A ESRP1/circPHGDH/miR-149/RAP1B positive feedback loop promotes the malignant behaviors and glycolysis of prostate cancer cell

TL;DR

A new feedback loop involving circPHGDH, miR-149, and RAP1B promotes prostate cancer growth and spread, offering potential new treatment targets.

Contribution

Discovery of a novel ESRP1/circPHGDH/miR-149/RAP1B feedback loop driving prostate cancer progression through glycolysis and lactylation.

Findings

circPHGDH is upregulated in prostate cancer tissues and promotes cell proliferation, migration, and glycolysis.

circPHGDH regulates miR-149, which targets RAP1B, forming a feedback loop with ESRP1 lactylation.

Inhibiting circPHGDH suppresses tumor growth and metastasis in animal models.

Abstract

Circular RNAs are implicated in the pathogenesis of prostate cancer (PCa). However, their functions, biogenesis and molecular mechanisms remain largely elusive. Here we aimed to investigate the role of circPHGDH in PCa. Cellular behaviors were assessed by the colony formation assay, Transwell analysis, western blotting and Seahorse assay. The underlying mechanisms were investigated using a luciferase reporter assay, RNA pull-down and real-time quantitative PCR. The lactylation of ESRP1 was examined by RNA immunoprecipitation, immunoprecipitation and western blotting. Our results revealed that circPHGDH expression was upregulated in PCa tissues and cells. Furthermore, the knockdown of circPHGDH inhibited PCa cell proliferation, migration, invasion, epithelial–mesenchymal transition and glycolysis. Mechanistically, circPHGDH functioned as a sponge for miR-149, which in turn directly targeted RAP1B. The biogenesis of circPHGDH was regulated by the splicing factor ESRP1. The glycolytic product lactate stabilized ESRP1 by promoting its lactylation at the K43 site; conversely, circPHGDH knockdown suppressed ESRP1 lactylation. Moreover, the silencing of circPHGDH inhibited tumor growth and metastasis in vivo via the miR-149/RAP1B axis, whereas circPHGDH facilitated tumor progression. In conclusion, the lactylation-modified ESRP1/circPHGDH/miR-149/RAP1B axis drives the progression of PCa. These findings provide novel insights into the pathogenesis of PCa and suggest promising therapeutic targets for its treatment.

Prostate cancer is a common disease affecting older men, but younger men are increasingly being diagnosed. Early detection is difficult because the disease often shows no symptoms until it is advanced. Current treatments such as surgery and chemotherapy help, but challenges remain owing to the cancer’s complexity and resistance to treatment. Researchers explored a molecule called circPHGDH in prostate cancer. They found that circPHGDH is more common in cancerous tissues and helps cancer cells grow and spread. The study involved analyzing tissue samples from 51 patients and conducting various lab experiments to understand how circPHGDH works. They discovered that circPHGDH interacts with other molecules such as miR-149 and RAP1B, which are involved in cancer progression. By targeting this interaction, they hope to find new ways to treat prostate cancer.

This summary was initially drafted using artificial intelligence, then revised and fact-checked by the author.

Figures

a, miR-149 expression was elevated after mimic transfection and was downregulated after inhibitor transfection, confirming their efficacy. Notably, the downregulation of miR-149 by an inhibitor counteracted the inhibition of proliferation, migration and invasion of PCa cells induced by the circPHGDH knockdown (Supplementary Fig. 1b–e). Similarly, the effects of knockdown of circPHGDH, which upregulated E-cadherin levels and downregulated N-cadherin levels, were rescued by the miR-149 inhibitor (Supplementary Fig. 1f). Moreover, the inhibition of miR-149 abrogated the reduction of ECAR and the

a). Crucially, the enforced expression of RAP1B counteracted the suppression of cellular proliferation, migration and invasion induced by miR-149 overexpression (Supplementary Fig. 2b–e). Moreover, the effect of miR-149, which increased E-cadherin levels and downregulated N-cadherin levels, was partly abolished by RAP1B overexpression (Supplementary Fig. 2f). Moreover, the decrease in ECAR and the increase in OCR caused by miR-149 were reversed by RAP1B overexpression (Supplementary Fig. 2g, h). In addition, miR-149 expression and RAP1B expression was negatively correlated in 22Rv1 and VCaP ce

CircPHGDH is a miR-149 sponge in PCa cells.a Targeted miRNAs were chosen by a luciferase reporter analysis. b The predicted complementary sequences of circPHGDH and miR-149. c, d A dual-luciferase reporter assay (c) and RNA pull-down analysis (d) were determined for targeting relationship confirmation. e A qPCR was applied to test miR-149 expression in PCa cells after sh-circPHGDH or oe-circPHGDH transfection. f The MiR-149 expression in PCa tissues and para-PCa tissues was examined by qPCR (*n* = 51). g The correlation of miR-149 and circPHGDH expression in PCa tissues was analyzed using Pear

a). Then, these cells were injected into mice to establish a xenograft tumor model. The mean fluorescence intensity was higher in the lv-circPHGDH group than that in the lv-nc group (Supplementary Fig. 4b, c), indicating that circPHGDH accelerates tumor metastasis. Moreover, we isolated xenograft tumors and found that circPHGDH increased tumor volume and weight (Supplementary Fig. 4d–f). The IHC results showed that the overexpression of circPHGDH induced the upregulation of ki67 expression, whereas the hematoxylin and eosin (H&E) assay results showed that circPHGDH promoted the deterioration o

ESRP1 promotes circPHGDH biogenesis.a The effect of splicing factors on ESRP1 expression was evaluated qPCR. b The Pearson correlation coefficient was used to analyze the correlation between circPHGDH and ESRP1 expression in PCa tissues (*n* = 51). c, d RIP (c) and (d) in vitro RNA pull-down assays were conducted to determine the interaction between the ESRP1 protein and circPHGDH. e ESRP1 motifs were found in the flanking of circPHGDH. f The binding relationship between ESRP1 and circPHGDH was tested using RIP assay. g A qPCR was utilized to detect circPHGDH expression after silencing of ESRP

Knockdown of circPHGDH inhibits lactylation of ESRP1 at K43 site in PCa cells.a PCa cells were treated with LA or transfected with sh-circPHGDH, and the protein and lactylation levels of ESRP1 were measured. b The protein stability of ESRP1 was detected using western blotting after CHX treatment for 0, 6, 12 and 24 h. c The quantification results of ESRP1 protein stability. d ESRP1-WT vectors or ESRP1 mutated plasmids at the lactylation sites (K43R, K252R, K507R and K511R) were transfected into the PCa cells, and the effect of LA on ESRP1 lactylation levels was detected using IP and western bl

ESRP1 promotes the malignant behaviors of PCa cells by increasing circPHGDH expression.a–g After the ESRP1 overexpression and circPHGDH knockdown, the cell proliferation (a) was estimated by colony formation assay; Transwell assay was performed to evaluate migration (b) and invasion (d); the quantification results of cell migration are shown in (c); E-cadherin and N-cadherin levels (e) were examined by western blotting; and ECAR (f) and OCR (g) were detected using Seahorse assay. *n* = 3. *P* < 0.01 versus the vector group; ^#^*P* < 0.05 and ^##^*P* < 0.01 versus the ESRP1 + sh-nc group.

Silencing of circPHGDH inhibits tumor growth and metastasis in vivo by regulating the miR-149/RAP1B axis.a The 22Rv1 cells were stably transfected with lv-sh-circPHGDH and lv-sh-nc, and qPCR was performed to detect circPHGDH expression. b The 22Rv1 cells were stably transfected with lv-RAP1B and lv-nc, and RAP1B was measured using qPCR. A xenograft tumor model was generated using lentivirus-transfected cells, and antago-miR-149 or antago-miR-nc were injected into the tumors. c The tumor metastasis was observed using bioluminescence imaging measurements. d The quantification of bioluminescence

Full text

Introduction

Prostate cancer (PCa) is a heterogeneous epithelial malignancy of the prostate gland. The disease predominantly affects older men, with the highest incidence over the age of 65 years^1^. However, an increasing number of men under the age of 55 years are also being diagnosed with PCa. The incidence of PCa varies significantly on the basis of race, familial inheritance and geographical location^2^. Localized PCa is often asymptomatic; however, by the time symptoms appear, the disease is usually at an advanced and often incurable stage. Early diagnosis remains a challenge owing to the asymptomatic nature of the localized disease. Although biomarker testing, such as for prostate-specific antigen (PSA), has improved diagnosis, its utility is controversial owing to concerns regarding overdiagnosis^3^. The current treatment modalities for PCa include surgery, ablative radiotherapy androgen deprivation therapy and chemotherapy^4^. However, although these techniques have significantly improved the overall survival of patients, disease heterogeneity, treatment resistance and recurrence remain significant barriers to better clinical outcomes^5,6^. Currently, although multiple biomarkers have been identified as being implicated in PCa pathogenesis and proposed for treatment^7^, the clinical efficacy of targeting them remains suboptimal. Therefore, a deeper understanding of the pathogenesis is crucial for discovering more novel and effective therapeutic targets.

Unlike conventional linear RNAs, circular RNAs (circRNAs) are covalently closed noncoding RNA molecules that exhibit greater stability. CircRNAs are found in nearly all species and are known to regulate transcription, splicing, RNA stability, translation and signaling pathways^8^. Emerging evidence shows that circRNAs can function as microRNA (miRNA) sponges or interact with RNA-binding proteins^9,10^. In malignancies, circRNAs are commonly dysregulated and can regulate tumor progression, demonstrating their potential as diagnostic, prognostic and therapeutic targets^11,12^. However, the role of numerous circRNAs in PCa remains largely unexplored.

Abnormal splicing patterns, often driven by altered expression or activity of splicing factors, are a common feature of tumors^13^. Epithelial splicing regulatory protein 1 (ESRP1), also known as RBM35A, is an RNA-binding protein that functions as a key splicing factor^14^. ESRP1 is dysregulated in various malignancies, such as breast and colon cancers, and is critically involved in the epithelial–mesenchymal transition (EMT) process^15,16^. Moreover, ESRP1 has been shown to regulate the circularization and biogenesis of circRNAs, which are formed through the back-splicing of exons^17^. Therefore, investigating the regulatory relationship between ESRP1 and circRNAs may elucidate the potential functions of circRNAs in PCa.

Metabolic reprogramming is an important hallmark of cancers. Tumor cells often exhibit a heightened reliance on glycolysis to acquire energy and macromolecules—a phenomenon known as the ‘Warburg effect’—which promotes tumor progression and chemoradiotherapy resistance^18^. Lactate, the end product of glycolysis, has been identified as a crucial substrate for histone lactylation, a novel posttranslational modification that expands our understanding of the intricate relationship between metabolism and epigenetics in cancer^19^. Beyond histones, a growing number of non-histone proteins has also been shown to be modified by lactylation^20^. The lactylation of both histone and non-histone proteins is implicated in the regulation of gene expression and protein function, thereby influencing tumorigenesis and development^21^. However, research on protein lactylation is still in its infancy, and its precise regulatory mechanisms in tumor metabolism, growth and metastasis remain largely to be elucidated.

In this study, we screened for differentially expressed circRNAs in PCa and selected circPHGDH for further investigation. We then investigated the effects of circPHGDH on PCa progression and explored the underlying mechanisms. We identified that the biogenesis of circPHGDH is regulated by ESRP1 in a lactylation-dependent manner and that circPHGDH accelerated the malignant progression of PCa via the miR-149/RAP1B axis. Our findings highlight this axis as a source of novel therapeutic targets for PCa.

Materials and methods

Microarray

The GSE113153 dataset was obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). Differentially expressed circRNAs were analyzed in R, and the results were visualized as a heat map and volcano plot based on the criteria |log(fold change)| ≥2 and P < 0.05.

Clinical specimens

This study was approved by the Ethics Committee of Tianjin Medical University. PCa tissues and adjacent normal tissues (para-PCa) were collected from 51 patients diagnosed with PCa at our hospital. None of the patients had received any anticancer treatment before tissue collection. Fresh tissues were stored at –80 °C. Written informed consent was obtained from all patients before surgery.

Cell culture

Prostatic epithelial cells (WPMY1) and PCa cell lines (22Rv1, DU145, DuCaP, VCaP, LNCaP and C4-2), obtained from the American Type Culture Collection (ATCC), were cultured in Dulbecco’s modified Eagle medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37 °C in a 5% CO_2_.

Real-time qPCR

The total RNA was extracted using TRIzol reagent (Invitrogen, Thermo Fisher Scientific). The first-strand complementary DNA (cDNA) was synthesized from 1 μg of total RNA using reverse transcriptase (Vazyme). Genomic DNA (gDNA) was isolated with a Genome DNA Extraction Kit (TIANGEN). Quantitative PCR (qPCR) was performed using SYBR qPCR Master Mix (Vazyme) on an ABI 7500 Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). The gene expression was quantified using the 2^−ΔΔCT^ method. GAPDH was used as the internal control for circPHGDH and mRNAs and U6 for miR-149. All primers were designed and supplied by Genscript.

Subcellular isolation localization

Nuclear and cytoplasmic fractions of 22Rv1 and VCaP cells were separated using the Nuclear/Cytosolic Fractionation Kit (Cell Biolabs) according to the manufacturer’s instructions. The expression of U6, GAPDH and circPHGDH in each fraction was quantified by qPCR. U6 and GAPDH served as the nuclear and cytoplasmic fraction controls, respectively.

FISH

Fluorescence in situ hybridization (FISH) was performed using an RNA FISH Kit (GenePharma) according to the manufacturer’s protocol. In brief, 22Rv1 and VCaP cells (1 × 10^4 ^cells per well) seeded on coverslips were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X-100 for 15 min at room temperature. After washing twice with PBS, the cells were incubated with 2× saline sodium citrate for 30 min. Fluorescently labeled probes for circPHGDH and miR-149 were denatured at 37 °C for 5 min, added to the cells and hybridized overnight at 37 °C. After washing with 2× saline sodium citrate, the cells were counterstained with DAPI at 4 °C for 5 min and imaged using a FV1000 laser scanning confocal microscope (Olympus).

RNase R and actinomycin D treatment

For the RNase R stability assay, total RNA (2 µg) was treated with 3 U/µg RNase R (Epicentre) for 30 min at 37 °C. In addition, to assess the RNA half-life, the 22Rv1 and VCaP cells were treated with 2 µg/ml actinomycin D (Sigma-Aldrich) for 1, 2, 4, 8 and 12 h. The expression levels of circPHGDH and linear PHGDH were then measured by qPCR.

Cell transfection

The 22Rv1 and VCaP cells were seeded in six-well plates. When the cells reached approximately 80% confluence, they were transfected with the following constructs using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer’s protocol: short hairpin RNA (shRNA) targeting circPHGDH (sh-circPHGDH), shRNA targeting RAP1B (sh-RAP1B) or a negative control shRNA (sh-NC); a circPHGDH overexpression plasmid (oe-circPHGDH), a RAP1B overexpression vector (RAP1B); an ESRP1 overexpression vector (ESRP1) or an empty vector control (vector); an miR-149 mimic or a negative control mimic (NC mimic); a miR-149 inhibitor or a negative control inhibitor (NC inhibitor); and shRNAs targeting various splicing factors. All constructs and oligonucleotides were synthesized by GenePharma. At 48 h post transfection, cells were collected for subsequent analyses.

Determination of cell colony formation

Cells were seeded into six-well plates and incubated for 14 days, during which the culture medium was replaced every 3 days. Subsequently, the cells were fixed with 4% paraformaldehyde (Sigma-Aldrich) for 1 h and stained with 0.1% crystal violet (Sigma-Aldrich) for 15 min. Finally, the number of visible colonies was counted.

Determination of invasion and migration

Cell invasion and migration were assessed using 24-well Transwell chambers (Corning). For the invasion assay, chambers were precoated with Matrigel, whereas uncoated chambers were used for the migration assay. Cells were resuspended in serum-free medium, and a total of 1 × 10^5 ^cells were added to the upper chamber of each well. The lower chamber was filled with complete medium (600 µl) as a chemoattractant. Following incubation for 24 h, non-migrated or invaded cells on the upper surface of the membrane were removed. The cells that had migrated or invaded to the lower surface of the membrane were stained with a 0.1% crystal violet solution for 20 min and imaged using a BX51M light microscope (Olympus).

Seahorse assay

Cellular metabolic flux was evaluated by detecting the extracellular acidification rate (ECAR) and oxygen consumption rate (OCR). In brief, PCa cells (5 × 10^4 ^cells per well) were seeded into 24-well XF cell culture plates (Seahorse Bioscience). After the overnight incubation, ECAR and OCR were measured using the Seahorse XFe24 analyzer (Seahorse Bioscience).

Western blotting

Total protein was extracted using radioimmunoprecipitation assay lysis buffer (Thermo Fisher Scientific). Equal amounts of protein (30 µg) were separated by 10% SDS–polyacrylamide gel electrophoresis and subsequently transferred to polyvinylidene difluoride membranes (Merck Millipore). The membranes were then incubated with primary antibodies at 4 °C overnight, followed by incubation with a corresponding secondary antibody at 25 °C for 2 h. The protein bands were visualized using an enhanced chemiluminescence reagent (Thermo Fisher Scientific). The following antibodies were used: anti-E-cadherin (no. 3195, 1:1000, Cell Signaling Technology), anti-N-cadherin (no. 13116, 1:1000, Cell Signaling Technology), anti-β-actin (no. 4967, 1:1000, Cell Signaling Technology), anti-RAP1B (no. 2326, 1:1000, Cell Signaling Technology), anti-phosphorylated (p)-PI3K (no. 13857, 1:1000, Cell Signaling Technology), PI3K (no. 4292, 1:1000, Cell Signaling Technology), p-AKT (no. 9271, 1:1000, Cell Signaling Technology), AKT (no. 9272, 1:1000, Cell Signaling Technology), anti-Pan-lactylation lysine (Kla) (PTM-1401, 1:500, PTMBIO), anti-ESRP1 (PA5-21109, 1:2000, Invitrogen) and HRP-linked anti-rabbit IgG (no. 7074, 1:1000, Cell Signaling Technology).

Bioinformatic analysis

The potential targeting relationship between circPHGDH and miR-149 was predicted using the starBase database. Putative targets of miR-149 were predicted using the miRDB, starBase and TargetScan databases. The potential biological functions of miR-149 target genes were subsequently assessed by Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. RAP1B-associated genes were predicted using the LinkedOmics database. The signaling pathways involved in the correlated genes were analyzed by KEGG enrichment analysis.

Luciferase activity detection

Wild-type (wt) and mutant (mut) sequences of circPHGDH and the 3′-UTR of RAP1B were amplified and cloned into the psiCHECK-2 vector (Promega). For the assay, cells were co-transfected with the respective wt or mut reporter plasmids together with either the NC mimic or the miR-149 mimic using Lipofectamine 2000. A dual-luciferase analysis was performed 48 h after transfection. Relative luciferase activity was calculated as the ratio of Renilla to firefly luciferase activity.

In vitro RNA pull-down assay

Biotinylated miR-149 and NC probes were synthesized and incubated with streptavidin magnetic beads (Beyotime) for 3 h to generate probe-coated beads. The 22Rv1 and VCaP cells were lysed, and the resulting lysates were incubated with the probe-coated beads at 4 °C overnight. Following several washing steps, the bound RNA was extracted. The enrichment of circPHGDH and RAP1B was examined by qPCR.

An in vitro transcription and RNA pull-down assay was performed to assess the interaction between circPHGDH and RAP1B. In brief, circPHGDH overexpression plasmids were constructed as the template. The template was amplified using PCR with specific primers (T7 promoter sequences were added to the 5′ end of the forward primer). The PCR products were recovered and purified for in vitro transcription with T7 RNA polymerase using TranscriptAid T7 high yield transcription kit (Thermo Fisher Scientific). The transcribed sense and antisense circPHGDH were labeled with biotin using the Pierc RNA 3′ End desthiobiotinylation kit (Thermo Fisher Scientific). Then, RNA pull-down was performed. The recombinant ESRP1 protein was prepared by CUSABIO (no. CSB-YP007832HU). Streptavidin magnetic beads were combined with biotinylated sense or antisense circPHGDH at room temperature for 30 min. Subsequently, 300 ng of purified recombinant recombinant were incubated with RNA-beads complex at 4 °C for 1 h with gentle rotation.. After washing, the proteins were extracted from the RNA–protein complexes on the magnetic beads by boiling at 100 °C for 10 min. Western blotting was performed to detect ESRP1 expression.

RIP

A circPHGDH minigene vector was generated as previously described^22^. RNA-binding protein immunoprecipitation (RIP) was performed using an RNA Immunoprecipitation (IP) Kit (GENESEED) according to the manufacturer’s protocol. PCa cells were lysed using RIP lysis buffer for 10 min on ice. Subsequently, 5 µg of either anti-IgG (no. 2729, Cell Signaling Technology) (as a negative control) or anti-ESRP1 (PA5-21109, Invitrogen) were incubated with protein A/G beads at 4 °C for 2 h to prepare antibody-conjugated beads. Then, the cell lysate was incubated with the antibody-conjugated beads at 4 °C overnight. After washing with RIP buffer, co-immunoprecipitated RNA was isolated, and circPHGDH expression was quantified by qPCR.

IP

PCa cells were lysed using IP lysis buffer for 30 min on ice. The resulting cell lysate was incubated with 2 µg of anti-ESRP1 (PA5-21109, Invitrogen) and protein A/G magnetic beads (MedChem Express) at 4 °C overnight. After being pelleted by centrifugation at 3,000g for 5 min, the magnetic beads were washed three times with lysis buffer. Then, the protein levels of ESRP1 in the immunoprecipitates were examined by western blotting.

Measurement of protein stability

To measure protein stability, 22Rv1 and VCaP cells were first either treated with 10 mM L-lactic acid (LA; Sigma-Aldrich) for 24 h or transfected with sh-NC and sh-circPHGDH. Subsequently, these cells were treated with cycloheximide (CHX; Sigma-Aldrich) and collected at 0, 6, 12 and 24 h. At the indicated time points, proteins were extracted and the levels of ESRP1 were detected by western blotting.

Xenograft tumor model

The animal study was approved by the Experimental Animal Welfare Ethics Committee of Beijing MDKN (code: MDKN-2025-072), and all operations were performed according to the Guidelines for the Care and Use of Laboratory Animals. BALB/c nude mice (n = 15) were obtained from the JSJ-lab. Sh-circPHGDH lentivirus (lv-sh-circPHGDH), lv-sh-nc, circPHGDH overexpression lentivirus (lv-circPHGDH), lv-RAP1B and lv-nc were transfected into 22Rv1 cells. Stably transfected cells were selected using the puromycin. These transfected cells (1 × 10^6 ^cells) were injected subcutaneously into the dorsal flanks of the mice to establish a xenograft model. Tumor length and width were measured every week after injection. Tumor volume was calculated by the formula: 0.5 × (length × width^2^). After 7 days of injection, antago-miR-149 or antago-nc was injected into the xenograft tumors in a multi-site injection manner (20 μg per mouse). After measuring the volume on day 28, the tumors were visualized using the IVIS Lumina III imaging system (PerkinElmer) following an intraperitoneal injection of 150 mg/kg of D-luciferin. Following imaging, the mice were euthanized by an intraperitoneal injection of an overdose of pentobarbital sodium. The tumors were then excised, imaged and weighed.

IHC

The tumor tissues were fixed in formalin for 24 h, paraffin-embedded and sectioned at 5 µm. The sections were then dewaxed with xylene and rehydrated through a graded ethanol series. Antigen retrieval was conducted by heating the sections in 0.01 M sodium citrate buffer. After blocking with 5% normal goat serum, the sections were incubated with anti-ki67 (no. 9449, 1:800, Cell Signaling Technology) at 4 °C overnight, followed by incubation with an immunohistochemistry (IHC) detection reagent (no. 8125, Cell Signaling Technology) at 25 °C for 30 min. The results were visualized and captured under a light microscope.

H&E assay

The paraffin sections were dewaxed and rehydrated as mentioned above. Subsequently, the sections were stained with hematoxylin for 5 min and eosin for 1 min. After dehydration and clearing, the sections were examined under a light microscope.

Data analysis

All experiments were performed in triplicate. The results were presented as the mean ± standard deviation (SD) and analyzed using GraphPad Prism 7 software. Student’s t-test and one-way analysis of variance were used to analyze the comparisons between two groups and multiple groups, respectively. Correlation analysis was conducted using the Pearson correlation coefficient. A P value < 0.05 was considered to indicate statistical significance.

Results

Identification of circPHGDH in PCa

To identify key circRNAs involved in PCa progression, we first performed microarray analysis, which revealed that multiple circRNAs were differentially expressed in high-grade versus low-grade PCa (Fig. 1a, b). Among these dysregulated circRNAs, the role of circ_0000121 remains unknown, and we chose it for further investigation. The structure of circ_0000121 (hereafter referred to as circPHGDH) is shown in Fig. 1c. It was generated by the back-splicing of two exons of the PHGDH gene, located at chr1: 120285606-120286575, and its back-splice junction was confirmed using Sanger sequencing. Subsequently, the characteristics of circPHGDH were evaluated. As expected for a circular transcript, RNase R treatment induced the degradation of linear PHGDH mRNA but had little impact on circPHGDH (Fig. 1d). After actinomycin D treatment, the half-life of linear PHGDH mRNA was approximately 4 h, whereas that of circPHGDH exceeded 12 h (Fig. 1e). Furthermore, as shown in Fig. 1f, the back-splice junction of circPHGDH could be amplified from cDNA but not from gDNA. Moreover, subcellular fractionation and FISH assays showed that circPHGDH was mainly localized in the cytoplasm of 22Rv1 and VCaP cells (Fig. 1g, h). Collectively, these results indicate that circPHGDH is a stable and canonical circRNA in PCa cells.Fig. 1. Identification of circPHGDH in PCa cells.a The differentially expressed circRNAs in areas of high-grade and low-grade PCa were shown using a heat map. Orange: high expression; blue: low expression. b The differentially expressed circRNAs were shown in the volcano plot. Purple: downregulated circRNAs; red: upregulated circRNAs. c The genomic loci of circ_0000121 (circPHGDH). The back-splice junction was confirmed using Sanger sequencing. d The abundance of circRNA and mRNA were assessed using qPCR after RNase R treatment. e The abundance of circRNA and mRNA were assessed using qPCR after 22Rv1 and VCaP cells treating with actinomycin D for 1, 2, 4, 8 and 12 h. f The existence of circPHGDH was assessed using qPCR and electrophoresis from cDNA and gDNA. g circPHGDH, U6 (nuclear control) and GAPDH (cytoplasmic control) were evaluated using qPCR in the cytoplasm and nucleus of 22Rv1 and VCaP cells. h The location of circPHGDH in PCa cells was assessed using FISH. n = 3. **P < 0.01.

CircPHGDH promotes PCa cell proliferation, migration, invasion, EMT and glycolysis

Subsequently, we analyzed the role of circPHGDH in PCa. We first measured the circPHGDH expression in clinical samples, and the results of qPCR indicated that circPHGDH expression was significantly increased in PCa tissues compared with para-PCa tissues (Fig. 2a). The patients were divided into circPHGDH high-expression and low-expression groups. As presented in Table 1, no significant difference was observed in age and tumor number between the two groups. Moreover, high circPHGDH expression was related to large tumor size, high T stage and distant metastasis. These results suggest that circPHGDH might be a key regulator in PCa progression. Similarly, the expression of circPHGDH was elevated in a panel of PCa cells compared with the immortalized normal prostatic epithelial cell line WPMY1 (Fig. 2b). The 22Rv1 and VCaP cells were chosen for further study owing to their relatively high endogenous expression of circPHGDH. We then proceeded to overexpress or knock down circPHGDH in these PCa cells. The efficiency of these manipulations was confirmed by qPCR, showing that circPHGDH expression was downregulated following sh-circPHGDH transfection and upregulated following oe-circPHGDH transfection (Fig. 2c). Functionally, we found that the knockdown of circPHGDH significantly suppressed the proliferation, migration and invasion of 22Rv1 and VCaP cells, whereas the overexpression of circPHGDH promoted these malignant biological behaviors (Fig. 2d–f). Consistent with these findings, western blotting results showed that the silencing of circPHGDH increased E-cadherin levels and reduced N-cadherin levels, whereas the overexpression of circPHGDH had the opposite effects on their levels (Fig. 2g). In addition, circPHGDH knockdown resulted in a lower ECAR and a higher OCR, whereas circPHGDH overexpression was associated with a higher ECAR and a lower OCR (Fig. 2h, i). These results demonstrate that circPHGDH acts as an oncogene in PCa by promoting the proliferation, migration, invasion, EMT and glycolysis of PCa cells.Fig. 2. CircPHGDH promotes PCa cell proliferation, migration, invasion, EMT and glycolysis.a A qPCR was applied to detect circPHGDH expression in paired PCa and para-PCa tissues (n = 51). b The circPHGDH expression was measured by qPCR in normal cells (WPMY1) and PCa cell lines (22Rv1, DU145, DuCaP, VCaP, LNCaP and C4-2). c The circPHGDH expression detected by qPCR to confirm successful transfection. d–i After circPHGDH was knocked down or overexpressed, cell proliferation (d) was analyzed by colony formation analysis; migration (e) and invasion (f) were estimated by Transwell assay; E-cadherin and N-cadherin protein levels (g) were examined using western blotting; and ECAR (h) and OCR (i) were measured using Seahorse assay to reflect glycolytic fluxes. n = 3. P < 0.05, **P < 0.01, **P < 0.001, ^##^P < 0.01 versus the oe-nc group.Table 1. Correlation analysis of circPHGDH level and clinical features.SubjectsnLow (n = 26)High (n = 25)χ²P valueAge (years)1.7430.6763 <60261412 ≥60251213Tumor size4.2690.0388* <50 mm311912 ≥50 mm927Tumor number2.4220.1196 1321913 ≥219712T Stage8.2290.0415* T11082 T219118 T31569 T4716Distant metastasis4.7470.0294* YES17512 NO342113Note: *P < 0.05.

CircPHGDH is a miR-149 sponge in PCa cells

To elucidate the underlying mechanism of circPHGDH, we investigated its potential miRNA targets using a dual-luciferase reporter screen (Fig. 3a). MiR-149 was selected for further study because it exerted the most significant inhibitory effect on luciferase activity. Bioinformatic analysis predicted a putative binding site for miR-149 within the circPHGDH sequence (Fig. 3b). To validate this prediction, the overexpression of miR-149 significantly decreased the luciferase activity of the reporter containing wt-circPHGDH but did not affect the luciferase activity of the reporter with mut-circPHGDH (Fig. 3c). Moreover, a direct interaction was confirmed by RNA pull-down assays, which showed that circPHGDH could be specifically pulled down by a biotinylated miR-149 probe (Fig. 3d). Furthermore, miR-149 expression was increased after circPHGDH knockdown and was reduced following circPHGDH overexpression (Fig. 3e). These findings confirm that miR-149 is a target of circPHGDH. Next, we measured miR-149 expression in clinical samples and found that the levels of miR-149 were downregulated in PCa tissues (Fig. 3f). The Pearson correlation analysis results showed that circPHGDH expression was negatively correlated with miR-149 expression in PCa tissues (r = −0.631, P < 0.001; Fig. 3g). In addition, miR-149 expression was lower in a panel of PCa cell lines than that in WPMY1 cells (Fig. 3h). Finally, the results of FISH showed that circPHGDH and miR-149 colocalized in the cytoplasm of PCa cells (Fig. 3i). Taken together, these data demonstrate that circPHGDH functions as a sponge of miR-149 in PCa cells.Fig. 3. CircPHGDH is a miR-149 sponge in PCa cells.a Targeted miRNAs were chosen by a luciferase reporter analysis. b The predicted complementary sequences of circPHGDH and miR-149. c, d A dual-luciferase reporter assay (c) and RNA pull-down analysis (d) were determined for targeting relationship confirmation. e A qPCR was applied to test miR-149 expression in PCa cells after sh-circPHGDH or oe-circPHGDH transfection. f The MiR-149 expression in PCa tissues and para-PCa tissues was examined by qPCR (n = 51). g The correlation of miR-149 and circPHGDH expression in PCa tissues was analyzed using Pearson correlation coefficient. h The MiR-149 expression in WPMY1 and PCa cell lines was tested by qPCR. i The location of circPHGDH and miR-149 in PCa cells was observed using FISH. n = 3. *P < 0.05, **P < 0.01, ***P < 0.001,^##^P < 0.01 versus the oe-nc group.

Downregulated miR-149 abrogates the effects on cellular behaviors induced by circPHGDH knockdown

To determine if circPHGDH exerts its function through miR-149, rescue experiments were performed. As shown in Supplementary Fig. 1a, miR-149 expression was elevated after mimic transfection and was downregulated after inhibitor transfection, confirming their efficacy. Notably, the downregulation of miR-149 by an inhibitor counteracted the inhibition of proliferation, migration and invasion of PCa cells induced by the circPHGDH knockdown (Supplementary Fig. 1b–e). Similarly, the effects of knockdown of circPHGDH, which upregulated E-cadherin levels and downregulated N-cadherin levels, were rescued by the miR-149 inhibitor (Supplementary Fig. 1f). Moreover, the inhibition of miR-149 abrogated the reduction of ECAR and the increase of OCR caused by circPHGDH silencing (Supplementary Fig. 1g, h). In addition, the expression of circPHGDH and miR-149 was negatively correlated in 22Rv1 and VCaP cells (Supplementary Fig. 1i). In summary, these findings indicate that knockdown of circPHGDH suppresses the malignant behaviors of PCa cells by sponging miR-149.

MiR-149 targets RAP1B in PCa cells

To further elucidate the underlying mechanism, the downstream targets of miR-149 were predicted using the miRDB, starBase and TargetScan databases, resulting in a total of 153 common candidate targets (Fig. 4a). The KEGG enrichment analysis revealed that these targets were enriched in several pathways, including the Rap1 signaling pathway (Fig. 4b). It has been indicated that the activation of the Rap1 pathway promotes the migration and invasion of PCa cells, as well as facilitates tumorigenesis and PCa metastasis^23^. RAP1B, as a member of this pathway, was selected for further study. The predicted binding sites between RAP1B and miR-149 are shown in Fig. 4c. Experimentally, the overexpression of miR-149 decreased the luciferase activity of a reporter containing the wt-RAP1B 3′-UTR but not a reporter with a mut binding site (Fig. 4d). Furthermore, an RNA pull-down assay confirmed a direct interaction, as a biotin-labeled miR-149 probe successfully pulled down RAP1B mRNA (Fig. 4e). To link this to the upstream axis, we found that the knockdown of circPHGDH reduced the mRNA and protein levels of RAP1B, whereas the downregulation of miR-149 rescued this reduction (Fig. 4f, g). The expression of RAP1B in clinical samples was detected using qPCR, and the results showed that RAP1B was highly expressed in PCa tissues compared with para-PCa tissues (Fig. 4h). Moreover, RAP1B expression was negatively related to miR-149 expression in PCa tissues (r = −0.629, P < 0.001; Fig. 4i). Consistent with a potential oncogenic role, both qPCR and western blotting indicated that RAP1B was upregulated in PCa cell lines (Fig. 4j, k). Taken together, these data confirm that RAP1B is a direct target of miR-149 in PCa cells.Fig. 4. MiR-149 targets RAP1B in PCa cells.a The targets of miR-149 predicted using the miRDB, starbase and TargetScan databases are shown using Venn diagram. b The KEGG enrichment analysis of these targeted genes. c The predicted complementary sequences of RAP1B and miR-149. d, e Luciferase reporter (d) and RNA pull-down assays (e) were used to determine the targeting relationship. f, g qPCR (f) and western blotting (g) were applied to detect RAP1B expression after circPHGDH and miR-149 knockdown. h The RAP1B levels in PCa tissues and para-PCa tissues were detected by qPCR (n = 51). i The correlation of miR-149 and RAP1B expression in PCa tissues were analyzed using Pearson correlation coefficient. j, k The RAP1B levels in WPMY1 and PCa cell lines were tested by qPCR (j) and western blotting (k). l The RAP1B-correlated genes were predicted using the LinkedOmics database. m The KEGG enrichment analysis of all RAP1B-correlated genes. n The 22Rv1 and VCap cells were transfected RAP1B overexpression plasmids, the protein levels of RAP1B, phosphorylated (p)-PI3K, PI3K, p-AKT and AKT were detected using western blotting. n = 3. *P < 0.05, **P < 0.01, ***P < 0.001, ^##^P < 0.01 versus the sh-circPHGDH + nc inhibitor group.

To investigate the downstream potential molecular mechanisms of RAP1B, we analyzed RAP1B-associated genes. As shown in Fig. 4l, we found various genes that were positively and negatively correlated with RAP1B. The KEGG enrichment analysis results showed that these correlated genes were enriched in many pathways (Fig. 4m). Among these pathways, PI3K/AKT pathway activation is the most common event in PCa. This pathway promotes the growth, invasion and metastasis of PCa cells and thus plays a critical role in tumor formation, disease progression and therapeutic resistance^24,25^. Thus, we evaluated whether RAP1B regulated this pathway. The western blotting results showed that overexpression of RAP1B elevated the phosphorylation level of PI3K and AKT; however, the protein levels of PI3K and AKT were did not affected by RAP1B (Fig. 4n). The results suggest that PAR1B may regulate the PI3K/AKT pathway to serve its function in PCa.

RAP1B abrogates the effects on biological functions of PCa cells induced by miR-149

The functions of RAP1B were subsequently studied in the context of miR-149 activity. Following RAP1B overexpression vector transfection, the expression of RAP1B was significantly elevated, confirming successful transfection (Supplementary Fig. 2a). Crucially, the enforced expression of RAP1B counteracted the suppression of cellular proliferation, migration and invasion induced by miR-149 overexpression (Supplementary Fig. 2b–e). Moreover, the effect of miR-149, which increased E-cadherin levels and downregulated N-cadherin levels, was partly abolished by RAP1B overexpression (Supplementary Fig. 2f). Moreover, the decrease in ECAR and the increase in OCR caused by miR-149 were reversed by RAP1B overexpression (Supplementary Fig. 2g, h). In addition, miR-149 expression and RAP1B expression was negatively correlated in 22Rv1 and VCaP cells (Supplementary Fig. 2i). Collectively, these results indicate that miR-149 suppresses PCa cell proliferation, migration, invasion, EMT and glycolysis by directly targeting RAP1B.

Knockdown of RAP1B inhibits the malignant biological functions of PCa cells

The role of RAP1B alone in vitro was analyzed. The 22Rv1 and VCaP cells were transfected with sh-RAP1B and sh-nc, and qPCR results showed that RAP1B expression was significantly downregulated after sh-RAP1B transfection (Supplementary Fig. 3a), suggesting a successful transfection. The results of the functional experiments showed that the knockdown of RAP1B inhibited PCa cell colonies, migration and invasion (Supplementary Fig. 3b–e). The protein levels of E-cadherin were upregulated, whereas N-cadherin levels were downregulated after RAP1B knockdown (Supplementary Fig. 3f). In addition, the silencing of RAP1B reduced ECAR and enhanced OCR (Supplementary Fig. 3g, h). Together, RAP1B acts as an oncogene that promotes PCa cell proliferation, migration, invasion, EMT and glycolysis.

ESRP1 promotes circPHGDH biogenesis

Given that splicing factors regulate the cyclization of circRNAs, we next explored which splicing factor affected the biogenesis of circPHGDH. A functional screen revealed that circPHGDH expression was specifically reduced after the knockdown of ESRP1, whereas the silencing of other splicing factors did not affect its expression (Fig. 5a). Therefore, ESRP1 was selected for further study. In PCa tissues, ESRP1 expression was positively related to circPHGDH expression (r = 0.670, P < 0.001; Fig. 5b). RIP and in vitro RNA pull-down experiments showed that ESRP1 protein interacted with circPHGDH (Fig. 5c, d). Subsequently, we identified several putative ESRP1-binding motifs in the introns flanking the circPHGDH-producing exons (Fig. 5e). The results of a RIP assay demonstrated that ESRP1 could bind to the A, B, C and E regions of the PHGDH pre-mRNA (Fig. 5f). To confirm the functional importance of these motifs, we found that although the knockdown of ESRP1 significantly decreased circPHGDH expression from a wt minigene construct, this effect was abrogated when the A/B/C/E, A/B or C/E motifs were mutated (Fig. 5g). Moreover, the linear PHGDH expression was markedly elevated after the knockdown of ESRP (Fig. 5h). These results show that ESRP1 targets the A, B, C and E motifs by binding the flanking introns to promote the biogenesis of circPHGDH.Fig. 5ESRP1 promotes circPHGDH biogenesis.a The effect of splicing factors on ESRP1 expression was evaluated qPCR. b The Pearson correlation coefficient was used to analyze the correlation between circPHGDH and ESRP1 expression in PCa tissues (n = 51). c, d RIP (c) and (d) in vitro RNA pull-down assays were conducted to determine the interaction between the ESRP1 protein and circPHGDH. e ESRP1 motifs were found in the flanking of circPHGDH. f The binding relationship between ESRP1 and circPHGDH was tested using RIP assay. g A qPCR was utilized to detect circPHGDH expression after silencing of ESRP1 with these mut sequences. h The effect of the ESPR1 knockdown on the expression of circPHGDH and linear PHGDH was evaluated by qPCR. n = 3. **P < 0.01.

Knockdown of circPHGDH inhibits lactylation of ESRP1 at K43 site in PCa cells

Lactate produced by tumor glycolysis is a substrate for the lactylation modification of proteins, which can aggravate the malignant progression of a tumor. In this study, we focused on the lactylation of the splicing factor ESRP1. We found that treatment with LA increased both the protein levels of ESRP1 and its lactylation levels (Fig. 6a). As mentioned above, the knockdown of circPHGDH inhibits glycolysis. Consequently, we assessed the effect of circPHGDH on ESRP1 lactylation. Compared with the sh-NC group, the knockdown of circPHGDH decreased the protein and lactylation levels of ESRP1 (Fig. 6a). The stability of the ESRP1 protein was then evaluated. The results of a CHX chase assay indicated that LA prolonged the half-life of ESRP1, whereas the circPHGDH knockdown reduced its half-life (Fig. 6b, c). Moreover, the lactylation sites of ESRP1 were predicted and verified by site-directed mutagenesis. The results showed that although LA treatment increased the lactylation and protein levels of wt ESRP1 (ESRP1-WT), this effect was abolished when the K43 site was mutated (K43R) but not after the transfection of the K252R, K507R or K511R plasmids (Fig. 6d). Together, these findings suggest that the inhibition of glycolysis caused by circPHGDH knockdown suppresses the lactylation of ESRP1 at the K43 site, thereby reducing the stability of ESRP1, whereas LA treatment produced the opposite results.Fig. 6. Knockdown of circPHGDH inhibits lactylation of ESRP1 at K43 site in PCa cells.a PCa cells were treated with LA or transfected with sh-circPHGDH, and the protein and lactylation levels of ESRP1 were measured. b The protein stability of ESRP1 was detected using western blotting after CHX treatment for 0, 6, 12 and 24 h. c The quantification results of ESRP1 protein stability. d ESRP1-WT vectors or ESRP1 mutated plasmids at the lactylation sites (K43R, K252R, K507R and K511R) were transfected into the PCa cells, and the effect of LA on ESRP1 lactylation levels was detected using IP and western blotting, as well as the ESRP1 protein levels were measured by western blotting. n = 3. **P < 0.01 versus the NC group, ^##^P < 0.01 versus the sh-NC group.

ESRP1 promotes the malignant behaviors of PCa cells by increasing circPHGDH expression

Next, the effect of ESRP1 on cellular processes was assessed. The overexpression of ESRP1 promoted the colony formation, migration and invasion of PCa cells, an effect that was abrogated by circPHGDH knockdown (Fig. 7a–d). Consistent with this, ESRP1 downregulated the protein levels of E-cadherin and upregulated N-cadherin levels, whereas circPHGDH knockdown reversed these effects (Fig. 7e). In addition, the overexpression of ESRP1 enhanced ECAR and reduced OCR, which was also rescued by circPHGDH knockdown (Fig. 7f, g). These results indicate that ESRP1 acts as a tumor promoter in PCa by increasing circPHGDH expression.Fig. 7ESRP1 promotes the malignant behaviors of PCa cells by increasing circPHGDH expression.a–g After the ESRP1 overexpression and circPHGDH knockdown, the cell proliferation (a) was estimated by colony formation assay; Transwell assay was performed to evaluate migration (b) and invasion (d); the quantification results of cell migration are shown in (c); E-cadherin and N-cadherin levels (e) were examined by western blotting; and ECAR (f) and OCR (g) were detected using Seahorse assay. n = 3. **P < 0.01 versus the vector group; ^#^P < 0.05 and ^##^P < 0.01 versus the ESRP1 + sh-nc group.

Silencing of circPHGDH inhibits tumor growth and metastasis in vivo by regulating the miR-149/RAP1B axis

A xenograft mouse model was established to explore the role of the circPHGDH and RAP1B axis in vivo. The qPCR results showed that circPHGDH expression was downregulated after lv-sh-circPHGDH were stably transfected, and RAP1B expression was upregulated after lv-RAP1B were stably transfected into 22Rv1 cells (Fig. 8a, b). These transfected cells were subcutaneously injected into mice. After the tumor became visible, antago-nc or antago-miR-149 was injected into the tumors. The results showed that the downregulation of circPHGDH reduced mean fluorescence intensity, suggesting that the tumor metastasis was inhibited. The inhibition of miR-149 or enforced RAP1B expression abrogated the inhibition of tumor metastasis induced by circPHGDH knockdown (Fig. 8c, d). Moreover, the knockdown of circPHGDH reduced tumor size, weight and volume, whereas the inhibition of miR-149 or overexpression of RAP1B rescued these effects (Fig. 8e–g). The IHC results showed that the silencing of circPHGDH downregulated ki67 levels in tumors, an effect that was counteracted by antago-miR-149 or RAP1B overexpression (Fig. 8h). In addition, the histological analysis revealed that the malignant pathology of tumor tissues was attenuated by circPHGDH knockdown, which was reversed by miR-149 downregulation or RAP1B overexpression (Fig. 8h). Taken together, these in vivo findings demonstrate that circPHGDH accelerates tumor growth and metastasis by sponging the miR-149/RAP1B axis.Fig. 8. Silencing of circPHGDH inhibits tumor growth and metastasis in vivo by regulating the miR-149/RAP1B axis.a The 22Rv1 cells were stably transfected with lv-sh-circPHGDH and lv-sh-nc, and qPCR was performed to detect circPHGDH expression. b The 22Rv1 cells were stably transfected with lv-RAP1B and lv-nc, and RAP1B was measured using qPCR. A xenograft tumor model was generated using lentivirus-transfected cells, and antago-miR-149 or antago-miR-nc were injected into the tumors. c The tumor metastasis was observed using bioluminescence imaging measurements. d The quantification of bioluminescence imaging results. e The representative images of tumors isolated from mice in each group. f The tumor weight of xenograft tumors. g The tumor volume was detected every week after injection. h The levels of ki67 in tumors were detected by IHC assay, and the histopathology of tumors was assessed by H&E assay. n = 5. **P < 0.01 versus the lv-sh-nc or lv-nc group; ^##^P < 0.01 versus the lv-sh-circPHGDH + antago-nc group; ^&&^P < 0.01 versus the lv-sh-circPHGDH + lv-nc group.

Overexpression of circPHGDH accelerates tumor metastasis and growth

The role of circPHGDH overexpression in vivo was also explored. 22Rv1 cells were stably transfected lv-circPHGDH, and circPHGDH expression was elevated (Supplementary Fig. 4a). Then, these cells were injected into mice to establish a xenograft tumor model. The mean fluorescence intensity was higher in the lv-circPHGDH group than that in the lv-nc group (Supplementary Fig. 4b, c), indicating that circPHGDH accelerates tumor metastasis. Moreover, we isolated xenograft tumors and found that circPHGDH increased tumor volume and weight (Supplementary Fig. 4d–f). The IHC results showed that the overexpression of circPHGDH induced the upregulation of ki67 expression, whereas the hematoxylin and eosin (H&E) assay results showed that circPHGDH promoted the deterioration of the pathological condition of tumor tissues (Supplementary Fig. 4g). In short, circPHGDH promotes tumor growth and metastasis.

Discussion

In the present study, we explored the role and underlying mechanisms of circPHGDH in PCa. We found that the lactylation-modified ESRP1/circPHGDH/miR-149/RAP1B axis was instrumental in promoting the malignant progression of PCa.

Cell migration and invasion are key hallmarks of PCa, and EMT is a prerequisite for tumor cell migration^26^. During the EMT process, the expression of the epithelial marker E-cadherin is reduced, leading to a loss of cell adhesion, whereas the expression of mesenchymal markers such as vimentin and N-cadherin is increased, enabling tumor cells to migrate to other organs^27^. In addition, glycolysis is closely related to tumor cell survival, growth and chemoradiotherapy resistance^18^. Therefore, the inhibition of cell migration, invasion, EMT and glycolysis is critical to attenuating PCa progression. CircRNAs have been demonstrated to regulate the development and progression of PCa. For example, Shen et al.^28^ have reported that circFoxo3 decelerates PCa progression by inhibiting cell survival, migration and invasion. Li et al.^29^ have revealed that circ_0016068 promotes PCa cell growth and metastasis. In this study, we discovered a previously uncharacterized circRNA, circPHGDH, and investigated its role in PCa. Our results showed that circPHGDH expression was upregulated in PCa. The silencing of circPHGDH suppressed PCa cellular proliferation, migration, invasion, EMT and glycolysis in vitro, as well as impeded tumor growth and metastasis in vivo. Moreover, the overexpression of circPHGDH facilitated tumor growth and metastasis. These findings suggest that circPHGDH functions as an oncogenic circRNA in PCa, which may provide a new target for PCa treatment. Interestingly, a previous study has revealed that a circRNA also derived from the PHGDH gene promotes proliferation and invasion in papillary thyroid carcinoma by regulating the miR-122-5p/PKM2 axis^30^. However, it is important to note that the two circRNAs are different isoforms; we studied circ_0000121, which is cyclized from PHGDH exons 11 and 12, whereas they studied circ_0013768 that is cyclized from six exons from the PHGDH gene. This highlights that the formation of circRNAs from the PHGDH locus is diversified.

Given that circPHGDH was predominantly localized in the cytoplasm of PCa cells, we speculated that it regulated PCa progression through the competing endogenous RNA (ceRNA) mechanism. We focused on miRNAs involved in PCa progression that could be potential targets of circPHGDH and found that circPHGDH exhibited the strongest binding ability to miR-149. We then further confirmed that circPHGDH functions as a sponge for miR-149. MiR-149 is known to be a double-edged sword in malignancies, acting as either an oncogene or a tumor suppressor, depending on different cancer types^31,32^. Interestingly, miR-149 has also been reported to have both tumor-promoting and inhibitory effects in PCa^33,34^. It regulates cell proliferation, apoptosis, invasion and migration to participate in tumor development^35^. Our results indicated that miR-149 expression was decreased in PCa, and the downregulation of miR-149 reversed the suppressive effects on PCa cell behaviors and tumor progression in vivo induced by circPHGDH knockdown, supporting its role as a tumor suppressor in PCa^33,36^ Moreover, these findings suggest that circPHGDH promotes the progression of PCa by sponging miR-149.

The paradoxical role of miR-149 in promoting and suppressing cancer may depend on the cellular environment or be related to the balance of its target genes^37^. CircRNAs can competitively bind miRNAs to affect the expression of targeted mRNAs posttranscriptionally, according to the theory of the ceRNA mechanism. Multiple mRNAs were predicted to be targeted by miR-149, and in this study, we verified that RAP1B was a direct miR-149 target. Accumulating data have indicated that the abnormal activation of RAP1B is associated with malignancies. RAP1B participates in regulating cell morphology, cell adhesion, the immune response, as well as angiogenesis^38^. RAP1B is a recognized oncogene whose high expression promotes tumor progression in cancers such as gastric cancer^39^, liver cancer^40^ and thyroid cancer^41^ and is associated with poor prognosis^42^. In PCa, a previous study has shown that RAP1B is enriched in DU145 cells^43^. However, the role of RAP1B in PCa remains largely uncharacterized. Herein, we demonstrated that RAP1B levels were elevated in PCa. Moreover, the enforced expression of RAP1B abolished the tumor-suppressive phenotypes induced by miR-149 overexpression, and the knockdown of RAP1B inhibited malignant cellular behaviors. These data demonstrate that miR-149 suppresses the progression of PCa by targeting RAP1B. In addition, our in vivo data showed that RAP1B overexpression counteracted the inhibition of tumor growth and metastasis caused by circPHGDH knockdown. Taken together, our study establishes that the circPHGDH/miR-149/RAP1B axis accelerates the malignant progression of PCa. Moreover, we found that RAP1B activated the PI3K/AKT pathway, which plays a critical tumor-promoting role in PCa, suggesting that RAP1B functions as an oncogene through activating the PI3K/AKT pathway.

Subsequently, we explored the upstream regulatory mechanisms of circPHGDH. It is well-established that circRNAs are formed by the back-splicing of linear pre-mRNAs, which has led researchers to investigate what factors regulate this splicing process. The formation of numerous circRNAs is facilitated by the presence of long introns flanking the back-spliced exons^44^. Herein, circPHGDH possesses long flanking introns (introns 10 and 12) and a short intron between exons 11 and 12, a structural feature that may favor its circularization. Moreover, we found multiple ESRP1-binding motifs in the introns flanking circPHGDH, suggesting that ESRP1 may promote the biogenesis of circPHGDH. ESRP1 is an important splicing factor known to affect circRNA biogenesis. Previous studies have reported that ESRP1 regulates the circularization and biogenesis of circRNAs, such as circUHRF1, circ-NOLC1 and circ-TNPO3^45–47^. Our findings are consistent with these reports, identifying circPHGDH as a novel downstream target of ESRP1. Moreover, ESRP1 is known to accelerate tumor progression by promoting cell proliferation and EMT^48,49^. Consistent with these studies, our results showed that the overexpression of ESRP1 facilitated the proliferation, migration, invasion, EMT and glycolysis of PCa cells, suggesting ESRP1 promotes the progression of PCa, at least in part, by upregulating circPHGDH. However, because we have not yet successfully established cell lines stably expressing sh-circPHGDH and ESRP1, the role of ESRP1 in vivo in the context of circPHGDH knockdown remains unknown. This is a limitation of this study, and we will continue to explore it in our future work.

Lactate is known to accumulate in the tumor microenvironment owing to the glycolytic activity of tumor cells. Protein lactylation is regulated by glucose and lactate-mediated metabolic dynamics^20^. Growing evidence has revealed that the lactylation of proteins regulates tumor development. For example, HIF1α lactylation promotes the angiogenesis of PCa in vitro and in vivo by facilitating KIAA1199 transcription^50^. Moreover, MOESIN lactylation mediates regulatory T-cell growth to participate in tumorigenesis^51^. Herein, we focused on the lactylation of ESRP1. Our results indicated that LA promoted ESRP1 lactylation at the K43 site in PCa cells, thus enhancing ESRP1 protein stability, whereas the knockdown of circPHGDH had the opposite impact. These findings suggest that circPHGDH promotes glycolysis by regulating the miR-149/RAP1B axis, and the resulting lactate production in turn stabilizes the ESRP1 protein by promoting its lactylation, thereby further promoting the biogenesis of circPHGDH in a positive feedback loop.

In conclusion, this study reveals that the splicing factor ESRP1 promotes the biogenesis of circPHGDH. This oncogenic circRNA in turn facilitates the glycolysis of PCa cells by regulating the miR-149/RAP1B axis, and the resulting lactate production stabilizes ESRP1 via lactylation. This novel positive feedback loop enhances PCa cell proliferation, migration, invasion and EMT, thus accelerating the progression of PCa. This study provides novel insights into the pathogenesis of PCa and suggests that targeting this regulatory axis may represent a promising therapeutic strategy.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Supplementary information

Supplementary Information