Taxane-induced attenuation of the CXCR2/BCL-2 axis sensitizes prostate cancer to platinum-based treatment.

Abstract

Background

Taxanes are the most active chemotherapy agents in metastatic castration-resistant prostate cancer (mCRPC) patients, yet resistance almost invariably occurs representing an important clinical challenge. Taxane-platinum combinations have shown clinical benefit in a subset of patients but the mechanistic basis and biomarkers remain elusive.

Objective

To identify mechanisms and response indicators for the antitumor efficacy of taxane-platinum combinations in mCRPC.

Design, setting, and participants

Transcriptomic data from a publicly available mCRPC dataset of taxane-exposed and naïve patients was analysed to identify response indicators and emerging vulnerabilities. Functional and preclinical validation was performed in taxane resistant mCRPC cell lines and genetically engineered mouse models (GEMM).

Intervention

mCRPC cells were treated with docetaxel, cisplatin, carboplatin, the CXCR2 antagonist SB265610 and the BCL-2 inhibitor Venetoclax. Gain and loss of function in culture of CXCR2 and BCL-2 was achieved by overexpression or siRNA-silencing. Preclinical assays in GEMM mice tested the anti-tumor efficacy of taxane-platinum combinations.

Outcome measurements and statistical analysis

Proliferation, apoptosis and colony assays measured drug activity in vitro. Preclinical endpoints in mice included growth, survival and histopathology. Changes in CXCR2, BCL-2 and chemokines were analysed by RT-qPCR and Western Blot. Human expression data was analysed using GSEA, hierarchical clustering and correlation studies. GraphPad Prism software, R-studio, were used for statistical and data analyses.

Results and limitations:

Transcriptomic data from taxane-exposed human mCRPC tumors correlates with a marked negative enrichment of apoptosis and inflammatory response pathways accompanied by a marked downregulation of CXCR2 and BCL-2. Mechanistically, we show that docetaxel inhibits CXCR2 and that BCL-2 downregulation occurs as a downstream effect. Further, we demonstrated in experimental models that the sensitivity to cisplatin is CXCR2 and BCL-2 dependent and that targeting them sensitizes prostate cancer (PC) cells to cisplatin. In vivo, taxane-platinum combinations are highly synergistic and previous exposure to taxanes sensitizes mCRPC tumors to second line cisplatin treatment.

Conclusions:

The hitherto unappreciated attenuation of the CXCR2/BCL-2 axis in taxane treated mCRPC patients is an acquired vulnerability with potential predictive activity for platinum-based treatments.

Patient summary:

A subset of patients with aggressive and therapy resistant PC benefits from taxane-platinum combination chemotherapy however, we lack the mechanistic understanding about how that synergistic effect occurs. Here, using patient data and preclinical models, we found that taxanes reduce cancer cell escape mechanisms to chemotherapy-induced cell death, hence turning these cells more vulnerable to additional platinum treatment.

Introduction:

Despite significant initial responses to androgen deprivation therapy, most metastatic patients progress to an incurable castration-resistant prostate cancer (mCRPC) [1, 2]. Many drugs have been approved for the treatment of mCRPC including taxanes, and androgen receptor (AR) signaling inhibitors [2, 3]. Sadly, in patients progressing to these agents very few therapeutic options are available although platinum-based treatments have demonstrated a limited benefit in patients with aggressive variant prostate cancer [4-6]. Predictive markers are needed to determine the best treatment for patients previously treated with taxanes or AR signaling inhibitors.

Taxanes bind tubulin inhibiting mitosis but also AR nuclear translocation, reducing AR signaling [7-9]. Several factors are associated with taxane resistance, including expression of β-tubulin isoforms and activation of drug efflux pumps. PTEN loss, and activation of PI3K/AKT/mTOR, MAPK and NF-kB have also been associated with taxane resistance [10, 11]. While platinum agents are not routine for the treatment of mCRPC, there is an increasing use of these agents, especially in patients with small-cell or neuroendocrine tumor variants [6]. In fact, some antitumor activity has been described for carboplatin, cisplatin, and satraplatin in mCRPC patients [12]. Unfortunately, molecular biomarkers to identify mCRPC patients that could benefit from these drug combinations remain elusive. The incomplete understanding of the molecular mechanisms of taxane resistance limits the identification of vulnerabilities and potential therapeutic targets.

Here, we sought to elucidate the mechanistic basis and response indicators for the synergistic antitumor efficacy of taxane-platinum combinations in advanced PC (Supplementary figure 1). Mining of human PC datasets to compare the transcriptomes of taxane- exposed and -naïve patients showed a marked downregulation of CXCR2 and BCL-2 in taxane-exposed patients. Mechanistically, we showed that taxanes induce CXCR2 and BCL-2 downregulation. Further, we demonstrated that CXCR2 and BCL-2 determine cisplatin sensitivity and that targeting them sensitizes PC cells to cisplatin. Finally, in vivo preclinical data testing show that taxane-platinum combinations are highly synergistic and that previous exposure to taxanes sensitizes mCRPC tumors to cisplatin. Together our data identify an acquired vulnerability in taxane-treated mCRPC patients with potential predictive value for platinum-based drug combinations.

Materials and Methods:

See Supplementary materials extended details.

Computational analysis of human prostate cancer data:

Human transcriptome data [13, 14] was used for gene expression (Supplementary table 1) and Pathway enrichment using GSEA (Supplementary table 2). Univariate analyses were performed using Spearman rank correlation tests. For bivariate analysis linear regressions were performed (Supplementary table 3). False Discovery Rate and bootstrapping were performed when appropriate to correct for multiple testing.

Functional assays in vitro:

Docetaxel resistant DU145-DR and PC3-DR human PC cells had been previously generated [15]. Docetaxel, cabazitaxel, cisplatin, carboplatin (MedChemExpress), CXCR2 antagonist SB265610 (Sigma Aldrich) and BCL-2 inhibitor Venetoclax (MedChemExpress) were prepared in DMSO. Western Blot, cell viability, apoptosis and gain and loss of function studies were performed as previously described [16]. Oligonucleotides and antibodies are listed in Supplementary Table 4.

Preclinical assays in vivo:

Animal studies were approved by the Institutional Review Board at IDIBELL. The NPK mouse model had been previously published [17]. For tumor growth and survival assays, allografted NPK tumor bearing mice were enrolled in preclinical studies.

Immunohistochemical analysis:

Immunostaining of mouse prostate tumor tissues was done as described previously [18] on formalin-fixed paraffin-embedded sections incubated with primary and secondary antibodies shown in Supplementary Table 4

Statistical analysis:

Statistical differences were calculated using non-lineal regression and F-test for viability or two-tailed Student’s t-test for clonogenic and apoptosis and considered significant < 0.05. One-way ANOVA was used to vehicle and each treatment group in in vivo preclinical assays. In survival analysis, p-values were calculated using a log-rank test. Radiographic progression was defined using Response Evaluation Criteria in Solid Tumors version 1.1. PSA decline was evaluated according to PCWG3 guidelines (PSA decline ≥50%) to the date of radiographic progression of disease. Radiographic progression-free survival (PFS) was estimated by the Kaplan-Meier method and compared using the log-rank or Fisher exact test. PSA decline was evaluated using Fisher’s exact test with a significance level of 5% was used to measure the association with PSA decline.

Results

The CXCR2/BCL-2 axis is attenuated in human PC exposed to taxane.

We investigated the transcriptomic changes in mCRPC tumors exposed to taxanes in a human PC dataset [13]. GSEA identified significantly enriched Hallmark Cancer Pathways (Supplementary figure 2A, Supplementary table 2) including “Androgen response” (NES = −2.002; FDR q-value = 8.16x10⁻⁵). Interestingly, “Apoptosis” (NES = −1.817; FDR q-value = 5.76x10⁻⁴) and “Inflammatory response” (NES = −1.933964; FDR q-value = 2.05x10⁻⁴) pathways were negatively enriched in taxane-exposed mCRPC tumors (Figure 1A, B) prompting us to hypothesize a role in platinum treatment response as we previously showed in other tumor types [16]. Comparing gene expression changes between the significantly enriched pathways showed only the “interferon alpha and gamma” pathways as marginally differentially expressed compared to “Apoptosis” and “Inflammatory response” (p = 0.017 and 0.038, respectively; Figure 1C). Moreover, the mean expression was significantly lower than that of a random model equally sized to the top differentially expressed genes (p < 0.001) between taxane naïve and exposed tumors, and similar to that of a multi-gene signature using all genes in the leading edge of the significantly enriched pathways (Supplementary figure 2B).

Figure 1. CXCR2 and BCL-2 are downregulated in human prostate cancer exposed to taxane.

(A, B) GSEA showing a significant negative enrichment for the Apoptosis (A) and Inflammatory response (B) pathways from the Hallmarks Pathways of the MSigDB in the differential gene expression signature from the taxane naïve vs. taxane exposed tumors in the SU2C dataset. (C) Density plots show the difference in the median gene expression change between the Apoptosis and Inflammatory response pathway (control) and between these two pathways and each of the top significantly enriched pathways by GSEA using non-parametric Kruskal-Wallis test. (D) Unsupervised clustering of patients according to the gene expression of the leading-edge genes from the GSEA in (B). Shown is the Fisher exact t-test p value for the association of taxane exposed status and downregulation of the signature as shown in the heatmap. (E) Linear regressions for the correlation between CXCR2 (left) and BCL-2 (right) expression and the Normalized Enrichment Score of the Apoptosis pathway. Taxane exposed and taxane naïve patients are shown separately. The p-value indicates the significance of the differential association between gene expression and pathway enrichment using T-test. (F) mRNA expression differences between taxane exposed and taxane naiïve patients in the FHCRC dataset for CXCR2 (left) and BCL-2 (right) using T-test.

Control for multiple testing was performed using False Discovery Rate adjustment on differential expression, pathway enrichment and bootstrapping for correlation analysis.

Unsupervised clustering of the GSEA leading-edge genes from the “Apoptosis” significantly segregated patients exposed to taxanes in the SU2C dataset (p = 0.004) (Figure 1D). Significantly downregulated genes included the anti-apoptotic master regulator BCL-2 (p = 0.007) as well as CXCL8 (p = 0.043) and CXCL6 (p = 0.007) (Supplementary table 1), known to bind the CXCR2 receptor and trigger anti-apoptosis programs [19]. In fact, CXCR2 and BCL-2 expression levels are significantly correlated with the negative enrichment of the “Apoptosis” pathway in taxane-exposed patients but not in taxane naïve ones (p = 0.04 and 0.030, respectively) (Figure 1E and Supplementary table 3) and associated to decreased AR and increased NEPC score (Supplementary figure 2C, D). Moreover, CXCR2 and BCL-2 mRNA levels were found significantly downregulated (p = 0.042 and 0.001, respectively) in an independent human dataset [14] (Figure 1F). Together, these data indicate that taxane treatment attenuates the anti-apoptotic signaling mediated by CXCR2 and BCL-2 in mCRPC patients.

Docetaxel induces CXCR2/BCL-2 downregulation in vitro and in vivo.

We next asked if CXCR2/BCL-2 downregulation was causally linked to taxane exposure. We observed an immediate downregulation of CXCR2 and BCL-2 in docetaxel treated DU145 and PC3 cells (Figure 2A). In agreement with the in vitro observation, tumors from the docetaxel-treated NPK mice showed a marked reduction in CXCR2 and BCL-2 compared to non-treated controls (Figure 2B). Importantly, this downregulation of CXCR2 and BCL-2 upon taxane exposure was also observed in p53 proficient LNCaP cells (Supplementary figure 3I).

Figure 2. Effect of docetaxel treatment on CXCR2/BCL-2 axis regulation.

(A) Representative western blot images (n=3) showing protein expression changes of CXCR2 and BCL-2 in DU145 and PC-3 cells after treatment with docetaxel (Doce) at 6.5 and 15 nM, respectively, for 72 hours. α-tubulin was used as endogenous control. (B) Experimental strategy: one month after tamoxifen-mediated induction of the cre-dependent recombination, tumor-bearing mice (n=4) were exposed to 4 cycles of docetaxel treatment (oral gavage, 2 mg/kg, Monday to Friday) or vehicle (top). Western blot showing CXCR2 and BCL-2 protein expression changes in tumour tissues from NPK GEMM exposed to docetaxel or vehicle. α-tubulin was used as endogenous control (bottom) (C) Experimental strategy: DU145 and PC3 cells were converted to docetaxel-resistant cells by exposing them to increasing doses of docetaxel in an intermittent regimen during 1 year and 6 months, respectively (top). Western blot analysis (n=3) of CXCR2 and BCL-2 basal protein expression in PC3/PC3-DR and DU145/DU145-DR cell lines. α-tubulin was used as endogenous control (bottom). (D) Representative western blot images (n=3) showing changes in CXCR2 and BCL-2 protein expression in PC3 and DU145 cells under negative control (siCtrl) and CXCR2 (siCXCR2) gene silencing. α-tubulin was used as endogenous control. (E) Dose response curves for DU145 and DU145-DR cells after SB265610 treatment at 0-100 μM for 72h (mean ± SEM). (F) Western blot analysis (n=3) of CXCR2 and BCL-2 protein expression changes under empty control (Crtl) and CXCR2 overexpression (CXCR2ov) in PC3-DR cells. α-tubulin was used as endogenous control. (G) Dose-response curves for PC3-DR cell line, after CXCR2 overexpression (CXCR2ov), treated with 0-50 nM docetaxel for 72 hours (mean ± SEM). (H, I) Dose-response curves for DU145-DR (H) and PC3-DR (I) cell lines, after BCL-2 overexpression (BCL-2ov), treated with 0-50 nM docetaxel for 72 hours. (J, K) Bar graph representing the percentage (mean ± SEM) of late apoptotic cells after 72h of treatment with docetaxel in DU145-DR (5 nM) (J) and PC3-DR (15 nM) (K) after BCL-2 overexpression (BCL-2ov). p-value relative to cells transfected with an empty vector (Empty Ctrl) treated with docetaxel (L) Representative colony assay images after treatment with docetaxel at 0.5 nM for 72h in DU145-DR and PC3-DR cells, after BCL-2 overexpression (BCL-2ov). (M) Bar graph representing the percentage (mean ± SEM) of colonies after 72h of treatment with docetaxel in DU145-DR (left) and PC3-DR (right) after BCL-2 overexpression (BCL-2ov). p-value relative to cells transfected with an empty vector (Empty Ctrl) treated with docetaxel.

Results were obtained from at least 3 independent biological replicates. P-values were calculated using non-lineal regression and F-test for viability or two-tailed Student’s t-test t for clonogenic and apoptosis and considered significant when < 0.05.

We next evaluated CXCR2 and BCL-2 levels in docetaxel resistant cell derivatives, namely DU145-DR and PC3-DR (Supplementary Figure 3A). Consistently, protein levels of CXCR2 and BCL-2 in untreated DU145-DR and PC3-DR cells were profoundly reduced (Figure 2C). In addition, CXCR2 ligands CXCL8 and CXCL6 were also significantly downregulated (Supplementary figure 3B), as was CXCR1, which is co-regulated with CXCR2 [20] (Supplementary figure 3C). Notably, reduced CXCR2 expression was not due to promoter hypermethylation as treatment with the 5-AZA did not increase protein expression (Supplementary figure 3C, D).

Depletion of CXCR2 or CXCL8/CXCL6 has been shown to promote BCL-2 downregulation in different tumor types [16, 21, 22]. In fact, silencing of CXCR2 in parental DU145 and PC3 cells (Supplementary figure 3E) induced a marked reduction in BCL-2 protein expression compared with control cells (Figure 2D). Additionally, DU145-DR cells, which are depleted of CXCR2, were 4-fold more resistant to CXCR2 inhibitor SB265610 than parental cells (p < 0.0001) (Figure 2E). Conversely, CXCR2 overexpression significantly sensitized PC3-DR cells to docetaxel (p = 0.019) (Figure 2F, G and Supplementary figure 3F). Similarly, BCL-2 overexpression in DU145-DR and PC3-DR cells (Supplementary figure 3G) resulted in a significant change in docetaxel sensitivity (p = 0.012 and < 0.0001, respectively) (Figure 2H, I), apoptosis (p = 0.040 and 0.021, respectively) (Figure 2J, K) and clonogenicity (p=0.021 in PC3-DR cells) (Figure 2L, M).

Collectively, these results demonstrate that docetaxel induces a downregulation of CXCR2 and BCL-2, in vitro and in vivo, and that this downregulation is maintained after docetaxel resistance is established.

Downregulation of CXCR2/BCL-2 sensitizes PC cells to platinum

Inhibition of CXCR2 signaling and its downstream effector BCL-2 increases the sensitivity of tumor cells to platinum drugs [16, 22, 23]. Based on our results indicating that docetaxel triggers a marked reduction in the CXCR2/BCL-2 axis, and the clinical benefit observed for the combination of platinums and taxanes in mCRPC [24], we hypothesized that the docetaxel-resistant PC cells sensitivity to platinum treatment might be CXCR2/BCL-2 dependent. As predicted, docetaxel-resistant DU145-DR and PC3-DR cells were significantly more sensitive to cisplatin and carboplatin that their respective parental cells (DU145-DR, p = 0.004 and <0.0001; PC3-DR p < 0.0001 and < 0.0001 for cisplatin and carboplatin, respectively) (Figure 3A, B). In agreement with the observed downregulation of anti-apoptotic signaling in the taxane treated patients of the SU2C human PC cohort, cisplatin induced a significant dose dependent increase in late apoptosis rates in the docetaxel resistant DU145-DR cells as compared with parental DU145 cell line ( p=0.041 and p=0.019 for 2.5 and 5 μM of cisplatin, respectively) (Figure 3C,D and Supplementary figure 3K, L), which was also confirmed in clonogenic assays (Figure 3E-G).

Figure 3. Docetaxel resistant cells are more sensitive to platinum treatment.

(A) Dose response curves for DU145/DU145-DR and PC3/PC3-DR cells after cisplatin treatment at 0-50 μM or carboplatin treatment at 0-75 μM for 72h (mean ± SEM). (B) Table showing cisplatin and carboplatin IC50 values indicated as mean (95% CI) in PC3/PC3-DR and DU145/DU145-DR cell lines. R.I: Resistance Index, calculated as the ratio between IC50 of resistant sublines and its corresponding sensitive cell lines. (C) Representative images of apoptosis activation after 72h-cisplatin treatment in DU145 and DU145-DR cells. (D) Bar graph representing the percentage (mean ± SEM) of late apoptotic cells after 72h of treatment with cisplatin in DU145 and DU145-DR cell lines at their corresponding IC50 doses, 2,5 and 5 μM, respectively. p-value relative to parental DU145 cells. (E) Representative colony assay images after treatment with cisplatin or carboplatin at the indicated doses for 24h in DU145/DU145-DR and PC3/PC3-DR cells. (F,G) Bar graph representing the percentage (mean ± SEM) of colonies after 24h of treatment with cisplatin or carboplatin at the indicated doses in DU145/DU145-DR (F) and PC3/PC3DR (G) cells. p-value relative to parental cells treated with cisplatin or carboplatin.

Results shown were obtained from at least 3 independent biological replicates. P-values were calculated using non-lineal regression and F-test for viability or two-tailed Student’s t-test t for clonogenic and apoptosis and considered significant when < 0.05.

To ascertain the role of CXCR2 and BCL-2 on cisplatin sensitivity, we silenced them and assessed the cytotoxicity of cisplatin. Silencing of CXCR2 (Figure 2D and Supplementary figure 3E) showed a significantly enhanced sensitivity for cisplatin compared to non-targeting siRNA control DU145 and PC3 cells (p < 0.0001 and < 0.0001, respectively) (Figure 4A). Similar results were found when BCL-2 was silenced (p = 0.016 and p = 0.0005, respectively) (Figure 4B and Supplementary figure 3H). Importantly, treatment of docetaxel sensitive DU145 cells with the CXCR2 inhibitor SB265610 significantly sensitized cells to cisplatin in cell viability (Figure 4C, D) and clonogenic assays (Figure 4E), which was confirmed in p53 proficient LNCaP cells (Supplementary figure 3J). Similarly, treatment of DU145 or PC3 cells with the BCL-2 inhibitor venetoclax significantly enhanced cisplatin sensitivity (p = 0.023 and < 0.001, respectively) (Figure 4F, G).

Figure 4. Effect of CXCR2 and BCL-2 downregulation and overexpression on platinum sensitivity in PC cells.

(A) Dose-response curves for PC3 and DU145 cell lines, after CXCR2 gene silencing (siCXCR2), treated with 0-50 μM cisplatin for 72h (mean ± SEM). siControl: cells transfected with a silencer negative transcription control. (B) Dose-response curves for PC3 and DU145 cell lines, after BCL-2 gene silencing (siBCL-2), treated with 0-50 μM cisplatin for 72h (mean ± SEM). siControl: cells transfected with a silencer negative transcription control. (C) Bar graphs representing mean ± SEM percentage of cell viability after a 72-hour treatment with cisplatin, SB265610 or their concomitant combination at the indicated doses in DU145 cells. p-value relative to cisplatin alone. (D) Dot plot representing Combination Index values calculated for each combined treatment dose. (E) Representative colony assay images (left) and bar graph (right) representing the percentage (mean ± SEM) of colonies in DU145 cells after treatment with cisplatin, SB265610 or its combination at the indicated doses. p-valuerelative to cisplatin alone. (F) Representative colony assay images (left) and bar graph (right) representing the percentage (mean ± SEM) of colonies in DU145 cells after treatment with cisplatin, Venetoclax or its combination at the indicated doses. p-value relative to cisplatin alone (G) Representative colony assay images (left) and bar graph (right) representing the percentage (mean ± SEM) of colonies in PC3 cells after treatment with cisplatin, Venetoclax or its combination at the indicated doses. p-value relative to cisplatin alone. (H) Dose-response curves for PC3-DR cells, after CXCR2 overexpression (CXCR2ov), treated with 0-50 μM cisplatin for 72h (mean ± SEM) (I) Representative colony assay images (left) and bar graph (right) representing the percentage (mean ± SEM) of colonies in PC3-DR cells, after CXCR2 overexpression (CXCR2ov), treated with cisplatin at 0.25 μM for 24h. p-value relative to Empty Control treated with cisplatin. (J) Bar graph representing the percentage (mean ± SEM) of late apoptotic cells after 72h of treatment with 10 μM cisplatin in PC3-DR cells after CXCR2 overexpression (CXCR2ov). p-value relative to Empty Control (Ctrl) cells treated with cisplatin (K) Dose-response curves for DU145-DR (top) and PC3-DR (bottom) cells, after BCL-2 overexpression (BCL2ov), treated with 0-50 μM cisplatin for 72h (mean ± SEM). (L) Representative colony assay images after treatment with cisplatin at 0.25 μM for 24h in DU145-DR and PC3-DR cells, after BCL-2 overexpression (BCL-2ov). (M) Bar graph representing the percentage (mean ± SEM) of colonies after 24h of treatment with cisplatin after BCL-2 overexpression (BCL-2ov). p-value relative to BCL-2ov cells treated with cisplatin. (N) Western blot analysis (left) and bar graph (right) of cleaved PARP protein expression changes after cisplatin treatment at the indicated doses in PC3-DR cells under empty control (Ctrl) and CXCR2 overexpression. B-actin was used as endogenous control.

All results were obtained from at least 3 independent biological replicates. P-values were calculated using non-lineal regression and F-test for viability or two-tailed Student’s t-test t for clonogenic and apoptosis and considered significant when < 0.05.

Conversely, overexpression of CXCR2 cells resulted in a significant reduction in cisplatin sensitivity (p < 0.0001), increase in clonogenic capacity (p = 0.042) and reduced apoptosis (p = 0.005) (Figure 4H-J). This was also observed upon BCL-2 overexpression in either DU145-DR or PC3-DR cells (p < 0.0001 and p < 0.0001, respectively) (Figure 4K) together with resistance to clonogenic inhibition (p = 0.048 and 0.002, respectively) (Figure 4L, M) and a dose dependent reduction in apoptosis (Figure 4N, O).

Taken together, these data indicate that docetaxel-resistant PC cells are more sensitive to cisplatin treatment, and suggest that this vulnerability is in part mediated, by the downregulation of the CXCR2/BCL-2 signaling.

Taxane treatment sensitizes a mouse CRPC model to cisplatin treatment

To test the hypothesis that platinum added to taxane may improve survival and response in aggressive mCRPC, we conducted “preclinical assay 1” in allografted NPK PC models (Figure 5A). No significant reduction in tumor growth was observed in mice treated with either docetaxel, cabazitaxel or cisplatin alone (Figure 4B) albeit a modest improve in survival on mice treated with docetaxel (Median survival 65 ± 4.9 vs. 55.5 ± 4 days, p < 0.001) and cabazitaxel (61 ± 3.14 days vs. 55.5 ± 4 days, p = 0.006) but not cisplatin (56 ± 1.7 days vs. 55.5 ± 4 days, p = 0.3) (Figure 4D). Comparable results in tumor growth inhibition were found with carboplatin (Supplementary figure 4A). Treatments were well tolerated in vivo and no signs of toxicity were identified at the indicated doses (Supplementary figure 4B). Importantly, a significant reduction in tumor growth was observed in docetaxel or cabazitaxel cisplatin combinations (p = 0.0111 and p = 0.007, respectively) and a remarkable improvement in survival [(2 of 10 reaching endpoint in docetaxel+cisplatin; median survival 75 ± 2.2 days vs. 55.5 ± 4 days, p = 1.2x10⁻¹⁰); (none reaching endpoint in cabazitaxel+cisplatin; median survival 75 ± 0 days vs. 55.5 ± 4 days p = 6.8x10⁻¹²)] (Figure 5B-D).

Figure 5. Preclinical validation.

(A) Scheme of the preclinical assays carried out. Preclinical Assay 1 assessed the anti-tumor efficacy of single agent docetaxel, cabazitaxel or cisplatin versus docetaxel plus cisplatin or cabazitaxel plus cisplatin. Single agent treated tumors from Preclinical assay 1 where shifted to cisplatin to test whether previous taxane exposure would sensitize for cisplatin treatment. (B) Tumor growth curves for NPK allografts treated with the indicated drugs or vehicle for 75 days. ANOVA is used to assess the significance of the differences in tumor growth. p-value relative to vehicle condition. (C) Tumor growth curves for single docetaxel, cabazitaxel or vehicle treated tumors from (B) under cisplatin treatment for 45 days. ANOVA is used to assess the significance of the differences in tumor growth. p-value relative to pre-treated tumors with vehicle in (B) . (D-E) Kaplan Meier survival curves for mice enrolled on Preclinical assay 1 (D) or Preclinical assay 2 (E). (F) Kaplan-Meier analysis for radiographic Progression-Free Survival on platinum-based chemotherapy treated patients in the Beltran dataset according to their BCL-2 expression by RNAseq (n=21). Curves where compared using the log-rank test with p-value considered significant when < 0.05 (G) Waterfall plot (n=8) showing the magnitude of PSA decline in the subset of mCRPC adenocarcinomas within the cohort according to their BCL-2 expression. Fisher’s exact test with a significance level of 5% was used to measure the association between Bcl2 expression and PSA decline

We next asked whether previous exposure to taxane sensitizes otherwise resistant tumors to cisplatin treatment. We engrafted mice with control, docetaxel-treated and cabazitaxel-treated NPK tumors from “preclinical assay 1” and treated them with cisplatin in “Preclinical assay 2” (Figure 5C). Notably, sequential cisplatin post docetaxel or cabazitaxel significantly reduced tumor growth (p = 0.038 and 0.029, respectively) and improved survival [(3 of 10 reaching endpoint in cisplatin post docetaxel; median survival 40.5 ± 5.5 days vs. 32.5 ± 2.3 days, p = 0.00491); (5 of 10 reaching endpoint in cisplatin post cabazitaxel; median survival 40.5 ± 4.8 days vs. 32.5 ± 2.3 days p = 0.0019)] (Figure 5E).

We next interrogated a unique retrospective cohort of mCRPC patients who had undergone platinum-based treatment, namely the Beltran dataset [25]. First, we used the SU2C dataset to stratify patients in high vs low markers expression. Given the high significant correlation between BCL-2 and CXCR2 expression (Supplementary Figure 4C) and wide range of gene expression, we used BCL-2 z-scores to define the high and low expression groups (Supplementary Figure 4D,E; z-scores < 0.5 as Low BCL-2 and z-scores > 0.5 as High BCL-2). Interrogation of the Beltran dataset showed a clear trend towards better radiographic Progression Free Survival in patients with low BCL-2 expression compared to high expression (Figure 5F; H.R.=3.41, 95% CI 0.74 - 15.67 , p = 0.09) together with improved PSA responses among the CRPC Adenocarcinoma patients (n = 8; p = 0.07) (Figure 5G). Interestingly, despite this trend was not found using CXCR2 expression alone, the combined use of CXCR2 and BCL-2 showed the best trend (p = 0.059) (Supplementary Figure 4F, G).

Analysis of tumor specimens (Figure 6A, D and Supplementary Figure 5A-C) confirmed that the improved antitumor activity of the platinum-based combinations compared to single taxane was associated with a significant reduction of cell proliferation (p = 0.0027 and 0.0012 for docetaxel and cabazitaxel, respectively) (Figure 6B) and antiapoptotic BCL-2 (p = 0.019 and 0.010) and CXCR2 (Figure 6C and Supplementary Figure 5A), together with a marked increase in apoptosis (p < 0.001 and p = 0.004) (Figure 6E). and reduced AR (p < 0.001 and p = 0.006) in combination treatments (Figure 6F). Notably, sequential cisplatin post taxane also showed a marked reduction in proliferation and antiapoptotic BCL-2 (Supplementary Figure 5B).

Figure 6. Histopathological analysis of preclinical cohorts.

(A) Haematoxylin-Eosin staining with Immunohistochemistry for AR and Ki-67 in tumors from Preclinical assay 1. Microphotogaphs are shown at 400X magnification and representative of 5 different mouse tumors. Insets correspond to 63X magnification. Quantification of (B) Ki-67 and (C) BCL-2. (D) Representative immunohistochemistry and immunofluorescence microphotographs for BCL-2 and cleaved PARP. (E) Quantification of cleaved PARP staining and (F) nuclear A. Shown is the percent of positive nuclear stain over total number of cells in 5 different 40X magnifications fields from at least 3 independent mouse tumors. p-value relative to docetaxel or cabazitaxel treatment alone. P-values were calculated using two-tailed Student’s t-test t and considered significant when < 0.05.

Taken together, these data strongly suggest that the attenuation of the CXCR2/BCL-2 axis represents a vulnerability in a subset of advanced PC patients and a potential predictive determinant of response to platinum-based chemotherapy that needs to be prospectively validated.

Discussion:

Taxane-platinum combinations have shown clinical benefit for a subgroup of mCRPC patients [6], highlighting the need for predictive response indicators. Exploiting publicly available datasets, we have found that in taxane-exposed PC tumors, apoptosis and inflammatory response pathways are dysregulated and that the CXCR2/ BCL-2 axis is markedly reduced. In vitro and in vivo assays showed that loss of CXCR2 and BCL-2 represent an emerging vulnerability to genotoxic platinum-based treatments. Accordingly, interrogation of a cohort of platinum treated mCRPC patients suggests an association with improved clinical response. Finally, in agreement with other studies [26], our preclinical trials in vivo confirm that taxane exposure sensitizes PC cells to platinum.

In addition to stabilizing microtubules, taxanes induce apoptosis by inhibiting antiapoptotic proteins, such as BCL-2 [27]. Accordingly, following treatment with paclitaxel, BCL-2 is phosphorylated in PC3 and LnCaP cells, inhibiting its anti-apoptotic action and enhancing cell death [28, 29]. Consistent with our results, BCL-2 downregulation after docetaxel treatment has been shown in breast cancer cells [30]. Downregulation of a drug’s target and signaling rewiring to elicit pro-survival pathways is a well-known mechanism of cancer treatment resistance [31]. Therefore, our data suggest that mCRPC patients lose BCL-2 expression, at least in part, through CXCR2 downregulation as a mechanism to evade docetaxel-induced apoptosis. In agreement with previous data [16, 22, 23], we demonstrated that inhibition of the CXCR2 chemokine receptor markedly increased the sensitivity of AR-dependent and -independent PC cells to cisplatin treatment, suggesting that the inverse relationship between docetaxel- and cisplatin-resistance could be mediated, at least in part, by CXCR2 downregulation after taxane exposure. Conversely, other studies also demonstrated that when cells become resistant to platinum they often become sensitive to taxanes [32, 33].

In line with findings from a recent Phase I/II trial [24], our in vivo preclinical results in an AR-indifferent PC model suggest that therapeutic responses to cisplatin-based chemotherapy in docetaxel-resistant mCRPC patients might be predicted by CXCR2/BCL-2 expression. While we show that taxane consistently reduces CXCR2 and BCL2, only a subset of patients is expected to respond to taxane-platinum combinations, likely reflecting the inherent limitation of preclinical models in recapitulating the spectrum of heterogeneity. Finally, clinical qualification of CXCR2 and/or BCL-2 as a predictive biomarker will first require determining the most suitable analytical test and biospecimen. Our data using transcriptomics from the SU2C cohort to define expression thresholds subsequently used in the retrospective Beltran dataset illustrate the feasibility of this general approach to estimate the subset of patients expected to benefit from platinum-based combinations. Ultimately, cut-off thresholds to classify patients as low or high BCL-2 or CXCR2 expression and their association to treatment response will need to be tested and validated in independent prospective trials.

Conclusions:

The CXCR2/BCL-2 anti-apoptotic axis is markedly reduced in taxane exposed mCRPC tumors representing an emerging vulnerability to genotoxic platinum-based treatments. Therefore, taxane-resistant tumors are sensitized to cisplatin treatment and taxane-cisplatin combinations have anti-tumor efficacy in aggressive mCRPC. Our findings also suggest that CXCR2/BCL-2 expression levels might predict therapeutic response to cisplatin-based treatment in docetaxel-resistant mCRPC patients.

Take home message:

CXCR2/BCL-2 anti-apoptotic axis is markedly reduced in taxane exposed mCRPC tumors representing an emerging vulnerability with potential predictive value for platinum-based treatments. Therefore, taxane resistant tumors are sensitized to cisplatin treatment and taxane-cisplatin combinations have antitumor efficacy in aggressive mCRPC mice models.