Regulation of eIF4F Translation Initiation Complex by the Peptidyl Prolyl Isomerase FKBP7 in Taxane-resistant Prostate Cancer

Molecular chaperones are upregulated and associated with resistance to treatment in castration-resistant prostate cancer (CRPC; refs. 1, 2), an advanced form of prostate cancer that develops when the disease progresses after initial treatment with surgery and/or medical castration using androgen deprivation therapy (ADT). Docetaxel and cabazitaxel are 2 taxane chemotherapies that are approved for treatment of metastatic CRPC (3). Recent studies have reported benefits when docetaxel was combined with ADT during the hormone-sensitive stage of the disease (3, 4). Unfortunately, approximately 50% of the patients do not respond to docetaxel. Taxane resistance mechanisms include decreased cellular drug accumulation due to membrane-bound efflux protein overexpression, tubulin isotope overexpression, and defects in apoptosis (5, 6). Many of these pathways involve molecular chaperones, but strategies for targeting them have encountered only limited efficacy. New therapeutic strategies are needed to overcome taxane resistance and improve outcomes for men with prostate cancer.

Several molecular chaperones, including the heatshock proteins Hsp27 and Hsp90, clusterin, and FKB506-binding proteins (FKBP), are involved in protein folding, cellular signaling, apoptosis and transcription, and are potential targets for cancer treatment (7, 8). FKBP12 (FKBP1A) was the first enzyme shown to bind FK506, a natural immunosuppressant, and rapamycin, a macrolide indicated in organ rejection prophylaxis after renal transplantation (9). FKBP12–rapamycin complex associates with the major downstream Akt mTOR-kinase (mammalian target of rapamycin kinase) and has immunosuppressive and antiproliferative properties. Larger FKBPs, such as FKBP52 (FKBP4) and FKBP51 (FKBP5), have also been shown to form rapamycin-induced ternary complexes that inhibit mTOR-kinase activity. FKBP51 and FKBP52 regulate the microtubule-associated protein, tau, and thus affect microtubule stability (10). FKBP7, a molecular chaperone cloned from mouse heart (11), is located in the endoplasmic reticulum (ER) and has been shown to suppress the ATPase activity of mouse ER chaperone HSPA5/GRP78/Bip by its prolyl-peptidyl isomerase activity (11, 12).

Here, we established 4 chemoresistant prostate cancer cell lines to explore mechanisms of taxane resistance, and investigated the FKBP7 signaling pathway and its potential role in taxane resistance in human patients with prostate cancer.

Parental and taxane-resistant IGR-CaP1, PC3, LNCaP, and 22RV1 human prostate cancer cell lines were maintained in RPMI1640 medium with 10% FBS. Pooled taxane-resistant populations were obtained by exposing cells to docetaxel (Sanofi-Aventis) or cabazitaxel (Selleckchem) in a dose-escalation manner as described for IGR-CaP1 (13). Resistant cell lines were treated monthly with the maximum dose of docetaxel or cabazitaxel to maintain the resistant phenotype. Other cell lines are detailed in the Supplementary Information.

Gene expression was profiled using a 4 × 44K human whole genome (G4112F) expression array (Agilent Technologies) with dual-color dye-swap competitive hybridization. Total RNA from untreated parental IGR-CaP1 cells was the RNA reference. Total RNA from IGR-CaP1 cells resistant to 5, 12, 25 50, 100, and 200 nmol/L of docetaxel were used as samples (2 replicates/sample). Image (Feature Extraction software: Agilent Technologies) and gene expression (Bioconductor) analyses was performed.

For resistant cell lines, log₁₀ ratios were computed against the relevant sensitive cell line. To select relevant genes, we combined 3 strategies. Genes permanently over/under-expressed in all resistant cell lines were tested using multiple t tests with bootstrap resampling–based analysis (10,000 samples with replacement). Resulting P values were adjusted using the Benjamini–Hochberg correction method (ref. 14; P_adj ≤ 0.05 considered as significant). Genes with monotonic increasing/decreasing expression over increasing doses were tested using 5-parameter logistic regression (14). The decision rule combined an absolute fold change of ≥2 between the upper and lower asymptotes, and a P_adj ≤ 1e−3, representing correlation quality between the fitted and observed values. Supplementary, potentially informative genes were selected using an information criterion method with reversed principal component analysis (probes considered as observations). An information criterion per gene was computed to quantify its ability to separate samples and 998 genes potentially implicated in docetaxel resistance were identified (592 upregulated and 406 downregulated). Image-based high-content siRNA screening, data analysis, and hit calling are detailed in Supplementary information.

Prostate tissue samples used for the tissue microarray (TMA) were obtained from the Vancouver Prostate Centre Tissue Bank. This study followed the ethical guidelines stated in the Declaration of Helsinki, specimens were obtained from patients with their informed written consent form following a protocol approved by the Institutional Review Board of the University of British Columbia (UBC; Vancouver, British Columbia, Canada). The hematoxylin and eosin slides were reviewed and the desired areas were identified.

Eight TMAs were constructed (Beecher Instruments) by punching duplicate 1-mm cores per sample. All specimens (n = 381) were obtained through radical prostatectomy, except CRPC samples were obtained via transurethral resection of the prostate.

Two TMAs were constructed from 90 patients who had received docetaxel after radical prostatectomy. Analysis was performed on 69 selected patients with cancer with good core integrity. Immunostaining was performed using an automatized technique (Biotin-Streptavidin system and solvent-resistant DAB Map kit) with a Discover XT Autostainer (Ventana Medical Systems). Slides were digitized with the SL801 autoloader and Leica SCN400 scanning system (Leica Microsystems; ×20 magnification). Clearly positive/negative and mixed positive/negative cores were identified. FKBP7 staining was analyzed (0: no staining; 1: faint or focal stain; 2 and 3: convincingly intense stain in most cells).

Cell transfection was performed for siRNA sequences (see Supplementary Information). Cells were plated in 96-well plates with 20 nmol/L siRNA for reverse transfection. Transfection efficiency was checked by Western blot analysis.

Immunoblots were prepared using whole-cell lysate with RIPA buffer, protease inhibitors (Roche), and phosphatase inhibitors (Sigma-Aldrich), then analyzed using an enhanced chemiluminescence-based detection kit (Pierce). See Supplementary Information for antibody sources.

Animal experiments were approved by the local ethics committee (CEEA IRCIV/IGR No. 1226.01, registered with the French Ministry of Research) and performed in compliance with EU Directive 63/2010. IGR Animal Resources holds a Department of Health and Human Services Animal Welfare Insurance (No. A5660-01) and complies with the Guide for the Care and Use of Laboratory Animals.

The docetaxel-resistant mouse model (Supplementary Fig. S4A) was established in nude mice (see Supplementary Information). Two shRNAs targeting FKBP7 (shFK-1 and shFK-2) were engineered and packaged using the GIPZ lentivirus delivery system (see Supplementary information for details). A total of 2 × 10⁶ IGR-CaP1-Rvivo cells transduced with sh-ctrl, skFK-1, or shFK-2 (in 100-μL PBS with 50% Matrigel) were injected subcutaneously in NOD SCID gamma mice purchased from the IGR Animal Resources. Tumor growth was monitored for 50–70 days. When tumors reached an average volume of 450—500 mm³, mice were injected intraperitoneally, 3 times with docetaxel at 30 mg/kg or vehicle, once every 3 weeks.

IGR-CaP1–docetaxel-resistant and RPE-1 cells were grown in 4 T150 flasks and harvested at approximately 80% confluency. Cells were lysed and approximately 10-mg proteins were extracted twice for 30 minutes in buffer (120 mmol/L NaCl, 20 mmol/L HEPES, 1 mmol/L EDTA, 5% glycerol, 0.5% NP40, and protease inhibitor). Pooled supernatants were incubated overnight with 6 μg of either control IgG or FKBP7 antibodies, and 50 μL of magnetic beads (Dynabeads Protein A, Thermo Fisher Scientific). After 4 washes, protein complexes were eluted twice in LDS buffer (Thermo Fisher Scientific) and loaded onto 10% Bis-Tris protein gel. After 5 minutes of migration, the band containing the proteins was excised and processed (standard protocol; ref. 15). Peptide mixtures were analyzed on EASY 1000nLC + Q-EXACTIVE (Thermo Fisher Scientific), using an EASY-Spray Nanocolumn (ES800 15 cm 75 μm), 300 nL/minute flow and 2-hour gradient of acetonitrile + 0.1% formic acid (5% starting and 35% final acetonitrile concentrations). Mass resolution for the full scan was set at 70,000 at 400 m/z. The 10 most intense precursor ions from a survey scan were selected for MS/MS fragmentation using high-energy collision dissociation fragmentation with 27% normalized collision energy (detected at mass resolution 17,500 at 400 m/z). Dynamic exclusion was set for 30 seconds with a 10-ppm mass window. Each sample was analyzed in triplicate. The acquired data were analyzed with Proteome Discoverer software using a Mascot search engine (Proteome ID UP000005640; 20253 sequences). MS/MS spectra were searched with a precursor mass tolerance of 10 ppm and fragment mass tolerance of 0.05 kDa. Trypsin was specified as protease with a maximum of 2 missed cleavages allowed. The minimum peptide length was specified as 6 amino acids. Data were searched against a decoy database, and the FDR was set at 1% of the peptide level. FKBP7 interactors were selected when proteins (compared with IgG control) were only identified in the FKBP7 immunoprecipitation or were enriched at least 3-fold in the specific immunoprecipitation. Specific protein interactors were processed [Ingenuity Pathway Analysis (IPA)].

IGR-CaP1–docetaxel-resistant or RPE-1 cells were adapted, respectively, to RPMI or DMEM:F12 stable isotope labeling by amino acids in cell culture (SILAC) media containing either 12C6, 14N4 l-arginine–HCl + 12C6 l-lysine–2HCl (light media) or 13C6, 15N4 l-arginine–HCl + 13C6 l-lysine–2HCl (heavy media) (Thermo Fisher Scientific) for a minimum of 5 cell doublings. Cells grown in heavy media were transfected with control siNT and cells cultured in light media were transfected with si-FKBP7-2. Equal amounts of cells were mixed and lysed in LDS sample buffer and reducing agent (Thermo Fisher Scientific). Samples were loaded onto NuPage 10% Bis-Tris protein gel and proteins were allowed to enter the gel (with application of 150 V) for only 5 minutes, then the protein-containing band was excised and processed (standard protocol; ref. 16). Analysis of the obtained peptide mixtures was performed as described above, except the H/L ratio per peptide, which was calculated by the quantitation node. The average light to heavy ratio was calculated per identified protein and the lists of proteins in IGR-CaP1–docetaxel-resistant and RPE-1 were processed through IPA.

FKBP7–eIF4G and eiF4E–eiF4G interactions were detected by in situ proximity ligation assay (PLA; Duolink, Sigma-Aldrich). Cells were fixed with paraformaldehyde, permeabilized, and the PLA protocol was performed (Olink Bioscience). After blocking, primary antibodies were incubated for 1 hour at 37°C. PLA probe secondary antibodies were incubated for 1 hour at 37°C. After ligation and DNA amplification, amplicons were detected using far red fluorescence, nuclei were stained with DAPI, and slides were mounted with Olink Mounting Medium. Images were acquired with a Virtual Slides VS120-SL microscope [Olympus; magnification 20×, air objective (0.75 NA), 10-ms exposure for the DAPI channel and 300-ms exposure for the Cy5 channel; 1 pixel = 0.32 μm] and the number of PLA signals/cell was counted using Image Analysis toolbox in Matlab (2011a).

For the TMA, FKBP7 staining was analyzed with the χ² test and the recurrence % was calculated with a Fisher test. Kaplan–Meier curve statistical significance P values were calculated using the Cox proportional hazard model. Cell proliferation curves and tumor growth curves were analyzed by a 2-way ANOVA with Bonferroni post tests. The significance of the mRNA coding level for eiF4G when FKBP7 is silenced was set with a Student t test. The significance of eIF4G–FKBP7 and eIF4E–eIF4G interactions was tested with the linear model or Wilcoxon rank test.

Microarray experiments were submitted to the Array Express data base (European Bioinformatics Institute; www.ebi.ac.uk/arrayexpress/), accession number E-MTAB-4869. MS data were deposited in PRIDE (www.ebi.ac.uk/pride/).

To decipher the mechanisms of taxane resistance in prostate cancer, we developed a series of 4 isogenic parental, docetaxel-resistant, and cabazitaxel-resistant cell lines representative of the different types of epithelial prostate cancer cells (Supplementary Fig. S1). By comparing the gene expression profiles of parental and docetaxel-resistant cells (13, 17), we generated a signature of 998 highly differentially expressed genes potentially correlating with chemoresistance (Supplementary Table S1A–B). Following image-based high-content screening in which the 593 upregulated genes were independently targeted with 4 siRNAs, we identified 34 genes required for cell survival of IGR-CaP1–docetaxel-resistant cells, for which at least 2 siRNAs showed robust Z scores >2 for G₀ cell-cycle arrest phenotype modification and cell proliferation (Supplementary Table S2A and S2B). We focused on chaperone proteins responsive to stress. In particular, as already reported (18, 19), the ER chaperone protein, HSPA5 (GRP78/BiP), was sorted as a candidate involved in chemoresistance in our model. Considering that HSPA5 has been shown to interact with FKBP7 chaperone in a mouse model (12), we focused on the potential role of the uncharacterized FKBP7 human protein, one of the hits during screening, in the mechanism of taxane resistance in prostate cancer.

FKBP7 protein levels were higher in the 4 parental prostate cancer cell lines compared with the RWPE-1 noncancerous prostate cells (Fig. 1A), and in all docetaxel-resistant and cabazitaxel-resistant cells compared with their respective parental cells, with an 8-fold change in taxane-resistant IGR-CaP1 cells. Increased FKBP7 protein levels were related to gene expression upregulation, as we found more FKBP7 mRNA in the taxane-resistant cells than in parental cells (Supplementary Fig. S2A). An increase in FKBP7 expression was an early response of cells observed after 48–72 hours of treatment with microtubule-targeting agents such as taxanes and nocodazole (Fig. 1B; Supplementary Fig. S2B). In contrast, FKBP7 was not overexpressed in enzalutamide-resistant 49F and 42D cells generated from an in vivo LNCaP model (20, 21), compared with the enzalutamide-sensitive V16D cells (Fig. 1C).

To investigate the physiologic relevance of FKBP7 expression in prostate cancer, we determined FKBP7 expression by IHC using tissue microarrays containing 381 prostate cancer and benign tissues obtained from radical prostatectomy or transurethral resection (Supplementary Table S3). Consistent with observations in noncancerous RWPE-1 cells versus prostate cancer cell lines in Fig. 1A, higher FKBP7 expression was seen in prostate cancers than in benign tissues (Fig. 1D). Quantification of staining intensities showed that FKBP7 levels were significantly higher in prostate cancers than in benign tissues (Fig. 1D).

We evaluated FKBP7 expression using TMAs comprising a subset of 69 prostate cancers from patients who had received docetaxel as neoadjuvant therapy. High FKBP7 levels correlated significantly with recurrence % in patients (postsurgical prostate cancer recurrence, metastasis, or death); 79% of patients with high FKBP7 levels developed recurrence compared with 42% of patients with low FKBP7 (Fig. 1E). High levels of FKBP7 were significantly associated with a shorter time-to-recurrence in patients (P = 0.004; HR for low FKBP7: 0.3846; Fig. 1F). Levels of FKBP7 during progression to lethal prostate cancer did not correspond to gene mutations in published data from whole exome sequencing in docetaxel-treated patients with CRPC (22). These results show strong correlations between FKBP7 levels and resistance to taxane treatment in prostate cancer.

To determine whether FKBP7 could be a therapeutic target in chemoresistant prostate cancer, we used 2 siRNA sequences targeting FKBP7 to knockdown FKBP7 expression in chemoresistant cells. A statistically significant decrease in chemoresistant cell growth was observed after FKBP7 silencing (Fig. 2A). Except for 22RV1 cells, a similar effect was observed in the parental cells (Supplementary Fig. S3A). FKBP7 silencing induced slight apoptosis of chemoresistant cells, and PARP cleavage was further increased after treatment with docetaxel and cabazitaxel (Fig. 2B). This chemosensitization effect was not observed by measuring cell growth over the same time-course (Supplementary Fig. S3B). Conversely, FKBP7 overexpression is not sufficient to render prostate cancer cells more resistant to taxanes as shown in parental IGR-CaP1 and 22R1 cells transduced with lentivirus expressing FKBP7 (Supplementary Fig. S3C). These results show that FKBP7 is not a primary event of the acquisition of the resistant state but show the importance of FKBP7 signaling in maintaining the chemoresistance state.

The IGR-CaP1 cell line was used to generate a chemoresistant metastatic CRPC mouse model. Various clones of IGR-CaP1 resistant to docetaxel were injected subcutaneously into nude mice. The 25 nmol/L–resistant emerging tumor was maintained in vivo for 5 successive passages to increase the tumorigenicity (Supplementary Fig. S4A). These IGR-CaP1-Rvivo tumors did not respond to 3 successive injections of 30 mg/kg docetaxel, therefore constituting a new docetaxel-resistant mouse model, whereas the parental IGR-CaP1 mouse model responded to docetaxel (Fig. 3A). A new IGR-CaP1-Rvivo cell line generated from the IGR-CaP1-Rvivo tumors showed chemoresistant characteristics (with IC₅₀ = 207 nmol/L toward docetaxel; Fig. 3B). In addition, and in agreement with results obtained in vitro and in human samples, the resistant IGR-CaP1-Rvivo cell line showed high levels of FKBP7 (Fig. 3B). Thus, to check whether FKBP7 could be a therapeutic target in docetaxel-resistant mice, we established 2 IGR-CaP1-Rvivo cell lines in which FKBP7 was stably silenced with 2 different shRNAs. A substantial reduction of FKBP7 level was achieved in cells that were stably expressing shRNAs targeting FKBP7 (shFK-a and shFK-b), compared with control shRNA–expressing cells (Fig. 3C). shRNA-transduced cell lines were subsequently injected subcutaneously in immunodeficient mice. In the absence of docetaxel, significantly reduced tumor growth was observed after FKBP7 depletion (68% and 47% inhibition of tumor growth in shFK-a- and -b–transduced cells, respectively) versus control shRNA-transduced cell line (Fig. 3C). This effect was more pronounced in the xenograft from the shFK-a cells showing a high reduction of FKBP7 to a level similar to that observed in the parental sensitive cells, suggesting that FKBP7 sustains the growth of chemoresistant tumors. Treatment of mice with docetaxel when tumors reached 450–500 mm³ strongly abrogated the growth of shFK-a–transduced tumors, whereas it slightly decreased the growth of control shRNA–transduced tumors (Fig. 3C). Consistently, in vitro proliferation data showed that shFK-a–transduced cells are more sensitive to higher dose of taxanes (Supplementary Fig. S4B). Thus, these results demonstrate that in vitro and in vivo FKBP7 silencing inhibits cell proliferation and sensitizes chemoresistant cells to taxanes, and that FKBP7 can be a relevant therapeutic target to overcome chemoresistance in prostate cancer.

We determined the levels of FKBP7 in noncancerous human cell lines of different origins (fibroblastic, myoblastic, epithelial, and endothelial cells). These noncancerous cells showed a high level of FKBP7, which was similar (in RPE-1) or even higher than that observed in IGR-CaP1–docetaxel-resistant cells (Supplementary Fig. S5A), but in contrast to the results obtained in chemoresistant cells, we observed no effect of FKBP7 inhibition on cell growth or viability (Supplementary Fig. S5B), suggesting that FKBP7 could have different functions in cancerous and noncancerous cells.

To identify molecular pathways in which FKBP7 functions in noncancerous and chemoresistant cells, we performed qualitative and quantitative mass spectrometry analyses. The protein interactome of FKBP7 was identified by immunoprecipitating endogenous FKBP7 with 2 antibodies in noncancerous RPE-1 and docetaxel-resistant IGR-CaP1 cell lines (Supplementary Table S4A and S4B). Protein network analysis through IPA indicated that in both cell lines, FKBP7 protein interactors were mainly distributed in the same 3 intracellular pathways associated with protein translation. Specifically, eIF2, eIF4, and mTOR signaling were the most represented pathways in the signatures (Fig. 4A), and many eukaryotic translation initiation factors were represented. To elucidate the molecular mechanisms by which FKBP7 depletion exerts its cytotoxic effects in chemoresistant cells only, we used SILAC in FKBP7-silenced RPE-1 and IGR-CaP1–docetaxel-resistant cells and in cells transfected with control siRNA, and we obtained lists of 910 and 1,223 differentially expressed proteins in RPE-1 and in IGR-CaP1–docetaxel-resistant cells, respectively (Supplementary Table S5A–S5B). Consistently with previous results, IPA analysis showed that the major pathways affected by FKBP7 silencing were the same 3 pathways implicated in protein translation control; based on calculated Z-score, these pathways were downregulated in IGR-CaP1–docetaxel-resistant cells, but mainly upregulated in noncancerous RPE-1 cells (Fig. 4B and C). All 3 individual components of the eIF4F cap-dependent mRNA translation complex (eIF4E, eIF4A, and eIF4G) were identified as protein partners of FKBP7 in resistant and RPE-1 cells. Coimmunoprecipitation experiments confirmed the interaction of FKBP7 with eIF4G (Fig. 4D; Supplementary Fig. S6A), but interaction of FKBP7 with eIF4A was not detectable (Supplementary Fig. S6A).

EIF4F is an interesting target, known to be involved in resistance mechanisms to many cancer therapies (23–26). We focused on eIF4G, the most downregulated eukaryotic translation factor in IGR-CaP1–docetaxel-resistant cells, but not in RPE-1 cells. FKBP7 silencing led to decreased eIF4G in IGR-CaP1–docetaxel-resistant cells, but to increased eIF4G in corresponding noncancerous RPE-1 cells (Fig. 5A). The level of eIF4G was also lower in other chemoresistant cellular models (Fig. 5B) and in parental cells (Supplementary Fig. S6B). In contrast, protein levels of eIF4A and eIF4E remained unchanged after FKBP7 silencing (Fig. 5A).

To understand how FKBP7 silencing decreased the eIF4G level, we analyzed the effect of FKBP7 depletion on eIF4G gene expression. Efficient FKBP7 silencing did not affect eIF4G mRNA levels in FKBP7-silenced chemoresistant IGR-CaP1 and 22RV1 cells (Supplementary Fig. S6C), so our results showed that although eIF4G is still transcribed, its protein level decreases when FKBP7 is silenced, thus leading to the hypothesis that FKBP7 regulates eIF4G expression at mRNA translation or protein stability levels. Treatment with the translation inhibitor, cycloheximide, combined with FKBP7 silencing showed that although the steady-state level of eIF4G was decreased upon FKBP7 silencing, the amount of eIF4G in control cells decreased over time, whereas it was stable in FKBP7-depleted cells supporting the hypothesis that FKBP7 regulates eIF4G expression at the protein level (Fig. 5C). However, we cannot exclude the possibility of an increase in eIF4G mRNA translation to compensate the protein loss observed when FKBP7 was silenced. We therefore investigated the ability of FKBP7 to interact with eIF4G by performing PLA, allowing the interaction to be quantitatively visualized as fluorescent dots. Our results revealed an interaction between FKBP7 and eIF4G in docetaxel-resistant IGR-CaP1 and docetaxel-resistant 22RV1 cells (Fig. 5E). This interaction was largely affected in FKBP7- and eIF4G-depleted cells, as shown by the lower numbers of dots/cell (43% and 80% reduction in IGR-CaP1–docetaxel-resistant cells, 93% and 86% in 22RV1 docetaxel-resistant cells, for FKBP7 and eIF4G silencing respectively; Fig. 5E); the silencing of eIF4G was very efficient (Fig. 5D) and the specificity of the FKBP7–eIF4G interaction was validated (Supplementary Fig. S6D).

As FKBP7 interacts with eIF4G and modulates eIF4G protein levels, we investigated effects of FKBP7 on eIF4F complex formation using PLA. Docetaxel treatment led to decreased formation of the eIF4E–eIF4G complex in docetaxel-sensitive IGR-CaP1 cells (Fig. 5F). Strikingly, but consistently with other studies showing that eIF4E–eIF4G formation determines the sensitivity to anticancer-targeted therapies (24), this effect was not observed in the IGR-CaP1–docetaxel-resistant cell line (Fig. 5F). By affecting the level of eIF4G, siRNA-mediated FKBP7 depletion induced decreased eIF4E–eIF4G complex formation in resistant cells after docetaxel treatment (Fig. 5F). Thus, the reduction of FKBP7 and subsequent decrease in eIF4G level altered the global translation level (Supplementary Fig. S6B), and mimicking this reduction of eIF4G level with eIF4G knockdown equally decreased the eIF4F assembly and eIF4G level (Supplementary Fig. S6E). Therefore, FKBP7, which is upregulated in chemoresistant cells, could increase eIF4F complex activity by directly upregulating eIF4G protein levels, leading to hyperactivation of cap-dependent translation and subsequent cell survival after taxane treatment. FKBP7 stabilized the eIF4F complex formation after taxane treatment and, thus, could be a novel eIF4F regulator.

As FKBP7 regulates eIF4F complex formation, we first target the eIF4F using eIF4G-targeted siRNA before docetaxel treatment on 2-chemoresistant cell lines (with different androgen-receptor expression). EIF4G silencing provoked a marked cytotoxic effect 5 days after transfection. When combined with docetaxel, chemosensitization was further observed in docetaxel-resistant 22RV1 cells (IC₅₀ = 62 nmol/L with siNT and IC₅₀ = 23 nmol/L with si4G; Fig. 6A). We next tested 3 small eIF4F inhibitors (silvestrol, FL3, and FL23), which are reported to target eIF4F complex by inhibiting helicase eIF4A (24, 27–29). Silvestrol induced a cytotoxic effect in the A375 melanoma cell line (24), but showed no effect on IGR-CaP1–docetaxel-resistant cells (Fig. 6B), maybe because silvestrol is a Pgp drug efflux pump substrate (30), which is highly expressed in IGR-CaP1–docetaxel-resistant cells (Supplementary Fig. S7A). In contrast, docetaxel-resistant IGR-CaP1 and 22RV1 cells were sensitive to low doses of FL3 (IC₅₀ = 19 nmol/L and 14 nmol/L, respectively) and FL23 (IC₅₀ = 17 nmol/L and 7 nmol/L, respectively; Fig. 6C). Parental cells were sensitive to the same order of flavagline concentrations. Both FL3 and FL23 translation inhibitors target eIF4A, but are not Pgp substrates (31). Their IC₅₀ values were lower than those observed in melanoma A375 cells (IC₅₀ = 35 and 25 nmol/L, respectively; Supplementary Fig. S7C). Results showed that FL3 and FL23 were able to kill parental and chemoresistant cells, but no synergy was observed when they were used in combination with docetaxel in resistant cells (Supplementary Fig. S7B and S7C). We confirmed their cytotoxic effect on parental, docetaxel-resistant, and cabazitaxel-resistant cell lines, as shown by the induction of PARP cleavage (Fig. 6D). FL3 and FL23 did not lead to apoptosis of RPE-1 cells (Fig. 6D), thus reinforcing the interest of targeting eIF4A in prostate cancer.

Resistance to chemotherapy represents a major challenge (32, 33). Understanding chemoresistance mechanisms and identifying biomarkers is crucial for developing new therapeutic strategies and overcoming drug resistance. In this study, we made the unprecedented observation that FKBP7 is an important determinant of prostate cancer cell response to taxanes. The acquisition of taxane resistance after long-term treatment of prostate cancer cells with docetaxel or cabazitaxel correlates with increased FKBP7 levels. Identification of the FKBP7 signaling network allowed us to link the participation of FKBP7 in chemoresistance to the initiation step of protein translation (eIF4F translation initiation complex). Using eIF4A inhibitors at nanomolar concentrations may help to circumvent docetaxel and cabazitaxel resistance in 2 prostate cancer chemoresistant cell lines.

Several FKBP proteins have been shown to participate in cancer progression and chemoresistance (7). FKBP5 regulates steroid receptor activation and prostate cancer progression (34). Its expression levels correlate with tumor-cell sensitivity to chemotherapeutic agents (35), and it negatively regulates Akt kinase, thus increasing chemosensitivity (36). FKBP5 is also involved in resistance to anthracycline in malignant melanoma (37) and in taxol resistance in ovarian cancer cells (38). FKBP5 overexpressed after cells had been treated with microtubule-targeting agents, but this was not observed with DNA-damaging agents, such as cisplatin (38). This specificity to microtubule-targeting agents may be related to the regulatory role of FKBPs on the cytoskeletal proteins observed for FKBP4 and FKBP5 (10). Other FKBPs, such as FKBPL, are known to have therapeutic and biomarker potential in cancer (39). The implication of endoplasmic reticulum FKBPs in carcinogenesis has been reported for FKBP10 (40–42) and FKBP14 (43, 44).

Our study revealed the functional role of FKBP7 chaperone in docetaxel and cabazitaxel resistance in prostate cancer, which is observed in AR-positive and AR-negative cells, and is independent of ABCB1/MDR1 drug efflux pump expression. Our clinical data showed that high FKBP7 expression is more frequent in tumors versus normal tissues, and correlates with a lower time-to-recurrence in patients receiving taxane neoadjuvant chemotherapy. High levels of FKBP7 correlate with a bad prognosis, which suggests that FKBP7 expression could be a relevant marker of taxane resistance. Preclinical evidence in a docetaxel-resistant mouse model confirmed that FKBP7 expression sustained taxane-resistant prostate cancer cell growth. Therefore, our results implicate a previously uncharacterized member of the peptidyl-prolyl isomerase (PPIase) family in the mechanism of resistance to microtubule-targeting agents.

Although high expression of FKBP7 is observed both in noncancerous and chemoresistant cells, FKBP7 silencing triggers cell death in taxane-resistant tumor cells only, suggesting that the survival of resistant cells may be attributed to a novel function of FKBP7. The comparison of resistant and noncancerous cells in a large proteomic approach allowed us to identify eukaryotic translational initiation factor and mTOR pathways as the main FKBP7 regulation networks. These pathways are deregulated in resistant tumor cells versus normal cells. We also identified the eIF4G component of the eukaryotic initiation factor, eIF4F, as a major downstream target of FKBP7 and showed that FKBP7–eIF4G interaction controls the eIF4G protein level. Translation initiation is a highly regulated biological process hijacked by tumor cells to increase the protein synthesis rate for specific genes and promote cell survival. EIF4G is a large scaffolding protein that binds eIF4A helicase and eIF4E cap-binding protein to form the heterotrimeric eIF4F complex for mRNA translation. Evidence showed that increased eIF4F activity contributes to the malignant transformation process via increased translation of a limited set of pro-oncogenic mRNA transcripts (45). As described for the cis-trans prolyl isomerase Pin1 (46), the regulation of eIF4G level in resistant cells may be attributed to the PPIase activity of FKBP7, which may act as a dynamic switch allowing the maintenance of eIF4G in an activated but unstable conformation. This acceleration of isomerization may account for high eIF4G level in resistant cells. When FKBP7 is decreased by RNAi silencing, eIF4G is expressed at a low level but the protein seems to be more stable. This increased stability, which might be part of a feedback loop to overcome the decreased level of eIF4G, may be achieved by a basal chaperone activity of FKBP7. Thus, our data suggest that chemoresistant cell survival may be attributed to overstimulation of eIF4F, one of the downstream effectors of the Akt–mTOR pathway, mediated by FKBP7 overexpression (Supplementary Fig. S7D).

We showed that FKBP7 plays an important role in survival networks protecting cancer cells against therapeutic agents, so FKBP7 could be an interesting target in prostate cancer. No structural analysis of the full-length FKBP7 protein is currently available for the design of inhibitory molecules, and although screening strategies have been developed to discover new drugs inhibiting FKBP activity, designing isoform-specific inhibitors is still challenging (47). Thus, we will now aim our research towards the design of small-molecule ligands with FKBP7 specificity. Discovery of the FKBP7–eIF4G interaction led us to assess the efficacy of inhibitors directly targeting the eIF4F complex. We believe that targeting translation machinery is a promising strategy for minimizing acquired resistance (45).

We tested silvestrol, an eIF4A inhibitor that is known to provide therapeutic benefits in prostate cancer xenografts (28). Unfortunately, silvestrol showed no obvious efficacy towards docetaxel-resistant IGR-CaP1 cell proliferation, which could be attributed to high expression of ABCB/MDR1 in this model. In contrast, flavaglines FL3 and FL23, which also target eIF4A, are highly cytotoxic in docetaxel-resistant models. These synthetic compounds (48, 49) are known to overcome multidrug resistance in vitro (31) and alleviate resistance to vemurafemib in melanoma (24).

Our findings open a new avenue in the field of chemoresistance, and our study reveals the critical role played by chaperone FKBP7 in acquired taxane resistance in prostate cancer and its potential for development as a predictor of chemoresistance. By targeting FKBP7 or the eIF4F complex, we also found novel therapeutic strategies that would help to manage taxane-resistant prostate cancer.

Y. Loriot is a consultant/advisory board member for Astellas, AstraZeneca, Bristol-Myers Squibb, Clovis, Incyte, Janssen, MSD, Pfizer, Roche, Sanofi, and Seattle Genetics. K. Fizazi is a consultant/advisory board member for Amgen, Astellas, AstraZeneca, Bayer, Janssen, Orion, Roche, and Sanofi. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.F. Garrido, J. Camonis, F. Perez, K. Fizazi, A. Chauchereau

Development of methodology: M.F. Garrido, N.J.-P. Martin, N. Al Nakouzi, E. Del Nery, F. Perez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.F. Garrido, N.J.-P. Martin, M. Bertrand, C. Gaudin, E. Del Nery, S. Lerondel, A. Le Pape, M. Gleave

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.F. Garrido, F. Commo, E. Del Nery, Y. Loriot, K. Fizazi, A. Chauchereau

Writing, review, and/or revision of the manuscript: M.F. Garrido, N. Al Nakouzi, E. Del Nery, J. Camonis, M. Gleave, Y. Loriot, L. Désaubry, K. Fizazi, A. Chauchereau

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Compagno, L. Désaubry

Study supervision: A. Chauchereau

Other (execution of complementary experiments): N. El Kalaany

Other (pathology): L.Fazli

Other (provided scientific research and feedback): S. Vagner

We gratefully thank A. Lescure and S. Tessier, who are Biophenics staff, the platform of preclinical evaluation AMMICA, the bioinformatic Core Facility, the Development in Pathology Group (UMR981), and S. Shen (UMR981). We thank and pay tribute to Vasily Ogryzko (UMR8126), who gave us precious help with proteomics but passed away before this publication. RWPE-1 was a kind gift from G. Mouchiroud (University Claude Bernard, Lyon, Villeurbanne, France), HK-2 was provided by S. Gad-Lapiteau (UMR1186, Villejuif, France), and HUVEC was provided by S. Rodrigues-Ferreira (UMR981, Villejuif, France). We also thanks Joanna Moore, ELS (Citoxlab France), for assistance in the preparation of the manuscript. M.F. Garrido was supported by the Idex Paris-Saclay fellowship and the Association pour la Recherche sur les tumeurs de la Prostate (ARTP). N.J.-P. Martin, M. Bertrand, and N. El Kalaany were supported by the PARRAINAGE CHERCHEUR charity program of Gustave Roussy. This work was supported by grants from: INSERM, the Université Paris-Sud11, the grant PAIR PROSTATE program No. 2010-1-PRO-03 from the INCA, the ARC Foundation, the Ligue contre le cancer, the ECOS-Sud A10S03 program, AMGEN, the Paris Alliance of Cancer Research Institutes program, “Investissements d'Avenir” an initiative of the French Government implemented by ANR under reference ANR-11-PHUC-002, Taxe d'apprentissage Gustave Roussy P21NNGE for genomic and 2014MG for proteomic analyses, and Terry Fox New Frontiers Program Project Grant TFF116129.

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.