The mechanism of resistance to CDK4/6 inhibition and novel combination therapy with RNR inhibition for chemo-resistant bladder cancer

Bladder cancer (BCa) is the most prevalent urological cancer worldwide [1]. A significant proportion of BCa (89%) exhibits molecular alterations in the cell cycle pathway, and targeting cyclin-dependent kinases 4 and 6 (CDK4/6) is deemed as a promising therapeutic strategy [2]. Selective inhibitors of CDK4/6 (CDK4/6is) have been approved by the US Food and Drug Administration (FDA) [3]. They could induce cell cycle arrest in BCa immediately, and after this “sensitive stage”, unknown compensatory mechanism may cause acquired resistance [4, 5]. To address this issue, our study employed multi-omics and identified ribonucleotide reductase regulatory subunit M2 (RRM2), a crucial component of the ribonucleotide reductase (RNR) complex [6], as a key mediator in conferring acquired resistance. We further investigated whether Palbociclib activates proteolysis of RRM2 by the ubiquitin-proteasome system (UPS) and the ubiquitin-like proteins (UBLs) during the sensitive stage. Additionally, we explored whether RRM2 is controlled by E2F transcription factor 3 (E2F3) when acquired resistance is established. Interestingly, upregulation of RRM2 may also cause chemotherapy resistance [7]. Thus, we verified if concurrent inhibition of RNR and CDK4/6 holds promise as a novel therapeutic strategy for BCa patients, especially those exhibit resistance to chemotherapy. All the study designs and methods are described in the Supplementary file.

Retinoblastoma (RB)-positive BCa elicits a sequential progression from sensitivity to resistance to Palbociclib [8]. We utilized multi-omics to identify key regulators of this process (Figure 1A, Supplementary Tables S1-S5). The only candidate matching all three high-throughput screening approaches was RRM2, and pathway analysis further demonstrated related mechanisms (Supplementary Figures S1-S2). To validate this finding, we examined the cell cycle distribution and expression levels of the other RNR subunit RRM1 and RRM2 in a time kinetic (Figure 1B-C, Supplementary Figure S3A-D). Transcript levels were initially downregulated, followed by a partial recovery, while the decline and recovery pattern of proteins mirrored this. We then transduced single-guide RNAs of RRM1 and RRM2 into T24 synergistic activation mediator (SAM) cells and confirmed partial resistance (Figure 1D-E, Supplementary Figure S3E). However, degradation of RRM2 was still observed at early time points (Supplementary Figure S3F), indicating that proteolysis might be essential for therapy response. We then applied the proteasome inhibitors Epoxomicin/MG-132 in combination with Palbociclib. As shown, protein degradation of RRM2 was effectively blocked, but only partially for RRM1 (Figure 1F, Supplementary Figure S4A-B). We next tested the combination of Palbociclib with ubiquitin-like proteins (UBLs) inhibitor MLN4924 and proved that the initial degradation of RRM1/2 was UPS-dependent but UBL-independent (Figure 1G, Supplementary Figure S4C-E).

To further investigate the regulation network of RRM2, we interrogated RRM2 in The Cancer Genome Atlas (TCGA) bladder carcinoma study for mRNA co-expression (Supplementary Table S6) and performed network analysis using Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (Supplementary Figure S5A). Additionally, by analyzing spatial transcriptomics data on Spatial Omics DataBase (SODB) and ConsensusPathDB (CPDB), we found that E2F1 was spatially co-expressed with RRM2 (Supplementary Figure S5B-C). We confirmed that Palbociclib mainly affected the expression levels of E2F1 and E2F3, and notably, E2F3 was consistently upregulated (Supplementary Figure S5D-E). By knocking down E2F1, we observed a significant re-expression of RRM2, whereas blocking E2F3 maintained its downregulation (Figure 1H-I). Surprisingly, after long-term suppression of RRM2 expression by siE2F3, we observed an upregulation of E2F1 and blocking of cell cycle re-entry (Figure 1I, Supplementary Figure S5F). Double knockdown of E2F1 and E2F3 slightly prevented the recovery of RRM2 expression levels (Supplementary Figure S5G). These findings collectively suggest that E2F1 might have a suppressive function, while E2F3 drives the RRM2 re-expression.

We conducted comprehensive tests with the novel validated RNR inhibitor COH29 [9] in combination with Palbociclib to overcome acquired resistance. The combination successfully blocked long-term colony formation (Figure 1J, Supplementary Figure S6A). Additionally, cell viability assays confirmed synergism (Figure 1K, Supplementary Figure S6B). It is worth noting that the combination partially blocked the recovery of RRM2 in RT112 and UM-UC-3 cells, but not in T24 and 253J cells (Figure 1L, Supplementary Figure S6C). These suggested that the synergism is independent of RRM2 expression level, as COH29 blocked the function of RNR (Supplementary Figure S6D). Furthermore, synergism and senescent morphology were also observed using a living cell monitoring system in the combination groups (Supplementary Figure S7A). However, apoptosis through Caspase3/7 could not be observed (Supplementary Figure S7B). Subsequently, the combination was tested in two models: the chicken chorioallantoic membrane (CAM) and the murine subcutaneous xenograft model. Both monotherapies suppressed tumor growth, while the combination induced more potent effect (Figure 1M, Supplementary Figure S8A-E). This effect was attributed to proliferation as indicated by a significant decrease in Ki-67 expression (Supplementary Figure S8F).

By utilizing Gene Expression Profiling Interactive Analysis (GEPIA), we found that RRM2 was significantly upregulated in tumor tissues, while RRM1 and RB showed no difference (Figure 1N, Supplementary Figure S9A). Higher expression of them was related with significantly worse disease-free survival (DFS) (Supplementary Figure S9B). We also identified 4 patients from the Cornell/Trento cohort [10] who had primary and paired advanced tumors and were treated with chemotherapy. Among them, 2 patients (WCM088 and WCM117) showed amplified RRM2 only in the metastasis (Supplementary Figure S9C), suggesting that amplification of RRM2 may contribute to chemo-resistance. In addition, we retrospectively collected 20 muscle-invasive BCa specimens from patients who underwent radical cystectomy (Supplementary Table S7). These specimens were divided into three subgroups according to their response to chemotherapy. We observed lower RRM2 levels in the chemo-responded group compared to chemotherapy-naïve patients, but an increase in the chemo-resistant group (Supplementary Figure S9D-E). We tested the combination therapy in chemo-resistant BCa cell lines and found that they responded synergistically (Figure 1O, Supplementary Figure S9F-L).

In conclusion, we identified RRM2 as key mediator in conferring acquired resistance to Palbociclib in BCa (Figure 1P). Dual inhibition of RNR and CDK4/6 could effectively overcome this acquired resistance and holds promise as a potential therapeutic strategy for chemo-resistant BCa.

Zhichao Tong, Wanhai Xu, Roman Nawroth and Jürgen E. Gschwend designed the study and wrote the manuscript, Zhichao Tong, Yubo Zhao, Lou Lienhard, Shiyu Bai, Ziqi Wang, Yuling Zhao, Thilo Bracht, Barbara Sitek performed the experiments, Qi Pan, Pengyu Guo, Benedikt Ebner analyzed the data.

The authors declare no competing interests.

This work was supported by the National Natural Science Foundation Fund of China (82002688; U20A20385), China Postdoctoral Science Foundation (2021M693828), Postdoctoral Scientific Research Development Fund of Heilongjiang Province (LBH-Z22030), National Key Research and Development Program of China (2021YFB3801000).

The research procedures were approved by the Committees for Ethical Review of Harbin Medical University (2022-DWSYLLCZ-38; KY2023-62).