Exosomes derived from myeloid-derived suppressor cells facilitate castration-resistant prostate cancer progression via S100A9/circMID1/miR-506-3p/MID1

Castration-resistant prostate cancer (CRPC) is a major cause of recurrence and mortality among prostate cancer (PCa) patients. Myeloid-derived suppressor cells (MDSCs) regulate castration resistance in PCa. Previously, it was shown that intercellular communication was efficiently mediated by exosomes (Exos), but the role and the mechanism of MDSC-derived Exos in CRPC progression was unclear. In this study, the circRNA expression profiles in PC3 cells treated with MDSC-Exo and control cells were investigated using a circRNA microarray. The data showed that circMID1 (hsa_circ_0007718) expression was elevated in PC3 cells treated with MDSC-Exo. Moreover, high circMID1 expression was found in PCa compared with benign prostatic hyperplasia (BPH) tissues and in CRPC patients compared with hormone sensitive prostate cancer (HSPC) patients. Further studies showed that MDSC-Exo accelerated PCa cell proliferation, migration, and invasion, while circMID1 deficiency inhibited MDSC-Exo-regulated CRPC progression in vitro and in vivo. Mechanistically, MDSC-derived exosomal S100A9 increased circMID1 expression to sponge miR-506-3p, leading to increased MID1 expression and accelerated tumor progression. Together, our results showed that a S100A9/circMID1/miR-506-3p/MID1 axis existed in MDSC-Exo-regulated CRPC progression, which provided novel insights into MDSC-Exo regulatory mechanisms in CRPC progression.

Page 2 of 17 Gao et al. Journal of Translational Medicine (2022) 20:346 mechanisms of androgen-independent signalling pathways could identify more effective therapies for CRPC. Cancer cells might weaken the immune defense by activating immunosuppressive mechanisms or inhibiting immune responses, which form a tumor microenvironment leading to cancer cell growth, survival, and metastatic spread. Myeloid-derived suppressor cells (MDSCs) comprise a prominent immune cell subset infiltrating the CRPC microenvironment [6,7]. Elevated levels of MDSCs are detected in tumor tissue and peripheral blood of PCa patients, and the levels correlate with disease progression [8,9]. MDSCs are immature myeloid cells with potent immunosuppressive mechanisms in tumor microenvironments (TME). They mediate immune suppression by polarizing macrophages to a tumor-promoting phenotype [10] and preventing tumorreactive T lymphocyte activation, which promote tumor angiogenesis [11]. Among these mechanisms, some demand interactions between target cells and MDSCs, or soluble mediator release. IL23 produced by MDSCs can activate the androgen receptor (AR) pathway in PCa cells, which promote cell proliferation and survival in androgen-deprived conditions. Antibody-mediated IL-23 inactivation restores sensitivity to ADT in mice, suggesting a MDSC-mediated castration resistance mechanism [6]. However, the MDSCs-mediated resistance mechanism to castration remains largely unknown.
It is accepted that intercellular communication is not related to soluble mediators or cell-to-cell contact, and could instead be effectively mediated by exosomes (Exos) [12,13]. Exos derived from immune cells are major components of the TME, which lead to heterogeneous and complex interplay between tumor cells and immune cells. In fact, MDSCs-shed Exos contain proinflammatory mediators such as S100A9 and S100A8 [14], which promote tumor progression and suppress anti-tumor responses. Recent reports have shown that Exos derived from MDSCs had the biological ability to regulate chemotherapy resistance and tumor progression [15]. Despite an increasing number of investigations indicating MDSC-released exosomes function in immune suppression, it is not clear if MDSCs-derived Exos have specific biological activities regarding castration resistance and the regulation of prostate tumor progression. Identifying new molecular mechanisms of MDSC-Exo-regulated CRPC progression may therefore provide promising ways to improve CRPC treatment. In addition, exosomes are emerging as a new type of cancer biomarkers. Numerous studies have now shown that the unequivocal markers for tumor-derived exosomes may represent a very sensitive diagnostic approach for clinical evaluation of disease progression [16][17][18]. Rashid's research has shown that the differential in vivo distribution of radioisotope 131 I-labeled exosomes from diverse cellular origins, e.g., tumor cells with or without treatments, MDSCs and endothelial progenitor cells [19]. Alterations in protein or nucleic acid profiles of exosomes are correlate with pathological processes of many diseases including cancer, thus, tracking this change could be utilized to monitor disease progression, metastasis, and exosome-based targeted therapy.
Circular RNAs (circRNAs) have been recently shown to be endogenous noncoding RNAs, which function significantly in many cancer progressions, including Pca [20,21]. One of the main ways that circRNAs regulate gene expression is by functioning as a miRNA "sponge" to modify miRNA activity via sequestration, which alters mRNA target gene expression [22]. The circRNAs from the AR gene have been identified in CRPC [23], which suggests a potential role as biomarkers or as therapeutic targets. Wu et al. reported that circRNA17 altered the enzalutamide sensitivity and CRPC cell invasion as suppressors, by regulating Arv7 expression by sponging miR-181c-5p [24]. Nevertheless, the functions and roles of circRNAs in PCa progression are still unknown.
In the current study, using a circRNA microarray profiling, we showed that circMID1 (hsa_circ_0007718) expression was markedly elevated in PC3 cells treated with Exos from MDSCs. Moreover, high circMID1 expression was found in PCa compared with benign prostatic hyperplasia tissues and in CRPC patients compared with hormone sensitive prostate cancer (HSPC) patients. In vitro and in vivo studies showed that circMID1 deficiency inhibited MDSC-Exo-regulated CRPC progression. Mechanistically, we showed that MDSC-derived exosomal S100A9 increased circMID1 expression to sponge miR-506-3p, which might then lead to increased MID1 expression and accelerate tumor progression. The results provided novel insights into the MDSC-Exo regulatory mechanisms in CRPC progression.

Cell culture
An androgen-independent human PCa cell line (PC3) and DU145 were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and maintained according to ATCC guidelines. We cultured PC3 cells in RPMI-1640 medium with 100 IU/mL penicillin, 10% fetal bovine serum (FBS), and 100 μg/mL of streptomycin, and maintained the cells in a humidified incubator with an atmosphere comprised of 5% CO 2 and 95% air at 37 °C.

Isolation and characterization of Exos derived from MDSCs
We generated human MDSCs from human peripheral blood mononuclear cells (PBMCs) in vitro as previously described [6,25]. Briefly, we isolated PBMCs from healthy donors by venipuncture, followed by differential density gradient separation (Ficoll HyPaque; Sigma-Aldrich, St. Louis, MO, USA). The PBMCs (1 × 10 6 cells/mL) were cultured in T25 flasks with RPMI 1640 medium containing 10% heat-inactivated FBS with granulocyte-macrophage colony-stimulating factor (Pep-roTech, Rocky Hill, NJ, USA) at 10 ng/mL and IL-6 at 10 ng/mL for 1 week. In vitro-generated polymorphonuclear MDSCs (PMN-MDSCs) were characterized by flow cytometry for expressions of CD15, CD11b, and CD14. The PMN-MDSCs were CD15 + CD14 −/low CD11b + (the sorting strategy is indicated by green boxes in Additional file 1: Fig S1A). We collected and used the cells for further experiments.
We cultured isolated MDSCs (2 × 10 6 cells/mL) in RPMI-1640 medium containing 10% exosome-depleted FBS for 2 days. FBS was depleted from contaminating bovine exosomes by ultracentrifugation for at least half of a day at 100,000×g. We collected the culture medium using a 0.22 μm filter (BD Falcon; Corning, Corning, NY, USA), and extracted Exos via the ExoQuick Exosome Precipitation Solution (System Biosciences, Palo Alto, CA, USA) following standard procedures. Exos were resuspended in 200 μL of medium. Electron micrographs were observed by transmission electron microscopy (Tecnai-12; Philips, Amsterdam, Netherlands). The BCA protein assay kit was used to quantify the exosomes. Protein markers, CD9, CD63, Hap70, calnexin were determined by immunoblotting. The size and concentration of exosomes were directly measured using the Nanopar-ticleTracking Analyzer (NTA, Malvern Panalytical, Malvern, UK). Identification of MDSC-exosomes was shown in Additional file 1: Fig. S1.

Microarray analysis of circRNAs
We performed circRNA microarray analysis using the Arraystar Human circRNA Array, version 2.0 (Arraystar, Rockville, MD, USA). We performed RNA extraction and microarray hybridization following standard procedures. We extracted total RNA from PC3 cells treated with phosphate-buffered saline (PBS) or MDSC-derived Exos (50 μg/mL) at 48-h using TRIzol reagent (Life Technologies, Carlsbad, CA, USA) following standard procedures. We digested total RNA using RNAse R (Epicentre, Madison, WI, USA) to eliminate linear RNAs and enrich cir-cRNAs. Then, the enriched circRNAs was amplified and transcribed into fluorescent cRNAs using the random priming method using the Arraystar Super RNA Labeling Kit (Arraystar), and hybridizing the cRNAs using the Arraystar Human circRNA Array, version 2 (8 × 15 K). Finally, the slides were washed and scanned using a Scanner G2505C (Agilent, Santa Clara, CA, USA) and the acquired array images were analyzed. Quantile normalization and subsequent data processing was performed using the R software limma package. CircRNAs differentially expressed with statistical significance between MDSC-Exo-PC3 and control-PBS-PC3 (fold-change (FC) ≥ 2 and p ≤ 0.05) were identified through Volcano Plot filtering. Hierarchical clustering was performed to show the distinguishable expression pattern of circRNAs among samples.

Clinical prostate specimens
Patients (ages ranged from 49 to 78 years with a median age of 64 years) undergoing radical prostatectomy for localized PCa (n = 56) and benign prostatic hyperplasia (n = 31) undergoing transurethral resection of the prostate were collected from the Hangzhou Hospital of Traditional Chinese Medicine. All tissue specimens were pathologically confirmed as PCa by two experienced pathologists. The Ethics Committee of the Hangzhou Hospital of Traditional Chinese Medicine approved this study, and we obtained written consent from each participant prior to the study.

Quantitative real-time PCR
We extracted RNA and miRNAs from the cultured cells and tissues using TRIzol reagent (Invitrogen, Karlsruhe, Germany) and an mirVana miRNA isolation kit (Ambion, Austin, TX, USA) following standard instructions. We then synthesized cDNAs using a High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, MA, USA). We performed qPCR analyses using an ABI PRISM 7500 Sequence Detection System (Life Technologies, Grand Island, NY, USA). We used U6 snRNA as an endogenous control of miRNAs, and normalized circRNAs levels to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). We calculated relative gene expression levels using the 2 −ΔΔCt method. The primers used for RT-PCR are included in Table 1.

Cell proliferation assay
We assessed cell proliferation using the Cell Counting Kit-8 (CCK-8) assay (Dojindo, Shanghai, China). In brief, we seeded cells (3 × 10 4 ) into plates with 96 wells and incubated them at 37 °C for 1 day before transfection. We added CCK-8 solution (10 μL) to every well 2 days after transfection. At the indicated time points, we measured absorbance at a wavelength of 450 nm for 5 days using a Spectra Max 250 spectrophotometer (Molecular Devices, San Jose, CA, USA). The experiments were performed in triplicate.

Migration and invasion assays
We performed Transwell assays for PC3 and DU145 cell invasion and migration in 8.0 µm Boyden chambers with 24 wells. For the invasion assays, the chambers were precoated with Matrigel, and cells (3 × 10 4 ) were seeded in serum-free medium into the upper level of the chamber. We added full culture medium to the lower chamber as a chemoattractant. After 1 day of incubating at 37 °C with 5% CO 2 , the membranes were fixed with methanol and stained with 0.1% Crystal Violet. We randomly selected five visual fields (200×) and counted the number of the invading cells using a light microscope. We also carried out migration assays without precoating the chambers.

Cellular uptake of exosomes
Exosomes (20 mg) derived from MDSCs were labeled with CFSE dye, which was diluted at a ratio of 1:1000. After 15 min incubation at 37 °C, the exosomes were harvested and washed with PBS by centrifugation (100,000×g, 70 min). CFSE-labeled exosomes or unlabeled exosomes were added to PC3 and DU145 cells and incubated for 48 h at 37 °C, and the cellular uptake was observed using a confocal fluorescence microscope. The sample was treated in the same manner and used as a control for any free CFSE dye left in the solution. Immunofluorescence microscopy was done as previously described [26]. Primary antibody was mouse IgG antialpha-tubulin DM1A (1:2000) (Sigma). Secondary antibody was donkey anti-mouse IgG AlexaFluor 594 (1:500) (Molecular Probes).

Luciferase reporter assay
The constructs containing wild-type (WT) or mutant (MUT) cirMID1 sequences or the MID1 3'-UTR containing the WT or mutated miR-506-3p binding site were subcloned into the luciferase gene by using a pmir-GLO vector (Promega, Madison, WI, USA). We co-transfected HEK293 or PCa cells with 100 ng of luciferase reporter vectors combined with miRNAs mimics/NC or a miR-506-3p mimics/NC (200 pmol) using Lipofectamine 2000 (Invitrogen and Thermo Fisher Scientific). After 2 days of transfection, we detected firefly and Renilla luciferase activities using a dual luciferase reporter assay system (Promega), and expressed the specific activity as the relative activity of firefly to Renilla luciferase. The experiments were conducted in triplicate.

RNA immunoprecipitation (RIP) assays
We performed RIP assays using a Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Bedford, MA, USA) following the standard protocol. PC3 cells (2 × 10 7 ) were lysed in RIP lysis buffer and the cell lysates incubated, dividing them into two equivalent parts, with either isotype-matched anti-IgG antibody (Millipore) or 5 μg human anti-Argonaute2 (AGO2) antibody (Millipore) with rotation at 4 °C overnight. We added magnetic  beads and continued incubation at 4 °C for 1 h. We immunoprecipitated the samples with proteinase K at 55 °C for 1 h, then treated the complexes of RNA with TRIzol reagent (Life Technologies) for further purification. We utilized purified RNA to detect circMID1 and miR-506-3p expression levels using qRT-PCR.

Mouse xenograft model assays
All male BALB/c (nu/nu) mice were purchased from the Laboratory Animal Center of Zhejiang University. Animals were maintained in standard conditions with water and food provided ad libitum. The Institutional Animal Care and Use Committee at of the Second Affiliated Hospital in Zhejiang University approved the animal studies. For xenograft animal assays, 4 ~ 6-week-old BALB/c nude mice were injected subcutaneously with a total of 5 × 10 6 cells into the right flank (five mice per group). We measured the tumor size every 5 days for 1 month, and the volume was determined using the formula: volume = (length × width 2 ) × 0.5. After 1 month, we euthanized the mice by carbon dioxide asphyxiation and removed the tumors and weighed them. Finally, the tumor tissues were immediately flash-frozen in liquid nitrogen and stored at − 80 °C until analysis.

Immunohistochemistry (IHC) staining
The tumors were resected and fixed in 4% paraformaldehyde, embedded with paraffin, and cut into 5 μm sections. The slices were dewaxed with xylene and dehydrated in graded ethanol, followed by blocking endogenous peroxidase with 3% hydrogen peroxide. After antigen retrieval, the sections were incubated with primary antibodies against MMP-2 (1:1000; Abcam, Cambridge, UK) or Ki67 (1:1000; Abcam) overnight at 4 °C. Then, the slides were incubated with horseradish peroxidase-labeled secondary antibody (1:1000, Abcam) for 1 h. The slides were developed with a diaminobenzidine (DAB) detection kit (Dako Cytomation, Glustrop, Denmark). After being counterstained with hematoxylin, the samples were observed using a light microscope (Olympus, Tokyo, Japan).

Statistical analysis
Data are denoted as the mean ± standard deviation (SD). Student's t-test was used to compare two groups. Comparisons of data among multiple groups were analyzed using one-way analysis of variance (ANOVA), followed by Bonferroni post hoc test. We calculated correlations through Pearson's correlation. P-values < 0.05 indicated a statistical significance. We performed all experiments at least in triplicate.

Expression profiles of circRNAs in PC3 cells treated with exosomes derived from myeloid-derived suppressor cells
To determine the role and molecular mechanisms underlying MDSC-Exo in PCa progression, PC3 cells were cocultured with MDSC-Exo or control-PBS and then PC3 cells were collected for their circRNA profiles analysis by using circRNA microarrays. We detected a total of 4248 circRNAs. Among them, 90 circRNAs were differentially expressed (fold-changes > 2.0; P < 0.05 and FDR < 0.05) between MDSC-Exo-PC3 cells and control-PBS-PC3 cells (Fig. 1A). Of the 90 circRNAs, 12 circRNAs were upregulated and 78 were downregulated in MDSC-Exo-PC3 cells compared with controls. To further evaluate the microarray data, we selected the four most upregulated and four most downregulated circRNAs to confirm their expressions in PC3 cells with different treatments by using qRT-PCR (Fig. 1B, C). The hsa_circ_0007718 was the top upregulated circRNA in MDSC-Exo-PC3 cells. By browsing the hg19/GRCh37 reference genome, hsa_circ_0007718, located at chrX:10491131-10535643, was assumed to derive from the MID1 gene (midline 1), which is located on chromosome Xp22.2. In this manner, we termed hsa_circ_0007718 as "circMID1. " To confirm that circMID1 was differentially expressed in PCa, we examined prostatectomy specimens of PCa patients and benign prostatic hyperplasia (BPH) using qRT-PCR. We discovered that the circMID1 expression was significantly higher in PCa compared to BPH tissues (p < 0.01, Fig. 1D). The circMID1 expression was also higher in CRPC patients compared to HSPC patients (p = 0.008, Fig. 1E). In summary, the results suggested that circMID1 upregulation might be relevant to PCa progression.

Exosomal S100A9 from MDSC promoted circMID1 expression in PC3 cells
We next investigated the underlying mechanism of the circMID1 increase in PC3 cells. The proinflammatory protein, S100A9, was previously shown to be abundant in MDSC-shed exosomes and to have a chemotactic function for MDSCs [27]. As shown in Fig. 3A, abundant S100A9 protein was detected in MDSC-Exo. Then, S100A9 expression was knocked down in MDSCs using siRNA. MDSC-Exo si-S100A9 , which had a lower expression of S100A9 in exosomes, was isolated and purified (Fig. 3B). The knockdown of S100A9 significantly reduced circMID1 expression levels in MDSC-Exotreated PC3 and DU145 cells (Fig. 3C). As RNA binding proteins (RBPs) are necessary for circRNA [28], we then examined whether RBPs DHX9 and QKI, which regulate the biogenesis of a number of circRNAs [29,30], were responsible for circMID1 biogenesis. The knockdown of S100A9 significantly increased DHX9 protein expression levels in MDSC-Exo-treated PC3 and DU145 cells (Fig. 3D). DHX9 is a well-known RNA helicase that can interact with inverted Alu repeats (IARs) to decrease the expression of a subset of circRNAs [29]. Abundant IARs are found within the MID gene locus, some of which are located around exons 7 to 8 (Fig. 3E). The results confirmed that circMID1 abundance increased after silencing DHX9 (Fig. 3F, G), suggesting that DHX9 could be a potential regulator. Additionally, to study the internalization of MDSCderived exosome S100A9 by PCa cells, MDSC-Exo was labeled with carboxyfluorescin diacetate succinimidyl ester (CFSE) fluorescent dye. The uptake of MDSC-Exo S100A9 was observed using a confocal microscope. We found that CFSE-labeled exosomes were localized in the cytoplasm, the uptake of CFSE exosomes by PCa cells was evident at 48 h, suggesting that the number of CFSE exosomes absorbed by PC3 and DU145 cells was observed, but not in the control cells treated with free CFSE dye or incubated unlabeled exosomes (Fig. 3H).
The S100A9 levels in PC3 cells after 48 h incubation with CFSE exosomes was then detected (Fig. 3I), indicating that S100A9 was highly expressed in MDSC-Exo-treated PC3 and DU145 cells.

CircMID1 functioned as a miR-506-3p sponge
To investigate the potential miRNAs associated with circMID1, the possible binding partners of circMID1 were predicted by bioinformatics analyses using two databases in the public domain: STARBASE version 3.0 and CIRCBASE. Bioinformatics data demonstrated that circMID1 interacted with several miRNAs including miR-27a-3p, miR-513b-5p, miR-888-5p, miR-320, miR-22-3p, miR-506-3p, miR-124-3p and so on. The luciferase reporter vector including the circMID1 sequence was constructed, which was transfected with various miRNA mimics into HEK293 cells. The results showed that miR-506-3p exclusively and significantly reduced fluorescein intensity (Fig. 4A). circMID1 had a binding site for miR-506-3p (Fig. 4B), suggesting that circMID1 Exosomal S100A9 from MDSC promoted circMID1 expression in PC3 cells. A A representative western blot of S100A9 expression in MDSC-Exo, and quantitative analysis of S100A9 expression. GAPDH is control protein for cells, Hsp70 is control protein for exosomes. **P < 0.01, NS: no significance. B A representative western blot of S100A9 expression in MDSCs and MDSC-Exo treated with control siRNA (si-NC) or S100A9 siRNA (si-S100A9), and quantitative analysis of S100A9 expression. **P < 0.01, * vs. si-NC. C Relative expression of circMID1 in PC3 and DU145 cells treated with MDSC-Exo or MDSC-Exo si-S100A9 . ** , ## P < 0.01, * vs. control, # vs. Exo. D The expressions of RNA-binding protein DHX9 and Quaking (QKI) were determined by western blotting in PC3 and DU145 cells treated with MDSC-Exo or MDSC-Exo si-S100A9 . The analysis of result of western blot for DHX9 and QKI. ** , ## P < 0.01, * vs. control, # vs. Exo. E The distribution of IARs within the MID1 gene locus. The upper line shows the whole length of MID1, while the lower line shows ± 500 bp of the region between exon 7 and exon 8. F The expression levels of DHX9 were evaluated by western blotting after transfection with DHX9 siRNAs in PC3 and DU145 cells, and quantitative analysis of S100A9 expression. **P < 0.01, * vs. si-NC. G The relative levels of circMID1 in MDSC-Exo-treated PC3 and DU145 cells were measured after DHX9 was silenced. * , # P < 0.05; ** , ## P < 0.01, * vs. control + si-NC. H The uptake of MDSC-derived exosomes by PC3 and DU145 cells at 48 h. The free CFSE treated cells was used as a control (Control). Green, CFSE-labeled exosomes; red, alpha-tubulin, a microtubule marker (scale bar, 20 μm). I The level of S100A9 in PC3 and DU145 cells after 48-h incubation with MDSC-Exo was determined by western blotting. The analysis of result of western blot for S100A9. * P < 0.05 had a stable interaction with miR-506-3p. Therefore, we conducted dual-luciferase reporter assays to determine whether miR-506-3p functioned as a circMID1 target in PCa cells. The luciferase activity was markedly reduced as cells were co-transfected with luciferase reporters, circMID1-WT and miR-506-3p mimics, meanwhile, co-transfection with mutated luciferase reporter circMID1-MUT and miR-506-3p mimics did not significantly affect luciferase activity in PC3 and DU145 cells (Fig. 4C). Previous results suggested that Argonaute 2 (Ago2) protein binds with miRNAs and circRNAs to form an RNA-induced silencing complex. We therefore performed RNA immunoprecipitation (RIP) assays to pull down RNA transcripts bound to Ago2 in PC3 cells. We found that miR-506-5p and circ-MID1 were pulled down efficiently by anti-Ago2, but not by the nonspecific anti-IgG antibody (Fig. 4D). Subsequently, we verified the circMID1 effect on PCa cell miR-506-3p expression. Knockdown results showed that si-circMID1 upregulated miR-506-3p expression Fig. 4 The circMID1 acts as a sponge for miR-506-3p. A HEK293 cells co-transfected with miRNAs mimics or control miRNAs (miR-NC) and luciferase reporter containing circMID1 (circMID1-WT) or mutant construct (circMID1-MUT). Dual luciferase reporter assays were performed. B The predicted binding sites of miR-506-3p and wild type (WT) or mutant (MUT, red) circMID1 are shown. C PC3 and DU145 cells co-transfected with miR-506-3p mimics or control miR-506-3p (miR-NC) and luciferase reporter containing circMID1 (circMID1-WT) or mutant construct (circMID1-MUT). Dual luciferase reporter assays were performed ( ** P < 0.01). * vs. circMID1-WT + miR-NC. D The Ago2 RIP showed that Ago2 significantly enriched circMID1 and miR-506-3p ( ** P < 0.01). E The expression level of miR-506-3p was determined by RT-PCR in PC3 and DU145 cells with si-control or si-circMID1 after MDSC-Exo treatments. ** , ## P < 0.01, * vs. control, # vs. Exo + si-control. in PC3 and DU145 cells without or with MDCS-Exo treatment (Fig. 4E). To further validate the direct binding between circMID1 and miR-506-3p, we performed MS2bp-MS2bs based RIP assay in PC3 cells. As shown in Fig. 4F, miR-506-3p was significantly enriched in RNAs retrieved from MS2bs-circMID1 compared with that from MS2bs-circMID1mt or control MS2bs-Rluc, indicating the specific interaction between circMID1 and miR-506-3p. Moreover, the miR-506-3p expression levels in the PCa samples were significantly lower than that of BPH tissues (Fig. 4G), and miR-506-5p was negatively associated with circMID1 in PCa tissues (Fig. 4H). The results agreed with the observed circ-MID1 overexpression in PCa and demonstrated that circMID1 was a miR-506-3p sponge in PCa.

CircMID1 and MID1 functioned as ceRNAs in PCa through miR-506-3p regulation
To detect if circMID1 acted as a ceRNA to liberate the MID1 expression and sequester miR-506-3p, we then characterized MID1 expression in PCa tissues. The data verified that MID1 was upregulated in PCa tissues (Fig. 5A). Pearson's correlation illustrated that MID1 was negatively correlated with miR-506-3p and positively correlated with circMID1 (Fig. 5B, C). In addition, we employed TargetScan to indicate the miR-506-3p putative target genes, which showed that MID1 was predicted (Fig. 5D). To verify this finding, we performed luciferase reporter assays, which showed that luciferase activity decreased significantly when being co-transfected with miR-506-3p mimics and a luciferase reporter containing WT MID1 3′-UTR, but not with a mutant luciferase reporter in PC3 and DU145 cells (Fig. 5E). Western blotting and qRT-PCR verified that MID1 protein and mRNA expression levels were decreased by miR-506-3p mimics and increased with miR-506-3p depletion by its antagomir in PC3 and DU145 cells (Fig. 5F, G). Together, these results suggested that MID1 was a direct miR-506-3p target, which might also sequester miR-506-3p. We then characterized circMID1 expression after MID1 knockdown and found circMID1 repression (Fig. 5H). To validate if MID1 acted as a ceRNA, we co-transfected si-MID1 and miR-506-3p inhibitors and discovered that circMID1 repression was reversed in PC3 and DU145 cells (Fig. 5H). We then detected MID1 mRNA and protein expressions after circMID1 knockdown and observed reduced expression of both. The expression was reversed after co-transfection with miR-506-3p inhibitors in PC3 and DU145 cells (Fig. 5I,  J). These data suggested that MID1 and circMID1 functioned as ceRNAs by harboring miR-506-3p.

Discussion
Evidence suggests that crosstalk between cancer cells and MDSCs induces cancer progression [31,32]. MDSC-derived exosomes are found to function as intercellular messengers. Studies have analyzed the Exo activities derived from MDSCs under tumor environments [27,33], but the underlying mechanisms remain poorly characterized. Our results showed, in agreement and MID1 in PC3 cells transfected with over-MID1. ** P < 0.01, * vs. control. E, F PC3 cells were transfected with over-circMID1, anti-miR-506-3p and over-MID1, and then treated with MDSC-Exo si−S100A9 . Cell proliferation (E) was measured by the CCK-8 assay, and cell migration and invasion (F) were evaluated by Transwell assays of these PC3 cells with the indicated treatments. **,## P < 0.01, * vs. Exo, # vs. Exo si−S100A9 . All experiments were conducted at least three times with previous findings of unbiased mass spectrometry analyses [27], that S100A9 was abundant in MDSC-Exo. Furthermore, the knockdown of S100A9 inhibited MDSC-Exo-treated PC3 cell migration, proliferation, and invasion. It has been reported that MDSC-derived Exos polarized macrophages to a phenotype that promoted tumor progression, and possessed S100A8/ A9 chemotactic activity [27,33], which suggested that MDSC shed exosomes, which functioned as communicators in the tumor microenvironment. Following ADT, most PCa develops into CRPC, which is a major clinical problem because the recurrent disease does not respond well to alternative therapies. Therefore, identification of the molecular mechanisms of androgen-independent signalling pathways using current genomic approaches would help to improve therapies for this disease.
So, the androgen independent PCa cell lines PC3 and DU145 were choose in this study. However, this can be a limitation. The existing research for understanding of this mechanism of PCa progression is not comprehensive due to lack of available data from one androgen dependent PCa cell line. This research analyzed the circRNA expression profiles in MDSC-Exo-PC3 cells compared with control cells by microarray analyses, and focused on the role and underlying mechanism of the increased circMID1 expression in MDSC-Exo-regulated CRPC progression. Collectively, our results suggested that exosome-mediated transfer of S100A9 from MDSCs to PCa cells promoted proliferation, migration, and invasion of PCa cells via upregulation of circMID1, ultimately accelerating the progression of CRPC.
CircRNAs, as a novel class of noncoding RNAs, are known to participate in cancer progression [34,35]. . F The expression of invasion-related factor MMP-2 in each group, detected by immunohistochemistry (scale bar, 50 μM). **,## P < 0.01, * vs. control, # vs. Exo. The measurement data were expressed as the mean ± standard deviation (SD). All experiments were repeated at least three times  [38], which is consistent with our findings. However, our study also has a limitation. Because we performed bioinformatics analysis to screen out circMID1 only using the two samples and there is a big variability between the two control samples, other important circRNAs may be ignored. This study showed that circMID1 expression was decreased while DHX9 expression was increased in PC3 cells treated with S100A9-silenced MDSC-Exo. DHX9, an abundant nuclear RNA helicase, binds specifically to IARs, and loss of DHX9 increases the formation of a subclass of circRNAs [29,39]. Alu elements are well-known retrotransposons, members of the short interspersed nuclear element (SINE) family of repetitive elements, and IARs around exons promote circRNA formation [40,41]. We found that multiple IARs were located within the MID1 gene, some of which were around exons 7 to 8, where circMID1 was formed. CircMID1 expression was further elevated by silencing DHX9, suggesting that DHX9 affected the circularization of the MID1 gene in an Alu element-dependent manner.
The circRNA-miRNA-mRNA axis is an indispensable regulatory model and circRNAs could function as ceRNAs by regulating miRNAs to play a critical role in transcriptional control [22,42]. Based on bioinformatics analyses, it was assumed that the circMID1/ miR-506-3p/MID1 axis functioned essentially in MDSC-Exo-regulated CRPC progression. Bioinformatics analyses and luciferase reporter assays indicated that circMID1 sponged miR-506-3p, and MID1 was the miR-506-3p target. Cytological functional experiments validated that circMID1 and miR-506-3p could reverse the effects on the cell phenotype. A miR-506-3p antagomir could rescue biological alterations induced by circMID1 silencing. Specifically, we showed mechanistically that circMID1 was involved in MDSC-Exo-regulated CRPC progression by acting as a miR-506-3p sponge. However, circMID1 may also interact with other miRNAs according to bioinformatics analysis. Thus, whether other miRNAs have a role in circMID1-regulated tumorigenesis needs to be investigated in the future.
The upregulated circMID1 was accompanied by downregulated miR-506-3p and upregulated MID1 in PCa tissues. Studies have verified that MID1 is overexpressed in PCa in a stage-dependent manner, and AR is promoted by sustained MID1 upregulation in ADT, causing PCa progression into castration resistance [43]. In this study, MID1 was overexpressed in PCa, and miR-506-3p bound to MID1 and inhibited MID expression in PC3 cells. The circMID1 regulated MID1 expression by sponging miR-506-3p because miR-506-3p inhibition released the inhibitory effect of circMID1 deficiency upon MID1 expression. The circRNA enhanced PCa progression through MID1 because MID1 overexpression rescued the progression defect mediated by circMID1 deficiency in MDSC-Exo-PC3 cells. Furthermore, circMID1 overexpression or miR-506-3p inhibition or MID1 overexpression rescued the inhibitory effect of S100A9-knockdown MDSC-Exo on PC3 and DU145 cells. Thus, the study validated that a S100A9/circMID1/miR-506-3p/MID1 axis existed in MDSC-Exo-regulated CRPC progression.

Conclusion
In conclusion, our data suggested that circMID1 was upregulated in MDSC-Exo-regulated CRPC progression, and that circMID1 knockdown inhibited growth and prostate cancer cell invasion and migration, with MDSC-Exo. The identification of the novel alteration in circR-NAs expression is the first key step toward the better understanding of the role of circRNAs in CRPC. Then, deciphering the exact molecular mechanisms of circRNA function will be critical for understanding CRPC progression and developing new potential therapeutic targets. Exosomal S100A9 from MDSC promoted circMID1 expression in PC3 cells, and the circMID1 acted as a ceRNA to regulate MID1 expression through miR-506-3p (Fig. 8). We discovered an essential S100A9/circMID1-miR-506-3p-MID1 axis in MDSC-Exo-regulated CRPC progression and found a critical function of circRNAs in disease progression.