and ChIP assays was performed according to the instruction (CUT&RUN Assay Kit, Cell Signaling Technology, USA). Lyse the cells and fragment the chromatin by sonicating the input samples using M220 focused-ultrasonicator (Covaris, USA) device. The micro-TUBE-50 AFA Fiber Screw-Cap (Covaris, USA) was used for sample holder and 150 bp target base pair peak program was chosen for chromatin fragmentation. The peak incident power was 75 W and cycles per burst was set to 200 cpb with the total treatment time was 520s. DNA were purified from input and enriched chromatin samples using the Chromatin IP DNA Purification Kit (Active Motif, USA). Quantitative PCR reactions were performed using the ChamQ SYBR Color qPCR Master Mix (Vazyme, Nanjing, China) on the ABI STEPONE system (Applied Biosystems, Foster City, CA, USA). The antibodies used in this procedure are listed in key resources table and the primers used are listed in Table S1. All ChIP assays were repeated three times and individual qPCR reactions performed in triplicates with results presented as mean G SD.

Cell proliferation assay
The Cell Counting Kit-8 (CCK-8) assay was used to evaluate cell proliferative potential. Cells were seeded and cultured in 96-well plates at a cell density of 1,500 cells per well. After adhesion, 10 mL CCK-8 (Cell Counting Kit-8, Solarbio, Beijing, China) reagent was added to each well for 2 extra hours incubation and the absorbance was measured at 450 nm using a Microplate reader (Bio-Rad Laboratories, Inc.).

Colony formation assay
Cells were seeded in 6-well plates (300 cells/ well) and incubated in a humidified air containing 5% CO 2 at 37 C for 10-14 days. Cells were washed with PBS, fixed with 4% paraformaldehyde for 15 min and stained with crystal violet for 15 min at room temperature. After staining, images were captured and the number of colonies (containing cells >50 cells) was counted manually.

Wound healing assay
A total of 2310 5 cells/well were seeded and cultured in 12-well plates to create a confluent monolayer. After adhesion, a horizontal scratch was made using a sterile 200-mL microliter pipette tip. Cells were cultured for 48 h in a medium containing 2% (wt/vol) FBS, and images of the ''scratch closure'' were photographed with a light microscope (Leica Microsystems, Inc.).

Transwell invasion assay
For transwell invasion assay, a total of 1310 4 cells in serum-free DMEM medium were seeded in the upper chamber pre-coated with Matrigel (8 mm Transwell inserts, BD Biosciences). The lower chambers were filled with DMEM with 10% FBS as an attractant. After 24 hours of incubation, cells on the bottom surface were fixed with 4% paraformaldehyde for 15 min and stained using 0.1% crystal violet for 15 min at room temperature. Stained cells were visualized under a light microscope (Leica, USA). For each insert, at least three selected fields were photographed and counted.