RNA binding protein PRRC2B mediates translation of specific mRNAs and regulates cell cycle progression

Abstract Accumulating evidence suggests that posttranscriptional control of gene expression, including RNA splicing, transport, modification, translation and degradation, primarily relies on RNA binding proteins (RBPs). However, the functions of many RBPs remain understudied. Here, we characterized the function of a novel RBP, Proline-Rich Coiled-coil 2B (PRRC2B). Through photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation and sequencing (PAR-CLIP-seq), we identified transcriptome-wide CU- or GA-rich PRRC2B binding sites near the translation initiation codon on a specific cohort of mRNAs in HEK293T cells. These mRNAs, including oncogenes and cell cycle regulators such as CCND2 (cyclin D2), exhibited decreased translation upon PRRC2B knockdown as revealed by polysome-associated RNA-seq, resulting in reduced G1/S phase transition and cell proliferation. Antisense oligonucleotides blocking PRRC2B interactions with CCND2 mRNA decreased its translation, thus inhibiting G1/S transition and cell proliferation. Mechanistically, PRRC2B interactome analysis revealed RNA-independent interactions with eukaryotic translation initiation factors 3 (eIF3) and 4G2 (eIF4G2). The interaction with translation initiation factors is essential for PRRC2B function since the eIF3/eIF4G2-interacting defective mutant, unlike wild-type PRRC2B, failed to rescue the translation deficiency or cell proliferation inhibition caused by PRRC2B knockdown. Altogether, our findings reveal that PRRC2B is essential for efficiently translating specific proteins required for cell cycle progression and cell proliferation.


INTRODUCTION
RNA-binding proteins (RBPs) are vital regula tors af fecting the fate of nearly all classes of RNA molecules ( 1 , 2 ). Through one or multiple RNA-binding domains ( 3 , 4 ), RBPs interact selecti v el y or globall y with RN As throughout their lifespan with temporal, spatial and functional dynamics ( 5 ). Once recruited to RNAs, RBPs participate in posttranscriptional regulation of gene expression, including RNA splicing, transport, modification, translation and degradation ( 6 , 7 ). Accumulating evidence has demonstrated the essentiality of RBP-directed posttranscriptional regula tions a t cellular and organismal le v els ( 7 ) with alterations in the abundance or functionality of RBPs linked with various human disorders ( 8 , 9 ). Thus, exploring the role of RBPs in posttranscriptional regulation of gene expression contributes significantly to our understanding of basic biology and human diseases. mRNA translation is one of the posttranscriptional processes affected by RBPs e xtensi v ely ( 10 ). As the most energy-consuming and precisely regulated process in cells, translation is tightly controlled through cis-acting RNA elements such as terminal oligopyrimidine (TOP) motifs and CA-rich elements (11)(12)(13) and through transacting factors such as mRNA binding proteins (mRBPs) ( 2 , 10 , 14 ). The latter affects all steps in global or transcriptspecific translation ( 10 ). Numerous mRBPs and (m)RNAbinding domains have been identified through massspectrometry-based 'mRN A interactome ca pture' methods w here probes, mostl y deoxythymidine oligonucleotides (oligo(dT)), were used to ca pture pol y(A)-bearing mR-N As to gether with their interacting mRBPs after ultraviolet (UV) light crosslinking (15)(16)(17)(18). These methods identify mRBPs physically bound to sequence-or structurespecific elements in mRNAs that modulate gene expression at the posttranscriptional le v el ( 6 ). Howe v er, despite continuous attempts at functional characterization of individual mRBPs ( 5 ), most newly identified mRBPs remain unstudied.
One of the functionally unannotated mRBPs that may regula te transla tion is Pr o-r ich C oiled-coil 2B (PRRC2B). PRRC2B is identified as an mRNA-binding protein based on its enrichment with the oligo(dT) captured mRNAs across multiple different cell types ( 15 , 18 ) and its putati v e arginine-gly cine (RG)-rich domains documented to interact with RNA ( 19 , 20 ). Previous studies suggested that PRRC2B is part of the eukaryotic initiation factor 4G2 (eIF4G2)-media ted transla tion initia tion complex (21)(22)(23). This is based on its co-immunoprecipitation with eIF4G2 and eukaryotic translation initiation factor 3 (eIF3) in human breast cancer cells and mouse embryonic stem cells (21)(22)(23). Howe v er, the e xact function and mRNA targets of PRRC2B in translational regulation remain unknown.
Although understudied at the mechanistic le v el, PRRC2B has been linked with various human cancers. PRRC2B is highly expressed in fast-proliferating tumor cells such as Wilms' tumor, and its high expression is tightly associated with poor ov erall survi val of patients ( 24 , 25 ). In addition, fusion genes of PRRC2B with MGMT, DEK, and ALK have been reported in multiple different tumors (26)(27)(28)(29)(30). Ther efor e, understanding the function of PRRC2B may shed light on translational regulation with relevance to tumor cell proliferation and human diseases.
In this study, we perform a comprehensi v e characterization of PRRC2B in translational regulation of gene expression in human embryonic kidney (HEK293T) cells. We demonstra te tha t PRRC2B and a 750 amino acid-long fragment of PRRC2B bind to GA-or CU-rich elements near the transla tion initia tion codon of the main open reading frame on a subset of mRNAs, promoting their translation via interacting with translation initiation factors. Loss of PRRC2B binding leads to decreased translation efficiency of PRRC2B-bound mRNAs encoding oncogenes and cell cycle regulators such as cyclin D2, resulting in reduced cell cycle pro gression. Alto gether, our study re v ealed the func-tion of PRRC2B in translational regulation of specific proteins affecting cell proliferation.

Molecular cloning
DNA fragments were obtained and amplified by PCR from Genomic DNA or cDNA extracted from HEK293T cells by Trizol (ThermoFisher) according to manufacturer's instructions. Amplified DNA fragments were separated on 1% agarose gel, purified, digested by restriction endonucleases together with the backbone plasmid at 37 • C overnight, and ligated to the cleaved backbone by T4 DNA ligase (NEB). The ligated products were transformed to Esc heric hia coli competent cells (DH5 ␣). Miniprep was performed on single colonies to purify the plasmids. Plasmid sequences were then confirmed by Sanger sequencing.

Cell culture and transfections
Human HEK293T cells were propagated in Dulbecco's modified Eagle's medium (DMEM; Corning) supplemented with 10% fetal bovine serum. Cells were transfected with plasmids using Lipofectamine 3000 (Invitrogen) as described by the manufacturer.

Lucifer ase r eporter construction
The 5 UTR of CCND2 mRNA was cloned into the pGL3-TK-5UTR-BsmBI-Luciferase reporter plasmid purchased from Addgene ( https://www.addgene.org/114670/ ). The DNA fragment was amplified from cDNA pr epar ed from HEK293T using primers with the extra 5 end corresponding to the BsmbI cut sites in the plasmid: 5 -AACGT CT CCACAC-3 f or the f orward primer and 5 -AACGT CT CT CTT CCAT-3 for the r e v erse primer. The DNA fragment and backbone were then cleaved via the restriction enzyme BsmbI for 1 hour at 55 • C and then ligated using T4 DNA ligase at room temperature, followed by selection on X-Gal coated plates. White colonies were cultured and verified using Sanger sequencing. To construct mutant 5 UTRs, Q5 high fidelity polymerase was used to introduce the desired mutations. 0.5 l of the PCR reaction is then incubated with a mixture of T4 DNA ligase, T4 PNK, and DpnI in T4 DNA ligase buffer for 1 h (37 • C, 15 min; 25 • C,30 min; 37 • C, 15 min). 5 l of this reaction is transformed into competent cells and then plated on ampicillin agarose plates.

PAR-CLIP
FLAG-tagged full-length and fragments of PRRC2B were cloned into 5 -KpnI and 3 -NotI sites in pcDNA3.1 plasmid for ov ere xpression in HEK293T cells. PAR-CLIP was perf ormed essentially f ollowing the protocol previously described ( 31 ) with some modifications. Briefly, cells were treated with 800 mM 4sU 16 h before crosslinking by 312 nm ultraviolet exposure using Stratagene Stratalinker UV 1800. Crosslinked cells were lysed in a nati v e lysis buffer with 50 mM HEPES, pH 7.5, 150 mM KCl, 2 mM EDTA, 1 mM NaF, 0.1% (v / v) Nonidet P -40, 1 mM DTT, 25 l / ml protease inhibitor cocktail for mammalian tissues (Sigma-Aldrich), and 1 U / l RNase T1. Immunoprecipitation was performed using anti-DYKDDDDK magnetic agarose (ThermoFisher) followed by on-bead RNA digestion with 5 U / l RNase T1, on-bead dephosphorylation with calf intestinal alkaline phosphatase (CIP), and onbead phosphorylation with T4 PNK and 32 [P]-ATP. Radiolabeled RNA-protein complexes were extracted from beads by boiling in 70 l of SDS loading buffer (Bio-Rad), separated on a 4-12% Bis-Tris gel, and imaged with phosphorimager. Bands corresponding to the protein size were cut off from the gel and subjected to proteinase K digestion and RNA purification by phenol-chloroform. The purified RNAs were run on 15% denaturing polyacrylamide gel and purified for deep sequencing. Sequencing libraries were constructed using the NEBNext ® Small RNA Library Prep Set following the manufacturer's instructions. Sequencing was performed on Illumina HiSeq2500 sequencer with 75 bp single-end sequencing.

PAR-CLIP data analysis
We applied the CLIP Tool Kit (CTK) to perform all steps of the analysis from raw reads to binding sites and target genes as previously described ( 32 ). Briefly, raw reads were filtered based on the mean quality score with CTK builtin function and processed by cutadapt / b1 ( 33 ) to remove adapters. Processed reads with the identical sequences were labeled as PCR duplicates and removed from downstream anal ysis. Ma pping was performed using bowtie / 1.3.1 ( 34 ) with the -best -strata option against the Homo sapiens r efer ence genome (GRCh38, hg38) downloaded from the UCSC genome browser ( http://genome.ucsc.edu ). One mismatch for each alignment was allowed to tolerate the T to C mutation introduced by 4sU crosslinking. Clustering and peak calling were performed on aligned reads with CTK built-in functions using the 'valley seeking' algorithm ( 32 ). Peak significance was first assigned based on whether the observed peak height is more than expected by chance using background models and scan statistics, then further evaluated based on the presence of T to C mutations. Significant peaks around T to C mutations were then treated as binding sites. Sequences flanking the T-to-C mutation sites (-20 to + 20 nt) in the overlapped binding sites between fulllength PRRC2B and P2 on mRNA exons were submitted to MEME ( 35 ) for de novo motif discovery with settings to find the longest possible motif enriched against a randomized set of sequences of the same length. Motifs with E -values < 0.05 were reported.

RNA purification and RT-qPCR
Cells were lysed by 1000 l of Trizol (Qiagen) and mixed with 200 l chloroform. The mixture was spun down at 16,000 g for 10 min. RNA was precipitated from the aqueous layer by adding two volumes of isopropanol and spinning down at 16,000 g for 10 min. The pellet was washed twice with 70% ethanol, left to dry, and resuspended in nuclease-free water. DNase I was used to remove genomic DNA contamina tion. For quantita tion of mRN A, cDN As wer e pr epar ed using iScript master mix RT Kit (Biorad) and subsequently qPCR-amplified using SYBR Primer Assay kits (Biorad). Notabl y, w hen a primer set was first used, the identity of the resulting PCR product was confirmed by cloning and Sanger sequencing. Once confirmed, melting curves were used in each subsequent PCR to verify that each primer set reproducibly and specifically generates the same PCR product. We included all the detailed information r equir ed f or MIQE in supplemental inf ormation and Table S8.
Polysome profiling and RNA extraction 1 × 10 9 cells were incubated with cy clohe ximide (100 g / ml; Sigma) for 10 min and harvested using a native lysis buffer with 100 mM KCl, 5 mM MgCl 2 , 10 mM HEPES, pH 7.0, 0.5% Nonidet P -40, 1 mM DTT, 100 U / ml RNasin RNase inhibitor (Promega), 2 mM vanadyl ribonucleoside complexes solution (Sigma-Aldrich (Fluka BioChemika)), 25 l / ml protease inhibitor cocktail for mammalian tissues (Sigma-Aldrich), cy clohe ximide (100 g / ml). The lysate was spun down at 1500 g for 5 min to pellet the nuclei. The supernatant was loaded onto a 10-50% sucrose gradient and spun in an ultracentrifuge at 150,000 g for 2 h. The gradients were then transferred to a fractionator coupled to an ultraviolet absorbance detector that outputs an electr onic trace acr oss the gradient. Using a 60% sucrose chase solution, the gradient was pumped into the fractionator and divided equally into 9 or 10 fractions. For subsequent RNA extraction, 400 l of each fraction was mixed with an equal amount of chloroform: phenol: isoamyl alcohol and 0.1 × of 3 M sodium acetate (pH 5.2), and then spun down at 16,000 g for 10 min. The upper aqueous layer was mixed with 3 volumes of 100% ethanol and 0.1 volume of 3M sodium acetate (pH 5.2) and incubated at -20 • C overnight. The solution was then spun at maximum speed for 30 min to pellet the RNA, which was washed twice with 70% ethanol and resuspended in nuclease-free water. DNase I was then added to remove genomic DNA contamination followed by another round of RNA purification. For RNA deep sequencing, RNA purified from fractions #6-10 was combined as polysome-associated RNA.

Mass spectrometry analysis
Mass spectrometry (MS) was performed by the Mass Spectrometry Resource Lab of the Uni v ersity of Rochester Medical Center. Briefly, the IP elution was loaded onto a 4-12% gradient SDS-PAGE gel, run for 10 min, and stained with SimplyBlue SafeStain (Invitrogen). The stained regions were then excised, cut into 1mm cubes, de-stained, reduced, and alkylated with DTT and IAA (Sigma), dehydrated with acetonitrile, and incubated with trypsin (Promega) at 37 • C overnight. Peptides were then extracted by 0.1% TFA and 50% acetonitrile, dried down in a CentriVap concentrator (Labconco), and injected onto a homemade 30 cm C18 column with 1.8 m beads (Sepax) on an Easy nLC-1000 HPLC (ThermoFisher Scientific) connected to a Q Exacti v e Plus mass spectrometer (ThermoFisher Scientific) for MS. Raw data from MS experiments were mapped against Swis-sProt human database using the SEQUEST search engine within the Proteome Discover er softwar e platform, version 2.2 (ThermoFisher Scientific). The Minora node was used to determine relati v e protein abundance using the default settings. P ercolator w as used as the FDR calculator, and peptides with a q -value > 0.1 were filtered out. Biological replicates and two IPs using rabbit pre-immune IgG (negati v e control) were included for calculating SAINT probability. The calculation was performed by the CRAPome w e b tool ( 43 ) according to the user guide.

Western blotting
Cells were lysed in RIPA buffer (Thermo Fisher Scientific), and total cell proteins were separated in a 10% denaturing polyacrylamide gel, transferred to polyvinylidene difluoride membranes (PVDF; Amersham Biosciences), probed using primary antibodies, and incubated with either a mouse or rabbit secondary antibod y conjuga ted with horseradish peroxidase (GE Biosciences). According to the manufacturer's suggestions, protein abundance was quantified by the Pierce BCA Protein Assay kit (Thermo, UL294765) for quantitati v e western b lotting. If applicab le, the same protein mass was loaded, and ␤-actin was used as an internal control.

Dual-lucifer ase r eporter assay
Control or PRRC2B knockdown HEK293T cells were grown in 96-well plates until 80% confluency. The cells were then transfected with an equal amount of experimental FLuc reporter plasmid and a control renilla luciferase (RLuc) plasmid for 18 h. The cells were then incubated with Dual-Glo lucifer ase substr ate (Promega) according to the manufactur er's r ecommendations. The final r eadings of the FLuc were normalized to RLuc to obtain the relati v e luminescence reading. To compare between shCtrl and sh-PRRC2B, the relati v e luminescence was further normalized against reporter mRNA abundance ( FLuc / RLuc ).

Cellular phenotyping
For cell proliferation, cells were seeded at 500 / well in 96well plates as biological triplicates. MTT assay was performed using MTT Cell Proliferation Kit I (Sigma-Aldrich) Nucleic Acids Research, 2023, Vol. 51, No. 11 5835 per the manufacturer's recommendations e v ery 24 h until day 7. Cell-doubling time was calculated for cells in the log phase of the growth curve. For the cell cycle, flow cytometry was performed with propidium iodide (PI) stained cells according to the standardized protocol ( 44 ). Biological triplicates were perf ormed f or flow cytometry. G0 / G1, S,and G2 / M peaks were autodetected and quantified by FlowJo, and percentages of total cells detected by flow cytometry wer e r eported.

Statistics
All quantitati v e data wer e pr esented as mean ± SD and analyzed using Excel (Microsoft Of fice). Sta tistical analyses were performed using the two-tailed Student's t test with P < 0.05 considered significant.

PRRC2B directly interacts with a subset of mRNAs
We first applied photoacti vatab le ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) ( 31 ) to identify transcriptome-wide PRRC2B binding sites in HEK293T cells. During PAR-CLIP, RBPs were crosslinked to their bound RNAs by 4-thiouridine (4sU) and UV exposur e, enriched by immunopr ecipitation (IP) after RNase T1 digestion, and visualized on Bis-Tris gel by radioacti v e labeling of bound RNAs (Figure 1 A, Supplementary Figure  S1A). Following this method, we enriched FLAG-tagged full-length and three truncated variants of PRRC2B (P1, P2, and P3) (Figure 1 B). While the extent of enrichment is comparable (Supplementary Figure S1B), only full-length and 750 amino acid-long RG-rich motif containing P2 variant of PRRC2B showed significant RNA signal at the position of the protein with corresponding size (Figure 1 C). Crosslinked RNAs (mostly 19-35 nt) were then purified from PRRC2B and P2, visualized on polyacrylamide gel (Figure 1 D), and subjected to library construction and deep sequencing. Following an established analysis pipeline with in-house modifications, we identified a total of 1062 binding sites for full-length PRRC2B (Supplementary Table S1) and 749 binding sites for P2 (Supplementary Table S2) from two biological replica tes tha t highly correla te with each other (Supplementary Figure S1C, D). Among PRRC2B and P2 CLIP sites, 462 overlapped binding sites were found (Supplementary Tables S1, S2), accounting for ∼44% of the fulllength binding sites and ∼62% of the P2 binding sites (Figure 1 E). The majority of the overlapped binding sites were on mRNAs (326, ∼71%), with a small portion on long noncoding RN A (lncRN A) (62, ∼13%), miRN A (25, ∼5%), snoRNA (3, ∼1%), and other RNA species (2, < 1%) (Figure 1 F). Similar distributions were also observed in fulllength PRRC2B and P2 binding sites when analyzed separately (Supplementary Figure S1E). Further examination of the binding sites on mRNAs re v ealed tha t 5 untransla ted region (5 UTR) possessed the highest density (binding site per 1000 nt), followed by the coding sequence (CDS), 3 untranslated region (3 UTR), and intron (Figure 1 G, S1F). The enrichment in mature mRNAs (5 UTR, CDS, 3 UTR) versus introns was further supported by the fact that the majority of PRRC2B are localized in the cytosol where mature mRNAs (5 UTR, CDS, 3 UTR) are dominant (Supplementary Figure S1G). Interestingly, we observed a highest peak of binding sites close to the start codon region of the main open reading frames (mORFs) (Figure 1 H, Supplementary Figure S1H). A more detailed examination re v eals that majority of the peaks in 5 UTR are at 70 nt upstream of the start codon of mORFs (Figure 1 I). Subsequent scanning of consensus sequences in the overlapped binding sites across intact mRNAs and in the sub-regions (5 UTR, CDS, 3 UTR) on mRNAs by MEME ( 35 ) showed motifs that are either GA-or CU-rich (Figure 1 J, S1I), w hich hardl y accrue within the same binding site. Together, these results suggest that PRRC2B directly binds to a select cohort of mRNAs at the region near translation initiation sites.

Loss of PRRC2B decreases translation of its bound mRNAs
The enrichment of PRRC2B binding sites near the transla tion initia tion sites suggested its potential involvement in transla tional regula tion. Accor dingly, we e xamined changes in transla tion ef ficiency a t the transcriptomic le v el upon PRRC2B knockdown by polysome profiling and RNA sequencing (Figure 2 A) ( 45 , 46 ). A doxy cy cline (Dox)inducible short hairpin RN A (shRN A) targeting PRRC2B (shPRRC2B) was used to knockdown PRRC2B with a non-targeting shRNA as a control (shCtrl) (Figure 2 B, C). RNA-seq was performed on polysome-associated mR-NAs (associated to > 2 ribosomes, considered acti v ely translated) and total mRNAs from cells 36 h after Dox induction of PRRC2B knockdown to calculate translation efficiency (TE; ratio of FPKM value of polysome-associated RNA to that of total RNA) for each mRNA in PRRC2B knockdown (shPRRC2B) and control (shCtrl) cells. In the sequencing results, a total of 25,537 mRNAs were detected of which 1455 mRNAs exhibited increased TE (Log 2 (TE shPRRC2B / TE shCtrl ) > 0.5; adjusted P value < 0.05; TE-up) upon PRRC2B knockdown while 3197 mRNAs exhibited decreased TE (Log 2 (TE shPRRC2B / TE shCtrl ) < -0.5; adjusted P value < 0.05; TE-down) (Figure 2 D, Supplementary Table S3). Gene Ontology (GO) analysis re v eals enrichment of the TE-up genes in 'pattern specification process' and 'mitochondrial transport' while enrichment of TE-down genes in 'cytosolic transport' and 'regulation of mRNA processing' (Supplementary Figure S2A, B, Supplementary Table S4). To further elucidate the TE changes caused directly by loss of PRRC2B binding, we focused on mRNAs with the overlapped binding sites (PRRC2Bbound mRNAs) (Supplementary Table S5). We observed significantly changed TE in 98 out of total 223 PRRC2Bbound mRNAs, among which only 18 mRNAs exhibited increased TE (Log 2 (TE shPRRC2B / TE shCtrl ) > 0.5; adjusted P value < 0.05; TE-up) while 80 mRNAs showed decreased TE (Log 2 (TE shPRRC2B / TE shCtrl ) < -0.5; adjusted P value < 0.05; TE-down) (Figure 2 E). The TE-up PRRC2Bbound mRNAs were functionally enriched in 'cytoplasmic translation' and 'ribosome assembl y' w hile TE-down PRRC2B-bound mRNAs were enriched in 'stem cell population maintenance' and 'maintenance of cell number' (Supplementary Figure S2C, D, Supplementary Table S6).
Since d ysregula ted TE in unbound mRNAs were detected, we suspected that secondary effects might have been The longest significant consensus motifs that MEME can identify from the sequences flanking the T-to-C mutation sites (-20 to + 20 nt) in the 326 overlapped binding sites on mRNAs. Significance was tested against randomized sequences according to the MEME user guide. E-value estimates the expected number of motifs found in a similarly sized set of random sequences ( 35 ). E -value < 0.05 was considered statistically significant. triggered by the primary changes due to PRRC2B knockdown. To distinguish the primary and secondary changes, we examined the translational changes of some PRRC2Bbound and unbound mRNAs at earlier time points after Dox induction by performing re v erse transcriptionquantitati v e PCR (RT-qPCR) on RNAs extracted from polysome profiling fractions (Figure 2 A). All tested PRRC2B-bound mRNAs were first validated to interact with the endogenous PRRC2B in cells by RNA-binding protein immunoprecipitation (RIP) and RT-qPCR (Supplementary Figure S2E). All tested TE-down PRRC2Bbound mRNAs showed shifts towards lighter polysome or pre-polysome fractions (with smaller number of ribosomes on mRNAs) from heavier polysomes fractions (with larger number of ribosomes on mRNAs) at 18 or 24 h after Dox induction, suggesting reduced transla tion ef ficiency (Figure 2 F, Supplementary Figure S2F Figure S3B). Together, these results suggest that loss of PRRC2B primarily decreases translation of its bound mRNAs.

PRRC2B binding to CCND2 mRNA facilitates CCND2 protein expression
To further elucidate the role of PRRC2B binding on mR-NAs, we selected cyclin D2 ( CCND2 ) mRNA, which has a CU-rich PRRC2B binding site in 5 UTR (Figure 3 A), as an example to examine the relationship between PRRC2B binding and protein expression. Antisense oligonucleotides (ASO) base-paring with the PRRC2B binding site on CCND2 mRNA (ASO1) were used to specifically block PRRC2B binding by forming double-stranded RNA at the PRRC2B binding site (Figure 3 ). ASO1 significantly inhibited the binding of PRRC2B, while ASO base-paring with the region adjacent to binding sites (ASO2) (Figure 3 ) and ASO without targets in human cells (Ctrl ASO) had minor effects (Figure 3 B). While none of the ASOs tested caused a significant alteration in CCND2 mRNA abundance (Figure 3 C), ASO1 significantly decreased CCND2 protein expression in wild-type cells (Figure 3 D). Notably, reduced CCND2 protein and mRNA expression were also observed in ASO1-treated shCtrl-expressing cells while not observed in ASO1-tr eated shPRRC2B-expr essing cells at 36 h after Dox induction of PRRC2B knockdown (Supplementary Figure S3C, D), confirming the requirement of PRRC2B for the mechanism of action of ASO1.
We next investigated the importance of the CU-rich elements in PRRC2B binding to CCND2 mRNA by dualluciferase reporter assays. Wild-type (WT-FLuc) and mu- To test whether the observed luciferase expression differences are PRRC2B-dependent rather than just sequencedependent, we measured the reporter expression in shCtrl cells and shPRRC2B cells at 36 h after Dox induction of PRRC2B (Supplementary Figure S3F). Compared with WT-FLuc or CT2GA-FLuc, we observed that the expression of Rep-FLuc and C2G-FLuc decreased in shCtrl cells while increased in shPRRC2B cells ( Supplementary Figure S3F, orange and dark gray bar). Although the exact reason for the increased expression in shPRRC2B cells is unclear (possibly due to secondary or compensatory effect triggered after PRRC2B knockdown), our data suggests that PRRC2B plays a role in the transla tion regula tion of C2G-FLuc and T2A-FLuc mRNAs. T2A-FLuc, on the other hand, showed little expression in both shCtrl and sh-PRRC2B cells (Supplementary Figure S3F, light gray bar), suggesting that the 5 UTR sequence changes (T2A) themselves might decrease luciferase expression independent of PRRC2B to some extent. Interestingly, we still observed significant luminescence of WT-FLuc and CT2GA-FLuc in shPRRC2B cell (Supplementary Figure S3F, blue bars). We suspect that this might result from other mechanisms tha t can transla te PRRC2B mRNA targets in the absence Relati v e mRNA abundance (normalized by 18S rRNA and RLuc mRNA) of reporter mRNAs detected by RT-qPCR. For all the histograms in this figure, biological triplicates were performed. Data were expressed as mean ± SD, and the Student's t test was used to calculate the significance. NS, not significant; * P < 0.05; ** P < 0.01; *** P < 0.001. All western blot results are presented with quantification plots and analyzed by Student's t test. Significance was calculated by comparing to Ctrl ASO in (D), * P < 0.05; ** P < 0.01; *** P < 0.001. of PRRC2B and the long half-life of the luciferase protein.
Together, these results suggest that PRRC2B binding to the target mRNA, CCND2 , promotes protein expression without changing the mRNA abundance.

PRRC2B knockdown inhibits cell proliferation
We next examined the biological consequence of PRRC2B binding on target mRNAs by focusing on cellular changes upon PRRC2B knockdown. As 'maintenance of cell number' is the top functionally enriched GO term of TEdown PRRC2B-bound mRNAs (Supplementary Figure  S2D, Supplementary Table S6), changes in cell proliferation were first examined. Decreased cell proliferation and increased doubling time were observed from 1 to 7 days after induction of PRRC2B knockdown (Figure 4 A, B). This change could be partially explained by the observed decrease in the G1 / S phase transition (Figure 4 C, D). Of note, similar changes were not observed in cells without induction of PRRC2B knockdown while confirmed in cells treated with two additional shRNAs targeting PRRC2B (Supplementary Figure S4A-H). The decreased G1 / S phase transition could result from the reduced expression of cyclin D2, which has been established to facilitate the G1 / S phase transition ( 47 , 48 ). Accordingly, decreased G1 / S phase transition was observed when PRRC2B binding to CCND2 mRNA was blocked by ASO1 (Figure 4 E, F). Together, these results suggest that PRRC2B affects cell proliferation Data wer e expr essed as mean ± SD in A, B, D and F, and the Student's t test was used to calculate the significance. NS, not significant; * P < 0.05; ** P < 0.01; *** P < 0.001. through regulating the translation of cell cycle regulators such as cyclin D2.

PRRC2B interacts with eukaryotic initiation factors eIF4G2 and eIF3
To explore mechanistically how PRRC2B regulates the translation of PRRC2B-bound mRNAs, we performed IP followed by liquid chromato gra phy-mass spectrometry analysis to capture the interactome of endogenous PRRC2B (Figure 5 A). Ninety-fiv e high-confidence interactions were identified (SAINT (Significance Analysis of IN-Teractome) probability > 0.95) ( 49 ), of which translation initiation factors eIF4G2 and eIF3 subunits were predominant (Figure 5 B, Supplementary Table S7). The 3 UTR-binding protein FXR1 ( 50 ) known to co-IP with eIF4G2 ( 23 ) was also detected (Figure 5 B, Supplementary Table  S7). These highly confident interactions were validated by western blot following IP of either endogenous or FLAGtagged PRRC2B (Figure 5 C, Supplementary Figure S5A). They were further confirmed to be RNA-independent since they were retained following RNase T1 digestion of RNA to less than 35-nt long (Supplementary Figure S5B, C). In contr ast, no inter action of PRRC2B with eIF4G1 or capbinding protein eIF4E was observ ed e v en though they are highly abundant and dri v e canonical cap-dependent transla tion initia tion (Figure 5 B , C , Supplementary Figure S5A, C). This suggests that PRRC2B may participate specifically in the eIF4G2-mediated translation initiation ( 22 ). This notion is further supported by the co-presence of PRRC2B, eIF4G2, eIF3 and FXR1 in the 40S and 60S fractions where transla tion initia tion is happening (Figure 5 D). Surprisingly, a significant amount of PRRC2B was observed in the 80S and disome fractions (Figure 5 D). Subsequent IP experiments showed that P2 and P3 variants of PRRC2B contribute to its interaction with FXR1, while P1 is primarily responsible for interactions with eIF4G2 and eIF3 (Supplementary Figure S5D). A more detailed IP-based mapping of truncated P1 showed that the N-terminal 1-150 amino acids of PRRC2B contain the eIF3 binding domain while the 1-450 amino acids region interacts with eIF4G2 (Supplementary Figure S5E, F). Sequences in these N-terminal r egions ar e highl y conserved among PRRC2B ortholo gues among vertebrates (from zebrafish to humans), highlight-ing its functional importance (Extended Data 1). Together, these results suggest that PRRC2B interacts with translation initiation factors on translated mRNAs.

PRRC2B functions in translational regulation by interacting with translation initiation factors eIF3 and eIF4G2
Subsequentl y, we examined w hether PRRC2B functions through interacting with translation initiation factors eIF3 and eIF4G2. As eIF4G2 is a noncanonical translation initiation factor, we first examined its role in translating three PRRC2B-bound mRNAs ( CCND2 , CRKL , YWHAZ ). We observed that all three proteins showed decreased expression upon siRNA-mediated eIF4G2 knockdown, suggesting that eIF4G2 is involved in the translation of PRRC2B-bound mRNAs (Supplementary Figure S6A). This notion is further supported by the fact that increased expression of the three proteins was not observed in eIF4G2 knockdown cells but in wild-type cells upon ov ere xpressing PRRC2B (Figure 6 A). We then asked whether interacting with eIF4G2 and eIF3 is essential for the PRRC2B function. We constructed a truncated PRRC2B variant with no eIF4G2 or eIF3 binding ability (del450) (Figure 6 B, C) and tested whether del450 could facilitate the translation of PRRC2B-bound mRNAs. Our rescue experiments showed that del450 failed to fully rescue the protein expression and translational changes on CCND2 , CRKL and YWHAZ mRNAs caused by PRRC2B knockdown, while the full length PRRC2B rescued the protein expression successfully (Figure 6 D, S6B), suggesting del450 cannot facilitate translation of PRRC2B-bound mRNAs. Similar changes were also observed when rescuing inhibitions in cell proliferation and G1 / S phase transition by wild-type PRRC2B or del450 ( Figure 6 E-G, Supplementary Figure S6C). Together, these findings demonstrate that interaction with translation initiation factors eIF4G2 and eIF3 is essential for the function of PRRC2B in translational regulation.

DISCUSSION
In this study, we demonstrate the role of PRRC2B, a novel RNA-binding protein, in the translational regulation of PRRC2B-bound mRNAs. We discover that PRRC2B binds to CU-or GA-rich sequences near the translation initiation site, facilitating the synthesis of a group of proteins related to cell proliferation. This effect can be inhibited by antisense oligonucleotide targeting PRRC2B binding sites in PRRC2B-bound mRNAs such as CCND2 mRNA. Mechanistically, the function of PRRC2B primarily depends on its interaction with eIF4G2 and eIF3, suggesting a positi v e regulatory role of PRRC2B in eIF4G2-mediated translation. Altogether, this study re v eals a nov el translation regula tory pa thway tha t maintains ef ficient cell prolifera tion and can be targeted to diminish cell cycle progression.
Using PAR-CLIP ( 31 ), we demonstrated the binding of PRRC2B to mature mRNAs. Unlike many promiscuous RNA-binding proteins, PRRC2B binds to both CUand GA-rich RNA elements. Although CU-and GA-rich RNA sequences tend to be complementary, we assume that PRRC2B binds to single-stranded RNA since these elements seldom appear adjacent to each other, and ther efor e are unlikely to base pair to form double-stranded RNA. Our mutagenesis assays demonstra ted tha t muta ting a CUrich binding site to GA-rich has a minor effect on PRRC2B binding and function. We conclude that PRRC2B binds to both motifs with a similar affinity. Although CU-and GArich motifs are found throughout mature mRNAs (in 5 UTR, CDS, and 3 UTR), PRRC2B predominately binds the motif sequence near the translation initiation sites where it interacts with transla tion initia tion factors eIF4G2 and eIF3. Despite the fact that eIF4G2 has been accounted for tr anscript selectivity ( 23 , 51 ), inter acting with it hardly affects PRRC2B binding since the PAR-CLIP for P2, the 751-1500 amino acid-containing fragment of PRRC2B not interacting with eIF4G2 or eIF3, exhibited similar binding profile and target sequences as the full-length PRRC2B. Based on this, we conclude that the P2 region of PRRC2B dicta tes a t least some degree of transcript selectivity in an eIF4G2-and eIF3-independent manner. eIF4G2 has been evident to dri v e noncanonical eIF4G1and eIF4E-independent translation initiation ( 22 , 23 , 52 ) together with eIF3D, which replaces eIF4E in cap-binding ( 52 ), and FXR1, which binds to mRNA 3 UTR and facilita tes mRNA circulariza tion ( 50 ). We conclude from the PRRC2B interactome ( Figure 5 , S5) that PRRC2B selecti v ely participates in eIF4G2-mediated translation initiation but not the canonical eIF4G1-media ted transla tion initiation. This discovery echoes previous studies on eIF4G2 ( 22 , 23 ) and is essential for understanding the PRRC2B function ( Figure 6 ). We specula te tha t PRRC2B may act as a scaffold recruiting other factors since it interacts with mRNA, eIF4G2, eIF3, and FXR1 through distinct regions with little dependency on each other (Figure 1 , Supplementary Figure S5). Besides, considering that PRRC2B dictates transcript selectivity, it may navigate eIF4G2 to specific mRNAs. Howe v er, gi v en the mRNA abundance of eIF4G2 is roughly three times of PRRC2B in HEK293T cells (FPKM value detected by our RNA-seq), PRRC2B likel y onl y participa tes in a fraction of eIF4G2-media ted translation. In addition, PRRC2B does not completely cosediment with translation initiation complexes in the 40S to 80S fractions ( Figure 5 D), suggesting that it may have other functions beyond translation initiation (e.g. initiationto-elongation transition). To establish the role of PRRC2B in eIF4G2-mediated translation initiation, future studies involving in vitro translation are required. Howe v er, obtaining recombinant PRRC2B will be significantly challenging gi v en that full-length PRRC2B is 243 KDa and highly disordered, as predicted by AlphaFold ( 53 ).
Although the mechanism is not fully defined, PRRC2B facilita tes the eIF4G2-media ted transla tion of its bound mRNAs (Figures 2 , 3 ). These mRNAs include multiple oncogenes and cell cycle regulators such as CCND2 ( 47 , 48 ) and CRKL (54)(55)(56), agreeing with the link between PRRC2B and various types of human cancer (24)(25)(26)(27)(28)(29)(30). Our results suggest that PRRC2B promotes cell proliferation by facilitating the translation of CCND2 mRNA, which encodes the D-type cyclin facilitating G1 / S transition in cancer and stem cells ( 47 , 57 ). Blocking PRRC2B binding to CCND2 mRNA by ASOs significantly decreases the CCND2 expression in cells expressing PRRC2B ( Figure  3 ), thereby inhibiting cell proliferation. This discovery inspired us to inhibit the expression of oncogene and cell cycle proteins by ASOs e xclusi v ely in cells highly expressing PRR C2B. Because PRR C2B is highly expressed in cancer cells such as large B cell lymphoma (DLBC), thymus cancer (THYM) ( 25 ), and Wilms' tumor ( 24 ), these ASOs can be applied for potential anti-cancer treatment, which deserves future investigation in biological or disease-relevant models. In addition, se v eral other validated PRRC2B target mRNAs (Supplementary Figure S2E, F) encode oncogenic or pr o-pr oliferation pr oteins that may contribute to PRRC2B-media ted regula tion of cell cycle progression, including SERBP1 ( 58 ), CTC1 ( 59 ), ATAD5 ( 60 ), and BPTF ( 61 ). Moreov er, multiple Torin 1-sensiti v e mTORC1 (mammalian target of rapamycin complex 1) pathway Data wer e expr essed as mean ± SD in (E, F, G), and the Student's t test was used to calculate the significance. NS, not significant; * P < 0.05; ** P < 0.01; *** P < 0.001. All western blot r esults ar e pr esented with quantification plots and analyzed by Student's t test. Significance was calculated by comparing to shCtrl in (A) and shCtrl + FLAG OE in (H). * P < 0.05; ** P < 0.01; *** P < 0.001. downstream mRNAs ( 62 ) were identified as PRRC2Bbound target transcripts by PAR-CLIP-seq and polysomeseq, such as EEF2 , HSP90AB1 , and HSPA8 (Figure 2 E,  Table S5), indicating potential crosstalk between PRRC2B-eIF4G2 and mTORC1 pathways, which may need further studies.
While PRRC2B is evident to facilitate the translation of its bound mRNAs, we still found targets to have unchanged or increased transla tion ef ficiency upon PRRC2B knockdown. The unaltered TE could indicate that the translation of PRRC2B-bound mRNAs may not entirely depend on PRRC2B-mediated translation, as there could be other mechanisms sufficient to carry out the translation when PRRC2B is absent. The increased TE is probably due to secondary or compensatory effects since these mRNAs, similar to the TE-changed unbound mR-NAs, exhibited no translational changes at earlier time points after induction of PRRC2B knockdown (Supplementary Figure S2G). These secondary or compensatory effects could result from the primary changes in TE of PRRC2B-bound mRNAs or the altered cell cycle progression upon PRRC2B knockdown ( Figure 4 ) since translation varies dramatically in different phases of the cell cycle ( 63 ).
The PRRC2 proteins have been repeatedly linked with mRNA binding and posttranscriptional regulation ( 21-23 , 64-67 ). A recent study suggests that PRRC2 proteins facilitate leaky scanning of upstream open reading frames (uORF) ( 67 ). Howe v er, our results showed both mRNAs with and without uORFs are bound and regulated by PRRC2B , suggesting tha t the ef fect of PRRC2B is not limited to uORFs but determined by the presence of GA-and CU-rich binding sites. Besides, previous studies showed PRRC2A interacts with N6-methyladenosine (m 6 A) around the stop codon to stabilize mRNA ( 64 ), while PRRC2C facilities the efficient formation of stress granules ( 65 , 66 ). Although PRRC2B shares sequence similarities with PRR C2A and PRR C2C (Extended Data 2), the reported function of PRRC2A and PRRC2C were not recapitulated in PRRC2B (Figure 1 H, Supplementary Figures S1H, 2I, S2H, data not shown). On the other hand, PRRC2B loss-of-function mutations have been linked with human congenital heart defects ( 68 ) and a variety of mouse de v elopmental defects, including car diovascular de v elopmental defects ( 69 ). Ther efor e, the functional di v ersity, r edundancy of the thr ee members of the PRRC2 protein family, and the biological role of PRRC2B in an in vivo system, warrant future elucidation.

DA T A A V AILABILITY
Data supporting the findings of this manuscript are available from the corresponding author upon request. The sequencing data raw files and processed files were deposited in the GEO database and are accessible under accession numbers GSE220057, GSE220058 and GSE220059.