Double-Stranded RNA Structural Elements Holding the Key to Translational Regulation in Cancer: The Case of Editing in RNA-Binding Motif Protein 8A

Mesothelioma is an aggressive cancer associated with asbestos exposure. RNA-binding motif protein 8a (RBM8A) mRNA editing increases in mouse tissues upon asbestos exposure. The aim of this study was to further characterize the role of RBM8A in mesothelioma and the consequences of its mRNA editing. RBM8A protein expression was higher in mesothelioma compared to mesothelial cells. Silencing RBM8A changed splicing patterns in mesothelial and mesothelioma cells but drastically reduced viability only in mesothelioma cells. In the tissues of asbestos-exposed mice, editing of Rbm8a mRNA was associated with increased protein immunoreactivity, with no change in mRNA levels. Increased adenosine deaminase acting on dsRNA (ADAR)-dependent editing of Alu elements in the RBM8A 3′UTR was observed in mesothelioma cells compared to mesothelial cells. Editing stabilized protein expression. The unedited RBM8A 3′UTR had a stronger interaction with Musashi (MSI) compared to the edited form. The silencing of MSI2 in mesothelioma or overexpression of Adar2 in mesothelial cells resulted in increased RBM8A protein levels. Therefore, ADAR-dependent editing contributes to maintaining elevated RBM8A protein levels in mesothelioma by counteracting MSI2-driven downregulation. A wider implication of this mechanism for the translational control of protein expression is suggested by the editing of similarly structured Alu elements in several other transcripts.


Introduction
Malignant mesothelioma (reviewed in [1,2]) is a rapidly fatal tumor arising in the mesothelium, which has mesodermal origins and covers many important internal organs, such as the lungs (pleural mesothelioma), peritoneal cavities (peritoneal mesothelioma), the sacs surrounding the heart (pericardial mesothelioma), and the testis (tunica vaginalis mesothelioma). Although mesothelioma is a rare cancer, its incidence is still rising. Exposure to asbestos has been clearly identified as a cause of mesothelioma, and although the  [19] were cultured as previously described [3] at 37 • C in a humidified 5% CO 2 incubator. SPC111, SPC212, ZL5, ZL34, ZL55, SDM85, SDM77, SDM58 and SDM104 were established in our laboratory [19].

RNA Interference
In order to downregulate RBM8A, ADARs, and MSI2 expression, ON-TARGETplus SMARTpool small interfering RNAs (siRNAs, Dharmacon, Horizon Discovery, Cambridge, UK) against RBM8A, ADAR1, ADAR2, and MSI2 were used. siGENOME non-targeting siRNA pool #2 was used as a control. siRNA dissolved in 1× siRNA buffer (Dharmacon) was combined with transfection reagent DharmaFECT 1 dissolved in OptiMEM (final concentration 0.084%) and incubated for 20 min. Then, cells resuspended in normal growth medium were added to the siRNA/DharmaFECT 1 mixture and seeded onto plates, allowing for a final siRNA concentration of 10 nM. A total of 8 × 10 4 cells (12-well plate) were plated for whole-cell protein lysates as well as RNA extraction 48 h later. For the cell viability assay, MPM and normal mesothelial cells were seeded at a density of 500 cells per well in a 384-well plate, and viability after silencing was determined 5 days later by the addition of 6 µL of CellTiter-Blue (Promega, Madison, WI, USA) to each well and incubation for 4 h at 37 • C. Afterwards, 15 µL of 6 SDS was added to each well, and absorption was measured using a BioTek CYTATION5 Imaging reader. siGENOME non-targeting control pool #2 and the siGENOME UBB SMARTpool siRNA (Dharmacon) were used as negative and positive controls, respectively.
Synthesized cDNA was diluted at 1:60 and used for real-time quantitative PCR (RT-qPCR). SYBR Green (Thermo Fisher, Reinach, Switzerland, cat. no. 4367659) and gene-specific primers (sequences listed in Supplementary Table S1) were used for PCR amplification and detection on a 7500 FAST Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific) or QuantStudio 5 Real-Time PCR System. Relative mRNA levels were determined by comparing the PCR cycle thresholds between cDNA of a specific gene and human histone 3 or mouse tubulin beta 4A class IVa (∆Ct). In order to compare the abundance of different RBM8A transcripts, primers were used on equivalent copy numbers of plasmids with only RBM8A cDNA or pmirGlo-RBM8A-3'UTR AluJo-Y 240-2996 (Addgene, Cambridge, USA, 158106 and 171131, respectively).

Adar2 Cloning, Sequencing, and Transfection
Mouse Adar2 cDNA amplified from the RN5 cell line was cloned into the NheI and EcoRI sites of the pCI-puro vector, which contains a puromycin resistance gene [3]. The insert was validated by sequencing (performed by Microsynth AG, Balgach, Switzerland), and all primers are indicated in Supplementary Table S1. LP9/TERT-1 cells were transfected with pCI-puro_Adar2 (Addgene #158111) using Lipofectamine 3000 reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. LP9/TERT-1-transfected cells were selected with puromycin.

Analysis of Mitochondrial Ribosomal Protein L3 Alternative Splicing
The PCR reaction was performed as previously described [22] (primer sequence in Supplementary Table S1) with GoTaq G2 DNA polymerase kit using 40 ng of cDNA for 29 cycles at 51 • C annealing temperature. PCR products were loaded on a 4% agarose gel.

Editing Quantification
The localization of editing sites was retrieved from REDIportal [23], which contains editing sites from different databases: RADAR [24], DARNED [25], and Atlas [26]. We considered only sites present in at least two databases. For PCR of different Alu regions, cDNA was used. PCR was performed in a total volume of 50 µL containing 1× Green Go TaqR Flexi Buffer (Promega), 2 mM MgCl2 solution, 0.2 mM PCR Nucleotide Mix, 0.2 mM of each primer, 1.25 U of Go TaqR G2 Hot Start Polymerase (Promega), and 25 ng of cDNA. The primers used are listed in Supplementary Table S1. Products were confirmed by electrophoresis on a 2% agarose gel and excised. After purification according to the Macherey-Nagel NucleoSpin ® Gel and PCR clean-up protocol, products were sent for Sanger sequencing (performed by Microsynth AG, Balgach, Switzerland). Raw sequencing outputs were quantified with ImageJ software.

Dual-Luciferase Assay
The dual-luciferase assay was performed in different mesothelial and MPM cells. A total of 10 5 cells/well were seeded in a 12-well plate, and the following day, 200 ng of each reporter plasmid described above along with 2.5 µL of Lipofectamine 3000 reagent mixed in 800 µL of OptiMEM was added to the corresponding wells and incubated for 9 h. Then, OptiMEM medium was replaced by normal culture medium. After 48 h, transfected cells were lysed, and reporter activity was measured using the dual-luciferase reporter assay according to the manufacturer's instructions (Promega, Madison, WI, USA).

RNA Pull-Down
The RNA pull-down assay was adapted from previously published studies [27] or protocols [28]. Unedited and edited RBM8A-3'UTR AluY-Sz6 2598-4066 were subcloned from the reporter plasmids described above into pSP72 (Promega) (Addgene #175587 and 175586, respectively). Previously described pSelectp53 [29] was used to generate p53 mRNA. These plasmids were linearized at the 3 UTR in order to prepare the template DNAs for in vitro transcription.
All biotin-labeled RNA transcripts were produced in vitro using the SP6 (Promega, #P1280) or T7 (Thermo Fisher #EP0111) kit with biotin-16-UTP (Roche, 11388908910) and purified according to the protocol provided by the manufacturer. Five picomoles of biotinylated RNAs was heated to 65 • C for 10 min and then slowly cooled to 4 • C. RNAs were mixed with 100 µg of precleared nuclear extracts from SPC111, ZL55, or ACC-Meso4 mesothelioma cells in binding buffer (150 mM KCl, 25 mM Tris pH 7.4, 5 mM EDTA, 0.5 mM DTT, 0.5% NP40, protease inhibitors, 1 mM PMSF, 100 U/mL Superasin) and incubated at 4 • C for 60 min. Fifty microlitres of washed Streptavidin Magnetic Beads (NEB S1420S) was added to each binding reaction and further incubated at 4 • C for 60 min. Beads were washed three times with the binding buffer and then boiled in 1× loading buffer for 10 min. The retrieved proteins were subjected to SDS-PAGE and further visualized by immunoblotting assay.

RNA Analysis
Rbm8a mRNA editing in crocidolite (blue asbestos)-exposed mice vs. sham mice was assessed as previously described [15]. Mice were repeatedly injected intraperitoneally over a time course of 21 weeks with crocidolite or sham-exposed and sacrificed at 33 weeks, 12 weeks after the last crocidolite exposure. Tissues, including mesothelium and tumor masses, were collected from euthanized mice and consecutively processed for RNA-seq analysis. The sacrificed mice corresponded to three groups: sham, crocidoliteexposed mice with preneoplastic lesions, and crocidolite-exposed mice bearing tumors. The RNA isolation, library generation, and RNA-seq analysis pipelines were previously described [15]. RNA-seq data are deposited in the European Nucleotide Archive (ENA), accession no. PRJEB15230.
Variant analysis was performed using the Genome Analysis Toolkit (GATK, https:// gatk.broadinstitute.org/hc/en-us, accessed on 25 June 2017) software following guidelines for processing RNA-seq data. Specifically, we looked at variants in transcript regions and ignored sites with fewer than 2 reads supporting the SNV. We also excluded variants in immunoglobulin (IG) loci since they are generated by well-known somatic hypermutation Cells 2021, 10, 3543 6 of 20 mechanisms, which would be a potential confounding effect. In order to identify A to G mutations due to RNA-editing events, predicted A to I sites from http://rnaedit.com/ download/ (accessed on 20 July 2017) were translated to mm10 coordinates. Due to the limited availability of samples, all crocidolite samples (tumor and inflamed mesothelium) were pooled and compared to samples from sham-treated mice.
To measure the quality of predicted microRNA (miRNA)-target interactions, target scores for miRDB records [30] were checked by the software. MiRNA-target interactions with scores ≤80.0 were considered not relevant.

Statistical Analysis
The figures represent the mean values from at least three independent experiments. Paired and unpaired t-test, Mann-Whitney, Kruskal Wallis, Fisher's exact test, or one-way as well as two-way ANOVA tests were used and are specified when used. Error bars indicate the standard error of the mean. Statistical analysis was performed using Prism 8 (GraphPad 8.0.0).

Mesothelioma Cells Are More Sensitive to RBM8A Deficiency Compared to Mesothelial Cells
We analyzed RBM8A protein expression in several mesothelioma cell lines. RBM8A protein is heterogeneously expressed in mesothelioma cell lines ( Figure 1A), and the average levels are 4-fold higher in mesothelioma cell lines (n = 14) when compared to normal mesothelial cells (n = 5) ( Figure 1A,B; p < 0.05). The SV40-immortalized Met5A line has intermediate RBM8A expression and was not included in the normal mesothelial cell group because it bears genetic mutations [31] and has a generally altered phenotype [32]. Data mining from https://discover.nci.nih.gov/rsconnect/cellminercdb/ (accessed on 28 April 2020) [33] revealed a weak (r < 0.7) although significant correlation between RBM8A mRNA and protein in cancer cells (Supplementary Figure S1A), and we observed similar results in our collection (Supplementary Figure S1B), indicating the involvement of post-transcriptional processes in regulating protein levels.
To investigate whether RBM8A deficiency affects MPM vs. mesothelial cells differently, RBM8A was silenced in six MPM cell lines and two mesothelial cell lines, resulting in decreased protein levels, as expected ( Figure 1C). This was accompanied by exon 4 skipping in MRPL3, a previously described readout of EJC function [22], in all cell lines, confirming EJC loss of function ( Figure 1D). However, RBM8A silencing was better tolerated in mesothelial cells, which maintained 92% of cell viability, while MPM survival decreased to 46% ( Figure 1E, p < 0.0001). Although we previously observed in a genetically reconstituted model that BAP1-proficient MPM cells are more sensitive to RBM8A silencing, the small number of investigated MPM cell lines did not allow us to further investigate this aspect.
Nevertheless, taken together, these data indicate that higher expression levels of RBM8A in MPM are paralleled by a more essential function of EJC in cancer cells when compared to mesothelial cells.

RBM8A mRNA Has Higher Editing Levels in Mesothelioma Compared to Mesothelial Cells
In our previous studies [14,15] investigating mesothelioma development in mice exposed to crocidolite (blue asbestos), we observed increased RNA editing, and one of the genes that were significantly more edited in asbestos-exposed mice was Rbm8a (Figure 2A), suggesting that one possible post-transcriptional process controlling Rbm8a expression is RNA editing. Indeed, A-to-I editing can affect RNA stability, miRNA-or RNA-binding protein binding ability, and translational activity (recently reviewed in [16]). While no significant increase in Rbm8a mRNA expression was detected in asbestos-exposed mesothe-lium (Supplementary Figure S2A), we observed increased nuclear Rbm8a immunoreactivity in mesothelial cells and mesothelioma tumors upon asbestos exposure ( Figure 2B), indicating that RNA editing coincided with increased protein levels. The edited sites (mouse mm10 chr 3: 96632713 and 96632460) are consistent with known editing occurring in the repetitive sequences present in the Rbm8a 3 UTR ( Figure 2C). Indeed, they represent two out of three editing sites retrieved from REDIportal [23]. confirming EJC loss of function ( Figure 1D). However, RBM8A silencing was better tolerated in mesothelial cells, which maintained 92% of cell viability, while MPM survival decreased to 46% ( Figure 1E, p < 0.0001). Although we previously observed in a genetically reconstituted model that BAP1-proficient MPM cells are more sensitive to RBM8A silencing, the small number of investigated MPM cell lines did not allow us to further investigate this aspect.  Since we observed increased Rbm8a editing in the mouse model of mesothelioma development, we investigated editing levels in cDNA from three MPM cell lines and three mesothelial cell lines (LP9/TERT-1, SDM104, and SDM85) using Sanger sequencing. As inosine is recognized as guanosine, A-to-I RNA editing sites are identified as overlapping peaks of adenosine and guanosine. Editing was detected in 2 out of 3, 11 out of 21, and 17  Figure 2C). In order to investigate the relative abundance of the two splice variants, we used two sets of primers (Supplementary Table S1). In normal mesothelial cells (Supplementary Figure S2B), only 25% of the overall RBM8A expression corresponds to the RBM8A-204 transcript, which is similar to its abundance observed in tumors, including mesothelioma (Supplementary Figure S2C), and human tissues in general [4,35], although heterogeneity was observed. However, the RBM8A-204 transcript is selectively enriched in the human mesothelioma translatome [36] compared to normal mesothelium, indicating that the length of the 3 UTR might be important for the translational regulation of RBM8A.
Since we observed increased Rbm8a editing in the mouse model of mesothelioma development, we investigated editing levels in cDNA from three MPM cell lines and three mesothelial cell lines (LP9/TERT-1, SDM104, and SDM85) using Sanger sequencing. As inosine is recognized as guanosine, A-to-I RNA editing sites are identified as overlapping peaks of adenosine and guanosine. Editing was detected in 2 out of 3, 11 out of 21, and 17 out of 30 sites in AluJo, AluY, and AluSz6, respectively. The editing levels were significantly higher at selected sites located in AluJo and AluSz6 in MPM cells compared to mesothelial cells ( Figure 2D). High editing levels were observed in both AluY (Supplementary Table  S2) and AluSz6. A-to-I edits were recently described to be enriched in regions with high structure scores [37], which is consistent with inverted Alu element pairs being able to form dsRNA, hence acting as a substrate for RNA-editing enzymes. Secondary structure prediction of the 3 UTR sequence using ViennaRNA software [38] (http://rna.tbi.univie.ac. at/cgi-bin/RNAWebSuite/RNAfold.cgi, accessed on 27 April 2021) confirmed the potential for dsRNA structure formation by AluY-AluSz6 sequences, where most editing sites are located ( Figure 2E). Interestingly, in the predicted secondary structure, 7 of the AluSz6 editing sites are in A.C mismatches and 10 are in A:U pairs, while 2 of the AluY editing sites are in A.C mismatches and 9 are in A:U pairs (Supplementary Figure S3), consistent with known ADAR preferences [39,40]. The significantly more edited site in mesothelioma is at chr1:145512955 and is in an A:U base pair.
Altogether, these data indicate that Alu elements in RBM8A in the 3 UTR are bona fide targets of RNA editing and that higher editing levels are associated with pathological conditions.

RBM8A 3 UTR Editing Is Mediated by Both ADAR1 and ADAR2
In order to investigate whether both ADAR enzymes, ADAR1 and ADAR2, are able to edit adenosines in the RBM8A 3 UTR, we analyzed AluSz6 editing levels after silencing ADAR1 or ADAR2 in four different MPM cell lines with different levels of ADAR1 and ADAR2 (Supplementary Figure S4), namely, SPC111, Mero82, Mero95, and NCI-H226 cells. The latter two cell lines have the highest expression of ADAR2, consistent with heterogeneous ADAR2 expression in mesothelioma (Hariharan, in preparation), and in all of them, ADAR1 expression is more abundant compared to ADAR2. Silencing of ADAR1 and ADAR2 was efficient in all four cell lines ( Figure 3A).
The RNA editing levels of all AluSz6 sites were significantly reduced by ADAR1 knockdown, but decreased editing due to ADAR2 knockdown was observed only in four selected sites (chr1:145512937, chr1:145512955, chr1:145513031, chr1:145513034) and only in cell lines with higher ADAR2 expression levels ( Figure 3B and Supplementary Table  S3). These results indicate that ADAR1 and ADAR2 both play a critical role in editing the 3 UTR of RBM8A, but with different characteristics. Cells 2021, 10, x FOR PEER REVIEW 11 of 21

RBM8A 3 UTR Editing Increases Protein Expression
In order to test the functional properties of Alu elements in the RBM8A 3 UTR, five different constructs, containing either single Alu elements or a pair of inverted Alu (AluJo-AluY and AluY-AluSz6) elements, were generated ( Figure 4A, left panel) by inserting them downstream of the Firefly luciferase reporter gene (pmirGLO vector). These constructs or the vector alone was transiently transfected into either mesothelial (SDM104, LP9/TERT-1, and SDM85) or MPM (ACC-Meso4, NCI-H226, and Mero82) cells, and reporter activity was recorded. The RNA sequence containing the inverted AluY and AluSz6 elements had the same predicted dsRNA structure (as analyzed using ViennaRNA software; Figure 4A, right panel) as was identified using the entire RBM8A 3 UTR sequence ( Figure 2E).
The RNA editing levels of all AluSz6 sites were significantly reduced by ADAR1 knockdown, but decreased editing due to ADAR2 knockdown was observed only in four selected sites (chr1:145512937, chr1:145512955, chr1:145513031, chr1:145513034) and only in cell lines with higher ADAR2 expression levels ( Figure 3B and Supplementary Table  S3). These results indicate that ADAR1 and ADAR2 both play a critical role in editing the 3′UTR of RBM8A, but with different characteristics.

RBM8A 3′UTR Editing Increases Protein Expression
In order to test the functional properties of Alu elements in the RBM8A 3′UTR, five different constructs, containing either single Alu elements or a pair of inverted Alu (AluJo-AluY and AluY-AluSz6) elements, were generated ( Figure 4A, left panel) by inserting them downstream of the Firefly luciferase reporter gene (pmirGLO vector). These constructs or the vector alone was transiently transfected into either mesothelial (SDM104, LP9/TERT-1, and SDM85) or MPM (ACC-Meso4, NCI-H226, and Mero82) cells, and reporter activity was recorded. The RNA sequence containing the inverted AluY and AluSz6 elements had the same predicted dsRNA structure (as analyzed using ViennaRNA software; Figure 4A, right panel) as was identified using the entire RBM8A 3′UTR sequence ( Figure 2E).  AluY alone significantly decreased ( Figure 4B) the reporter protein expression by 62% and 37% in mesothelial cells and MPM, respectively, indicative of its silencing function. AluSz6 alone had a weaker although significant downregulatory effect on the reporter activity ( Figure 4B) in both mesothelial and MPM cells. Interestingly, this paired inverted Alu element had a greater destabilizing effect compared to the single elements. This destabilizing effect was significantly greater in mesothelial cells compared to MPM cells ( Figure 4B). Since we observed a significantly higher level of RNA editing in AluSz6 in MPM compared to mesothelial cells, we hypothesized that the reduced destabilizing effect of inverted Alus in MPM was due to RNA editing in this sequence.
In order to address this question, we took advantage of the fact that while cloning the RBM8A AluY-AluSz6 fragment using cDNA, we obtained a construct harboring some mutations due to editing ( Figure 4C, upper panel). These mutations are sufficient to alter the secondary structure (Supplementary Figure S5), and transient transfection of the reporter carrying the mutated AluY-AluSz6 sequence resulted in a stabilizing effect compared to the wild-type sequence ( Figure 4C, lower panel) in both MPM and mesothelial cells, suggesting that editing levels might indeed be responsible for the stabilizing effect.
The AluJo-AluY pair was also investigated, although based on the predicted structure of the full-length RBM8A 3 UTR, this sequence is less likely to form dsRNA. AluJo alone did not significantly affect the activity of the reporter ( Figure 4D) in either mesothelial or MPM cells, and the reporter containing the AluJo-AluY pair did not result in a significant difference in downregulation between mesothelial and MPM cells ( Figure 4D).
Taken together, our data indicate that editing of the AluY-AluSz6 element of the 3 -UTR in RBM8A mRNA contributes to post-transcriptional regulation and control of RBM8A protein levels.

MSI2 Preferentially Interacts with Unedited RBM8A 3 UTR, and Silencing MSI2 Results in Increased RBM8A Levels
It is known that trans-acting factors, such as RNA-binding proteins (RBPs) and mi-croRNAs (miRNAs), regulate translation through direct or indirect interaction with a basic cap-binding complex to modulate the translation initiation complex (reviewed in [41]). Therefore, we analyzed whether editing affects the targeting of the AluY-AluSz6 element by miRNA. Analysis of the unedited and edited forms of AluY-AluSz6 using MiRDB software [30] resulted in differences in recognition by miRNAs with prediction scores lower than 67 (Supplementary Table S4), suggesting that stabilization is not mediated by altering miRNA target sequences.
To further investigate the mechanism of the RBM8A 3 UTR in protein stabilization, we sought RNA-binding proteins known to bind to it. The RBM8A 3 UTR was recently documented to interact with Musashi-2 (MSI2) in leukemic stem cells [42]. The Musashi (MSI) family of RNA-binding proteins, including MSI1 and MSI2, function by binding to the 3 UTRs of some target mRNAs [43,44] at a consensus sequence and then block translation by hindering access of the poly-A-binding protein to the elongation initiation complex [45]. We determined that MSI2 is more abundantly expressed compared to MSI1 in mesothelioma and mesothelial cells (Supplementary Figure S6). There are 38 potential MSI binding sites in the AluY-AluSz6 sequence of the RBM8A-204 transcript (http://rbpmap.technion.ac.il/index_DEV.html, accessed on 27 October 2021), which is in line with the knowledge of cooperative interaction acting on 3 UTR sequences [5]. To identify binding proteins, purified biotinylated unedited or edited sense AluY-AluSz6 RBM8A RNA generated by in vitro transcription was incubated with nuclear extracts of mesothelioma cells. Unedited RBM8A AluY-AluSz6 bound to more MSI2 when compared to the edited version ( Figure 5A). As a control, we used TP53 cDNA, where we identified a single MSI2 potential binding site. As expected, no MSI2 binding was observed in TP53 cDNA, since multiple sites are necessary for binding. The presence of dsRNA was detected by ADAR2, which was used as a control.
ing proteins, purified biotinylated unedited or edited sense AluY-AluSz6 RBM8A RNA generated by in vitro transcription was incubated with nuclear extracts of mesothelioma cells. Unedited RBM8A AluY-AluSz6 bound to more MSI2 when compared to the edited version ( Figure 5A). As a control, we used TP53 cDNA, where we identified a single MSI2 potential binding site. As expected, no MSI2 binding was observed in TP53 cDNA, since multiple sites are necessary for binding. The presence of dsRNA was detected by ADAR2, which was used as a control. If MSI2 preferentially binds to the unedited RBM8A 3′UTR to decrease its translation, then decreasing the levels of MSI2 is expected to increase RBM8A protein levels. We tested this hypothesis in NCI-H226 and Mero95 cells and observed that, indeed, silencing MSI2 If MSI2 preferentially binds to the unedited RBM8A 3 UTR to decrease its translation, then decreasing the levels of MSI2 is expected to increase RBM8A protein levels. We tested this hypothesis in NCI-H226 and Mero95 cells and observed that, indeed, silencing MSI2 resulted in increased RBM8A protein levels, while RBM8A mRNA levels were not significantly changed. Collectively, these experiments indicate that unedited AluY-AluSz6 RBM8A is preferentially physically associated with MSI2 either directly or indirectly and that MSI2 expression decreases RBM8A protein levels.

ADAR-Mediated Editing Increases RBM8A Protein Levels
We next investigated whether an increase in RBM8A mRNA editing would result in increased protein levels. To this aim, we transfected mesothelial cells with a cDNA encoding Adar2 ( Figure 6A). The choice of Adar2 was dictated by the fact that although we observed that both enzymes are able to edit the RBM8A 3 UTR, ADAR1 levels are not significantly different between mesothelial and MPM cells (Hariharan et al., ms in preparation), and in the mouse model of mesothelioma development, we previously observed that Adar2 levels specifically increase during tumor development [15]. The functionality of Adar2 expression in mesothelial cells was verified by investigating the editing of the codon I164V of Coatomer Protein Complex subunit α (COPA) mRNA, a specific ADAR2 substrate [46] ( Figure 6B). Expression of Adar2 resulted in significantly increased RBM8A editing levels in AluY and AluSz6 elements at positions chr1:145511983, chr1: 145512059, chr1:145512937, and chr1:145512955 ( Figure 6C). As predicted by previous data, this was accompanied by increased RBM8A protein levels in cells collected at different passages ( Figure 6D), while no change in mRNA levels was observed. Taken together, these data indicate that the overexpression of Adar2 in mesothelial cells recapitulates a characteristic of RBM8A observed in MPM, namely, the increased editing of its 3′UTR associated with increased protein levels.

Discussion
In this study, we show that RBM8A protein levels are higher in mesothelioma cells Taken together, these data indicate that the overexpression of Adar2 in mesothelial cells recapitulates a characteristic of RBM8A observed in MPM, namely, the increased editing of its 3 UTR associated with increased protein levels.

Discussion
In this study, we show that RBM8A protein levels are higher in mesothelioma cells compared to normal mesothelial cells. This is consistent with the knowledge that increased levels of mRNA-binding proteins such as RBM8A are a hallmark of cancer [47]; accordingly, silencing RBM8A drastically reduces cell growth in mesothelioma but not in mesothelial cells. Similarly, silencing RBM8A killed non-small-cell lung cancer (NSCLC) cells and not nonmalignant lung cells [48], suggesting that RBM8A could be a cancer-selective target.
In addition, we demonstrate that higher protein levels of RBM8A in mesothelioma are due to increased RNA editing of its 3 UTR, which protects it from MSI2 binding and negative translational regulation (Figure 7). In addition, we demonstrate that higher protein levels of RBM8A in mesothelioma are due to increased RNA editing of its 3′UTR, which protects it from MSI2 binding and negative translational regulation (Figure 7). This is one example where translational control explains the poor relationship between mRNA and protein expression, which has been observed in multiple tissues [49], the NCI-60 human cancer cell line panel [50], and cancer tissues [51][52][53]. It is consistent with RNA processing being one of the top pathways altered in mesothelioma [10,54]. To date, this has been only marginally explored, despite the known importance of translational regulation in MPM [55,56].
Translational control is known to add an important layer of regulation of patterning of the mesoderm, the tissue of origin of the mesothelium, where, intriguingly, the most significant networks of translationally regulated mRNA were shown to belong to signaling pathways important to mesothelioma, such as Hedgehog, Hippo, and FGF [57]. In that study, transcripts with low translational efficiency overall had longer 5′UTRs but not 3′UTRs and contained a significantly higher number of upstream AUGs (uAUGs) in their 5′UTRs. However, only slight differences in the 5′UTR have been detected as a cause for the specific enrichment of selected mRNAs in a recent translatome study [36], and in the case of RBM8A, there are no uAUGs in the 5′UTR.
According to our study, it appears that the 3′UTR has an important regulatory role. This is supported by the fact that two transcripts, one with a long and one with a short RBM8A 3′UTR, are conserved among primates, and the longer 3′UTR contains an AluY and an AluSz6 sequence (Supplementary Figure S7), suggesting a functional role. It is intriguing that only RBM8A-204 transcript levels, but not those of the shorter RBM8-201, are higher in polysomes from mesothelioma cells compared to normal mesothelium [36]. The two different transcripts likely arise from alternative polyadenylation (APA). Although not investigated in this study, RBM8A is among the 1346 genes with recurrent and tumorspecific APA in cancer [58], where the majority of events lead to a shorter 3′UTR through This is one example where translational control explains the poor relationship between mRNA and protein expression, which has been observed in multiple tissues [49], the NCI-60 human cancer cell line panel [50], and cancer tissues [51][52][53]. It is consistent with RNA processing being one of the top pathways altered in mesothelioma [10,54]. To date, this has been only marginally explored, despite the known importance of translational regulation in MPM [55,56].
Translational control is known to add an important layer of regulation of patterning of the mesoderm, the tissue of origin of the mesothelium, where, intriguingly, the most significant networks of translationally regulated mRNA were shown to belong to signaling pathways important to mesothelioma, such as Hedgehog, Hippo, and FGF [57]. In that study, transcripts with low translational efficiency overall had longer 5 UTRs but not 3 UTRs and contained a significantly higher number of upstream AUGs (uAUGs) in their 5 UTRs. However, only slight differences in the 5 UTR have been detected as a cause for the specific enrichment of selected mRNAs in a recent translatome study [36], and in the case of RBM8A, there are no uAUGs in the 5 UTR.
According to our study, it appears that the 3 UTR has an important regulatory role. This is supported by the fact that two transcripts, one with a long and one with a short RBM8A 3 UTR, are conserved among primates, and the longer 3 UTR contains an AluY and an AluSz6 sequence (Supplementary Figure S7), suggesting a functional role. It is intriguing that only RBM8A-204 transcript levels, but not those of the shorter RBM8-201, are higher in polysomes from mesothelioma cells compared to normal mesothelium [36]. The two different transcripts likely arise from alternative polyadenylation (APA). Although not investigated in this study, RBM8A is among the 1346 genes with recurrent and tumorspecific APA in cancer [58], where the majority of events lead to a shorter 3 UTR through the CstF64-mediated usage of proximal APAs.
The mesothelioma polysome transcription profile is different from the total transcriptome profile [36], but the reasons for this difference have not been elucidated yet. Our findings on the interaction with RBP and its modulation by RNA editing may shed light on the underlying mechanisms.
We cannot exclude that other types of control, such as silencing by a miRNA, may contribute to the post-transcriptional regulation of RBM8A. Indeed, miR-29a downregulates Rbm8a in retinal progenitors [59], and miR-29 levels are lower in mesothelioma cells compared to mesothelial cells [60]. In addition, overexpression of active ADAR2 in glioblastoma cells significantly decreased the levels of miR-29a and miR-29b [61]. The miR-29a site is located between the AluJo and AluY repetitive sequences. Therefore, the lack of difference in the AluJo-AluY reporter between mesothelial and mesothelioma cells rules out the involvement of miR-29a in increased RBM8A levels. We demonstrate here that RNA editing is an important mechanism in preventing negative translational control, and we put forward the hypothesis that this may occur as early as the initial stages of tumor development.
Indeed, ADAR-dependent editing increases upon mesothelioma development [14,15], and in this study, we demonstrate that it participates in translational control in mesothelioma. To our knowledge, this is the first time that RBM8A 3 UTR editing has been functionally characterized. Unedited RBM8A 3 UTR interacts with MSI and results in decreased protein levels. MSI was originally identified in Drosophila for controlling sensory organ development, and in mammals, the two MSI RBPs control cell fate in neural, hematopoietic, gastrointestinal, and breast systems (reviewed in [62]). MSI proteins have also been implicated in cancer stem cell generation and the progression of hematological malignancies and solid tumors (reviewed in [62]).
The MSI proteins downregulate the translation of their targeted mRNAs by interaction with poly(A) binding protein, hence competing with eIF4G for interaction [45], although this is context-dependent (reviewed in [62]).
Besides controlling cell fate, loss of Msi2 in primary mouse keratinocytes increases the migration of keratinocytes, and in wounded skin, Msi2 is strongly downregulated in epidermal stem/progenitor cells, located at the leading edge of the wound [63]. This may be the reason why we detected a significant (56%, p = 0.0008218, FDR = 0.003218) decrease in Msi2 in inflamed mesothelium in mice after their exposure to asbestos. The fact that the RBM8A 3 UTR has decreased interaction with MSI protein when edited is in line with the recent concept that RNA editing may regulate RBP function [64].
An additional hint indicating that this mechanism of regulation of protein expression might be shared by different transcripts is the fact that by using BLASTN (https://blast. ncbi.nlm.nih.gov/Blast.cgi, accessed on 17 September 2021) on the RBM8A AluY-AluSz6 sequence, we retrieved (Supplementary Table S5) several transcripts, including BRCA1 interacting helicase 1, leucine-rich repeat containing 28, leucine-rich repeat containing 57, and malic enzyme 2. These four transcripts are targets of RNA editing, bind MSI2 [42], and are enriched in the mesothelioma translatome [36], and the AluY-AluSz6 regions in these transcripts form a similar dsRNA structure (Supplementary Figure S8). Therefore, we put forward the hypothesis that RNA editing and MSI2 are likely to have a broad implication in the mesothelioma phenotype.
In addition, although editing of RBM8A was investigated in mesothelioma in this study, it is likely to occur in other cancer types as well. This work may contribute to elucidating the mechanism underlying RBM8A synthetic lethality in most cancer cell lines [12,13].
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cells10123543/s1, Figure S1: Correlation of RBM8A protein and RNA expression in different human cell lines. Figure S2: Rbm8a mRNA expression in crocidolite (blue asbestos)-exposed mice and abundance of RBM8A transcripts in mesothelial cells or various cancers. Figure S3: dsRNA stretch of RBM8A 3 UTR formed by AluY and AluSz6 elements. Figure S4: ADAR1/2 mRNA heterogeneous expression in four MPM cell lines. Figure S5: Editing of RBM8A 3 UTR AluY-AluSz6 entails structural changes. Figure S6: MSI2 expression is higher compared to MSI1 in MPM. Figure S7: AluY-AluSz6 sequence is highly conserved between humans and primates. Figure S8: AluY-AluSz6 dsRNA structure is conserved in other transcripts enriched in the MPM translatome. Table S1: Sequences of primers. Table S2: Percentage of A to G editing in RBM8A AluY element. Table S3: Statistical analysis of AluSz6 editing after ADAR1/2 silencing in Mero-82, Mero-95, NCI-H226, and SPC111 MPM cells. Table S4: Prediction of miR targeting in unedited vs. edited Alu Sz6 sequence.