Flightless I Homolog Reverses Enzalutamide Resistance through PD-L1–Mediated Immune Evasion in Prostate Cancer

Prostate cancer was the most frequently diagnosed malignancy among U.S. males and the second most commonly diagnosed malignancy among men worldwide in 2018 (1, 2). Androgen receptor (AR) signaling drives prostate tumorigenesis and tumor progression (3), and androgen-deprivation therapy (ADT) is the main treatment for prostate cancer. Prostate cancer naturally progresses from hormone-sensitive to hormone-resistant and eventually develops into castration-resistant prostate cancer (CRPC; ref. 4). Currently, endocrine therapies for CRPC include the second-generation AR inhibitors abiraterone and enzalutamide. However, the optimal treatment regimen for patients who have CRPC remains unclear and further research and clinical trials are needed.

In the past few years, immunotherapy has emerged as an exciting form of cancer treatment. Although prostate cancer is an immunogenic tumor (5), it can evade immune responses by producing immunosuppressive cytokines such as TGFβ, inducing T-cell apoptosis, and increasing the number of regulatory T cells (Treg) in the tumor microenvironment (TME; ref. 6). Overall, prostate cancer is considered a “cold” cancer, with minimal T-cell infiltration and minimal response to monotherapy with immune checkpoint inhibitors (ICI; refs. 7–10). The lack of response to ICI monotherapy may be due to the extremely low expression of PD-L1 in prostate cancer (11). Combination therapy is showing promising (12), with several large studies indicating that combining ICIs with radiotherapy, DNA-damaging agents, hormonal therapy, or chemotherapy can enhance immune responses and induce long-lasting clinical responses without obvious toxic side effects (13–15). Therefore, it may be possible to enhance the benefit of the current endocrine therapies for CRPC may by combining them with ICIs.

The flightless I homolog (FLII) protein has a C-terminal gelsolin-like actin binding domain (GLD) and an N-terminal leucine-rich repeat (LRR) protein–protein interaction domain. It interacts with various proteins to regulate intracellular signaling in tumors (16–18). Mammalian FLII functions as a coactivator of nuclear hormone receptor signaling, including signaling via estrogen receptors and thyroid hormone receptors (19). We previously showed that FLII antagonizes dihydrotestosterone-induced nuclear localization of AR, thereby reversing resistance to traditional ADT (bicalutamide) and enhancing sensitivity to enzalutamide (20). However, the role of FLII in enzalutamide-resistant CRPC remains unclear.

In this study, we found that PD-L1 expression was upregulated in enzalutamide-resistant prostate cancer cells (C4-2-R cells) compared with parental cells (C4-2 cells). The drug resistance gene YBX1 transcriptionally regulated expression of PDL1 in human and murine prostate cancer cells. Through binding to the YBX1 protein, FLII antagonized the entry of YBX1 into the cell nucleus, thereby inhibiting PD-L1 transcription and enhancing the immune response to prostate cancer. Restoration of FLII expression enhanced the sensitivity of CRPC cells to enzalutamide treatment and activated T cells by blocking the immune checkpoint PD-L1. In summary, we found that FLII inhibits tumor progression in part by repressing YBX1/PD-L1–mediated immune evasion. FLII may be a therapeutic target for CRPC.

The human prostate cancer cell lines C4-2 and PC3 were purchased from ATCC in 2018 (21, 22). Enzalutamide-resistant C4-2 (C4-2-R) cells were generated in our lab through continuous enzalutamide stimulation in 2018 (22). HEK 293T were purchased from The Cell Bank of Chinese Academy of Science. The mouse prostate cancer cell line RM-1 was purchased from Shanghai GeneChem Co., Ltd in 2019.

All the cells were cultured for less than 1 month continuously, and fresh frozen stocks were used at the initiation of each experiment. All the prostate cancer cell lines were authenticated by the manufacturers and routinely authenticated by growth, morphology, antibiotic resistance (where appropriate). They also were routinely tested for Mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza, catalog no. LT07–318). Cells were maintained in RPMI1640 medium (Servicebio, catalog no. G4530–500ML) containing 10% FBS (Servicebio. #G8001–500) and 1% penicillin/streptomycin (Thermo Scientific, catalog no. 15140122) and were cultured in a 5% CO₂ incubator at 37°C.

Lentiviral particles containing GV341-FLAG-YBX1 (catalog no. GOSL0109744) and GV348-HA-FLII (catalog no. GOSL0151830) were purchased from Shanghai GeneChem and used to infect cells with polybrene according to the manufacturer's recommendations to overexpress YBX1 and FLII, respectively. Small hairpin RNA (sh-RNA) against human YBX1 (catalog no. GIEL0104897) or mouse YBX1 (sh-YBX1; catalog no. GCI0107249), FLII (sh-FLII) (catalog no. GIEL0104894) and the corresponding vector with nonsense sequences (catalog no. GIEL0105343) were purchased from Shanghai GeneChem. C4-2-R cells were engineered to either stably express FLII or to stably express FLII and have YBX1 stably knocked down.

The p-FLAG-YBX1-CSD (residues 51–128 aa), p-FLAG-YBX1-C-term (residues 129–324 aa), p-HA-FLII-LRR (residues 1–450 aa), and p-HA-FLII-GLD (residues 451–1268 aa) plasmids were kindly provided by Professor Chen (20). Human (catalog no. GOSE0104889) and mouse (catalog no. GOSE0106640) luciferase (LUC) reporters driven by the PD-L1 promoter region were purchased from Shanghai GeneChem Co., Ltd. PD-L1-A-LUC contains a sequence −1∼−0.5 kb upstream of the TSS, lacking the Y-box (ATTGG); PD-L1-B-LUC contains a sequence −1∼−1 kb upstream of the TSS, lacking the Y-box; PD-L1-C-LUC contains a sequence −1∼−1.5 kb upstream of the TSS, containing the Y-box; PD-L1-M-LUC contains a sequence −1∼−1.5 kb upstream of the TSS, containing the Y-box mutation (AAAAA). Lipofectamine 2000 (Thermo Scientific, catalog no. 11668019) was used for plasmid transfection as recommended by the manufacturer.

Human prostate cancer tissue microarrays (TMA) were purchased from Wuhan Servicebio Biotechnology Co., Ltd (catalog no. PC1601). All information, including pathologic diagnosis and clinical stage, was downloaded from the online database (http://www.servicebio.cn/).

Normalized prostate cancer mRNA expression datasets were downloaded from cBioPortal—Prostate Adenocarcinoma [The Cancer Genome Atlas (TCGA), Provisional], GSE77930 (23), GSE21032 (24), and phs000447.v1.p1 (25)—and used to assess the association between FLII and PDL1 transcripts. All information, including pathologic diagnosis and clinical stage, was downloaded from the cBioPortal website (http://www.cbioportal.org/). The FLII and PDL1 mRNA expression z-scores relative to all samples (log RNA Seq V2 RSEM) were obtained. The correlation between the FLII and PDL1 mRNA expression was analyzed using the spearman rank correlation test. Statistical analysis was performed with GraphPad Prism 7 software.

Total cellular RNA was extracted using TRIzol reagent (Thermo Scientific, catalog no. AM9738) according to the manufacturer's recommendations. Specific experimental protocols were performed as described previously (26). First-strand cDNA synthesis was performed using Prime Script RT Master Mix (TaKaRa, catalog no. RR014A). qRT-PCR was performed using a standard SYBR Green PCR kit (Thermo Fisher Scientific, catalog no. 4402953). Relative expression of genes was calculated using the power formula: 2^−ΔΔC_t. qRT-PCR was performed on a Roche LightCycler 480 system. All experiments were performed in triplicate. Levels of PDL1, FLII, and AR splice variant 7 (AR-V7) were normalized to GAPDH expression. The primer sequences used are as follows:

• PD-L1:

• Forward, 5′-GTACCTTGGCTTTGCCACAT-3′,

• Reverse, 5′-CCAACACCACAAGGAGGAGT -3′;

• FLII:

• Forward, 5′- GCAGGACTGCTACGTCTTCC-3′,

• Reverse, 5′- TTCTTTTGCAGGCTGAAGGT-3′,

• AR-V7:

• Forward, 5′-CCATCTTGTCGTCTTCGGAAATGTTA-3′,

• Reverse, 5′-TTTGAATGAGGCAAGTCAGCCTTTCT-3′;

• GAPDH:

• Forward, 5′-GAGTCAACGGATTTGGTCGT-3′,

• Reverse, 5′-GACAAGCTTCCCGTTCTCAG-3′.

Prostate cancer TMA was purchased from Wuhan Servicebio Biotechnology Co., Ltd (catalog no. PC1601). TMA sections were fixed in 10% formalin, dehydrated, and embedded in paraffin sequentially. IHC analysis of human prostate cancer tissues was performed using a monoclonal FLII-specific antibody (Santa Cruz Biotechnology, catalog no. sc-21716), a monoclonal PD-L1–specific antibody (Abcam, catalog no. ab205921) and a monoclonal CD8A-specific antibody (Abcam, catalog no. ab17147). Stained sections were evaluated by two independent pathologists with no knowledge of the tissue sample information. The immunoreactive score (IRS) ranges from 0 to 12, which is a result of multiplication of the staining intensity score (0–3) and the positive cell proportion score (0–4). The IRS (score 0–1, negative; score 2–3, mild; score 4–8, moderate; and score 9–12, strong positive) gauged the expression of FLII, PD-L1, and CD8A.

Immunofluorescence (IF) staining was performed using a monoclonal HA-specific antibody (Cell Signaling Technology, catalog no. 2367) and a polyclonal YBX1-specific antibody (Proteintech, catalog no. 20339–1-AP). Briefly, adherent cells were fixed with 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, blocked using 3% BSA for 1 hour at 37°C, followed by incubation with primary antibodies and fluorescent secondary antibodies (Beyotime Institute of Biotechnology, catalog no. A0468 and A0473). Glass coverslips were analyzed under a fluorescence microscope of NIS-Element Viewer.

A truncated human and mouse PD-L1-promoter luciferase reporter and a control reporter were constructed by Shanghai GeneChem (as described in Lentiviral particles, plasmids, reporter genes, and small hairpin RNAs for cell line manipulation). A Renilla luciferase reporter and the truncated PD-L1 promoter reporter were cotransfected into cells overexpressing FLII and with YBX1 knocked down or cells overexpressing YBX1 using Lipofectamine 2000. Luciferase activity was measured with a microplate reader at 48 hours after transfection using Renilla luciferase for normalization.

Cells were lysed with RIPA buffer containing proteinase inhibitors (MedChemExpress, catalog no. HY-K0010). The lysate protein concentration was measured using a BCA Kit (Beyotime Institute of Biotechnology catalog no. P0011) according to manufacturer's instructions. Forty micrograms of total protein was subjected to SDS-PAGE and then transferred to polyvinylidene difluoride (PVDF) membranes. The PVDF membranes were blocked with 5% milk and incubated with primary antibodies overnight at 4°C. The next day, the membranes were washed with PBS and incubated with secondary antibodies (Beyotime Institute of Biotechnology, catalog no. A0208 and A0216). The proteins were visualized and quantified using ChemiDoc-XRS+. Immunoprecipitation (IP) was performed by adding the primary antibody and magnetic beads (Beyotime Institute of Biotechnology, catalog no. P2108–5ml) to the lysed cell supernatant. This was incubated overnight at 4°C in a shaker and then the centrifuged cell supernatant was subjected to SDS-PAGE. The subsequent steps were the same as those in the Western blotting experiments.

The antibodies used for Western blotting and IP were: FLII-specific antibody (Santa Cruz Biotechnology, catalog no. sc-21716), PD-L1–specific antibody (Abcam, catalog no. ab205921), HA-specific antibody (Cell Signaling Technology, catalog no. 2367), YBX1-specific antibody (Proteintech, catalog no. 20339–1-AP), FLAG-specific antibody (Proteintech, catalog no. 66008–2-Ig, 20543–1-AP), GAPDH-specific antibody (Santa Cruz Biotechnology, catalog no. sc-47724), Lamin A/C–specific antibody (Cell Signaling Technology, catalog no. #4777), and Lamin A–specific antibody (Cell Signaling Technology, catalog no. 86846).

Cells were lysed using the Nuclear Protein and Cytoplasmic Protein Extraction Kit (Beyotime Institute of Biotechnology, catalog no. P0028). The extracted nuclear protein and plasma protein were subjected to SDS-PAGE for Western blot analysis.

Direct binding of proteins in vitro was performed as described previously (20). The proteins released were resolved by SDS-PAGE followed by Western blot analysis.

Chromatin immunoprecipitation (ChIP) assays were performed according to protocols described previously (27). YBX1-specific (ProteinTech, catalog no. 20339–1-AP) and FLII-specific (Santa Cruz Biotechnology, catalog no. sc-21716) antibodies were used to perform IP, and normal IgG (Beyotime Institute of Biotechnology, catalog no. A7001) was used as a negative control. Two percent of the original DNA was used as a positive input control. After elution of chromatin, reversal of cross-links and DNA purification, the immunoprecipitated DNA was analyzed using qRT-PCR.

Peripheral blood mononuclear cells (PBMC) from three healthy people were segregated by Lymphoprep density gradient centrifugation (Blossom Biotechnologies Inc. catalog no. PG-1114547), as per the manufacturer's instructions. The collected peripheral blood was directly used for subsequent experiments after centrifugation. All study procedures were approved by the scientific council and ethical committee of Huazhong University of Science and Technology and all participants provided written informed consent.

PBMCs were plated at a density of 2 × 10⁶ per well in 6-well plates and stimulated with anti-CD28 (2 μg/mL; Santa Cruz Biotechnology, catalog no. sc-70612) and anti-CD3e (10 μg/mL; Santa Cruz Biotechnology, catalog no. sc-20047) for 48 hours to facilitate T-cell activation. C4-2-R cells stably overexpressing FLII or vector were treated with or without anti–PD-L1. After 24 hours, T cells were cocultured with mitomycin C–treated C4-2-R cells for 24 hours. The T-cell activation protocols were provided by eBioscience (http://www.ebioscience.com/cell-type/t-cells.htm). Stimulated PBMCs were then harvested and purified by Lymphoprep density gradient centrifugation and cocultured with C4-2-R cells at a 10:1 ratio for 16 hours. The coculture media were analyzed for TNFα, IFNγ, IL2, IL1β, TGFβ, and IL10 using an ELISA kit (ExCell Bio, catalog no. EH001–96) according to the manufacturers' recommendations. MHC class II–specific antibody (anti-MHCII, Invitrogen, catalog no. 25–5321–82) was used to assess the source of cytokines. Enzyme activity was examined at 450 nm using an ELISA reader (Highcreation).

RM-1 cells were cultured, harvested, and suspended in PBS. For flank subcutaneous injections, a total of 0.1 mL containing 2 × 10⁶ cells was injected subcutaneously into the flanks of 6- to 8-week-old C57BL/6 mice (Beijing HFK Bio-technology Co., Ltd). Ten days after injection, the mice were treated with enzalutamide (Selleck, catalog no. S1250; 10 mg/kg) twice per week. Tumor growth and the survival of tumor-bearing mice were evaluated. RM-1 tumors were harvested and assessed for cell-surface markers of CD8⁺ T cells, myeloid-derived suppressor cells (MDSC; CD11b⁺Gr-1⁺), and Tregs (CD4⁺CD25⁺Foxp3⁺) by flow cytometry. The mouse serum was analyzed for IL10, TNFα, TGFβ, and IFNγ using a cytokine ELISA kit (ExCell Bio, catalog no. EM008–96) according to the manufacturer's recommendations. Enzyme activity was examined at 450 nm. All the experimental and animal care protocols were approved by the Institutional Review Board of Tongji Medical College, Huazhong University of Science and Technology.

Flow cytometry assays were performed according to protocols as described previously (28). Tumor tissues were digested using 1 mg/mL collagenase IV (Sigma-Aldrich, catalog no. C4-28–100MG) and 0.2 mg/mL DNaseI (Life Technologies, catalog no. EN0521) at 37°C for 30 minutes. Dissociated cells were passed through a 70-μm cell strainer twice and washed three times in DMEM (Servicebio, catalog no. G4510–500ML). Red blood cells were lysed using ACK Lysis Buffer (Life Technologies, catalog no. A1049201). All samples were pretreated with CD16/CD32 FcR blocker (Thermo Fisher Scientific, catalog no. 16–0161–81) and LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Thermo Scientific, catalog no. L34963) before staining. Cells were stained with the following antibodies obtained from eBioscience: CD11b-PE (clone M1/70), Gr1-APC (clone RB6–8C5), F4/80 (clone BM8), CD8a-PE (clone 53–6.7), mouse Treg staining kit (clone FJK-16s, cat #88–8118); Biolegend: CD45-PerCP/Cy5.5 (clone 30-F11), CD8a-FITC (clone 53–6.7), CD8a-Alexa Fluor 594 (clone 53–6.7), CD4-FITC (clone GK1.5), CD8a-Pacific blue (clone 53–6.7), CD11b-PE (clone M1/70), Ki-67-APC (clone Ki-67). For intracellular cytokine staining, cells were stimulated with 50 ng/mL PMA, 500 ng/mL ionomycin, and 10 μg/mL GolgiPlug (BD Biosciences, catalog no. 555029) at 37°C for 4 hours. Cells were then harvested, fixed and permeabilized using Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences, catalog no. 554715) and stained with IFNγ-FITC (eBioscience; clone XMG1.2), IFNγ-PE (BioLegend; 53–6.7), and TNFα-Alexa Fluor 647 (BioLegend; clone MP6-XT22) to identify cytokine secreting CD8⁺ T cells. Data was collected on a BD LSRII Flow Cytometer and analyzed using FlowJo software (Tree Star Inc.).

All data were analyzed using GraphPad Prism version 7 software. Normally distributed continuous variables were represented as the mean ± SEM, and nonnormally distributed continuous variables were represented as the median. For normally distributed variables, an independent Student t test was used to compare two groups. One-way ANOVA followed by Tukey multiple comparison test was used to compare multiple groups. The correlation between variables with a normal distribution was evaluated using the Pearson correlation test, whereas nonnormally distributed variables were evaluated using the Spearman correlation test. P < 0.05 was statistically significant.

PD-L1 expression is upregulated in enzalutamide-resistant CRPC cells (29). We found that PD-L1 protein and mRNA levels were upregulated in C4-2-R cells, whereas FLII protein and mRNA levels were downregulated (Fig. 1A; Supplementary Fig. S1A). Our previous study shows that FLII enhances the sensitivity of CRPC cells to enzalutamide treatment (20). To explore the mechanism underlying this, we expressed FLII in C4-2-R, C4-2, and PC-3 cells. As shown in Fig. 1A, this downregulated expression of PD-L1 protein. In contrast, knockdown of FLII upregulated PD-L1 protein expression (Supplementary Fig. S1B). To evaluate the correlation between FLII and PD-L1 expression, we performed a pooled in silico analysis Using multiple publicly available prostate cancer datasets. As shown in Fig. 1B, FLII negatively correlated with PD-L1 expression in all the datasets analyzed. Further analysis using a prostate cancer TCGA dataset comprising 496 prostate cancer patient samples showed that FLII expression positively correlated with the number of tumor-infiltrating lymphocytes (TIL), including B cells, CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils, and dendritic cells (Supplementary Fig. S1C).

Although the dataset results showed a negative correlation between FLII and PD-L1, the correlation was weak. To further validate the association between FLII and PD-L1 expression, we examined the expression levels of FLII, PD-L1, and CD8A using IHC in a prostate cancer TMA. Because CRPC is an advanced tumor of the prostate, and most patients with CRPC have already lost the opportunity for surgery, it is difficult to obtain CRPC tissue samples. Consequently, CRPC tissue samples were not included in the TMA. Consistent with our previous results, we found that FLII expression negatively correlated with PD-L1 (Spearman rank correlation test, r = −0.30, P = 0.026) and positively correlated with CD8A (Spearman rank correlation test, r = 0.42, P = 0.0012) in tumor and normal tissues (Fig. 1C; Supplementary Fig. S1D). The correlation between CD8⁺ T-cell infiltration and FLII expression in the TCGA dataset also was weak. Therefore, we examined the expression levels of FLII and CD8A using IHC in the prostate cancer TMA. As shown in Fig. 1C; Supplementary Fig. S1D, FLII expression was positively correlated with CD8A (Spearman rank correlation test, r = 0.42, P = 0.0012) in tumor and normal tissues.

The transcription factor YBX1 plays a role in multidrug resistance in cancer therapy (30–33). In our previous study, we show that YBX1 transcriptionally regulates PD-L1 expression in doxorubicin-resistant HepG2 cells (27). Therefore, we examined the expression of YBX1 in C4-2-R cells. As shown in Fig. 2A, the proportion of YBX1 entering the nucleus was significantly higher in C4-2-R cells than in parental cells. Knockdown of YBX1 downregulated PD-L1 protein expression in C4-2-R and C4-2 cells (Fig. 2A; Supplementary Fig. S2A).

YBX1 binds to the Y-box sequence (ATTGG) in target gene promoters (31). We searched the proximal promoter region of human PDL1 (−1∼−1.5 kb upstream of the transcription start site (TSS)) and found a Y-box sequence (Fig. 2B). To determine its relevance, we performed a luciferase reporter assay. C4-2-R and PC3 cells were cotransfected with YBX1 overexpression plasmids and luciferase reporter plasmids containing different regions of the PD-L1 promoter (Fig. 2B). As shown in Fig. 2B, luciferase activity was highest in cells in which luciferase activity was driven by the entire proximal promoter region of human PD-L1, including the Y-box. Cells in which luciferase activity was driven by truncated versions of the PD-L1 proximal promoter that lacked the Y-box or by the entire PDL1 proximal promoter that contained a Y-box mutation had background levels of luciferase activity. To confirm the luciferase reporter assay results, ChIP was performed using C4-2-R and PC3 cells. As shown in Fig. 2C, YBX1 bound to the region of the PDL1 promoter containing the Y-box (P6: −1, 121 kb∼−1, 301 kb) and activated PD-L1 transcription.

Because FLII negatively correlated with PD-L1 expression, we investigated whether FLII affects PD-L1 transcription. We cotransfected FLII overexpression plasmid and the plasmid in which luciferase activity was driven by the entire PD-L1 proximal promoter into C4-2-R cells with or without stable knockdown of YBX1. As shown in Fig. 2D, overexpression of FLII in C4-2-R cells significantly decreased luciferase activity. FLII overexpression did not decrease PD-L1 expression in C4-2-R cells with stable knockdown of YBX1 (Fig. 2D). ChIP experiments with FLII-specific antibody showed that FLII did not bind to the PD-L1 promoter region (Supplementary Fig. S2B). Moreover, overexpression of FLII inhibited YBX1 binding to the PD-L1 promoter region (Supplementary Fig. S2C). Similarly, overexpression of FLII significantly decreased PD-L1 protein expression in C4-2-R cells, whereas overexpression of FLII with stable knockdown of YBX1 did not enhance the FLII-mediated decrease in PD-L1 protein expression in C4-2-R cells (Fig. 2E). Overall, these results indicate that FLII regulates the expression of PD-L1 via a YBX1-dependent pathway.

To explore the mechanism by which FLII regulates YBX1 signaling and thereby PD-L1 expression, we performed IF experiments and found that FLII and YBX1 colocalized in the cytoplasm of C4-2 cells (Fig. 3A). IP experiments using human C4-2 cells and murine RM-1 cells showed that FLII and YBX1 bound to each other (Fig. 3B). By expressing truncated forms of FLII and YBX1 in HEK 293T cells, we found that full-length FLII and the GLD region of FLII (residues 451–1268 aa) coimmunoprecipitated with full-length YBX1 (Fig. 3C). Conversely, full-length YBX1 and the C-terminal region of YBX1 coimmunoprecipitated with full-length FLII (Fig. 3C). We also performed direct protein binding with recombinant proteins. The results of GST pull-down experiments showed that YBX1 directly bound to FLII (Supplementary Fig. S3A). Overall, these results indicate that FLII physically interacts with YBX1 through an interface between the GLD region of FLII and the C-terminal region of YBX1.

The proportion of YBX1 entering the nucleus increases in multidrug-resistant cells (33). To explore whether the proportion of YBX1 entering the nucleus increased in C4-2-R cells, we performed immunofluorescence experiments. As shown in Fig. 3D, full-length FLII, but not the LRR region of FLII that did not bind to YBX1, sequestered YBX1 in the cytoplasm. In addition, the reintroduction of FLII did not affect the expression of YBX1 protein in C4-2-R cells (Supplementary Fig. S3B). To further confirm the results of IF staining suggesting that FLII interaction with YBX1 sequesters YBX1 in the cytoplasm, we performed cytoplasmic and nuclear protein extraction experiments for Western blot analysis. As shown in Fig. 3E, nuclear accumulation of YBX1 in C4-2 and C4-2-R cells was reduced when full-length FLII was overexpressed, suggesting that FLII partly blocked the entry of YBX1 into the nucleus. Consistent with our hypothesis, nuclear accumulation of YBX1 was increased when FLII was knocked down in C4-2 and C4-2-R cells (Supplementary Fig. S3C). Taken together, these results demonstrate that FLII antagonizes the nuclear accumulation of YBX1 in C4-2-R cells and may affect the transcriptional expression of YBX1 target genes.

High expression of PD-L1 in enzalutamide-resistant CRPC cells and the consequent reduction in dendritic-cell infiltration into tumors might promote immune evasion by CRPC cells (29). The PD-L1/PD-1 inhibitory pathway regulates T-cell activation and cytokine secretion in the TME (34, 35), and the secretion of cytokines in the TME is critical for many biological functions of T cells (36). Therefore, we assessed whether FLII-regulated PD-L1 expression affected cytokine secretion in a C4-2-R/T cell coculture model. When T cells were cultured alone, culture medium levels of TNFα, IFNγ, IL2, and IL1β were higher than when C4-2-R cells were cultured alone, and levels of TGFβ and IL10 were lower (Fig. 4A). In the coculture wells, stable overexpression of FLII in C4-2-R cells enhanced levels of TNFα, IFNγ, and IL2 in culture medium and decreased levels of TGFβ, IL10, and IL1β relative to wells with parental C4-2-R cells and T cells (Fig. 4A). When PD-L1 activity was blocked with a PD-L1–specific antibody, FLII overexpression had not effect on secretion of cytokines (Fig. 4A).

Given that FLII inhibited the transcriptional expression of PD-L1 by antagonizing nuclear accumulation of YBX1 (Fig. 3), we hypothesized that the YBX1 pathway might play a crucial role in the effects of FLII on cytokine secretion in the C4-2-R/T-cell coculture model. As shown in Fig. 4B, compared with the control group, overexpression of FLII, knockdown of YBX1, and treatment with PD-L1–specific antibody each had similar effects on cytokine secretion in C4-2-R/T-cell cocultures. In addition, combining overexpression of FLII and knockdown of YBX1 did not affect cytokine secretion compared with either manipulation alone, indicating that FLII-mediated effects on cytokine secretion were dependent on YBX1. Adding PD-L1-specific antibody treatment to overexpression of FLII, knockdown of YBX1 or the combination of the two did not alter cytokine secretion, indicating that cytokine secretion regulated by the FLII/YBX1 signaling pathway was dependent on PD-L1.

In addition, we performed intracellular cytokine staining when harvesting CD8⁺ T cells in the C4-2-R/T cell coculture model. As shown in Supplementary Fig. S4A, knockdown of YBX1 abolished FLII-regulated TNFα and IFNγ secretion by the CD8⁺ T cells. Blocking PD-L1 activity with a PD-L1-specific antibody abolished the effect of YBX1 knockdown on the secretion of cytokines (Supplementary Fig. S4A). To determine whether TGFβ, IL10 and IL1β came from T cells, we used the C4-2-R/T-cell coculture model with anti-MHCII treatment. As shown in Supplementary Fig. S4B, anti-MHCII treatment significantly decreased the secretion of TGFβ, IL10 and IL1β in coculture medium, suggesting that TGFβ, IL10, or IL1β are mainly coming from T cells. Overall, these results indicate that FLII regulates cytokine secretion by blocking a PD-L1–dependent pathway.

Having shown that YBX1 transcriptionally regulated PD-L1 expression in human prostate cancer cells, we investigated whether this also occurred in murine prostate cancer cells. We found that FLII overexpression in the RM-1 murine prostate cancer cell line downregulated PD-L1 protein expression (Fig. 5A). YBX1 knockdown also reduced PD-L1 protein expression and abolished the FLII-mediated decrease in PD-L1 protein expression (Fig. 5A). In luciferase reporter assays, FLII overexpression significantly reduced luciferase activity driven by a mouse Pdl1 promoter in RM-1 cells, as did YBX1 knockdown, which also abolished the FLII-mediated decrease in PD-L1–driven luciferase activity (Fig. 5B). Consistent with the results of the luciferase reporter assays, ChIP experiments indicated that YBX1 bound to the Y-box region of the Pdl1 promoter (P5: −1, 432 kb∼−1, 654 kb) and activated transcription of Pdl1 (Fig. 5C). In addition, in RM-1 cells overexpressing YBX1, luciferase activity was higher in cells in which luciferase activity was driven by a Pdl1 promoter region containing the Y-box than in cells in which luciferase activity was driven by a Pdl1 promoter region lacking the Y-box region (Fig. 5D).

CD8⁺ T cells play a key role in tumor immune surveillance (37). Tumor regression after PD1 blockade therapy requires preexisting CD8⁺ T cells, which are negatively modulated via PD-L1/PD-1–mediated immune inhibition (38). To investigate the function of FLII, YBX1, PD-L1, and CD8⁺ T cells in vivo, we transplanted RM-1 cells with stable overexpression of FLII or knockdown of YBX1 or vector into syngeneic C57BL/6 mice. These C57BL/6 mice were then treated with a CD8-specific antibody or a PD-L1–specific antibody. The tumor volume of tumor-bearing mice is shown in Fig. 5E and F and Supplementary Fig. S5A. Mice transplanted with tumors with FLII overexpression or YBX1 knockdown, or treated with anti–PD-L1 had significantly reduced tumor volume compared with controls, and CD8⁺ T-cell depletion disrupted the FLII-mediated tumor-suppressive effect (Tukey multiple comparison test, α = 0.05; **, P < 0.01; *, P < 0.05; ns, no significance). Survival of the mice is shown in Supplementary Fig. S5B. The mice with FLII-overexpressing or YBX1-knockdown tumors survived significantly longer than the mice in the control group. Flow cytometry analysis indicated that CD8 expression was lower in the anti-CD8–treated mice versus untreated mice (Supplementary Fig. S5C and S5D).

Next, to verify whether FLII regulated antitumor immunity through YBX1, we examined the infiltration of CD8⁺ T cells, MDSCs and Tregs in RM-1 tumor tissues. As shown in Supplementary Fig. S5E, FLII overexpression and YBX1 knockdown both increased tumor infiltration by CD8⁺ T cells and inhibited tumor infiltration by MDSCs and Tregs. These findings suggested that FLII regulated antitumor immunity through YBX1. We also performed the Ki-67 and annexin staining to detect proliferation and death, respectively, of CD8⁺ T cells in the C57BL/6 mice with HA-FLII–overexpressing or YBX1-knockdown tumors. As shown in Supplementary Fig. S5F, FLII overexpression and YBX1 knockdown both increased CD8⁺ T-cell proliferation and inhibited CD8⁺ T-cell apoptosis. These results indicate that CD8⁺ T cells are required for FLII overexpression-mediated function in tumor-bearing C57BL/6 mice.

To explore the combined therapeutic effects of FLII and enzalutamide in prostate cancer, we injected RM-1 cells with stable overexpression of FLII or vector control into syngeneic C57BL/6 mice and then treated the mice with enzalutamide. Enzalutamide treatment alone did not reduce the volume of RM-1 tumors compared with control, whereas the combination of FLII and enzalutamide significantly reduced the volume of RM-1 tumors (Tukey multiple comparison test, α = 0.05; **, P < 0.01; *, P < 0.05; ns, no significance; Fig. 6A; Supplementary Fig. S6A). There also was no significant trend toward improved survival of mice with enzalutamide treatment alone (Supplementary Fig. S6B). However, the mice with FLII-overexpressing tumors who received enzalutamide survived significantly longer than mice with FLII-overexpressing tumors that did not receive the treatment, which indicates that FLII cooperated with enzalutamide to suppress the progression of RM-1 tumors. To explore the mechanism of RM-1 cell resistance to enzalutamide, we generated an enzalutamide resistance curve for RM-1 cells. We found that RM-1 cells were naturally resistant to enzalutamide and that they expressed AR-V7 (Supplementary Fig. S6C). Flow cytometry analysis of tumor tissue showed that enzalutamide significantly increased PD-L1 expression of wild-type RM-1 tumors, whereas enzalutamide-treated FLII-overexpressing tumors had significantly reduced PD-L1 expression (Tukey multiple comparison test, α = 0.05; ***, P < 0.001; ns, no significance; Fig. 6B).

MDSCs and Tregs are the main components of the immunosuppressive TME (39). To explore the mechanism by which FLII reverses immune evasion, we examined the tumor-immune infiltrate, specifically, CD8⁺ T cells, MDSCs and Tregs, in RM-1 tumor tissues. Consistent with the results of the prostate cancer TCGA database analysis, enforced FLI1 expression increased infiltration of CD8⁺ T cells and inhibited infiltration of MDSCs and Tregs (Fig. 6C–E). Enzalutamide treatment alone did not affect infiltration of CD8⁺ T cells, MDSCs and Tregs, whereas the combination treatment of FLII overexpression and enzalutamide significantly increased infiltration of CD8⁺ T cells and suppressed infiltration of MDSCs and Tregs (Tukey multiple comparison test, α = 0.05; ***, P < 0.001; ns, no significance; Fig. 6C–E). Next, we examined the number of IFNγ⁺CD8⁺ T cells in RM-1 tumor tissue. As shown in Supplementary Fig. S6D, enzalutamide treatment alone did not increase or decrease the number of IFNγ⁺CD8⁺ T cells in vector control tumors. However, the combination of FLII overexpression and enzalutamide significantly increased the proportion of IFNγ⁺CD8⁺ T cells in C57BL/6 mice (Supplementary Fig. S6D). This result indicated that FLII enhanced the sensitivity to enzalutamide treatment in enzalutamide-resistant RM-1 tumors.

In addition, we injected RM-1 cells with stable knockdown of FLII or control vector into syngeneic C57BL/6 mice. As shown in Supplementary Fig. S6E, knockdown of FLII in RM-1 tumors increased the tumor volume. It also decreased the proportion of CD8⁺ T cells among tumor-infiltrating immune cells and increased the proportion of MDSCs and Tregs. These results demonstrated that FLII combated immune evasion by regulating immunocyte infiltration in the TME in enzalutamide-resistant RM-1 tumors. Analysis of T-cell cytokines in the serum of C57BL/6 mice showed that, enzalutamide treatment alone did not affect the levels of TNFα, IFNγ, IL10, and TGFβ in serum from C57BL/6 mice bearing RM-1 tumors (Fig. 6F). Enforced FLI1 expression in the RM-1 tumors significantly increased levels of TNFα and IFNγ and decreased levels of IL10 and TGFβ, and this effect was further enhanced by treatment with enzalutamide (Tukey multiple comparison test, α = 0.05; ***, P < 0.001; ns, no significance; Fig. 6F). On the basis of our in vitro and in vivo experimental results, we generated a schematic of the mechanisms of enzalutamide resistance–induced immune evasion mediated by the FLII/YBX1/PD-L1 signaling axis (Fig. 6G).

ICIs targeting the PD-L1/PD-1 and CTLA-4 pathways are a relatively new and therapeutically efficacious strategy in the treatment of various malignancies (40). ICIs reverse CD8⁺ TIL dysfunction and lead to tumor rejection in mice (41). Previous studies have shown that combining ICIs with hormonal therapy in CRPC can enhance immune system responses and induce long-lasting clinical responses without obvious toxic side effects (14). However, the specific mechanisms by which such combinations of benefits patients with hormonal therapy resistance remain unclear.

Endocrine therapy is the main clinical treatment modality for CRPC. However, resistance to such therapy remains a serious clinical problem. Endocrine therapy reagents for CRPC include abiraterone and enzalutamide. Prior studies show activation of the PD-L1/PD-1 pathway in CRPC that is resistant to endocrine therapy, and this pathway activation leads to a decrease in tumor-specific T-cell activity, which potentially promotes immune evasion by tumor cells (29). Consistent with these results, a small phase II clinical trial reported a rapid reduction in prostate-specific antigen (PSA) in 30% of patients with enzalutamide-resistant prostate cancer who were treated with the PD-1–specific antibody pembrolizumab (14). Therefore, there is some potential association between endocrine therapy and immunoresistance in CRPC. However, it remains unclear what biomolecules regulate the PD-L1/PD-1 pathway or how endocrine therapy modulates antitumor immunity and induces immunoresistance in enzalutamide-resistant CRPC.

The transcription factor YBX1, a member of the cold-shock domain protein family, induces chemotherapy-based immunoresistance in hepatocellular carcinoma (27). Nuclear localization and increased expression of YBX1 regulate the transcriptional expression of multidrug resistance gene (MDR1), thereby inducing multidrug resistance in tumor cells (42). Nevertheless, whether YBX1 has a role in the endocrine therapy–based immune evasion of CRPC remains unknown. In this study, we found that YBX1 and PD-L1 were associated with an immune evasion phenotype in enzalutamide-resistant CRPC. Our findings show that YBX1 directly regulated the transcriptional expression of PD-L1 in human and mouse prostate cancer cells. Knockdown of YBX1 reduced PD-L1 expression in enzalutamide-resistant CRPC cells. Moreover, YBX1 regulated cytokine secretion in the TME by upregulating expression of PD-L1. These results indicate that YBX1 plays a key role in enzlautamide resistance and induces tumor immunosuppression in the microenvironment and immune evasion in enzlautamide-resistant CRPC. Targeting the YBX1/PD-L1 signaling pathway, which could simultaneously reverse the immune evasion and multidrug resistance of enzlautamide-resistant CRPC cells, may benefit patients with CRPC. Similarly, targeting the YBX1/PD-L1 signaling pathway could reverse the chemotherapy resistance and immunoresistance of hepatocellular carcinoma (27), indicating that targeting YBX1/PD-L1 signaling axis has therapeutic potential in multiple tumor contexts.

FLII protein, first found in Drosophila melanogaster, contains GLD and LRR domains. In a previous study, we found that FLII reverses enzalutamide resistance by promoting AR protein degradation and inhibiting AR nuclear accumulation (20). We also showed that FLII inhibits tumor cell growth, migration, and invasion in prostate cancer through AR-dependent pathways, and that high expression of FLII predicts a better prognosis and reversed enzalutamide resistance in CRPC. Herein, we found that FLII was negatively correlated with PD-L1 expression and that low FLII expression predicted an immune-resistant phenotype in enzalutamide-resistant CRPC. FLII physically interacted with YBX1 and inhibited its nuclear localization in enzalutamide-resistant CRPC cells. FLII suppressed IFNγ-induced, PD-L1–mediated CD8⁺ T-cell apoptosis and modulated cytokine secretion in the TME by targeting the YBX1/PD-L1 pathway. Furthermore, CD8⁺ T cells were required for FLII overexpression–mediated tumor suppression in CRPC. In our previous study, results in an animal model show FLII inhibits the growth of enzalutamide-sensitive prostate cancer (C4-2 tumors), which do not depend on the expression of PD-L1 (20), suggesting that FLII affects the progression of prostate cancer through multiple mechanisms. In this study, FLII inhibited the growth of enzalutamide-resistant prostate cancer (RM-1 tumors) in vivo by downregulating the expression of PD-L1 (Fig. 5E and Supplementary Fig. S5A). These findings indicate that FLII plays a key role in tumor cell immunoresistance and immune evasion in enzalutamide-resistant CRPC and that inhibition of the YBX1/PD-L1 signaling pathway may be possible by increasing the expression levels of FLII, thereby enhancing the therapeutic effect of strategies targeting endocrine therapy–based immune evasion.

Blockade of the PD-L1/PD1 pathway enhances antitumor immunity and reduces the exhaustion of TILs by increasing proliferation of effector CD8⁺ T cells and inhibiting Tregs and MDSCs, thereby yielding more powerful tumor control (43). Increased infiltration of Tregs and MDSCs contributes to the immunosuppressive microenvironment of tumors (44). Herein, a combination of FLII overexpression and enzalutamide was very effective in enzalutamide-resistant CRPC. This effect was likely a result of the decreased PD-L1 expression mediated by FLII in RM-1 tumor cells. These findings suggest that PD-L1 expression is inversely correlated with TIL infiltration in the TME and that its upregulation contributes to tumor immunosuppression in enzalutamide-resistant CRPC. Although FLII reintroduction induced a modest increase in TIL infiltration, the number of MDSCs and Tregs was not significantly reduced in the TME after FLII reintroduction alone.

In conclusion, we found that FLII regulated the expression of PD-L1 via a YBX1-dependent signaling pathway. This is an important mechanism by which FLII regulates the PD-L1/PD-1 immune checkpoint signaling pathway through the YBX1 signaling axis in enzalutamide-resistant CRPC. Restoring FLII expression reversed endocrine therapy resistance via partial blockade of the YBX1/PD-L1 signaling pathway in vitro and in vivo. The synergistic effect of this endocrine therapy and immunotherapy was due to the proliferation of effector CD8⁺ T cells and decreased infiltration by Tregs and MDSCs. This study provides convincing support for a targeted therapy involving the functional interaction between the FLII and YBX1/PD-L1 signaling pathways to treat endocrine therapy–resistant CRPC.

L. Bao reports grants from National Natural Science Foundation of China during the conduct of the study. No disclosures were reported by the other authors.

H. Ruan: Resources, data curation, writing–original draft. L. Bao: Formal analysis, methodology, writing–original draft. Z. Tao: Conceptualization, supervision, writing–review and editing. K. Chen: Conceptualization, supervision, funding acquisition.

This work was supported by grants from National Natural Sciences Foundation of China (81672524 and 82002706); Hubei Provincial Natural Science Foundation of China (2018CFA038); Science, Technology and Innovation Commission of Shenzhen Municipality (JCYJ20180305164838833); and Fundamental Research Funds for the Central University (2019kfyRCP004).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.