Abstract

   Genetic and epigenetic alterations occur in many physiological and
   pathological processes. The existing knowledge regarding the
   association of PIWI‐interacting RNAs (piRNAs) and their genetic
   variants on risk and progression of prostate cancer (PCa) is limited.
   In this study, three genome‐wide association study datasets are
   combined, including 85,707 PCa cases and 166,247 controls, to uncover
   genetic variants in piRNAs. Functional investigations involved
   manipulating piRNA expression in cellular and mouse models to study its
   oncogenetic role in PCa. A specific genetic variant, rs17201241 is
   identified, associated with increased expression of PROPER (piRNA
   overexpressed in prostate cancer) in tumors and are located within the
   gene, conferring an increased risk and malignant progression of PCa.
   Mechanistically, PROPER coupled with YTHDF2 to recognize N
   ^6‐methyladenosine (m^6A) and facilitated RNA‐binding protein
   interactions between EIF2S3 at 5′‐untranslated region (UTR) and
   YTHDF2/YBX3 at 3′‐UTR to promote DUSP1 circularization. This
   m^6A‐dependent mRNA‐looping pattern enhanced DUSP1 degradation and
   inhibited DUSP1 translation, ultimately reducing DUSP1 expression and
   promoting PCa metastasis via the p38 mitogen‐activated protein kinase
   (MAPK) signaling pathway. Inhibition of PROPER expression using
   antagoPROPER effectively suppressed xenograft growth, suggesting its
   potential as a therapeutic target. Thus, targeting piRNA
   PROPER‐mediated genetic and epigenetic fine control is a promising
   strategy for the concurrent prevention and treatment of PCa.

   Keywords: N ^6‐methyladenosine, PIWI‐interacting RNA, Prostate cancer,
   Single nucleotide polymorphism
     __________________________________________________________________

   This study identified a prostate cancer risk‐associated SNP,
   rs17201241, located within an oncogenic piRNA called PROPER.
   PROPER‐YTHDF2/YBX3/EIF2S3‐DUSP1 signaling axis guides DUSP1 degradation
   and translation control based on an m^6A methylation‐regulated
   mRNA‐looping.

   graphic file with name ADVS-11-2402954-g004.jpg

1. Introduction

   Prostate cancer (PCa) is the second most common cancer in males and the
   fifth leading cause of cancer death worldwide.^[ [50]^1 ^] Emerging
   evidence has demonstrated that genetic and epigenetic events, including
   genetic variants,^[ [51]^2 , [52]^3 ^] non‐coding RNAs (ncRNAs),^[
   [53]^4 , [54]^5 ^] and N ^6‐methyladenosine (m^6A) methylation,^[
   [55]^6 ^] contribute to the development and progression of PCa.
   Genome‐wide association studies (GWASs) including our previous reports
   coupled with large‐scale collaborative replication efforts in Asians
   have identified over 200 genetic variants that contribute to the
   development of PCa.^[ [56]^7 , [57]^8 ^] As with most other diseases
   and traits, these risk‐associated single nucleotide polymorphisms
   (SNPs) map primarily to noncoding regions of the genome.

   Among different types of ncRNAs, PIWI‐interacting RNAs (piRNAs) are a
   new type of small ncRNAs that were initially identified with high
   abundance in germline cells, showing most well‐known and critical
   regulatory functions in silencing transposons to maintain genome
   integrity in gametogenesis and fertility.^[ [58]^9 , [59]^10 ^]
   Nevertheless, several studies have reported additional regulatory roles
   of piRNAs in cancers of different origins, such as germline cells or
   somatic tissues. For instance, piRNAs such as piR‐823, piR‐001773, and
   piR‐017184 were found to play critical roles in epigenetic regulation
   of cancer progression.^[ [60]^11 , [61]^12 ^] Our previous study showed
   that piRNAs were abundantly expressed and harbored a large number of
   genetic variants in various tumors, suggesting their potential roles in
   carcinogenesis.^[ [62]^13 ^] However, the molecular mechanisms
   underlying the causal and biological effects of these piRNA‐associated
   SNPs conferring the risk of cancers, including PCa, remain largely
   unknown.

   In addition, mounting epitranscriptomic evidence has elucidated the
   crucial role of m^6A in malignancies.^[ [63]^14 ^] In general, the m^6A
   modification is enriched around translation stop codons and in the
   3′‐untranslated region (UTR) to regulate the mRNA fate.^[ [64]^15 ^]
   The dynamically modified m^6A sites are recognized and executed by
   variable readers, such as YTH domain‐containing proteins and
   heterogeneous nuclear ribonucleoproteins (hnRNPs) family proteins,
   thereby regulating mRNA splicing, exporting, stability, and translation
   processes.^[ [65]^16 ^] As a crucial m^6A reader, the YTH domain family
   protein 2 (YTHDF2) preferentially recognizes m^6A and recruits
   RNA‐degrading enzymes or adaptor proteins to trigger rapid degradation
   of the m^6A‐containing mRNA, thereby interfering with mRNA
   translation.^[ [66]^17 ^] However, the mechanisms of ncRNA‐participated
   m^6A‐containing mRNA stability and translation regulation in PCa remain
   elusive.

   Given the complexity of post‐transcriptional modifications in
   regulating gene expression, it is necessary to combine these with
   classical regulatory layers, such as those involving the well‐explored
   mitogen‐activated protein kinase (MAPK) pathways, to explain the
   mechanisms of oncogenesis. Dual‐specificity phosphatases (DUSPs) are
   crucial regulators of cellular signaling pathways, particularly the
   MAPK pathways.^[ [67]^18 ^] These enzymes dephosphorylate both
   threonine and tyrosine residues on MAPKs, effectively modulating their
   activity and ensuring precise cellular responses to various stimuli.
   DUSPs play vital roles in controlling cell proliferation,
   differentiation, and apoptosis, thus maintaining cellular
   homeostasis.^[ [68]^19 ^] By regulating MAPK signaling, DUSP1 can
   impact processes essential for metastasis, such as cell migration,
   invasion, and the epithelial‐mesenchymal transition (EMT). In the
   context of PCa, DUSP1 is downregulated and is significant due to its
   ability to influence tumor progression and metastasis.^[ [69]^20 ,
   [70]^21 ^] However, the post‐transcriptional modification and
   regulation of DUSP1 itself in PCa has not been completely understood.

   In this study, combined GWASs enabled us to find the PCa
   risk‐associated variant rs17201241, residing in PROPER (piRNA
   overexpressed in prostate cancer), which is aberrantly highly expressed
   in advanced PCa. We identified an unexpected RNA epigenetic regulatory
   mechanism by which PROPER is assembled with m^6A binding protein YTHDF2
   and YBX3 to form a piRNA‐induced silencing complex (pi‐RISC), thereby
   inhibiting translation and promoting degradation of DUSP1 with the
   translation initiation factor EIF2S3 in an mRNA‐looping manner.
   Therefore, this study not only reports a previously unknown function of
   piRNA in post‐transcriptional regulation but also provides new
   therapeutic opportunities for targeting the piRNA‐mediated translation
   machinery in cancer.

2. Results

2.1. Identification of PCa Risk‐Associated piRNAs by GWAS Meta‐Analyses

   The flowchart of identifying PCa risk‐associated genetic variants
   residing in piRNAs is shown in Figure [71] 1A. A total of 13695
   differentially expressed piRNAs were observed from The Cancer Genome
   Atlas (TCGA) cohort. Briefly, a total of 169 SNPs were extracted
   according to the physical locations of the differentially expressed
   piRNAs, after standard quality control, only five SNPs being retained
   for further association analyses (Figure [72]1A).

Figure 1.

   Figure 1
   [73]Open in a new tab

   Identification and expression analyses of PCa risk‐associated piRNAs.
   A) Flowchart illustrating the SNP filtrating design. B) Sequences of
   rs17201241 (red arrow) resided piR‐hsa‐136611 derived from SNORD48 were
   analyzed and the conserved motifs were annotated. C) Confirmation of
   piR‐hsa‐136611 binding to PIWIL4 using biotinylated piR‐hsa‐136611
   (Mut) probe in PC‐3 and 22Rv1 cell by biotin‐streptavidin RNA pull‐down
   assay. D) Distribution of PROPER in nucleus and cytoplasm of PCa cells
   with U6 or GAPDH as nucleus or cytoplasm markers, respectively. Data
   are presented as mean ± s.d. (n = 4). E) The detection of PROPER in PCa
   cells were analyzed by fluorescence in situ hybridization using
   DIG‐labeled locked nucleic acid (LNA) probes to PROPER (red). The
   nucleus was counterstained with DAPI (blue). Scale bars, 20 µm. F,G)
   SNORD48 (F) or PROPER (G) expression levels in PCa tissues (n = 483)
   and normal tissues (n = 52) from TCGA database. P values were
   calculated using unpaired two‐sided Mann‐Whitney U‐tests. H,I) SNORD48
   (H) or PROPER (I) expression levels in 31 paired PCa tissues and normal
   tissues by RT‐qPCR. P values were calculated using paired two‐sided
   Mann‐Whitney U‐tests. J) Spearman's rank correlation coefficient
   analysis between SNORD48 and PROPER expression in tumors from TCGA
   database. K) Spearman's rank correlation coefficient analysis between
   SNORD48 and PROPER expression in tumors by RT‐qPCR. L) The association
   between PROPER expression level and Gleason score (GS) in TCGA. P
   values were calculated using unpaired two‐sided Mann‐Whitney U‐tests.
   M) The representative dot of RNA FISH of PROPER and H&E staining in
   prostate tissue microarray containing 90 paired PCa and adjacent normal
   tissues. N) Relative immunofluorescence density [arbitrary units
   (a.u.)] of RNA FISH between 90 paired tumor and adjacent normal
   tissues. Scale bars, 100 µm. P values were calculated using two‐sided
   paired Mann‐Whitney U‐tests. O) Relative immunofluorescence density
   [arbitrary units (a.u.)] of RNA FISH between aggressive and
   non‐aggressive PCa tissues. P values were calculated using unpaired
   two‐sided Mann‐Whitney U‐tests.

   We then combined three independent GWAS datasets through a
   meta‐analysis to detect the risk loci for PCa in Asians and Europeans.
   The results demonstrated that only rs17201241 C>T was significantly
   associated with a decreased PCa risk (OR = 0.94, 95% CI = 0.91‐0.97,
   P = 7.33×10^−5) (Table [74]S1, Supporting Information). We further
   evaluated the genetic effects of rs17201241 on PCa risk with clinical
   features in ChinaPCa, for which the characteristics of the subjects are
   summarized in Table [75]S2 (Supporting Information). There was low
   heterozygosity among clinical variables of N stage (I ^2 = 34%,
   P = 0.22) and M stage (I ^2 = 60%, P = 0.11) in the stratification
   analyses (Table [76]S3, Supporting Information), indicating that the
   rs17201241 T allele may play a stronger protective role in the
   non‐metastatic PCa.

   Next, we performed small RNA deep‐sequencing of PCa cells to find the
   potential PCa risk‐associated piRNA harboring the SNP rs17201241
   (Table [77]S4, Supporting Information). Unexpectedly, we found a peak
   around the size of 31 nts, corresponding to the length of piRNAs
   (Figure [78]S1A,B, Supporting Information). By mapping clean reads to
   the piRBase database and published small RNA deep‐sequencing data of
   germ cells,^[ [79]^22 , [80]^23 ^] we found piR‐hsa‐136611 that
   harboring rs17201241 was widely detected both in PCa tissues, human
   oocytes, and early embryos (Figure [81]S1C, Supporting Information).
   Notably, the sequence alignment analysis showed that piR‐hsa‐136611 was
   located in the same region with a small nucleolar RNA (snoRNA) named
   SNORD48 (Figure [82]1B). This piRNA retained the conserved
   characteristic sequences of snoRNA C’/D boxes (Figure [83]1B),
   following piRNA characteristics reported in previous studies.^[ [84]^24
   , [85]^25 ^]

   The abundance of piR‐hsa‐136611 was relatively high in PC‐3 and 22Rv1
   cells (Figure [86]S1D, Supporting Information). To confirm the
   existence of piR‐hsa‐136611, we isolated small RNAs from PCa cell lines
   and performed a northern blot assay, which confirmed the existence of
   piR‐hsa‐136611 (Figure [87]S1E, Supporting Information). We also
   detected the expression of PIWI proteins by western blotting using
   specific antibodies (Figure [88]S1F, Supporting Information). Next, the
   RNA immunoprecipitation (RIP) and RNA pull‐down assays confirmed the
   specific binding of piR‐hsa‐136611 to the PIWIL4 protein
   (Figure [89]1C; Figure [90]S1G, Supporting Information). Consistent
   with this, piR‐hsa‐136611 expression significantly decreased when
   knocking down PIWIL4 (Figure [91]S1H,I, Supporting Information).
   Furthermore, to examine the 2′‐O‐methylation modification status of 3′
   terminus of piR‐hsa‐136611, a typical feature of piRNA, we conducted
   sodium‐periodate‐mediated oxidation experiments.^[ [92]^26 ^] The
   results indicated that piR‐hsa‐136611 and positive control piR‐hsa‐1681
   exhibited resistance to periodate treatment, indicating their
   methylation (Figure [93]S1J, Supporting Information). In contrast, the
   negative control miRNA miR‐21 were oxidized after treatment and
   therefore degraded (Figure [94]S1J, Supporting Information). Together,
   these results suggest the real existence of piR‐hsa‐136611. As
   piR‐hsa‐136611 has not been documented in the Genebank, we named it
   PROPER, which stands for a piRNA overexpressed in prostate cancer,
   considering the follow‐up analysis.

2.2. PROPER is Up‐Regulated and Promotes Malignant Phenotypes of PCa

   Subcellular localization analysis revealed that PROPER is mainly
   distributed in the cytoplasm (Figure [95]1D,E). Importantly, both
   PROPER and its precursor SNORD48 exhibited higher expression in PCa
   tissues compared to normal tissues in the TCGA cohort
   (Figure [96]1F,G), which was consistent with our independent in‐house
   validation cohort (Figure [97]1H,I). Additionally, PROPER expression
   showed a significant positive correlation with SNORD48 in both the TCGA
   and our patient cohorts (Figure [98]1J,K). Northern blot assays also
   confirmed the elevated expression of PROPER in PCa (Figure [99]S1K,
   Supporting Information). Moreover, PROPER was also found to be highly
   expressed in various other tumors (Figure [100]S1L, Supporting
   Information). It is noteworthy that, increased PROPER expression was
   associated with higher Gleason scores and tumor progression
   (Figure [101]1L; Figure [102]S1M, Supporting Information). To further
   establish the clinical relevance of PROPER expression, we performed RNA
   FISH using DIG‐labeled locked nucleic acid (LNA) probes targeting
   PROPER on a tissue microarray containing 90 paired PCa and normal
   tissues (Table [103]S5, Supporting Information). The results confirmed
   higher expression levels of PROPER in PCa tissues compared to benign
   prostate samples (Figure [104]1M,N), particularly in aggressive PCa
   cases (Figure [105]1O).

   To assess the effects of PROPER on the malignant phenotype of PCa
   cells, we conducted a series of loss‐of‐function assays. The efficiency
   of shRNA‐mediated PROPER knockdown was shown in Figure [106]S2A
   (Supporting Information). PROPER knockdown substantially inhibited the
   proliferation and the colony formation ability of PCa cells
   (Figure [107] 2A,B). Flow cytometry analysis revealed a significant
   increase in apoptosis in PROPER knockdown cells (Figure [108]2C).
   Furthermore, knockdown of PROPER significantly suppressed the migration
   and invasion properties of PCa cells (Figure [109]2D,E). We also
   evaluated the impact of PROPER on tumor growth in vivo by constructing
   a xenograft mouse model (Figure [110]2F). We observed a marked reduced
   in tumor growth rate, fluorescence intensity, and tumor weight in mice
   transplanted with shRNA‐mediated PROPER knockdown cells after 6 weeks
   (Figure [111]2G–J). Subsequent H&E and IHC staining demonstrated that
   PROPER knockdown significantly decreased the Ki‐67 signal
   (Figure [112]S2B, Supporting Information). Moreover, a PCa bone
   metastasis model examined by micro‐computed tomography (micro‐CT)
   showed less bone destruction in PROPER knockdown mice compared to the
   control group (Figure [113]2K). The analysis also revealed a decreasing
   trend in tibial bone mineral density associated with high PROPER
   expression in tumors, with gradual thinning and loosening of trabecular
   bone micro‐architecture (Figure [114]2L–O; Figure [115]S2C, Supporting
   Information). Collectively, these results suggest that PROPER plays an
   oncogenic role in PCa and indicate an obvious link between high PROPER
   expression and aggressive prostate tumor phenotype, potentially
   impacting the clinical progression of PCa.

Figure 2.

   Figure 2
   [116]Open in a new tab

   PROPER functions as an oncogene to promote malignant phenotypes and
   metastasis of PCa in vitro and in vivo. A) Effects of PROPER expression
   on the proliferation of PC‐3 and 22Rv1 cells tested by EdU assay.
   Nuclei labeled with hoechst are in blue. The fluorescent thymidine
   analog EdU was used to identify proliferative cells by labeling their
   DNA in red. Representative images (left, middle) and quantitative
   statistics by flow cytometry (right). Data are presented as mean ± s.d.
   (n = 4); P values were calculated using unpaired two‐sided Mann‐Whitney
   U‐tests. B) Effects of PROPER expression on the colony formation
   capability of PCa cells. Representative images (left) and
   quantification (right) of colony formation results. Data are presented
   as mean ± s.d. (n = 4); P values were calculated using unpaired
   two‐sided Mann‐Whitney U‐tests. C) Effects of PROPER expression on PCa
   cell apoptosis. Data are presented as mean ± s.d. (n = 4); P values
   were calculated using unpaired two‐sided Mann‐Whitney U‐tests. D,E)
   Effects of PROPER on the capability of invasion (D) and migration (E)
   of PCa cells in vitro. Representative images of transwell assays (left
   panels) and quantitative statistics (right panels). Scale bars, 100 µm.
   Data are presented as mean ± s.d. (n = 4); P values were calculated
   using unpaired two‐sided Mann‐Whitney U‐tests. F, G) Representative
   luminescence images of nude mice (F) and xenograft tumors (G) (n = 5).
   H) The growth curves of xenograft tumors. P values were calculated
   using unpaired two‐tailed Student's t‐tests. I,J) Quantitative
   statistics of luminescence imaging and tumor weight in nude mice. Data
   are presented as mean ± s.d. (n = 5); P values were calculated using
   unpaired two‐tailed Student's t‐tests. K) Represents the micro‐CT
   images of the bone metastasis from sacrifice mice. Tibiae are shown in
   left and the section of the micro‐CT image are shown in right (n = 6).
   L–O) Quantitative analyses of bone volume/total volume (BV/TV, %) (L),
   trabecular number (Tb.N*) M), trabecular thickness (Tb.Th*) (N), and
   trabecular separation (Tb.Sp*) (O). Data are presented as mean ± s.d.
   (n  =  6); P values were calculated using unpaired two‐tailed Student's
   t‐tests.

2.3. The rs17201241 T Allele Affects SNORD48 Mature and PROPER Expression

   We proceeded to investigate whether there is an association between
   variation at rs17201241 and PROPER expression. An expression
   quantitative trait loci (eQTL) analysis in a collection of 90 human
   prostate tumors demonstrated that the protective rs17201241 T allele
   was significantly associated with decreased PROPER levels
   (Figure [117]S3A, Supporting Information). Moreover, the sequences of
   PROPER could be precisely aligned within the sequences of SNORD48, and
   the cleavage sites required for PROPER production were not random.^[
   [118]^24 , [119]^27 ^] Considering that Fibrillarin (FBL) or NOP56 are
   known to play critical roles in snoRNA maturation, stability, and small
   nucleolar ribonucleoproteins (snoRNPs) complex assembly
   (Figure [120]S3B, Supporting Information),^[ [121]^28 ^] we performed
   siRNA‐mediated knockdown for FBL and NOP56. As expected, this led to
   reduced expression of both SNORD48 and PROPER (Figure [122]S3C–F,
   Supporting Information). In addition, ectopic expression of SNORD48
   significantly increased PROPER expression (Figure [123]S3G, Supporting
   Information). These results further confirm that SNORD48 serves as the
   precursor of PROPER.

   Since rs17201241 is located within the kink‐turn (k‐turn) structure,
   which is known to be essential for snoRNP assembly,^[ [124]^29 ^] we
   hypothesized that rs17201241 could affect the binding affinity between
   SNORD48 and NOP56, thereby influencing SNORD48 maturation. To verify
   this, we modeled the 3D structure of NOP56 and SNORD48 carrying
   different alleles of rs17201241 (Figure [125]S3H,I, Supporting
   Information). Subsequently, we conducted an in‐slico RNA‐protein
   docking analysis of SNORD48 and NOP56. The results of the molecular
   docking simulation supported the perfect docking of SNORD48 with NOP56,
   with a ZRank score of −38.773 (Figure [126]S3J, Supporting
   Information). Conversely, rs17201241 resulted in a weaker interaction
   between SNORD48 and NOP56 with a ZRank score of −28.312
   (Figure [127]S3K, Supporting Information). Furthermore, when
   transfected SNORD48 plasmid harboring different alleles of rs17201241
   in PC‐3 cells after knocking down of endogenous SNORD48, we found the
   PROPER expression significantly increased in the C allele than in the T
   allele group (Figure [128]S3L, Supporting Information). These results
   suggest that the rs17201241 T allele may impact SNORD48 maturation and
   subsequently decrease the expression of SNORD48‐derived PROPER.

2.4. PROPER Affects YTHDF2‐Mediated mRNA Translation

   We aimed to investigate the molecular mechanisms underlying the role of
   PROPER in PCa. Since ncRNAs can repress target genes located at a
   nearby genomic locus via base‐pairing rules,^[ [129]^30 ^] we initially
   examined the mRNA level of nine genes within a 0.5 m‐base region
   centered around PROPER (Figure [130]S4A, Supporting Information).
   However, knockdown of PROPER did not alter the expression of these
   genes (Figure [131]S4B, Supporting Information), implying that PROPER
   does not function by regulating the expression of neighboring genes.

   Another potential function we speculated was that PROPER interacts with
   proteins to form RNA‐protein complexes. Therefore, we performed
   biotin‐streptavidin RNA pull‐down assays using biotinylated PROPER, and
   several protein bands specifically enriched in PROPER were subjected to
   mass spectrometry (Figure [132] 3A). A total of 14 potential
   PROPER‐interacting proteins were reproducibly identified based on
   unique peptides (Figure [133]3B; and Table [134]S6, Supporting
   Information). Notably, the most enriched pathways among these 14
   proteins were involved in the “Translation, ribosomal structure and
   biogenesis pathway” and “RNA processing and modification pathway”
   (Figure [135]3C). Intriguingly, YTHDFs, which are well‐characterized
   m^6A readers known to regulate RNA stability and translation, were
   found to be enriched by PROPER (Figure [136]S4C, Supporting
   Information). Among the YTHDFs proteins, western blotting indicated
   that YTHDF2 was pulled down to a greater extent compared to YTHDF1 or
   YTHDF3 (Figure [137]3D), which was further validated by RIP‐qPCR
   (Figure [138]S4D, Supporting Information). An RNA pull‐down assay using
   purified His‐tagged recombinant YTHDF2 proteins was performed to verify
   the direct interaction between PROPER and YTHDF2 (Figure [139]S4E,
   Supporting Information). To identify the functional regions of YTHDF2
   responsible for its binding to PROPER, we generated a series of
   constructs expressing truncated forms of YTHDF2 (Figure [140]3E). It
   demonstrated that the Pro/Gln/Asn(P/Q/N)‐rich domain, but not the YTH
   domain, was critical for the RNA‐binding activity of YTHDF2 to PROPER
   (Figure [141]3F,G). RT‐qPCR of PROPER after m^6A methylated RIP (MeRIP)
   and YTHDF2 knockdown indicated that piRNA PROPER does not carry m^6A
   modifications (Figure [142]S4F,G, Supporting Information). Clinical
   correlation analyses of TCGA data showed that YTHDF2 was highly
   expressed in PCa tissues and positively associated with the tumor stage
   and prognosis (Figure [143]S4H–K, Supporting Information). FISH and IF
   assays showed colocalization of PROPER with YTHDF2 in PCa tissues
   (Figure [144]3H). Moreover, PROPER and YTHDF2 were found to be
   colocalized in the cytoplasm at the cellular level (Figure [145]3I),
   which was consistent with our subcellular localization analysis
   (Figure [146]1D,E).

Figure 3.

   Figure 3
   [147]Open in a new tab

   PROPER interacts with YTHDF2 to regulate mRNA translation. A) Silver
   staining identification of the PROPER‐protein complex pulled down by
   PROPER sense probe or anti‐sense probe with protein extracts from PC‐3
   and 22Rv1 cells using a biotin‐based pull‐down assay. B) Venn diagram
   showed potential interacting RNA‐binding proteins (RBPs) of PROPER by
   mass spectrometry and bioinformatic analysis. C) KEGG pathway
   enrichment analysis of PROPER conjunct 14 proteins. D) Immunoblots of
   YTHDF1/2/3 enriched by PROPER probes in PC‐3 and 22Rv1 cells. IgG were
   set as negative controls. E) Diagrams of full‐length and
   domain‑truncated fragments of YTHDF2; the P/Q/N‐rich domain and YTH
   domain are indicated. F) RIP‐qPCR analysis of PROPER immunoprecipitated
   using an anti‐FLAG antibody in PC‐3 cells expressing the indicated
   YTHDF2 plasmids. The data were normalized to the input levels. Data are
   presented as mean ± s.d. (n = 4); P values were calculated using
   unpaired two‐tailed Student's t‐tests. G) RNA pull‐down of PROPER probe
   with FLAG‐tagged YTHDF2 domain‐deletion mutants. H,I) YTHDF2
   immunofluorescence (green) and FISH analysis of PROPER (red) in PCa
   tissues (H) (Scale bars, 20 µm) and PC‐3 cells (I) (Scale bars, 10 µm).
   The nucleus was counterstained with DAPI (blue). J) Schematic of SUnSET
   assay that based on puromycin incorporation to monitor protein
   synthesis. K,L) SUnSET analyses of PC‐3 and 22Rv1 cells showing the
   changes in protein synthesis among the indicated samples. M) Schematic
   of polysome profiling that is based on sucrose‐gradient separation of
   translated mRNAs, which are associated with polysomes, from
   untranslated ones. N,O) Polysome profiles and polysome‐to‐monosome
   (P/M) ratio changes upon PROPER knockdown (N) and PROPER carries
   different alleles (O). The x axis indicates the free ribosome subunit
   (40S/60S), monosome (80S), and polysome fractions separated by 10%−45%
   sucrose gradients. The y axis indicates the absorbance at 260 nm. Data
   are presented as mean ± s.d. (n = 4); P values were calculated using
   unpaired two‐sided Mann‐Whitney U‐tests.

   YTHDF2 functions by inducing degradation and altering the translation
   efficiency of target mRNAs in the cytoplasm through recognition of m^6A
   modification sites.^[ [148]^17 ^] Therefore, we performed a surface
   sensing of translation (SUnSET) assay to detect puromycin‐labeled
   polypeptides, which serves as a measure of global protein synthesis
   (Figure [149]3J). The results showed that knockdown of YTHDF2 or PROPER
   led to increased protein synthesis, and this effect was significantly
   enhanced when the complex was simultaneously knock down
   (Figure [150]3K). Consistently, overexpression of PROPER, but not
   PROPERmut, revised this effect (Figure [151]3L). In line with these
   findings, depletion of PROPER resulted in a significant change in the
   polysome profile and an increased polysome‐to‐monosome (P/M) ratio
   (P = 0.0286) (Figure [152]3M,N), indicating that PROPER suppressed
   protein translation by regulating the initiation step (Figure [153]S4L,
   Supporting Information). Subsequently, we generated PC‐3 cells with
   different alleles of rs17201241 using CRISPER/Cas9 (Figure [154]S4M,
   Supporting Information), and PROPER containing the T allele exhibited a
   higher P/M ratio than that containing the C allele (P = 0.0286)
   (Figure [155]3O). Taken together, these results suggest that the
   PROPER/YTHDF2 complex may regulate the stability and translation of
   specific target RNAs at the post‐transcriptional level.

2.5. Analysis of the m^6A Methylome and PROPER/YTHDF2 Downstream Targets

   The screening flowchart for the target genes of PROPER/YTHDF2 complex
   is depicted in Figure [156]S5 (Supporting Information). Initially, we
   performed MeRIP‐Seq in PC‐3 cells with stable knockdown of PROPER or
   control using shRNA. The analysis of m^6A peaks through the MEME
   algorithm revealed a consensus m^6A motif RRACH, confirming successful
   immunoprecipitation (Figure [157] 4A). Further analysis showed that
   observed m^6A modifications were mainly present in protein‐coding mRNAs
   (Figure [158]4B; Figure [159]S6A, Supporting Information). Consist with
   previous reports,^[ [160]^31 ^] m^6A peaks were predominantly
   distributed in 3′‐UTRs, coding sequences (CDSs), and near stop codons
   (Figure [161]4C). A comparison of m^6A abundance between PROPER
   knockdown and control cells identified 1724 upregulated m^6A peaks and
   1726 downregulated m^6A peaks (Figure [162]4D; Figure [163]S6B and
   Table [164]S7, Supporting Information). These differential peaks were
   predominantly located in 3′‐UTRs (Figure [165]4E). However, there was
   no significant difference in global m^6A levels between PROPER
   knockdown and control cells, as detected by dot blot assays
   (Figure [166]S6C, Supporting Information).

Figure 4.

   Figure 4
   [167]Open in a new tab

   DUSP1 is the target of the PROPER/YTHDF2 complex in PCa cells. A) The
   m^6A motif detected by the MEME motif analysis with m^6A‐Seq data. B)
   Percentages of various RNA species with different m^6A modifications.
   C) Metagene profiles of m^6A enrichment across mRNA transcriptome
   composed of 5′‐UTR, CDS, and 3′‐UTR, in PC‐3 cells treated with PEOPER
   knockdown and control. D) Volcano plot showing the m^6A enrichment in
   mRNAs of PEOPER knockdown and control groups in PC‐3 cells.
   m^6A‐containing mRNAs with 1724 increased (up) and 1726 decreased peak
   (down) enrichment are highlighted in red and green, respectively.
   Distribution of m^6A peaks and YTHDF2‐binding peaks across transcripts.
   E) Percentages of m^6A peaks in different transcript segments in PROPER
   knockdown and control groups in PC‐3 cells. F) Correlation between the
   level of gene expression (overall transcript) and changes in m^6A
   levels in PROPER knockdown and control groups in PC‐3 cells. The
   upregulated peaks were termed as the hypermethylated m^6A peaks and the
   downregulated peaks were termed as the hypomethylated m^6A peaks. G)
   Gene ontology (GO) enrichment analysis of differentially expressed
   genes. H) Venn diagram illustrated 208 overlapping peaks of
   PROPER/YTHDF2 complex targets identified by m^6A‐Seq, RNA‐Seq, and
   RIP‐Seq (data from [168]GSE49339) analysis. I) MeRIP‐Seq displays the
   distribution of m^6A peaks and YTHDF2‐binding peaks in DUSP1 mRNA of
   PROPER knockdown and control groups in PC‐3 cells. J) The
   m^6A‐containing YTHDF2 binding motif at the 3′‐UTR of DUSP1 mRNA (red,
   m^6A motif). K) In‐slico RNA‐protein docking analysis displays the
   crystal structure of YTH domain docking with the conserved A(m^6A)C of
   DUSP1.

   Next, we analyzed the correlation between m^6A modification and
   corresponding gene expression. The m^6A peak data was plotted against
   the RNA‐Seq data (MeRIP‐Seq input library) of gene expression
   (Figure [169]4F). We identified a total of 432 genes with differential
   m^6A‐modification and corresponding expression change (Table [170]S8,
   Supporting Information). These genes were primarily associated with
   “Immune system process”, “Cytoplasm”, and “Protein binding” based on
   gene ontology (GO) analysis (Figure [171]4G). Combining this with the
   YTHDF2 RIP‐Seq data,^[ [172]^32 ^] we identified 208 common peaks
   corresponding to 175 genes (Figure [173]4H), of which 138 genes were
   upregulated and 37 genes were downregulated (Figure [174]S6D,
   Supporting Information). We next used photoactivatable
   ribonucleoside‐enhanced crosslinking and immunoprecipitation sequencing
   (PAR‐CLIP‐Seq) data to explore any direct interactions within the above
   peaks between YTHDF2 and its targeted transcripts with adenosine
   methylation (Figure [175]4; and Table [176]S9, Supporting Information).
   To further narrow down the target genes that may be regulated by the
   PROPER/YTHDF2, we knocked down PROPER or YTHDF2 and found that only the
   expression of DUSP1 showed a significant increase that consistent with
   the MeRIP‐Seq results and the YTHDF2 RIP‐Seq data (Table [177]S9,
   Supporting Information). We observed the presence of m^6A‐containing
   YTH binding motif in the 3′‐UTR of DUSP1, which was within a
   hypermethylated m^6A peak upon PROPER knockdown (Figure [178]4I,J).
   Alignment with CLIP‐Seq data from multiple other cells constantly
   indicated the abundance of methylated adenosine bases in the 3′‐UTR of
   DUSP1 mRNA (Figure [179]S6E, Supporting Information). Furthermore,
   In‐slico RNA‐protein docking analysis confirmed the docking capability
   of the crystal structure of YTH‐YTHDF2 with the conserved A(m^6A)C of
   DUSP1 (Figure [180]4K). These results strongly indicate that the
   m^6A‐modified DUSP1 is an intriguing and potential regulatory target
   gene of the PROPER/YTHDF2 complex.

2.6. The PROPER/YTHDF2 Complex Mediates DUSP1 Downregulation in an
m^6A‐Dependent Manner

   The m^6A modification is mainly catalyzed by the heterodimer of
   methyltransferase‐like 3 (METTL3) and methyltransferase‐like 14
   (METTL14).^[ [181]^33 ^] We observed a significant decrease in overall
   RNA m^6A levels upon knockdown of METTL3/METTL14, as detected by dot
   blot assays (Figure [182]S7A‐D, Supporting Information). Using
   MeRIP‐qPCR assay, we found that the enrichment of m^6A in DUSP1 was
   drastically reduced upon METTL3/METTL14 knockdown (P = 0.0286)
   (Figure [183]5A), which led to increased expression levels of DUSP1
   (Figure [184]5B,C; Figure [185]S7E,F, Supporting Information).
   Moreover, the interaction between YTHDF2 and DUSP1 was also remarkably
   impaired upon METTL3/14 knockdown (Figure [186]5D). Consistently,
   knockdown of YTHDF2 also resulted in a substantial increase in DUSP1
   level (Figure [187]5E; Figure [188]S7G, Supporting Information). Next,
   we expressed truncated versions of YTHDF2 in PCa cells and found that
   deleting either the YTH domain or the P/Q/N abolished the binding
   ability of YTHDF2 to DUSP1 (Figure [189]5F). Additionally, DUSP1
   protein levels were significantly downregulated in cells co‐expressing
   the YTHDF2 shRNA and the full‐length YTHDF2 overexpression plasmids
   (Figure [190]5G). These results suggest that the interaction between
   the PROPER/YTHDF2 complex and DUSP1 is regulated via an m^6A‐dependent
   manner.

Figure 5.

   Figure 5
   [191]Open in a new tab

   The PROPER/YTHDF2 complex mediates DUSP1 downregulation. A)
   Gene‐specific MeRIP‐qPCR validation of the m^6A levels at DUSP1 in PC‐3
   cells. Data are presented as mean ± s.d. (n = 4). The data were
   normalized to the input levels. P values were calculated using unpaired
   two‐sided Mann‐Whitney U‐tests. B,C) Knockdown of METTL3 and METTL14 on
   the expression of DUSP1. Quantification of blotting intensity for
   indicated proteins is shown below protein bands (the one in NC is set
   as 1.0 after normalization with GAPDH blotting). D) METTL3/14 knockdown
   impaired the interaction between YTHDF2 and DUSP1 by RIP‐qPCR. The data
   were normalized to the input levels. P values were calculated using
   unpaired two‐sided Mann‐Whitney U‐tests. E) The effects of YTHDF2
   knockdown on the expression of DUSP1. Quantification of blotting
   intensity for indicated proteins is shown below protein bands (the one
   in NC is set as 1.0 after normalization with GAPDH blotting). F)
   RIP‐qPCR analysis of DUSP1 immunoprecipitated using an anti‐FLAG
   antibody in PC‐3 cells expressing the indicated YTHDF2 plasmids. The
   data were normalized to the input levels. Data are presented as
   mean ± s.d. (n = 4). P values were calculated using unpaired two‐sided
   Mann‐Whitney U‐tests. G) DUSP1 protein levels in cells co‐expressing
   the YTHDF2 shRNA and the YTHDF2‐truncated plasmids. Quantification of
   blotting intensity for indicated proteins is shown below protein bands
   (the one in Vector is set as 1.0 after normalization with GAPDH
   blotting). H) RNA decay assays indicate the effects of PROPER and
   YTHDF2 knockdown on the stability of DUSP1 RNA. I) RIP‐qPCR analysis of
   DUSP1 and PROPER immunoprecipitated by YTHDF2. Data are presented as
   mean ± s.d. (n = 4). P values were calculated using unpaired two‐sided
   Mann‐Whitney U‐tests. J) Immunoblotting of DUSP1 protein expression
   upon knockdown of PROPER and/or YTHDF2. Quantification of blotting
   intensity for indicated proteins is shown below protein bands (the one
   in line 1 is set as 1.0 after normalization with GAPDH blotting). K)
   The effects of PIWIL4 knockdown coupled with PROPER (Mut) transfection
   on the expression of DUSP1 in PC‐3 cells. Data are presented as
   mean ± s.d. (n = 4). n.s., not significant; P values were calculated
   using unpaired two‐sided Mann‐Whitney U‐tests. L) Targeting sites
   prediction of PROPER within the 3′‐UTRs of DUSP1 using the RNAhybrid.
   M) Predicted PEOPER regulatory elements at the 3′‐UTRs of DUSP1 and the
   synthetic PROPER, mutated PROPER, and wild‐type and mutated 3′‐UTR
   Renilla luciferase reporters. N) Reporter gene assays evaluate the
   functional consequence of the predicted piRNA:mRNA interactions. The
   reporters were co‐transfected along with cognate chemically synthesized
   piRNAs (containing phosphate at the 5′ terminus and 2′‐O‐methylation at
   the 3′ terminus) or mutant versions. Data are presented as mean ± s.d.
   (n = 4). n.s., not significant; P values were calculated using unpaired
   two‐sided Mann‐Whitney U‐tests. O) RT‐qPCR analysis of reporter mRNAs.
   Data are presented as mean ± s.d. (n = 4); n.s., not significant; P
   values were calculated using unpaired two‐sided Mann‐Whitney U‐tests.

   Considering that MeRIP‐Seq assays demonstrated increased m^6A
   modification and expression of DUSP1 after PROPER knockdown, we
   hypothesized that the highly expressed PROPER in PCa might synergize
   with YTHDF2 and facilitate its recognition of m^6A‐modified DUSP1,
   leading to its degradation. We performed RNA decay assays and found
   that knockdown of PROPER or YTHDF2 reduced the stability of DUSP1
   (Figure [192]5H). Moreover, knockdown of PROPER followed by
   YTHDF2/PROPER complex isolation in a RIP assay using YTHDF2 antibody
   revealed a significant attenuation of the interaction between YTHDF2
   and DUSP1 (Figure [193]5I; Figure [194]S7H, Supporting Information),
   resulting in increased DUSP1 expression levels (Figure [195]S7I,J,
   Supporting Information). This effect was further enhanced when PROPER
   and YTHDF2 were simultaneously knocked down (Figure [196]5J). In
   contrast, ectopic expression of PROPER profoundly reduced endogenous
   DUSP1 levels (Figure [197]S7K, Supporting Information). Since we had
   previously demonstrated the binding of PROPER with PIWIL4 protein
   (Figure [198]1C; Figure [199]S1G, Supporting Information) and the
   effects of PIWIL4 knockdown in suppressing PROPER expression
   (Figure [200]S1H,I, Supporting Information), we found that PIWIL4
   knockdown coupled with PROPERmut transfection marginally affected the
   DUSP1 level, likely due to the absence of endogenous PROPER caused by
   PIWIL4 knockdown (Figure [201]5K).

   Next, we investigated the mechanism by which PROPER guides YTHDF2 to
   recognize m^6A modification at the 3′‐UTR of DUSP1, thereby leading to
   DUSP1 downregulation. A previous study reported the involvement of
   piRNAs in instructing massive mRNA elimination in spermatids, similar
   to the action of miRNA sponges in somatic cells.^[ [202]^34 ^]
   Therefore, we speculated that the function of PROPER might be related
   to sequences in the 3′‐UTR of DUSP1. Using the RNAhybrid, a tool for
   finding the minimum free energy (MFE) hybridization of a long and a
   short RNA,^[ [203]^35 ^] we identified potential targeting sites of
   PROPER within the 3′‐UTR of DUSP1 with an MFE of −35.0 kcal mol^−1
   (Figure [204]5L). We performed a reporter gene assay to evaluate the
   functional consequence of the predicted piRNA:mRNA interactions
   (Figure [205]5M). Co‐transfection of the reporters with cognate
   chemically synthesized piRNAs resulted in significant repression of
   DUSP1 reporters but not the mutant versions by co‐transfected PROPER
   (Figure [206]5N). RT‐qPCR analysis of the reporter mRNAs confirmed
   piRNA‐targeting dependent responses, indicating that piRNAs may induce
   mRNA degradation (Figure [207]5O). Taken together, these results
   suggest that PROPER not only couples with YTHDF2 to promote
   m^6A‐dependent deadenylation and decay of DUSP1 but also triggers these
   effects through imperfect base‐pairing with specific sites in the
   3′‐UTRs of DUSP1.

2.7. YBX3 and EIF2S3 are Critical Partners of the PROPER/YTHDF2‐Ediated DUSP1
Decay

   Among the identified interacting proteins of PROPER, we also found YBX3
   and EIF2S3 that post‐transcriptionally controls mRNA stability or
   translation (Figure [208]3B; Figure [209]S8A,B, and Table [210]S6,
   Supporting Information). The RNA pull‐down assay confirmed the
   interaction of YBX3 and EIF2S3 with PROPER (Figure [211]S8C, Supporting
   Information). In addition, immunofluorescence analysis in PCa tissues
   showed colocalization of PROPER with YBX3 and EIF2S3 (Figure [212]S8D,
   Supporting Information).

   A recent report established YBX3 as a translation repressor of
   m^6A‐modified transcripts.^[ [213]^36 ^] A previous study has reported
   that YBX3 primarily binds to mRNAs toward their 3′‐UTR.^[ [214]^37 ^]
   Available eCLIP‐Seq data of YBX3‐RNA complexes also showed the
   distribution of multiple YBX3‐binding peaks across the 3′‐UTR of DUSP1
   (Figure [215]S8E, Supporting Information). Notably,
   co‐Immunoprecipitation (Co‐IP) assays indicated that YBX3 could
   interact with YTHDF2 (Figure [216] 6A; Figure [217]S8F, Supporting
   Information), which was further confirmed by an in situ proximity
   ligation assay (PLA), demonstrating a specific protein interaction
   between YBX3 and YTHDF2 in the cell cytoplasm in a PROPER‐dependent
   manner (Figure [218]6B; Figure [219]S8G, Supporting Information).
   Therefore, we hypothesized that YBX3 might serve as a cofactor of
   YTHDF2 in recognizing the m^6A modification of DUSP1 and regulating
   translation efficiency. To confirm this, we conducted RIP assays with
   the YBX3 antibody and revealed that the recognition of YBX3 to DUSP1
   m^6A sites was dependent on PROPER (Figure [220]S8H,I, Supporting
   Information). Notably, polysome profiling followed by RT‐qPCR further
   demonstrated that more DUSP1 was present in the polysome‐bound
   fractions in YBX3 and YTHDF2‐depleted cells compared to YTHDF2 or
   YBX3‐depleted cells and the controls (Figure [221]6C;
   Figure [222]S8J,K, Supporting Information). These results suggest that
   YBX3 acted as a cofactor of YTHDF2 in suppressing translation
   efficiency.

Figure 6.

   Figure 6
   [223]Open in a new tab

   YBX3 and EIF2S3 are required for the PROPER/YTHDF2 complex mediated
   DUSP1 circularization and translational suppression. A) Lysates
   prepared from PC‐3 and 22Rv1 cells were hybridized with biotinylated
   PROPER probe and subject to RNA pull‐down assays. The remaining lysates
   were subject to western blotting with antibodies against YBX3 and
   EIF2S3. B) Proximity ligation assay (PLA) demonstrate the interaction
   between YBX3 and YTHDF2 in PC‐3 cells; (n = 4). C) Polysome profiles
   and polysome‐to‐monosome (P/M) ratio changes upon YBX3 knockdown,
   YTHDF2 knockdown, YBX3 and YTHDF2 knockdown together, and control PC‐3
   cells; (n = 4). D) Co‐IP assays of the interaction between Myc‐EIF2S3
   and Flag‐YTHDF2 or HA‐YBX3 in PC‐3 cells. E) PLA demonstrate the
   interaction between EIF2S3 and YTHDF2 in PC‐3 cells; (n = 4). F) PLA
   demonstrate the interaction between EIF2S3 and YBX3 in PC‐3 cells;
   (n = 4). G) Myc‐EIF2S3 Co‐IP with Flag‐YTHDF2 in PROPER knockout, and
   PROPER knockout with ectopically expressed PROPER/PROPERmut of PC‐3
   cells. H) Myc‐EIF2S3 Co‐IP with HA‐YBX3 in PROPER knockout, and PROPER
   knockout with ectopically expressed PROPER/PROPERmut of PC‐3 cells. I)
   Schematic diagram illustrating the mechanism that PROPER functions as a
   rivet to enhance EIF2S3‐YTHDF2/YBX3 interaction. J) EIF2S3 tethered at
   the 3′‐UTR of mRNA augments translation of an upstream reporter gene by
   MS2 tethering assay. Data are presented as mean ± s.d. (n = 3); n.s.,
   not significant; P values were calculated using unpaired two‐tailed
   Student's t‐tests. K) EIF2S3 mediated DUSP1 mRNA circularization and
   translation was inhibited by PORPER. Data are presented as mean ± s.d.
   (n = 3); n.s., not significant; P values were calculated using unpaired
   two‐tailed Student's t‐tests.

   EIF2S3, a subunit of the eukaryotic initiation factor 2 (eIF2), is
   critical in eukaryotic translation initiation involved in the
   rate‐limiting step in protein synthesis and translation deregulation,
   which is known to be important in cancer.^[ [224]^38 ^] Our Co‐IP assay
   confirmed the protein interaction between EIF2S3 and YTHDF2/YBX3
   (Figure [225]6D; Figure [226]S8L,M, Supporting Information).
   Furthermore, the PLA assay revealed the proximity of EIF2S3‐YTHDF2 and
   EIF2S3‐YBX3 in living cells, and their interactions were markedly
   weakened after PROPER knockdown (Figure [227]6E,F; Figure [228]S8N,O,
   Supporting Information).

2.8. PROPER Promotes the EIF2S3‐YTHDF2/YBX3 Complex‐Mediated DUSP1
Circularization and Translational Suppression

   Notably, PROPER depletion significantly decreased the
   EIF2S3‐YTHDF2/YBX3 interaction, while ectopic expression of PROPER, but
   not PROPERmut, rescued the effect (Figure [229]6G,H), suggesting that
   these interactions depend on PROPER. Therefore, we propose that PROPER
   is a crucial member of the EIF2S3‐YTHDF2/YBX3 complex and functions as
   a “rivet” binding to EIF2S3 and enhancing their interaction
   (Figure [230]6I). Next, we performed MS2 tethering assays to evaluate
   the effect of PROPER on mRNA circularization. A dual‐luciferase vector
   was constructed, which included a Firefly reporter and a Renilla
   luciferase controlling for different transfection efficiencies
   (Figure [231]S8P,Q, Supporting Information). Consequently, the
   translation efficiency decreased when EIF2S3 was tethered to the 5′‐UTR
   of reporter mRNA (Figure [232]6J,K), indicating that PROPER promoted
   DUSP1 circularization and translational suppression. Collectively,
   these results indicate that the role of PROPER in translation
   suppression is to serve as a component of the translation initiation
   complex, promoting the EIF2S3‐YTHDF2/YBX3 complex‐mediated DUSP1
   circularization and translational suppression.

2.9. PROPER Regulates DUSP1 Expression that Contributes to PCa Progression

   To explore whether DUSP1 is associated with human PCa progression, we
   performed clinical correlation analyses in multiple independent
   clinical PCa datasets. The results indicated that DUSP1 expression
   levels were greatly downregulated in primary and metastatic prostate
   tumors compared to the normal prostate gland (Figure [233] 7A,B;
   Figure [234]S9A–D, Supporting Information). In addition, the
   downregulation of DUSP1 significantly correlated with a higher Gleason
   score, pre‐treatment prostate‐specific antigen (PSA) levels, and tumor
   progression in patients (Figure [235]7C,D; Figure [236]S9E, Supporting
   Information). Moreover, patients with lower mRNA levels of DUSP1 has an
   increased risk of postoperative biochemical recurrence and a shorter
   overall survival time (Figure [237]7E,F). We also observed that the
   expression of PROPER was negatively correlated with that of DUSP1
   (Figure [238]7G). Mechanistically, the role of DUSP1 in PCa progression
   and its impact on patient outcomes have been widely reported by
   negatively regulating p38 MAPK signaling pathway.^[ [239]^21 , [240]^39
   , [241]^40 ^] The bone metastasis model of PCa showed that knockdown of
   PROPER expression or elevation of DUSP1 expression significantly
   inhibited tumor cell‐induced bone erosion of proximal tibiae compared
   to the control group (Figure [242]7H; Figure [243]S9F, Supporting
   Information). However, DUSP1 knockdown significantly interfered with
   the inhibitory effect of PROPER knockdown on PCa metastasis
   (Figure [244]7H; Figure [245]S9F, Supporting Information). The
   knockdown of PROPER led to a decrease in the phosphorylation of p38 and
   ERK (Figure [246]7I).

Figure 7.

   Figure 7
   [247]Open in a new tab

   PROPER regulates DUSP1 expression that contributes to PCa progression.
   A,B) DUSP1 downregulated in primary PCa tissues (A) and metastasis PCa
   (B). P values were calculated using unpaired two‐tailed Student's
   t‐tests. C,D) DUSP1 downregulated in patients with high Gleason score
   (C) and high pre‐treatment PSA (D); GS, Gleason score. P values were
   calculated using one‐way ANOVA. E,F) Lower levels of DUSP1 correlate
   with increased risk of biochemical recurrence in two cohorts of PCa
   patients. P values were assessed by a log rank test. G) Spearman's rank
   correlation coefficient analysis between PROPER and DUSP1 expression in
   tumors from TCGA database. H) Representative Micro‐CT images of the
   hindlimbs of mice 9 weeks after inoculation with the PC‐3 cells. I)
   Western blotting analysis of the regulations of PROPER on p38 MAPK
   pathway. J) Timeline schematic for treatment of mouse tumors with
   antagoPROPER. Colored arrows indicate the times when different events
   occurred. K) Inhibitory effects of intra‐tumor administration of
   antagoPROPER on xenograft growth. Shown are pictures of tumors with
   (upper) or without (bottom) injection of antagoPROPER (n = 5).

   To test whether PROPER might be a therapeutic target for PCa treatment,
   we conducted experimental therapy in mouse subcutaneous xenograft
   models using antagoPROPER. We injected antagoPROPER or a scramble
   sequence directly into subcutaneous tumors (Figure [248]7J). After 5–8
   injections, the subcutaneous tumors treated with antagoPROPER showed
   significantly reduced tumor volume and weight compared to those treated
   with scramble controls (Figure [249]7K; Figure [250]S9G–I, Supporting
   Information). These results imply that targeted inhibition of PROPER
   expression may inhibit PCa progression.

3. Discussion

   In the present study, we have identified a PCa risk‐associated SNP,
   rs17201241, located within an oncogenic piRNA called PROPER, which is
   associated with malignant phenotypes and aggressiveness of PCa.
   Mechanistically, PROPER interacts with the m^6A reading proteins YTHDF2
   and YBX3 to form pi‐RISC. This pi‐RISC promotes the degradation and
   translational suppression of DUSP1 with the translation initiation
   factor EIF2S3 in an mRNA‐looping manner. These findings reveal a
   PROPER‐YTHDF2/YBX3/EIF2S3‐DUSP1 signaling axis and uncovered a novel
   post‐transcriptional epigenetic mechanism of piRNA that guides certain
   mRNA degradation and translation control, based on an m^6A
   methylation‐regulated mRNA‐looping.

   To date, GWAS‐identified risk‐associated SNPs mainly map to noncoding
   regions of the genome.^[ [251]^41 ^] However, few studies have focused
   on genetic variants of piRNAs in somatic tissues. Our previous work
   revealed that rs11776042 in a lncRNA‐derived piRNA, piR‐015551, was
   significantly associated with a decreased risk of colorectal cancer.^[
   [252]^42 ^] The present study has revealed for the first time that the
   SNP rs17201241 resides in a snoRNA‐derived piRNA cluster associated
   with PCa risk. Numerous functional piRNAs have been reported to be
   derived from snoRNAs.^[ [253]^27 , [254]^43 ^] By combining piRNA
   features and sequence alignment analysis of small RNA sequencing
   data,^[ [255]^44 ^] we successfully identified a novel piRNA called
   PROPER. Understanding the functional implications of PCa
   risk‐associated SNPs provides new opportunities to elucidate the
   mechanisms underlying PCa tumorigenesis.^[ [256]^45 ^] Through eQTL
   analysis, RNA‐protein docking analysis, and experimental studies, we
   confirmed that rs17201241 could affect the maturation of SNORD48,
   leading to high expression of PROPER in PCa tissues. Our previous work
   has systematically summarized the expression patterns of piRNAs in
   somatic tissues and their associations with clinical outcomes.^[
   [257]^43 ^] In this study, we demonstrated that PROPER is an oncogenic
   piRNA in association with aggressive PCa.

   Biologically, piRNAs were initially identified for their well‐known
   functions in repressing transposable elements in germline cells.^[
   [258]^46 ^] Liu and her colleagues also discovered that MIWI/piRNA has
   a dual function, instructing both massive mRNA elimination and
   activating mRNA translation during spermiogenesis.^[ [259]^34 ,
   [260]^47 ^] Intriguingly, our results uncovered three piRNA‐binding
   proteins, YTHDF2, YBX3, and EIF2S3, that are also significantly linked
   to translational regulation.^[ [261]^36 , [262]^48 ^] Dysregulation of
   the canonical translation machinery and abnormal activation of
   translation signaling pathways play critical roles in enhancing rapid
   translation of oncogenes or inhibiting tumor suppressor genes.^[
   [263]^49 ^] A closed‐loop model mediated by piRNA and multiple RBPs is
   proposed to regulate multiple rounds of mRNA translation. This is
   supported by functional and physical circularization between the capped
   5′ terminus and the polyadenylated 3′ terminus of mRNA mediated by the
   eIF4F‐PABPC1 complex,^[ [264]^50 ^] which maybe the critical hub for
   cancer development.^[ [265]^51 , [266]^52 , [267]^53 ^] Sun et al.
   reported a micropeptide, APPLE, in acute myeloid leukemia (AML).^[
   [268]^54 ^] APPLE enhances the PABPC1‐eIF4G interaction, promoting mRNA
   circularization and the assembly of the eIF4F initiation complex,
   thereby supporting a specific pro‐oncogenic translation program. Wei
   et al. reported a novel methylation‐independent function of METTL3 in
   the cytoplasm, where METTL3 promotes gastric cancer by upregulating
   mRNA translation of specific transcripts.^[ [269]^55 ^] METTL3
   interacts with PABPC1, stabilizing its association with the cap‐binding
   complex eIF4F, preferentially enhancing the translation of epigenetic
   factors without m^6A modifications. However, unlike typical
   eIF4G–PABPC1‐mediated mRNA looping, we presented the first evidence of
   this EIF2S3‐YTHDF2/YBX3 loop, which presumably suppresses translation
   through the piRNA‐guided m^6A‐modified mRNA degradation mediated by
   different translation initiation factors and patterns of epigenetic
   modification.

4. Conclusion

   Altogether, this study revealed a novel PCa risk‐associated genetic
   variant, rs17201241, located within the piRNA PROPER. These findings
   support the notion that PROPER exerts an oncogenic role in PCa by
   acting as a translation co‐regulator in the EIF2S3‐YTHDF2/YBX3 loop
   axis. The presence of this piRNA as a driver in cellular malignant
   transformation highlights the similarities between tumorigenesis and
   gametogenesis, providing valuable insights into our understanding of
   tumorigenesis and presenting new opportunities for targeting the
   translation mechanisms in cancer cells.

5. Experimental Section

Study Subjects

   Three GWASs were included in a meta‐analysis, consisting of East Asian
   (i.e., Chinese and Japanese) and European populations.^[ [270]^56 ,
   [271]^57 , [272]^58 ^] Briefly, the Chinese GWAS including 1151 cases
   and 1202 controls was part of the Consortium for Prostate Cancer
   Genetics (ChinaPCa);^[ [273]^8 ^] the cases were recruited from
   hospital‐based and pathologically diagnosed PCa patients and the
   cancer‐free controls from the communities or physical examination
   centers. The Japanese GWAS had 5408 cases and 103939 controls with data
   and samples available from the BioBank Japan Project (BBJ) that
   established in the Institute of Medical Science at the University of
   Tokyo.^[ [274]^59 ^] From the BBJ, pathologically proven PCa cases were
   selected. Control samples from BBJ were from four population‐based
   prospective cohorts, including the Tohoku University Tohoku Medical
   Megabank Organization (ToMMo),^[ [275]^60 ^] Iwate Medical University
   Iwate Tohoku Medical Megabank Organization (IMM), the Japan Public
   Health Center‐based Prospective Study,^[ [276]^61 ^] and the Japan
   Multi‐institutional Collaborative Cohort Study.^[ [277]^62 ^] The
   European GWAS from the Prostate Cancer Association Group to Investigate
   Cancer Associated Alterations in the Genome (PRACTICAL) is an
   international prostate cancer consortium that includes 79148 cases and
   61106 controls ([278]http://practical.ccge.medschl.cam.ac.uk/).^[
   [279]^58 ^]

   All participating studies obtained an informed consent from all
   participants by following the protocols approved by their institutional
   ethical committees before enrollment, and the ethical committee at each
   institute approved the project.^[ [280]^56 , [281]^57 , [282]^58 ^]

Clinical Specimens

   An additional 31 pairs of PCa and adjacent normal tissues were used to
   analyze RNA expression by RT‐qPCR, and another 90 pairs of PCa and
   adjacent normal tissues were analyzed using a high‐throughput tissue
   microarray analysis. The paired normal tissue and blood samples from
   these 90 patients were collected, and their RNA and DNA were extracted
   for the expression quantitative trait loci (eQTL) analysis. All the
   tissues were originally collected from patients who underwent surgical
   treatment and stored in liquid nitrogen until use. Serum was isolated
   from blood samples and stored at −80 °C. The present study was approved
   by the Nanjing Medical University Ethics Committee (Approval Number:
   IACUC‐2009008), and an informed consent was obtained from all patients
   and volunteers.

Polysome Profiling

   Cells were treated with 100 µg mL^−1 cycloheximide (CHX) (ACMEC) at
   37 °C for 15 min to inhibits protein synthesis by blocking translation
   elongation. Subsequently, the cells were washed twice with ice‐cold
   phosphate‐buffered saline (PBS) containing CHX and collected by cell
   scraping. Then cells were lysed in lysis buffer (20 mm Tris‐HCl, pH
   7.5, 5 mm MgCl[2], 100 mm KCl, 1% Triton X‐100) supplemented with
   50 U mL^−1 Ribonuclease Inhibitor (Promega), 100× Protease Inhibitor
   Cocktail EDTA Free (Abcam), 100 µg mL^−1 CHX. Cell lysates were
   centrifuged at 15,000 rpm at 4 °C for 10 min and the supernatant was
   loaded onto a linear 10% to 45% (wt/vol) sucrose gradients followed by
   ultracentrifugation with a SW40Ti rotor (Beckman) at 35,000 rpm for
   3 h. Absorbance at 260 nm was recorded using a BioComp Piston Gradient
   Fractionator equipped with a Bio‐Rad Econo UV Monitor. Polysome to
   monosome (P/M) ratios were calculated by comparing the areas under the
   monosome and polysome peaks. The extracted RNA and protein samples were
   then used for RT‐qPCR and western blotting analysis, respectively.
   Relative distribution of mRNA in each fraction was normalized by the
   total abundance of mRNA in all fractions marked as 100%.

Surface Sensing of Translation (SUnSET) Assay

   SUnSET assays were performed according to previously reported methods
   to monitor protein synthesis.^[ [283]^63 ^] Briefly, cells were
   incubated with 10 ug mL^−1 puromycin (InvivoGen) for 30 min at 37 °C,
   followed by washing with cold‐PBS and scraped into RIPA lysis buffer
   (Beyotime). Then, puromycin‐containing proteins were analyzed by
   western blotting using the puromycin antibody (Sigma‐Aldrich).

Small RNA Library Construction, Sequencing, and Data Analysis

   Total RNA was extracted from PC‐3 and 22Rv1 cells using TRIzol reagent
   (Invitrogen). The concentration and quality of RNA were assessed by a
   NanoDrop ND‐1000 (NanoDrop) and an Bioanalyzer 2100 (Agilent).
   Following passing the quality control tests, total RNAs were used for
   library preparation and sequencing. Briefly, small RNAs of 18–40 nt in
   length were purified from total RNA by size fractionation using 15%
   PAGE and sequentially ligated to 5′ and 3′ adaptors, followed by RT‐PCR
   amplification to produce sequencing libraries using the TruSeq Small
   RNA Sample Prep Kit (Illumina) according to the manufacturer's
   protocol. PCR products were gel‐purified and sequenced using Illumina
   HiSeq 2000 (Illumina). Clean reads were obtained by removing
   low‐quality reads and adaptor sequences from raw reads. Subsequently,
   the length distribution of the clean reads and common and specific
   sequences between these two samples were summarized. Small RNA reads
   were aligned with piRBase ([284]http://bigdata.ibp.ac.cn/piRBase/) to
   screen and annotate piRNAs using Bowtie.^[ [285]^64 ^]

m^6A RNA Immunoprecipitation (MeRIP)‐Seq and RNA‐Seq

   Total RNA from PC‐3 cells treated with PROPER knockdown and their
   negative controls was isolated using TRIzol reagent (Invitrogen), and
   the samples were quantified using NanoDrop ND‐1000 (NanoDrop). The RNA
   integrity was assessed by Bioanalyzer 2100 (Agilent) with RIN number
   >7.0, and confirmed by electrophoresis with denaturing agarose gel.
   Approximately more than 25 µg of total RNA was used to deplete
   ribosomal RNA according to the protocol of the Epicentre Ribo‐Zero Gold
   Kit (Illumina). Following purification, the ribosomal‐depleted RNA was
   fragmented into small pieces using Magnesium RNA Fragmentation Module
   (NEB) under 86 °C for 7 min. Then the cleaved RNA fragments were
   incubated for 2 h at 4 °C with m^6A‐specific antibody (Synaptic
   Systems) in IP buffer (50 mm Tris‐HCl, 750 mm NaCl and 0.5% Igepal
   CA‐630). Then the IP RNA was reverse‐transcribed to create the cDNA by
   SuperScript II Reverse Transcriptase (Invitrogen), which was next used
   to synthesise U‐labeled second‐stranded DNAs with E. coli DNA
   polymerase I (NEB), RNase H (NEB), and dUTP Solution (Thermo Fisher
   Scientific). The libraries were subjected to denaturation to obtain
   single‐stranded DNA molecules and captured on Illumina flow cells. An
   A‐base was then added to the blunt ends of each strand, preparing them
   for ligation to the indexed adapters. Each adapter contains a T‐base
   overhang for ligating the adapter to the A‐tailed fragmented DNA.
   Single‐ or dual‐index adapters were ligated to the fragments, and size
   selection was performed with AMPureXP beads. After the heat‐labile UDG
   enzyme (NEB) treatment of the U‐labeled second‐stranded DNAs, the
   ligated products were amplified with PCR. The average insert size for
   the final cDNA library was 300 ± 50 bp. At last, the 2 × 150 bp
   paired‐end sequencing (PE150) was performed on an illumina Novaseq 6000
   (LC‐Bio Technology) following the vendor's recommended protocol.

   The fastp software ([286]https://github.com/OpenGene/fastp) was used to
   remove the reads that contained adaptor contamination, low quality
   bases and undetermined bases with default parameter. Then sequence
   quality of IP and Input samples were also verified using fastp. HISAT2
   ([287]http://daehwankimlab.github.io/hisat2) was used to map reads to
   the reference genomeHomo sapiens (Version: v101). Mapped reads of IP
   and input libraries were provided for R package exomePeak
   ([288]https://bioconductor.org/packages/exomePeak), which identifies
   m^6A peaks with bed or bigwig format that could be adapted for
   visualization on the IGV software ([289]http://www.igv.org). MEME
   ([290]http://meme‐suite.org) and HOMER
   ([291]http://homer.ucsd.edu/homer/motif) were used for de novo and
   known motif finding followed by localization of the motif with respect
   to peak summit. Called peaks were annotated by intersection with gene
   architecture using R package ChIPseeker
   ([292]https://bioconductor.org/packages/ChIPseeker). Then StringTie
   ([293]https://ccb.jhu.edu/software/stringtie) was used to perform
   expression level for all mRNAs from input libraries by calculating FPKM
   (total exon fragments /mapped reads (millions) × exon length (kB)). The
   differentially expressed mRNAs were selected with log2 (fold change) >1
   or log2 (fold change) < ‐1 and P value < 0.05 by R package edgeR
   ([294]https://bioconductor.org/packages/edgeR).

CRISPR/Cas9‐Mediated Knockout

   For SNORD48 KO, the single guide RNAs (sgRNAs) were designed using the
   online CRISPR design tool (Red Cotton, Guangzhou, China,
   [295]https://en.rc‐crispr.com/). The pair of oligos for two targeting
   sites was annealed and ligated to the YKO‐RP006 vector (Ubigene). The
   YKO‐RP006‐hRABL6 [gRNA] plasmids containing each target sgRNA sequences
   were transfected into cells with Lipofectamine 3000 (Thermo Fisher
   Scientific). After 24–48 h, 1 µg mL^−1 puromycin (InvivoGen) was added
   to screen the cells; surviving cells were sorted into 96‐well plates
   and then expanded into 12‐well plates. Selection of single clones was
   performed after 2–4 weeks, and the selected SNORD48 KO clones were
   validated by PCR and Sanger sequencing. The sgRNAs and primers for
   CRISPR design are shown in Table [296]S10 (Supporting Information).

Immunoprecipitation (IP)

   Cell extraction was performed using IP Lysis Buffer (50 mm Tris‐HCl pH
   7.4, 1 mm EDTA, 50 mm NaCl, 1% Triton X‐100) supplemented with Halt
   Protease Inhibitor Cocktail (Thermo Fisher Scientific), and then
   incubated with primary antibody overnight at 4 °C. Next day, the
   antibody‐bound protein of interest in lysis buffer was incubated with
   40 µL of Protein A/G Beads (Thermo Fisher Scientific). After three
   washes with Wash Buffer (0.5 m Tris‐HCl pH 7.4, 1.5 m  NaCl),
   protein‐bound beads were mixed with 5 × loading buffer (Thermo Fisher
   Scientific) to the final concentration of 1 × loading buffer and boiled
   for 10 min at 95 °C. The samples were then stored at 20 °C or ready for
   sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE).
   To explore whether protein interaction was mediated by RNA, RNase A
   (Takara) was added into the cell lysate to the final concentration of
   20 mg mL^−1 and incubated for 30 min at 37 °C before IP procedures.

Proximity Ligation Assay (PLA) Assay

   PC‐3 cells were transfected with plasmids of FLAG‐YTHDF2, HA‐YBX3, and
   MYC‐EIF2S3. After 48 h, cells were incubated with primary antibodies
   (two antibodies raised in different species) in blocking solution at
   4 °C for 2 h. Cells were then washed in PBS plus 0.1% Tween 20 for
   2 × 5 min. Then, cells were incubated with PLA probes (anti‐Rabbit‐PLUS
   and anti‐Mouse‐MINUS) (Sigma, DUO92101) for 60 min at 37 °C. Cells were
   washed in 1× Wash Buffer A for 2 × 5 min at 37 °C. Then, dilute the
   Ligation stock and the Ligase according to the manufacturer's
   instruction. The mix was applied to cells and incubated for 30 min at
   37 °C. Cells were washed in 1x Wash Buffer A for 2 × 2 min. Then,
   dilute the amplification stock and the polymerase according to the
   manufacturer's instruction. Then, the mix was applied to to cells and
   incubated for 100 min at 37 °C. Cells were washed in 1x Wash Buffer A
   for 2×10 min and wash in 0.01× Wash Buffer B for 1 min. Mount the cells
   with Duolink In Situ Mounting Medium with DAPI, and cells were observed
   in a confocal microscope (Carl Zeiss LSM 710).

MS2 Tethering Assay

   The MS2 tethering assay was performed as previously described.^[
   [297]^65 ^] An artificial polypeptide eIF2S3 fused with MS2 was
   localized downstream of the reporter gene to generate
   pMS2‐pEGFP‐C1‐EIF2S3. Firefly luciferase (FLuc) represents a reporter
   RNA containing the firefly luciferase gene as a reporter. MS2‐binding
   sites (6, 12, or 24 copies) were inserted into the reporter RNA to
   generate Fluc‐6×MS2bs, Fluc‐12×MS2bs, and FLuc‐24×MS2bs, respectively.
   Translation of the reporter was greatly enhanced by the EIF2S3‐MS2
   fusion protein. Moreover, EIF2S3‐MS2 tethered at the 3′‐UTR enhanced
   translation. Dual luciferase assays were performed according to the
   manufacturer's instructions (Promega). The FLuc activity was normalized
   to Renilla luciferase activity to calculate the different transfection
   efficiencies.

Animal Models and In Vivo Imaging

   Animal experiments were carried out in compliance with approved
   protocols and guidelines. To examine the effects of PROPER on
   subcutaneous xenograft growth, male BALB/c nude mice (5 mice per group)
   were subcutaneously injected with 0.1 mL of cell suspension containing
   2 × 10^6 cells in the rear flank. When a tumor was palpable, it was
   measured every other day and its volume was calculated according to the
   formula volume = length × width^2 × 0.5.

   Subcutaneous xenografts were also treated with antagoPROPER, a
   synthesized PROPER inhibitor (Ribobio; Table [298]S10, Supporting
   Information) by intra‐tumor injection. Briefly, when tumors implanted
   in mouse flanks reached a certain volume, 50 µg of AntagoPROPER in
   50 µL phosphate buffered saline was directly injected into the tumor.
   The tumor was injected with AntagoPROPER NC (Ribobio; Table [299]S10,
   Supporting Information) as control. Treatment was delivered every three
   days for 22 days and tumor volume was measured before each injection.
   Mice were sacrificed at the end of the experiment and the xenografts
   were stripped and photographed. H&E staining was then performed for
   histological examination of prostate tumor.

   To generate a metastasis model, PC‐3 stable cells (2  × 10^6) which
   containing control vector or PROPER shRNA‐expressing vector were
   intratibial injected into the tibia of BALB/c nude mice (4‐week‐old).
   The mice were humanely sacrificed after 15 weeks. The tumor mass was
   removed from the tibia bone to detect its weight, and the body weight
   and survival time of animals in each group were recorded.

   Animal studies were approved by the Institutional Animal Care and Use
   Committee of Nanjing Medical University.

Statistical Analysis

   Multivariable logistic regression analyses were used to estimate odds
   ratios (ORs) and the corresponding 95% confidence intervals (CIs) for
   the associations of candidate SNPs with PCa risk after adjustment for
   covariates of age and the top principal components. The meta‐analysis
   of three GWAS dataset was performed using a fixed‐effect model. The
   heterogeneity of studies was tested by the Cochran's Q statistic and I
   ^2. The correlations between SNP genotypes and gene expression levels
   were evaluated using a linear regression model.

   All quantitative experiments were repeated using at least four
   independent biological repeats and were presented as mean ± s.d.
   Statistical comparisons between two groups were accessed by two‐tailed
   Student's t‐test and two‐sided Mann‐Whitney U‐test. Details of the
   statistical analysis for each experiment can be found in the relevant
   figure legends. Kaplan‐Meier survival data were analyzed using the
   log‐rank test. All image quantifications for migration and invasion
   experiments were performed using ImageJ2 (version 10.2). All
   statistical analyses were performed using RStudio (version 1.0.136)
   with R (version 3.3.3) and GraphPad Prism 6 software.

Ethics Approval

   The present study was approved by the Nanjing Medical University Ethics
   Committee (Approval Number: IACUC‐2009008), and an informed consent was
   obtained from all patients and volunteers.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   S.B., Z.T.D., and J.Y.X. contributed equally to this work. M.L.W. and
   G.C. conceived the study. S.B., Z.T.D., and J.Y.X. performed
   computational work. Z.T.D., L.L.F., and S.L.C. performed experimental
   work. G.C. and Y.F.C. contributed to sample collection. S.L.C. managed
   sample recruitment. F.L., M.L.D., Z.D.Z., and G.H.W. contributed to
   critical reading and revision of the manuscript. All authors
   participated in interpretation and writing the manuscript.

Supporting information

   Supporting Information
   [300]ADVS-11-2402954-s001.docx^ (10.1MB, docx)

Acknowledgements