CDK4/6 Inhibitor as a Novel Therapeutic Approach for Advanced Bladder Cancer Independently of RB1 Status

Bladder cancer is the most common malignancy of the urinary tract (1). At diagnosis, two major classes of bladder cancer are distinguished (2): approximately 75% of patients present a non-muscle–invasive disease (NMIBC, stage Ta, T1 or CIS), while the rest of the patients (25%) show a tumor already invading the muscle layers (MIBC stage T2 or higher). NMIBC is considered a relatively indolent tumor, and treated by transurethral resection (3). In the case of high-risk tumors, this can be followed by intravesical instillation with Bacillus Calmette-Guérin or mitomycin (3). Nevertheless, NMIBC displays one of the highest recurrence rates among all cancers, and a significant part of these recurrences shows tumor progression toward MIBC (4). The treatment of MIBC includes a radical cystectomy usually followed by cisplatin (CDDP)-based chemotherapy (5). Unfortunately, the disease becomes metastatic in a high proportion of the cases (50%–70%), leading to extremely low survival rates (5, 6). This scenario aggravates when considering the elevated mean age of patients at diagnosis, associated to frequent and severe comorbidities in almost 50% of patients with MIBC (6). Such patients, often called “unfit”, are not candidates to CDDP treatment having few therapeutic options (6). In spite of multiple trials aimed to develop new therapeutic options for patients with bladder cancer, few improvements have occurred in the last decades with the exception of immune checkpoint inhibitors. In this regard, although these inhibitors have demonstrated promising results in some cases, the proportion of patients who shows objective responses to these therapies is still very limited (20%–35%; refs. 7–9). Consequently, there is an urgent need of new avenues for the management of advanced bladder cancer.

The molecular portrait of bladder cancer has identified multiple alterations that could be actionable through molecularly targeted therapies (10, 11). Among them, the RB pathway is predominantly altered (10, 11). This can result from the inactivation of RB1 gene itself, or by mutation or amplification of different genes whose products mediate the functional inactivation of the pRb protein (10, 11). On the basis of this, drugs targeting these proteins are attractive therapeutic tools. Among them, CDK4/6 inhibitors are being tested in a large number of solid malignancies characterized by the presence of wt RB1 alleles and/or cyclin D or CDK4/6 amplification. In fact, they have been recently approved for treating hormone receptor–positive breast tumors in combination with compounds targeting ER-dependent signaling (12).

Here, we report the preclinical analysis of a CDK4/6 inhibitor in bladder cancer. Our current data show that this inhibitor exerts relevant antitumor effects in vitro and in vivo, affecting not only RB1 wt but also RB1-mutant tumors. In both cases, we find that such activities also rely on the inhibition of the FOXM1 phosphorylation and activation. In addition, treatment with CDK4/6 inhibitor sensitizes the cells to CDDP allowing a significant reduction in its effective dose with a consequent reduction in side effects. Therefore, these results collectively open a new avenue of possible bladder cancer therapy, especially for CDDP-unfit patients.

The bladder cancer cell lines (RT112, J82, 253J, 5637, UM-UC-1 and RT4), with known genomic characteristics (13) and validated by short tandem repeat typing, were maintained in DMEM GlutaMAX (Gibco-BRL Life Technologies) with 10% FBS (Hyclone) and 1% antibiotic–antimycotic (Gibco-BRL Life Technologies) at 37°C in a humidified atmosphere of 5% CO₂. Palbociclib isethionate (provided by Pfizer through a Compound Transfer Program grant) was dissolved in PBS. CDDP (Selleckchem) was dissolved in PBS (Sigma). Cell viability and cell-cycle assays are described in detail in Supplementary Material and Methods.

All the animal experiments were conducted in compliance with CIEMAT guidelines, and approved by the Animal Welfare Department of the Comunidad de Madrid (Spain; PROEX 088-15, PROEX 183/15). Palbociclib and CDDP concentrations were chosen based on previously published data (14, 15). After 15–30 days of treatment, mice were sacrificed and tumors were collected and processed. The transgenic mice were generated by breeding Rb1^F/F;Rbl1^−/− (16) and Trp53^F/F;Pten^F/F (17) mice. Adenovirus expressing Cre recombinase under keratin K5 promoter (18) was obtained from the Viral Vector Production Unit of the Autónoma University of Barcelona and surgically delivered into the bladder lumen as described previously (19, 20). Tumor appearance was routinely followed by visual inspection, hematuria, and abdominal palpation. At the time of sacrifice, tissues were collected and processed as reported previously (16, 21, 22). To test the effectiveness of CDDP plus palbociclib, tumor-bearing mice were treated for 28 days with vehicle (n = 20) or with daily palbociclib plus a once per week dose of CDDP (n = 20). Males and females were randomly adjusted to 50% on each group.

IHC and immunoblots were performed essentially as described previously (20). Antibodies used are listed in Supplementary Table S1 (more information in Supplementary Material and Methods).

All studies involving patients were conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. All patients provided written informed consent before study entry. For the University Hospital “12 de Octubre” cohort (Spain), tumor and nontumoral paired samples and medical records were analyzed from 87 patients (pathologic and clinical data are shown in Supplementary Table S2). Informed consent was obtained from all the patients and the study was approved by the Ethical Committee for Clinical Research of University Hospital 12 de Octubre (IRB ref. 10/050; refs. 23, 24). Resources from the EPICURO Study including 832 newly diagnosed urothelial bladder cancer (UBC) cases aged 22–80 years with a median follow-up of 70.7 months (range 0.7–117.7 months), and available tumor tissue were used to replicate previous results on FOXM1-P prognostic value. Informed consent was obtained from study participants in accordance with the Institutional Review Board of the Ethics Committees of participating hospitals that approved the study (IRB Hospital del Mar, ref. 2008/3296/1). More information is provided in Supplementary Materials and Methods.

Total RNA was isolated using miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions and DNA was eliminated (Rnase-Free Dnase Set, Qiagen). Reverse transcription was performed using the Omniscript RT Kit (Qiagen); TBP was used as reference gene using 50 ng of total RNA and specific primers (Supplementary Table S3). PCR was performed in a 7500 Fast Real Time PCR System using GoTaq PCR master mix (Promega). Melting curves were performed to verify specificity and absence of primer dimers. Reaction efficiency was calculated for each primer combination.

Control (PBS-treated) and palbociclib-treated (IC₅₀) RT112, J82 and 5637 cells, were analyzed using the GeneChip Human Transcriptome array (HTA) V2 Arrays (Affymetrix). Data are deposited in GEO (GSE105402). In the case of transgenic mouse samples, microdissections of the tumors were carried out before RNA extraction to avoid any possible contamination. Total RNA was extracted using miRNeasy FFPE kit (Qiagen). cDNAs from 12 ng of total RNA were generated, fragmented, biotinylated, and hybridized to the GeneChip Mouse Transcriptome Array 1.0 (MTA 1.0.), now known as Clarion D Assay Mouse (Affymetrix, Thermo Fisher Scientific). Data are deposited in GEO (GSE100716). The protocol followed and the normalization procedures for both datasets are described in the Supplementary Information.

Comparisons between groups were made using the Wilcoxon–Mann–Whitney test. Survival analyses (recurrence free or tumor progression in recurrence) according to various variables were performed using the Kaplan–Meier method and statistical differences between the patient groups were tested by the log-rank test. Contingency analyses were made using the Fisher exact test. Tumor growth differences were determined by ANOVA followed by Bonferroni multiple comparison tests. Significance of correlations was determined by Spearman rank order test. Discrimination between samples showing increased or decreased tumor/normal relative expression of either gene or miRNA expression was made using the median. SPSS 17.0 and GraphPad Prism 6.0 software were used.

Six different bladder cancer cell lines of known genomic characteristics [5637 and J82 (RB1 mutant) and RT112, 253J, UM-UC-1 and RT4 (RB1 wt)], and representative of various tumor stages and bladder cancer subtypes (13), displayed a similar sensitivity to CDK4/6 inhibitor in a range of concentrations similar to those reported for other tumor cells (ref. 25; Supplementary Fig. S1). We observed no significant differences in their IC₅₀ values depending on the RB1 status by XTT analyses (Fig. 1A). These findings were also supported by the analysis of the Cancer Cell line Encyclopedia collection (https://portals.broadinstitute.org/ccle) on urothelial cells or in small-cell lung cancer (SCLC, where RB1 mutations are prominent; Supplementary Fig. S2). We also observed that treated cells presented differences in their cell-cycle profiles: while RB1 wt cells had a marked reduction in the percentage of S-phase entry with cells arresting in G₁, RB1-mutant cells displayed a prominent reduction of G₂–M phase entry and a significant induction of apoptosis (sub-G₁ cells; Fig. 1B). Immunoblot studies showed that palbociclib caused inhibition of pRb phosphorylation in RB1 wt cells (Fig. 1C). In contrast, in RB1-mutant cells, besides the expected pRb absence, a reduction of Cyclin B and E2F3a expression was found (Fig. 1C). Finally, p27 induction was detected in all cells regardless their RB1 status (Fig. 1C). Tumor xenograft experiments revealed that CDK4/6 inhibition exerts similar activities in vivo, independently of RB1 status (Fig. 1D). IHC analyses showed a reduction of Ki67 in RB1 wt cell–derived tumors (Supplementary Fig. S3), in agreement with the in vitro cell-cycle profile data (Fig. 1B), whereas the decrease of phosphorylated histone H3 (pH3, indicative of mitosis) was more evident in RB1-mutant cell-derived tumors (Supplementary Fig. S3). Apoptosis, monitored by active caspase-3 expression (Supplementary Fig. S3), was only induced in RB1-mutant cell-derived tumors. As expected, the inhibition of pRb phosphorylation was only observed in tumors derived from RB1 wt cells (Supplementary Fig. S3).

To analyze the possible mechanism underlying the above described results, whole-transcriptome analyses were performed in RT112 (RB1 wt), J82 and 5637 (RB1 mutant) cells, upon treatment with palbociclib, finding prominent gene expression changes (Fig. 2A; Supplementary Table S4). A poor overlap of up- and downregulated genes was found among the three lines (Fig. 2B). In accordance, Gene Ontology analyses revealed few common functions affected by CDK4/6 inhibition (Supplementary Fig. S4). The putative binding motif enrichment from ChIP-Seq databases using Enrichr webtool (26) also revealed disparities among the deregulated genes in RB1 wt versus mutant cells (Supplementary Fig. S4). Interestingly, a significant fraction of the downregulated transcripts in both RB1 wt and RB1 mut cells showed binding sites for FOXM1 transcription factor (Fig. 2C), whereas those bound by E2F proteins appeared more significantly involved in RB1 wt cells (Fig. 2C). As CDK4 activates FOXM1 through phosphorylation in various residues including Thr600 (27), we analyzed the expression of the phosphorylated Thr600-FOXM1 in parallel with other oncogenic kinases upon palbociclib treatment (27). We observed a significant reduction of phosphorylated Thr600-FOXM1 in all cells, without significant changes in phospho-AKT, phospho-ERK1/2, or total FOXM1 levels (Fig. 2D). To validate this finding in vivo, we analyzed the expression of phosphorylated Thr600-FOXM1 in the xenografts derived from RB1 wt and mutant cells. In both cases, a significant inhibition was observed as a consequence of CDK4 inhibitor treatment in vivo (Fig. 2E). This was further supported by the reduction of phosphorylated Thr600-FOXM1 levels upon CDK4 knockdown (Fig. 2F) in RT112 and 5637 cells. On the other hand, CDK2 knockdown increased FOXM1 and phosphorylated Thr600-FOXM1 in RT112 (Fig. 2F′), whereas the moderate decrease in CDK2 levels decreased the phosphorylated Thr600-FOXM1 in 5637 cells (Fig. 2F′).

To analyze whether sensitivity to CDK4/6 inhibitor is dependent on FOXM1, we performed transfection experiments. We found that increased levels of FOXM1, which were accompanied with its increased phosphorylation, conferred higher sensitivity to CDK4/6 inhibitor, whereas the reduction of FOXM1 promoted resistance only in RB1-mutant cells (Supplementary Fig. S5). This is in line with FOXM1 as potential surrogate target of CDK4/6 inhibition, as the reduction or increased levels of a certain target may define resistance or sensitivity, respectively, to determined antitumor compound. In support of this, the levels of FOXM1 mRNA inversely correlated with sensitivity to palbociclib in RB1-mutant bladder cancer and SCLC (Supplementary Fig. S6). These results indicate that, in the absence of mutations, pRb is a primary determinant of palbociclib sensitivity. Then, we decided to determine whether reduction of FOXM1 could affect tumor growth in vivo. Xenograft experiments using parental or FOXM1-knockdown-RT112 (RB1 wt) and FOXM1-knockdown-5637 cells (RB1 mut) confirmed a significant inhibition of tumor growth as consequence of FOXM1 level reduction (Supplementary Fig. S5). Of note, we observed that the knockdown of FOXM1 induced features of senescence in RB1 wt cells, whereas it also promoted massive apoptosis in RB1-mutant cells (Supplementary Fig. S5), which is in agreement with the observations upon palbociclib treatment (Fig. 1 and data not shown).

On the basis of recent data indicating that increased expression or activity of FOXM1 can confer platinum resistance (28), we decided to check whether CDK4/6 inhibition may affect the response to CDDP. We performed sensitivity curves of CDDP alone or in the presence of palbociclib. A dramatic increase in CDDP sensitivity was observed in all palbociclib-treated cell lines (Supplementary Fig. S7). In fact, cooperativity studies using the Chou–Talalay method (29) showed a systematic synergy between CDDP and palbociclib, as demonstrated by the combination index < 1 in all cell lines (Fig. 3A). Importantly, the overexpression or knockdown of FOXM1 produced increased resistance or sensitivity to CDDP in RT112 and 5637 cells, respectively (Supplementary Fig. S8A). On the other hand, the knockdown of CDK4 had no effect on CDDP sensitivity, whereas it conferred resistance to palbociclib (Supplementary Fig. S8B). Given that the reduction of CDK4 levels accounted for decreased FOXM1 and phosphorylated FOXM1 levels (Fig. 2), these results suggest that the absence of CDK4, but not the inhibition of its kinase activity, might induce a FOXM1-independent mechanism that allows retaining CDDP sensitivity.

We next studied whether the synergy between CDDP and palbociclib could be affected by different FOXM1 levels. Therefore, we performed cooperative analyses using RT112 or 5637 cells upon overexpression or knockdown of FOXM1. The results (Supplementary Fig. S9) indicated that the synergy between palbociclib and CDDP is maintained irrespective of FOXM1 altered expression in 5637 cells, whereas the knockdown of FOXM1 accounted only for additive effect in RT112 cell line. These results can be explained by the altered independent sensitivity to CDDP or palbociclib promoted by alterations in FOXM1 levels, suggesting that increased resistance to CDDP promoted by increased expression of FOXM1 is counteracted by the increased sensitivity to palbociclib in RB wt and RB-mutant cells. On the other hand, in Rb wt, but not in Rb-mutant cells, the reduced FOXM1 levels accounted for a minor effect probably because the sensitivity to palbociclib is predominantly dictated by pRb phosphorylation in these cells (see also Supplementary Fig. S2A).

To confirm the synergy between palbociclib and CDDP shown in vitro, we treated xenografts from RT112 and 5637 cells with palbociclib, CDDP, or CDDP plus palbociclib finding that, while treatment with CDDP had no significant effects on tumor growth (Fig. 3B; Supplementary Fig. S9), the combined treatment produced a dramatic reduction on tumor growth (Fig. 3B), and in most cases evident tumor regression (Supplementary Fig. S10A and S10A′), both in RB1 wt and RB1-mutant cell-derived tumors (Fig. 3B). The histology of these tumors also showed a severe reduction in cellularity upon the combined treatment (Fig. 3C). Immunoblots revealed a significant reduction in phosphorylated FOXM1 in the tumors derived from both cell lines after palbociclib or palbociclib + CDDP treatments (Fig. 3D). IHC studies not only supported this FOXM1 phosphorylation reduction, but also demonstrated a substantial inhibition of proliferation and a large induction of apoptosis as a consequence of the combined treatment (Supplementary Fig. S10).

To explore whether palbociclib can be active in bladder cancer in a more physiologic in vivo context, we generated a novel metastatic bladder cancer transgenic mouse model. We used a previously validated model of invasive bladder cancer in which Pten and Trp53 tumor suppressor gene loss is induced by instillation of adenoviruses coding for Cre recombinase (19). To selectively induce recombination in urothelial epithelial cells, we replaced the general CMV promoter by the regulatory elements of keratin K5 (AdK5Cre). This drives the expression of the recombinase exclusively to basal urothelial cells as assessed by ROSA26 reporter analyses (not shown). In addition, we introduced Rb1^F/F; Rbl1^−/− alleles to induce Rb deficiency, and to circumvent possible compensatory mechanisms (30), obtaining quadruple knockout mice (QKO): Pten^F/F; Trp53^F/F; Rb1^F/F; Rbl1^−/−. After 90–120 days of the intravesical AdK5Cre inoculation, all the mice presented tumor lesions (Fig. 4A), characterized as highly aggressive urothelial tumors that invaded the muscle layer (Fig. 4B) and surrounding organs, even developing metastasis to the lungs (Fig. 4C), liver (not shown), and peritoneal cavity (Fig. 4C′).

The whole transcriptome (Supplementary Table S5) characterization of the primary tumors and their metastases compared with control bladder (siblings noninoculated with AdK5Cre) by Gene Set Enrichment Analysis (GSEA; ref. 31) displayed an increased expression of E2F and MYC target genes, induction of epithelial–mesenchymal transition, angiogenesis, and IFNγ response genes, as well as a decreased expression of Hedgehog signaling genes (Supplementary Fig. S11A). Moreover, analyses using Enrich web tool revealed upregulated cell cycle, DNA repair/replication, and chromatin-remodeling genes, primarily bound by E2F, MYC, and, remarkably, FOXM1 (Supplementary Fig. S11B, S11C′, and S11D′). The downregulated genes in these analyses were primarily involved in muscle organization (according to the invasive behavior of tumor cells reducing muscle component), VEGF signaling, negative regulation of intracellular signaling, and extracellular matrix organization (Supplementary Fig. S11B and S11C), and presented binding to various transcriptional regulators including Polycomb members (Supplementary Fig. S11D), which is in agreement with previous findings of the oncogenic activities of these chromatin-remodeling processes in bladder cancer (reviewed in ref. 32). Finally, we observed a significant overlap of the genes upregulated and downregulated with various series of infiltrating human bladder cancer datasets (Oncomine database), also with poor clinical outcome and RB1 mutation (Supplementary Table S6).

Given the characteristics of the mouse model and its similarities with advanced human bladder cancer samples, we decided to explore the effectiveness of the palbociclib plus CDDP combined treatment. We observed a significantly increased survival in the palbociclib + CDDP cohort (Fig. 4A). Necropsies showed the presence of bladder tumors (Fig. 4B, B′) and metastasis (Fig. 4C, C′, and D). However, whereas in the control group, all mice displayed aggressive invasive tumors as commented above (Fig. 4B), in the treatment group a large fraction of the animals showed only tumor remnants with reduced invasion of the surrounding muscle layer (Fig. 4B′), and a severely decreased metastatic burden (Fig. 4A). In the treated group, we further observed clear signs of tumor regression (Supplementary Fig. S12) with necrotic areas and massive lymphocyte infiltrates, which also affected the metastases (Fig. 4D; Supplementary Fig. S12). IHC studies revealed significant inhibition of proliferation, as demonstrated by the impaired BrdUrd incorporation and reduced pH3 (Fig. 4E and F), accompanied by apoptosis induction (Fig. 4G). Similar analyses also showed a significant inhibition of FOXM1 phosphorylation (Fig. 4H and M). There were no major effects on AKT, STAT3, or ERK activation in the treated tumors (Fig. 4I, J, K, and M). In contrast, the expression of EZH2 was reduced (Fig. 4L and M).

In the genomic characterization of these QKOs, unsupervised analysis showed that treated tumors clustered with control bladder samples. We also observed strong transcriptional changes as a consequence of the treatment (Fig. 4N). GSEA analyses revealed a reduced activation of E2F and MYC target genes in the treated tumors (Supplementary Fig. S13), along with G₂–M checkpoint and genes induced by VEGF (Supplementary Fig. S13). Moreover, we observed inhibition of K-Ras–dependent genes in treated tumors (Supplementary Fig. S13) and increased xenobiotic genes, according to the expected metabolism of antitumor drugs (Supplementary Fig. S13). Although no significant involvement of transcriptional regulators or gene ontologies were found in the case of upregulated transcripts, a preferential regulation by FOXM1 and E2F transcription factor was detected for the downregulated genes, along with reduced histone marks associated with transcriptional activation (Fig. 4O). These downregulated transcripts were primarily involved in cell cycle, telomere maintenance, chromosome organization, and DNA repair (Fig. 4P). Moreover, they presented a statistically significant overlap with the genes displaying reduced expression in a variety of human cancer cell lines upon treatment with palbociclib (Fig. 4Q).

The above results prompted us to analyze the possible relevance of FOXM1 in human bladder cancer. First, we studied gene expression data from a previous NMIBC-enriched patient series (Supplementary Table S2; ref. 23). The upregulated genes in primary tumors displaying recurrences (Fig. 5A) were typified by the binding of FOXM1, E2F, and MYC (Fig. 5B). They also displayed a significant overlap with those genes downregulated by palbociclib treatment on multiple cell lines representative of different cancer types (Fig. 5C). Remarkably, when samples were classified in an unsupervised manner according to the expression of FOXM1-bound genes (33, 34), a clear discrimination between normal and tumor samples was found and it discriminated those primary tumors corresponding to patients that suffered disease recurrence (Fig. 5D). The expression of the FOXM1 gene, in larger series analyzed by qRT-PCR, confirmed an increased expression in tumor tissues (Fig. 5E). Among tumors, increased FOXM1 expression was found in high-grade samples (Fig. 5G) and in those tumors of patients showing recurrence (Fig. 5H) or progression in recurrence (Fig. 5I). However, no differences were found between low-stage tumors (Ta–T1; Fig. 5F). By IHC (23), we observed that the expression of phosphorylated Thr600-FOXM1–discriminated tumors from patients showing early recurrence (Fig. 5J and K).

To further support these findings, we used a large epidemiologic series of primary bladder cancer encompassing the full spectrum of the disease (EPICURO study; ref. 35). The expression of phosphorylated Thr600-FOXM1 was found to be significantly higher in high-stage and high-grade tumors (Supplementary Table S7). Moreover, patients whose tumors showed high p-FOXM1 expression had a lower progression-free survival, although the disease-specific survival (DSS) was not statistically significantly different (Supplementary Table S8). In multivariate analysis adjusted for age, gender, stage, and grade, we also observed (Supplementary Table S7) that high p-FOXM1 was significantly associated with a higher risk of progression. In the stratified analysis by risk groups, the increase in risk of progression and DSS was restricted to the low-risk NMIBC (HR = 3.9; 95% CI, 1.6–9.31 and HR = 3.1; 95% CI, 0.9–9.6, respectively). In contrast, no differences in risk of progression and DSS were seen in patients with high-risk NMIBC and MIBC, suggesting that additional mechanisms may contribute to tumor aggressiveness in these disease groups.

We next used The Cancer Genome Atlas (TCGA) database, which is composed exclusively of MIBC (10). We found that high FOXM1 transcript levels were highly significantly associated with shorter disease-specific survival (Fig. 5L). Finally, we analyzed whether different TCGA bladder cancer subtypes displayed increased activity of FOXM1 through GSEA of those genes previously identified as FOXM1 targets by ChIP-seq (33, 34). This analysis revealed that TCGA groups III and IV displayed a significant positive enrichment, whereas group II showed negative enrichment (Fig. 5M). This observation was further supported by a supervised classification of TCGA subgroups by the expression of these same genes, showing increased expression in subgroups III and IV, and decreased expression in subgroup II (Fig. 5N). Finally, we monitored possible regulatory mechanisms between RB1, CDK4, and FOXM1 in bladder cancer. Using TCGA data, we found that upregulation of CDK4 and CDK6 mRNA tended to cooccur with the upregulation of FOXM1, which in turn was mutually exclusive with the upregulation of RB1 (Supplementary Fig. S14A). Overall, altered expression of these genes was associated with reduced DSS (Supplementary Fig. S14B). Similarly, using proteomic data from TCGA, we observed a significant negative correlation between FOXM1 and CyclinD1 or RB protein expression (Supplementary Fig. S14C), whereas no significant correlation was observed between FOXM1 and phosphorylated RB1 (Supplementary Fig. S14C).

Our analyses indicated a possible inverse regulation between RB and FOXM1. In agreement, the knockdown of RB1 gene in RT112 cells produced a significant increase in FOXM1 expression accompanied with increased CDK4 and Cyclin D3, and decreased Cyclin D1 (Fig. 6A). Because these changes were parallel to increased phosphorylated FOXM1 (Fig. 6A), these observations suggest that activation of FOXM1 might rely on CDK4/6/Cyclin D3 complexes in these cells. Remarkably, the sensitivity to palbociclib is not significantly affected in these cells by the reduction of Rb levels (Fig. 6B), which is in concordance with our above mentioned data.

Bladder cancer represents an important problem for society and health systems due to its increased incidence and prevalence. Moreover, it primarily affects elderly people with very limited therapeutic options. Immune checkpoint inhibitors are the only major improvements since the 1970s, but only a reduced percentage (20%–35%) of patients display objective benefit. In addition, although CDDP-based chemotherapy is effective in advanced bladder carcinoma, the improvement of DSS of these patients is of marginal relevance, as most patients with metastases die from the disease. Here, we present data supporting the basis for a novel approach to treat advanced bladder cancer: palbociclib, a CDK4/6 inhibitor, holds promise as a therapy in bladder cancer.

The rationale for using CDK4/6 inhibitors in the management of solid tumors has been usually associated with a wt RB1 gene status (12). Accordingly, the common functional inactivation of pRb, which resides in the activity of cyclin D–CDK4/6 complexes, is abrogated by the inhibitor leading to the proliferation arrest. CDK4/6 inhibitors have also gained clinical relevance because they can synergize with other targeted compounds, even being able of overcoming resistance to such treatments (36). Our results indicate that CDK4/6 inhibitor is active (in vitro and in vivo) in bladder cancer cells regardless their RB1 status, although the effects on expression of cell-cycle components and dynamics were different in RB1 wt and mutant cells. In this way, while RB1 wt cells displayed a predominant cell-cycle arrest in G1, RB1-mutant cells showed a reduced entry into G₂–M. To our knowledge, there are few examples of palbociclib antitumor effects regardless of RB1 status. For instance, in ovarian cancer cells, palbociclib exerts antiproliferative effects irrespective of RB1 status and cooperates with paclitaxel to induce cell death (37). In the case of RB1-mutated hepatoma cells, the sensitivity to palbociclib is due to the upregulation of the pRb relative p107 (38). This compensation has been previously reported regarding acute loss of Rb-related functions (30, 39). However, our analyses revealed no significant induction of p107 (RBL1) in RB1-mutated bladder cancer cells, while moderate induction of p130 (RBL2) was detected without significant changes in p16 (Fig. 1). Moreover, upon palbociclib treatment, we observed induced expression of p107 only in RB1 wt cells, indicating a different mechanism of action in RB1 wt versus mutant cells.

In addition to inducing proliferative arrest, palbociclib led to the death of RB1-mutant bladder cancer cells. This observation is in agreement with published reports showing that inhibition of cell-cycle kinases or downregulation of cyclins can trigger tumor cell apoptosis, in addition to cell-cycle arrest. Recently, Wang and colleagues have demonstrated that the inhibition of cyclin D3/CDK6 complex by palbociclib also promotes apoptosis irrespective of Rb family status in T-ALL cells (40). Interestingly, this seems to proceed through metabolic reprograming, driving glycolytic intermediates to the pentose phosphate and serine pathways (40). Whether similar changes occur in bladder cancer cells and how this might be affected by functional RB1 remains unknown. Curiously, our GSEA analyses unveiled metabolic datasets that are altered in transgenic tumors upon palbociclib + CDDP treatment (not shown). Therefore, the possibility that metabolic rewiring also participates in the therapeutic response observed is very attractive and will be subject of future research.

Although the analysis of genes regulated by palbociclib displayed strong differences between RB1-mutant and wt cells as described above, in silico analyses revealed that the downregulated genes were FOXM1 targets. FOXM1 is a transcriptional activator with multiple oncogenic roles (41) that was identified as a critical phosphorylation target in a proteomic analysis of potential substrates of CDK4 kinase (27). In agreement, we also observed that the treatment with palbociclib or knockdown of CDK4 led to reduced FOXM1 phosphorylation, supporting the relevance of FOXM1 to maintain the oncogenic properties of cancer cells and its relevance in solid tumors (41). In addition, we provide evidence indicating that inactivation of RB1 promotes the activation of FOXM1 in bladder cancer cells. It has been proposed that FOXM1 confers resistance to chemotherapeutic agents, including CDDP, as supported by our current data. Importantly, as CDDP is the gold standard for bladder cancer management (6), our findings suggest that the in vivo treatment with palbociclib would allow reducing CDDP doses, possibly allowing the treatment of the patients with “unfit” bladder cancer. In fact, using xenografts, we found a tumor regression in the combined treatment, while CDDP alone had no major effect. Similar findings also linking FOXM1 with CDDP resistance/sensitivity have been reported for other solid tumors (42). We further provide evidence that FOXM1 expression, and phosphorylated FOXM1 levels, dictates resistance or sensitivity to CDDP in vitro in bladder cancer cells. Of note, we also observed that the reduction of CDK4 levels did not affect CDDP sensitivity, indicating that under these conditions, CDDP sensitivity is not dependent on FOXM1, as we observed that the knockdown of CDK4 impinges FOXM1 phosphorylation and produces the reduction FOXM1 levels. Although the molecular mechanisms dependent on CDK4 levels or its kinase activity that may govern sensitivity to CDDP are presently unknown, it is possible to speculate that, in the absence of CDK4, cyclin D is released and may act through a CDK4-independent manner to sustain CDDP sensitivity. Importantly, the relationship between cyclin D levels and sensitivity to CDDP and other chemotherapeutic agents has been reported in other cell types (43–45). In addition, we observed that palbociclib induced p27 in all cell lines. Such induction, directly (46) or through the transcriptional modulation (47) of Fanconi genes (48) or AURKA (49), may contribute to this effect. In addition, the reduced EZH2 expression observed in the transgenic mouse model, or activity, as shown by reduced EZH2 Thr 487 phosphorylation, may increase CDDP sensitivity (50). These possibilities, in the context of FOXM1 expression and activity on its possible impact of CDDP and palbociclib sensitivity and the dependence of RB status, would be the subject of future research.

To further support these previous findings in cultured bladder cancer cells and their derived xenografts, we analyzed the treatment effects in a new in vivo metastatic bladder cancer mouse model (QKO), in which multiple genes involved in bladder cancer are inactivated (Rb1^F/F;Rbl1-/-; Trp53^F/F; Pten^F/F). The same approach using adenovirus coding for Cre recombinase has been previously validated showing that Pten and Trp53 loss is sufficient to cause invasive bladder cancer (19), whereas the complete ablation of retinoblastoma family (Rb, p107, and p130) induces high-risk NMIBC in mice (20). Using our QKO model, we found high-grade and invasive tumors with metastatic spreading, affecting the liver, lungs, and the peritoneal cavity, really similar to that observed in metastatic human bladder cancer (51). In agreement with our previous findings with cells, the combined treatment was highly effective. In addition, the transcriptome analyses revealed the reversion of gene patterns characteristics of the tumor cells. The transcriptome analyses also demonstrated a reduced EZH2 expression and a downregulation of FOXM1 phosphorylation. These findings could be attributed to the repression of activating E2Fs, and in particular E2F3a, already identified as main regulators of EZH2 expression in bladder cancer (20). In addition, treated tumors showed high immune cell infiltration. This observation has a great relevance given the recent evidences in breast cancer models, indicating that CDK4/6 inhibitors may trigger antitumor immune response (52). Moreover, there have been positive results in clinical trials with immune checkpoint inhibitors, although still limited to a low percentage of patients in bladder cancer (53). Therefore, it will be important to ascertain whether palbociclib, alone or in combination with CDDP, could confer increased sensitivity to immune checkpoint inhibitors, increasing the percentage of patients who could benefit from these therapies.

Finally, based on our results and to consider CDK4/6 inhibitors for the stratified management of bladder cancer, we also analyzed the association of FOXM1 expression and outcome. Our data revealed that the increased activity and expression of FOXM1 in NMIBC identifies primary tumors at high risk of recurrence, and the possibility that these tumors display invasive characteristics in the recurrences. These data support the previous FOXM1-dependent gene signature related to recurrence in bladder cancer (54). Moreover, integrative global genomic analyses of advanced bladder cancer supported its relevance (10, 11).

As a whole, our data represent the rational basis for possible future clinical trials in patients with bladder cancer. These observations collectively support the relevance of palbociclib, alone or in combination with other therapeutics, to be a tool for the clinical management of advanced bladder cancer. Our results also show the main role of FOXM1 in human patients with bladder cancer, and how its phosphorylation-mediated activation could identify patients of poor clinical outcome. In addition, our data open new possible avenues for combined treatments, particularly in cases where the patient's condition would not support the conventional use of CDDP-based chemotherapy and currently present limited treatment options.

D. Castellano is a consultant/advisory board member for Astra Zeneca, Bayer, Bristol-Myers Squibb, Janssen, Pfizer, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Rubio, F. Villacampa, D. Castellano, J.M. Paramio

Development of methodology: C. Rubio, M. Martínez-Fernández, C. Segovia, I. Lodewijk, C. Segrelles, E. Munera-Maravilla, M. Santos, F. Guerrero-Ramos, D. Castellano, J.M. Paramio

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Rubio, M. Martínez-Fernández, C. Segovia, I. Lodewijk, C. Suarez-Cabrera, F. López-Calderón, E. Munera-Maravilla, M. Santos, A. Bernardini, C. Lorz, M. José Gómez-Rodriguez, G. De Velasco, F. Villacampa, F. Guerrero-Ramos, S. Ruiz, F. de la Rosa, F.X. Real, M. Dueñas, J.M. Paramio

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Rubio, M. Martínez-Fernández, C. Suarez-Cabrera, C. Segrelles, F. López-Calderón, R. García-Escudero, G. De Velasco, S. Domínguez-Rodríguez, F.X. Real, N. Malats, D. Castellano, M. Dueñas, J.M. Paramio

Writing, review, and/or revision of the manuscript: C. Rubio, M. Martínez-Fernández, C. Segovia, I. Lodewijk, E. Munera-Maravilla, A. Bernardini, C. Lorz, M. José Gómez-Rodriguez, G. De Velasco, F. Villacampa, F.X. Real, N. Malats, D. Castellano, M. Dueñas, J.M. Paramio

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Rubio, C. Suarez-Cabrera, C. Segrelles, I. Otero, J.M. Paramio

Study supervision: C. Rubio, D. Castellano, J.M. Paramio

Other (review and revision of the manuscript): M. Santos

We express our deepest acknowledgement to patients and their families. We acknowledge Dr. A. Gandarillas [Institute of Research Marqués de Valdecilla (IDIVAL), Santander, Spain] for providing us FOXM1-coding plasmid and to Pilar Hernández, Raquel Ruiz-Palomares, and Jorge Peral for their help with experimental procedures. We also acknowledge Pfizer SLU for providing us palbociclib. This work was supported by FEDER cofounded MINECO grant SAF2015-66015-R, ISCIII-RETIC RD12/0036/0009, and PIE 15/00076 and CB/16/00228 (to J.M. Paramio); FEDER cofounded MINECO ISCIII grant PI15/00993 (to M. Santos); FEDER cofounded MINECO grants SAF2013-49147-P and SAF2016-80874-P and Ramon y Cajal contract RYC-2011-09242 (to S. Ruiz); and a grant from Asociación Española contra el Cáncer (AECC; to F.X. Real, D. Castellano, and N. Malats). This work was also partially supported by a Pfizer preclinical grant (CTP WI235570; to J.M. Paramio).

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.