LINC00152 promotes Ovarian tumor progression and Predicts Poor Prognosis by inhibiting the ubiquitination of BCL6

Background : Long non-coding RNA 00152 (LINC00152) has been found oncogenic in multiple somatic malignancies. But its roles in the ovarian cancer remain elusive. We aim to investigate its functions and mechanism in the ovarian tumor progression. Methods: We used RNAscope and RT-qPCR assays to detect the LINC00152 levels in 152 pairs of ovarian tumor and paratumorous tissues, conducted in vitro and in vivo experiments to evaluate its biological functions, and performed RNA immunoprecipitation, RNA-pulldown, MS/LS analysis, mutant construction and ubiquitination assays to explore the LINC00152-BCL6 binding and regulation. Results: LINC00152 was abnormally increased in the ovarian tumor tissues and its high expression predicted poorer survivals of patients; overexpression of LINC00152 promoted ovarian tumor proliferation and metastasis both in vitro and in vivo ; LINC00152 bond to Serine333 and Serine343 of BCL6 and then suppressed its ubiquitination; Finally, LINC00152 facilitated its oncogenic functions in a BCL6-mediated manner in ovarian cancer. Conclusions: LINC00152 is the independent prognostic predictor of ovarian cancer; and the abnormal expression of LINC00152 facilitates ovarian tumor proliferation and invasion by suppress the ubiquitination of BCL6. Our data might be helpful in exploring the molecular mechanisms of lncRNA-protein interactions, and might provide the potential target for the tumor pharmacology of ovarian cancer. spectrometric analyses to analyze the purified containing complex pulled down by 3 independent biotin RNA-protein pulldown


Introduction
Ovarian cancer is responsible for approximately 5% of total cancer death with the proportion of less than 3% of all malignancies in females [1] . Despite the declined incidence and mortality in the past 3 decades, inconspicuous symptoms and signs still lead to delayed diagnoses at advanced stages with pelvic dissemination, causing the relatively low 5-year survivals as 42% for stage III and 26% for stage IV in the United States [2] , revealing a highly aggressive nature that needs elaboration of underlying molecular process.
The long non-coding RNAs (lncRNAs, non-coding RNAs with sizes larger than 200bps) are associated with somatic tumor tumorigenesis and metastasis [3] . The non-coding RNA 00152 (LINC00152) is a lncRNA located at 2p11.2 [4] and has been involved in cell proliferation [5] , cell-cycle arrest [4] ,epithelial-mesenchymal transition [6] and cell invasion [7] to facilitate tumor initiation and progression. Our former study revealed that LINC00152 promoted tumor invasion and predicts poor prognosis in renal clear cell [8] and lung carcinomas [9] . Other studies have got similar findings in gastric cancer [10] and hepatic carcinoma [11] , suggesting that LINC00152 might serve as a powerful oncogenic lncRNA in somatic malignancies. Researchers have found that LINC00152 forms the loop feedback with FOXM1 to stimulate tumor growth [12] and might serve as the ceRNA (competing endogenous RNA) to regulate miRNAs to fulfil its biological functions [5,13,14] .
Although the impact of LINC00152 on ovarian cancer proliferation has been implicated recently [15] , its role and the exact molecular mechanism to regulate the ovarian tumor behavior need further investigation.
Ubiquitination is a posttranslational epigenetic mechanism that requires the conjugation of ubiquitin (Ub) to the lysine residues of specific substrate proteins via the enzymemediated transfer [16] . Now is known that ubiquitination is a crucial part of the molecular dysregulation in ovarian cancer [17] . And lncRNAs might facilitate tumorigenesis by regulating the process of the ubiquitination. It has been reported that lncRNA uc.134 represses CUL4A-mediated ubiquitination of LATS1 to inhibit the progression of hepatic cancer [18] . LncRNA-ANCR also has been reported to suppress the invasiveness of breast cancer cells by prompting the degradation of EZH2 [19] . But it is unknown whether LINC00152 could affects ovarian tumorigenesis via the ubiquitin-related mechanism.
In the process of searching the correlation between LINC00152 and the ubiquitinationproteosome system, we found that B-cell CLL/lymphoma 6 (BCL6) might be the potential target of LINC00152. BCL6 is a highly conserved zinc finger transcriptional factor [20] and its degradation is mediated via the ubiquitination-proteosome system [21,22] . We formerly have proved that BCL6 exerts oncogenic activities in ovarian cancer [23] . Based on these above, it is very likely that LINC00152-BCL6 regulation contributes to the ovarian tumor progression.
In this study, we aimed to investigate the role of LINC00152 in ovarian cancer via its possible regulation upon alterations of BCL6 degradation through the ubiquitinproteosome system, and the potential influences of the LINC00152-BCL6 interaction on the biological behaviors and clinical outcomes.
For the intraperitoneal tumor metastasis models, 8×10 6 SKOV3 cells stably infected with Lenti-NC or Lenti-LINC00152 were injected intraperitoneally (i.p.) (n= 3 per group). The abdominal circumferences of the mice were measured every 2 days. All the mice were anesthetized and sacrificed. The xenografts were excised and measured. The tumor volumes were calculated using the formula 1⁄2×r12×r2 (r1<r2). The intraperitoneal organs (livers, peritoneal membranes, and intestines, etc.) were excised and examined for the implanted lesions followed by sampling. All the tissues were fixed and embedded with paraffin. Hematoxylin& Eosin staining was used to observe the lesions through a microscope.

Biotin pull-down assays and mass spectrometry
The LINC00152 and its antisense plasmid were linearly cut, transcribed, and biotin-labeled

Statistical analysis
Each experiment was performed in triplicate, and survival data were shown as the median following time, and other data were presented as the mean ±SD. All statistical analyses were performed using SPSS 20.0 (IBM, SPSS, Chicago, IL, USA). Student's t-test and one-way ANOVA were used in either 2 or multiple groups for statistical significance.
Poison correlation and Spearman rank order were used to analyze the correlations; disease-free survival (DFS) and disease-specific survival (DSS) curves were calculated with the Kaplan-Meier method and were analyzed with the log-rank test. The DFS rate was calculated from the date of surgery to the date of progression (local and/or distal tumor recurrence) or the date of death. The DSS rate was defined as the length of time between the diagnosis and death or last follow-up. Univariate analysis and multivariate models were fit using a Cox proportional hazards regression model. All tests were 2-sided, and P<0.05 was considered statistically significant.
Other methods used in this study are listed the Supplementary Information.

LINC00152 is upregulated in ovarian cancer and predicts poor clinical outcomes
We first detected the levels of LINC00152 mRNA using RNAscope and RT-qPCR in 152 pairs of ovarian cancer and adjacent normal tissues (ANTs). The RNAscope showed that LINC00152 was positively stained in the tumor lesions while negative in the ANTs ( Figure   1A). 64.65% of high-grade serous, 72.00% of endometrioid, 23.81% of clear cell type and 28.57% of mucinous subtype showed positive expression of LINC00152 ( Figure 1A, Table   1). The RT-qPCR results showed that the mRNA level of LINC00152 was significantly higher in the 152 malignant lesions than in the paired paratumorous areas (P<0.001, Figure 1B Table 1) and lymph node metastasis (P<0.001, Table 1). Notably, the expression of LINC00152 detected by RT-qPCR was not correlated with the histological subtypes of ovarian cancer (high-grade serous, endometrioid, clear cell and mucinous types) ( Table   1). Within the endometrioid subtypes, the LINC00152 mRNA levels were not correlated with the tumor grades ( Table 1). When LINC00152 was divided into the high and low expression groups by the median value according to the previous studies [22,23] , the distributions of patients with advanced FIGO stage (P<0.001, Figure 1C We then elucidated the impacts of LINC00152 on patients' clinical outcomes. The total median follow-up time for the patients who were still alive at the endpoint for analysis was 31.50 months. The median follow-up time for the patients who were still alive at the endpoint was 29.50 months in the high LINC00152 expression group and 32.00 months in the low expression group. The LINC00152 levels were significantly increased in the patients with recurrence (P=0.010, Table 1). When dividing LINC00152 into the high and low expression groups by the median value, high LINC00152 expression was associated with recurrence (P<0. 001, Fig ure 1D). The Kaplan-Meier analysis showed that patients with high LINC00152 expression (n=76) had significantly shorter DFS (P=0.004; Figure   1E) and DSS rates (P=0.005; Figure 1F) compared with those with low expression (n=76).
These results corresponded to that of the cohorts from TCGA which illustrated that high   Figure 1F) and DSS (P<0.01, Figure 1F), but the combination of these two independent predictors presented superior prognostic value (P<0.01, Figure 1F). All these together suggest that abnormal LINC00152 expression predicts poor clinical outcomes of patients with ovarian cancer, and combined LINC00152 with FIGO stage could serve as a more sensitive and specific prognostic marker for ovarian cancer.

LINC00152 promotes ovarian cancer cell proliferation and metastasis in vitro and in vivo
To further study the biological effects of LINC00152 on ovarian cancer, we next

LINC00152 interacts with the co-localized BCL6 protein and elevates its level
Since we previously reported that lncRNAs interact with proteins to fulfill their biological functions, we first used LC-MS/MS spectrometric analyses to analyze the purified LINC00152 containing complex pulled down by 3 independent biotin RNA-protein pulldown assays in CAOV3 cells. The mass spectrometry identified that the complex contained the peptides of BCL6 (Figure 3A, Supplementary Figure S3A and Table S3), which was also confirmed by the western blotting result of RNA-protein pulldown products ( Figure   3A). Besides, we performed the RNA immunoprecipitation assay using CAOV3 cells and H293T cells, and again LINC00152 was found to be bound by the endogenous BCL6 ( Figure 3B), suggesting that LINC00152 can bind directly to BCL6 protein.
Then we analyzed the data of two cohorts in the TCGA and found that both LINC00152 and BCL6 genes were amplified (Supplementary Figure S3B), and the mutual exclusivity of these two genes was co-occurrence (P=0.011, Supplementary Figure S3C). Back to our study, overexpression of LINC00152 elevated the protein levels of BCL6 in A2780 and SKOV3 cells, while knockdown of LINC00152 reduced the protein levels of BCL6 in CAOV3 and ES-2 cells (Figure 3C). However, the interference of LINC00152 failed to affect the mRNA levels of BCL6 accordingly (Figure 3D), suggesting that the regulatory effect of LINC00152 on BCL6 is on the posttranslational level.
To further confirm the localization of LINC00152 and BCL6 protein, we performed RNA-Protein dual staining using the RNAscope technique (ACD, USA) in ovarian cancer tissues and tumor cell lines. It turned out that in the ovarian high-grade serous and endometrioid carcinoma tissues, the LINC00152 and BCL6 protein were expressed majorly in the nucleus ( Figure 3E), but the more sensitive immunofluorescence showed that there was still a minority of LINC00152 and BCL6 proteins localized in the cytoplasm in spite of the majority found in the nucleus (Figure 3F). The following immunofluorescence and nuclear-cytoplasmic RT-qPCR/western blotting assays in both CAOV3 and ES-2 cells also proved that LINC00152 and BCL6 protein were co-expressed mainly in the nucleus but slightly in the cytoplasm (Supplementary Figure S3D-S3E). Overexpression of LINC00152 in A2780 and SKOV3 cells enhanced the expressions of BCL6 protein both in the nucleus and cytoplasm by the immunofluorescence and RT-qPCR/western blotting assays (Figure 3G-3H). All these together suggest that LINC00152 interacts with the colocalized BCL6 protein and elevates its level in ovarian cancer.

LINC00152 elevates the level of BCL6 protein by binding its specific sites and suppressing its ubiquitination
Because LINC00152 changed the expression of BCL6 on the protein level but not the mRNA level (Figure 3C-D), we assumed that LINC00152 might regulate BCL6 posttranslationally. To confirmed this, we first used the protein inhibitor MG132 to successfully rescue the reduced levels of BCL6 protein caused by the knockdown of LINC00152 in CAOV3 and ES-2 cells (Figure 4A). In the meantime, we found that overexpression of  Figure S4C). Taken together, these data suggest that LINC00152 elevates the protein level of BCL6 by inhibiting its ubiquitination.
Since the previous studies have provided several potential sites correlated with degradation of BCL6 [24] , we constructed the corresponding plasmids of mutated and truncated BCL6 (Figure 4E). It turned out that MG132 inhibited the degradation of wildtype BCL6 and the truncated BCL6 proteins including BCL ΔN(1-121aa) , BCL Δ N(300-450aa) and BCL ΔN(520-680aa) but not the mutant BCL6 S333A/S343A ( Figure 4F). In contrast, under the influence of CHX, the wild-type BCL6 and all the truncated proteins of BCL6 (BCL ΔN(1-121aa) , BCL Δ N(300-450aa) and BCL ΔN(520-680aa) ) were degraded but not the mutant BCL6 S333A/S343A ( Figure 4G). Furthermore, overexpression of LINC00152 only failed to elevate the protein level of mutant BCL6 S333A/S343A ( Figure 4H). Finally, the RNA binding protein immunoprecipitation(RIP) assay showed that the binding between LINC00152 and mutant BCL6 S333A/S343A was significantly reduced compared with the wild-type BCL6 (P<0.01, Figure 4I). All these data suggest that LINC00152 binds to Ser333/Ser343 of BCL6 protein and elevates its protein level by suppressing its ubiquitination.

LINC00152 promotes ovarian cancer progression in a BCL6-mediated manner
To investigate whether LINC00152 exhibited its functions in ovarian cancer in a BCL6mediated manner, we first observed the effect of simultaneous overexpression of LINC00152 and knockdown of BCL6. It was shown that knockdown of BCL6 in the LINC00152-overexpressing A2780 and SKOV3 cells led to the reduction of metastasisrelated genes such as SNAIL and N-CADHERIN, as well as the proliferation-related gene CCNB1 ( Figure 5A). The CCK-8, colony-forming and EdU assays showed that overexpression of LINC00152 stimulated the proliferation of SKOV3 and A2780 cells but impaired by simultaneous knockdown of BCL6 (P<0.05, Figure 5B). The transwell assay proved that knockdown of BCL6 partially attenuated the effects of overexpression of LINC00152 on ovarian cancer cell invasion (P<0.05, Figure 5C). Consistently, the in vivo xenograft models confirmed that knockdown of BCL6 partially abolished the accelerated tumor growth and the increased tumor size/weight induced by overexpression of LINC00152 (P<0.05, Fig ure 5D).
In addition, we identified the indispensability of the interaction of LINC00152 with BCL6 on the biological functions of ovarian cancer. The CCK8, colony-forming and EdU assays revealed that the enhanced proliferation and invasion of SKOV3 cells induced by overexpression of LINC00152 could be further boosted by co-overexpression by the wildtype BCL6 but not the mutant BCL6 Ser333/Ser343 ( P<0.05, Figure 6A-C), and the xenograft models also indicated that the accelerated growth speed and increase size/weight of xenografts induced by overexpression of LINC00152 could be amplified by cooverexpression of the wild-type BCL6 but not the mutant BCL6 Ser333/Ser343 ( P<0.05, Figure 6D). All these data suggest that LINC00152 prompts ovarian tumor progression in a BCL6-mediated manner, and their interaction is essential to the functional fulfilment of LINC00152.

Discussion
In this study, we found that the mRNA level of LINC00152 was abnormally increased in the ovarian tumor tissues and cell lines accompanied with dysregulated biological functions, poorer clinical outcomes and facilitated cell proliferation and metastasis. Moreover, we provided the first-hand evidence of LINC00152 regulation on the degradation of BCL6, which directly promoted the progression of ovarian cancer. Thus, our study reveals the target molecules of LINC00152 and the molecular mechanisms of BCL6-related lncRNAs in the tumorigenesis of ovarian cancer.
Located at 2p11.2, LINC00152 has been reported as an oncogenic lncRNA in multiple somatic malignancies [11,27,28] . We formerly found that LINC00152 predicted negative prognosis and prompted tumor invasion in the renal clear cell carcinoma and lung adenocarcinoma [8,9] . In this study, we found the dysregulated increase of LINC00152 expression in ovarian cancer tissues as well as the relevance between its abnormal expression and poor prognosis. In addition, we found high LINC00152 expression was independently predicted the risks of poorer DFS and DSS, and the combination of LINC00152 and FIGO stage in ROC curve analysis could offer better predictive value of patients' survivals compared with FIGO staging or LINC00152 solely. Since the FIGO stage has been considered as the crucial risk factor for the prognosis of ovarian cancer [29] , our data suggested that even in the patient at the same FIGO stage, introduction of LINC00152 expression for stratified analysis might offer more precise prediction of risks of recurrence and mortality, which might also partially offer clues to the difference of survivals in patients at same FIGO stages. Based on the occult invasiveness of ovarian cancer, it would be of clinical value to apply LINC00152 as the regular monitoring factor postoperatively for the patients. Future studies need to focus on the consistence of LINC00152 expression in serum with that in tissues that might ease the detection and monitoring methods.
The previous studies have found that LINC00152 could compete as a ceRNA with miRs to fulfil its functions in hepatic and cervical cancers [11,15,30,31] . Besides, LINC00152 is related to the regulation of signaling pathways such as Wnt/β-Catenin pathway [12,32] . In contrast, Our MS/LS data based on RNA-pulldown assays and in vitro experiments suggested that oncogenic BCL6 was the posttranslational regulation target of LINC00152, and LINC00152 promoted ovarian tumor invasion and metastasis in a BCL6-mediated manner. We have reported that BCL6 facilitates ovarian tumor progression but its molecular regulation is unclear [23] . Although other studies have reported the possible links between BCL6 and lncRNAs [33] , the details of underlying mechanisms still remains elusive. In this study, we found that LINC00152 directly bond to the degradation sites of BCL6 thus inhibiting its ubiquitination, accumulating the BCL6 protein level to exhibit tumorigenic activities. The degradation of BCL6 protein has been referred to the MAPK1mediated phosphorylation of the Serine 333 and Serine 343 followed by the proteasome-mediated ubiquitination [26] . Alternatively, the BTB domain (N-terminal 32-99aa) might form dimers to induce the FBXL17-conducted degradation [34] . Based on these, we constructed truncated or point mutants of BCL6 in this study. It was interesting that LINC00152 had similar effects on the binding and accumulation of the wild-type BCL6 and the N-terminal truncated mutant, while LINC00152 nearly lost the ability of binding to the double-point mutant of BCL6 (S333A/S343A) and overexpression of LINC00152 failed to increase the protein level of BCL6 S333A/S343A . These results supported the possibility that LINC00152 might occupied the phosphorylated sites of BCL6 to inhibit its phosphorylation rather than interact with its BTB domain thus suppressed the degradation of BCL6.
Although our results could not exclude the other possibilities of regulation patterns, we still offer clues to the nature and laws of lncRNA-protein interaction. In future we shall continue to focus on the LINC00152-BCL6 centered networks to widen the potential links to additional miRNAs and other lncRNAs.

Conclusion
In summary, our study suggests that LINC00152 is the applicable independent prognostic predictor of ovarian cancer, and its abnormal expression facilitates ovarian tumor proliferation and invasion by binding to the degradation-required sites of BCL6 to suppress its ubiquitination. Our data might be helpful with the exploration of molecular mechanisms of lncRNA-protein interactions, and might provide the potential target for the tumor pharmacology of ovarian cancer. Commission and carried out in accordance with fudan university laboratory animalsadministration.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Disclosure of Potential Conflicts of Interest:
No potential conflicts of interest were disclosed.

Sponsorship
This study was supported by National Natural Science Foundation of China (No.

Author contributions
WYQ was the sponsor and designed the project. CTT, XMD, WP, LJ and LLH were responsible for the part of experiments and analyzed the data. WSN and WWW organized the data, composed the paper and substantively revised it.All authors read and approved the final manuscript.