Resistance to Platinum-Based Chemotherapy in Muscle-Invasive Bladder Cancer: Genetic and Immune Determinants and Implications for Treatment Sequencing — A Case Series

1. Introduction

Bladder and urinary tract urothelial cancer (UC) is a major health burden. In 2022 an estimated 614,000 new cases and 221,000 deaths were recorded worldwide, with approximately three-quarters of cases occurring in men and the highest incidence rates reported in Europe, with an age-standardised incidence of roughly 20–29 per 100,000 [1].

The systemic treatment landscape has evolved drastically with the introduction of immune checkpoint inhibitors (ICI) and antibody–drug conjugates (ADC) into international guidelines [2,3]. Although of considerable historical value, the place of platinum-based chemotherapy (PBC) has become uncertain, and the risk–benefit ratio of cisplatin or carboplatin is increasingly deemed unfavourable by both physicians and patients. In muscle-invasive bladder cancer (MIBC), neoadjuvant PBC was long the mainstay of treatment because of a clear overall survival (OS) benefit and a meaningful rate of pathologic complete response (pCR) [4,5,6].

This paradigm is now being directly challenged. The phase 3 KEYNOTE-B15/EV-304 trial, combining the ICI pembrolizumab with the ADC enfortumab vedotin (EV), outperformed PBC in cisplatin-eligible patients in event-free survival, OS and overall response. Compared with cisplatin–gemcitabine, the pCR rate was significantly higher for EV–pembrolizumab (55.8% vs. 32.5%; one-sided p < 0.0001) [7]. However, because the control arm received neither perioperative ICI (durvalumab, NIAGARA trial [8]), adjuvant ICI (nivolumab in CheckMate 274 [9]; pembrolizumab in AMBASSADOR [10]; atezolizumab in IMvigor011 [11]), nor a more effective PBC regimen such as dose-dense MVAC (dd-MVAC, VESPER trial [6]), the magnitude of this benefit may be overestimated. Ongoing randomised trials will more clearly define the residual value of PBC in the peri-operative setting.

In locally advanced and metastatic UC (la/mUC), PBC remains relevant from the second line onward, particularly in patients without FGFR alterations. The EV-302 trial demonstrated a clear first-line benefit of EV–pembrolizumab over PBC (median OS 31.5 vs. 16.1 months) [12]. In HER2-expressing la/mUC, disitamab vedotin combined with toripalimab has likewise outperformed PBC in the first line (median OS 31.5 vs. 16.9 months) [13], and trastuzumab deruxtecan has shown activity in HER2-expressing tumours [14]. Nevertheless, most patients ultimately progress on these therapies and are treated with PBC in second or later lines: in the EV-302 trial, 85.9% of patients in the EV–pembrolizumab arm who received any subsequent systemic therapy received PBC, and in the RC48-C016 trial [13] 74.2% of patients in the disitamab vedotin group did so.

In this research article we present an analysis of patients receiving dd-MVAC for MIBC with long follow-up. This includes the identification and characterisation of PBC-refractory patients, the underlying genetic alterations, the effect of dd-MVAC on these alterations and on other immune characteristics, subsequent lines of therapy and clinical outcomes. The aim is to better identify patients likely to benefit from PBC across the UC treatment trajectory and to inform the choice of therapy sequence.

2. Materials and Methods

2.1. Patient Selection and Data Extraction

After review and authorisation by the local ethics committee (Ghent University Hospital, reference 20171080), we identified 54 consecutive patients with MIBC who had received dd-MVAC in the neoadjuvant setting at Ghent University Hospital between November 2013 and November 2019. Patient characteristics and follow-up data were extracted from the electronic patient file and anonymised before analysis. Eligibility for cisplatin-based chemotherapy was assessed according to the Galsky criteria [15]. Pathologic complete response (pCR) was defined as the absence of residual invasive tumour and nodal involvement at radical cystectomy (ypT0 ypN0).

2.2. PD-L1 Determination

Pathology specimens from the diagnostic TURBT and from the radical cystectomy were stained for PD-L1 using two monoclonal antibodies: clone 22C3 (PD-L1 IHC 22C3 pharmDx, Dako/Agilent) and clone SP142 (VENTANA PD-L1 [SP142] Assay, Roche Diagnostics). Staining was performed on 4-µm formalin-fixed, paraffin-embedded sections according to the manufacturer’s protocol on the VENTANA BenchMark ULTRA platform; cross-platform use of Dako antibodies on the BenchMark ULTRA has been validated and shows interchangeable performance with the corresponding Dako Autostainer Link 48 [16]. For both assays, PD-L1 positivity was defined as ≥1% of viable tumour cells showing membranous staining of any intensity (tumour-cell [TC] score ≥1%). A uniform tumour-cell-based threshold was applied to both antibodies to allow direct head-to-head comparison; the conventional companion-diagnostic cut-offs (combined positive score ≥10 for 22C3 and immune-cell score ≥5% for SP142) were recorded as secondary categories. A combined variable (PD-L1 positive if either antibody and/or timepoint was positive) was used for the primary analysis. We note that the SP142 clone is known to show a distinctly different staining pattern from the 22C3, 28-8 and SP263 clones [17].

2.3. Immunohistochemistry and Antibody Staining

Sequential 4-µm sections from the same paraffin blocks were stained for CD3, CD8, FOXP3, PD-1, PD-L2 and NY-ESO-1 on the VENTANA BenchMark ULTRA platform. Immune infiltration was scored semi-quantitatively and independently by a dedicated uropathologist (S.V.) and the first author (D.D.M.), both blinded to clinical outcome; discordant scores were re-assessed jointly until consensus was reached. Antibody clones, sources, dilutions and detection conditions are summarised in Supplementary Table S1.

2.4. Genetic Analysis

Paired genetic analysis was possible for 13 MIBC patients with adequate material; 12 paired cases were evaluable for the alteration-level comparison. Of the remaining 41 patients, 12 did not undergo cystectomy (reasons: progressive disease on NAC, n = 4; transurethral resection alone, n = 4; primary radiochemotherapy, n = 2; primary radiotherapy, n = 1; upper-tract urothelial carcinoma, n = 1), 11 achieved a pCR, and 18 did not have adequate TURBT DNA material.

Paired genetic characteristics were analysed by DNA sequencing of TURBT and cystectomy specimens using both whole-exome sequencing (WES) and comprehensive genomic profiling (CGP) using the TruSight Oncology 500 (TSO500) panel. DNA extraction from FFPE biopsy material was performed using the QIAamp DNA FFPE Tissue Kit (QIAgen).

Target enrichment for WES was performed on 12 patients with paired samples by the KAPA HyperCap protocol using KAPA HyperExome Probes (Roche), followed by sequencing on NovaSeq 6000 using S4 Reagent Kit v1.5 (300 cycles) (Illumina). WES data were processed using an in-house bcbio pipeline including data mapping to the reference genome by BWA, variant calling using vardict and mutect2, and variant annotation & filtering with Ensembl Variant Effect Predictor (VEP) and BCFtools. Single-nucleotide variants and small insertions/deletions were determined in 64 bladder cancer specific gene panel - ACTB, ARID1A, ASXL1, ASXL2, ATM, BTG2, C3orf70, CASP8, CCND3, CDKN1A, CDKN2A, CREBBP, CUL1, ELF3, EP300, ERBB2, ERBB3, ERCC2, FAT1, FBXW7, FGFR3, FOXA1, FOXQ1, GNA13, HES1, HIST1H3B, HRAS, KANSL1, KDM6A, KLF5, KMT2A, KMT2C, KMT2D, KRAS, MB21D2, MBD1, METTL3, NFE2L2, NRAS, NUP93, PAIP1, PARD3, PIK3CA, PSIP1, PTEN, RB1, RBM10, RHOA, RHOB, RXRA, SF1, SF3B1, SPN, SPTAN1, SSH3, STAG2, TAF11, TMCO4, TSC1, TXNIP, USP28, ZBTB7B, ZFP36L1. Tumor mutational burden (TMB) was determined in the WES data using the following filtering of somatic variants (variant allele frequency [VAF] ≥ 5%, coverage ≥ 50x, EPR = pass (BCBio tumor-only somatic variant filter), BIOTYPE = protein_coding, not Existing_variation & BCFTools stats number of SNPs (count nt variant SNPs only)).

The TSO500 high-throughput (HT) library prep (Illumina) was performed on 7 of the 12 patients on paired samples according to manufacturer’s instructions and sequenced on NovaSeq 6000 using SP Reagent Kit v1.5 (200 cycles) (Illumina). TSO500 NGS data were analysed using the TSO500 Local App v2.1 (Illumina). Single-nucleotide variants, small insertions/deletions, focal amplifications, tumor mutational burden (TMB) and microsatellite-instability (MSI) status were determined in the TSO500 data sets. Variants of uncertain significance (VUS) were recorded separately from pathogenic/likely-pathogenic variants.

2.5. Statistical Analysis

All statistical analyses were performed in Stata (StataNow/BE 18.5 for Mac, Apple Silicon; revision 26 February 2025; StataCorp LLC, College Station, TX, USA). Baseline characteristics were compared between patients with and without pCR using Fisher exact tests for categorical variables and Mann–Whitney U tests for continuous and ordinal variables. Paired changes in immune markers and TMB between TURBT and cystectomy were tested with Wilcoxon signed-rank tests. Follow-up was calculated from the start of NAC to the last follow-up date. Overall survival (OS) and recurrence-free survival (RFS) were estimated by the Kaplan–Meier method and compared with the log-rank test across pathologic-response groups. A two-sided α of 0.05 was applied. Given the exploratory nature of the immune-marker panel, p-values are reported without correction for multiple comparisons and should be interpreted as hypothesis-generating.

3. Results

3.1. Patient Characteristics

Fifty-four patients were included. The cohort was predominantly male (43/54, 80%) with a median age at diagnosis of 67 years (IQR 56–71), and 44/54 (81%) completed the intended four cycles of dd-MVAC. Baseline characteristics according to pathologic response are shown in Table 1. Among the 42 operated patients, 11 (26%) achieved a pCR and 31 had residual disease. Hydronephrosis at diagnosis was strongly associated with the absence of pCR (1/11, 9% in the pCR group vs. 16/31, 52% in the no-pCR group; p = 0.016), and adjuvant radiotherapy was administered exclusively to patients without a pCR (0% vs. 42%; p = 0.009); two further patients in the no-surgery group received radiotherapy as their primary local treatment rather than as adjuvant therapy. Age, sex, smoking status, completion of four chemotherapy cycles and variant histology did not differ significantly between response groups.

3.2. Patient Trajectories and Survival

Individual patient trajectories are shown in the swimmer’s plot (Figure 1). Median follow-up from the start of NAC was 87 months (IQR 24–104) for the whole cohort and 94 months among survivors. Twenty-two patients (41%) experienced disease recurrence or progression and 17 (31%) died. Outcomes diverged sharply by pathologic response: no recurrences and a single death occurred among the 11 patients with a pCR (median follow-up 102 months), compared with 15 recurrences and 10 deaths among the 31 patients with residual disease (median follow-up 84 months). The 12 patients who did not undergo cystectomy had the poorest outcomes (7 recurrences, 6 deaths; median follow-up 45 months). Twenty patients received a second line of therapy, six a third line and three a fourth line, most commonly checkpoint inhibitors followed by platinum-based or taxane chemotherapy. Of the 12 patients who did not undergo cystectomy, three received primary radiotherapy (n = 1) or primary chemo-radiotherapy (n = 2) as their definitive local treatment.

Overall survival and recurrence-free survival differed significantly across pathologic-response groups (Figure 2; log-rank p = 0.040 and p = 0.026, respectively). Patients with a pCR had essentially no events, those with residual disease followed an intermediate trajectory, and the non-operated patients had the steepest early drop in both endpoints.

3.3. Immune-Related Characteristics and Effect of Neoadjuvant Chemotherapy

Baseline (TURBT) immune characteristics were not associated with pathologic response. Median CD3, CD8, FOXP3, PD-L2 and NY-ESO-1 scores were comparable between patients who did and did not achieve a pCR (all p > 0.05, Mann–Whitney U). PD-L1 positivity, assessed with the combined variable, was also not associated with pCR (PD-L1 positive in 3/7 evaluable pCR patients vs. 11/22 no-pCR patients; p = 1.000).

In contrast, neoadjuvant dd-MVAC significantly remodelled the tumour immune microenvironment. In paired TURBT–cystectomy comparisons, intratumoral FOXP3 (16/22 paired cases decreased; p < 0.001), NY-ESO-1 (10/21 decreased, none increased; p = 0.004), PD-L2 (10/22 decreased; p = 0.007) and CD3 (10/22 decreased; p = 0.029) all fell significantly after chemotherapy. CD3 score showed a borderline decrease (p = 0.053), whereas CD8, CD8 score and PD-1 did not change significantly. PD-L1 status was relatively stable: of 22 paired cases, 4 remained positive and 4 lost positivity, while 2 of 14 baseline-negative tumours gained PD-L1 expression after NAC.

3.4. Effect of Neoadjuvant Chemotherapy on Tumour Mutational Burden

TMB was assessed in paired samples by both CGP using TSO-500 and WES. TMB did not change significantly after neoadjuvant chemotherapy by either assay (TSO-500: median 7.1 pre-NAC vs. 6.5 post-NAC, n = 7 pairs, Wilcoxon p = 0.67; WES: median 23 vs. 25, n = 10 pairs, p = 0.86; Figure 3 and Figure 4). Baseline TMB values from the two assays were positively correlated (Pearson r = 0.79). Changes in TMB after NAC were heterogeneous and bidirectional, with no consistent direction of effect at the cohort level.

3.5. Effect of Neoadjuvant Chemotherapy on Genetic Alterations

Paired pre- and post-NAC mutational profiles were evaluable for 12 cases (Table 2). TP53 was the most frequently altered gene, present in the majority of cases, followed by alterations in chromatin-remodelling genes (KDM6A, KMT2C, KMT2D) and in PIK3CA. An activating FGFR3 alteration was identified in one case, and an ERCC2 mutation — associated with cisplatin sensitivity — in another. Focal amplifications (MYC, CCNE1, EGFR, MET, CCND1) were detected by TSO-500 in three cases, with copy numbers amplification generally decreasing in the post-NAC sample. Overall, the mutational landscape was largely concordant between TURBT and cystectomy specimens: the dominant clonal driver mutations were retained after chemotherapy, and most discordances involved gain or loss of VAF or uncertain-significance variants rather than truncal drivers, consistent with the TCGA molecular characterisation of MIBC [18].

4. Discussion

In this single-centre cohort of 54 MIBC patients treated with neoadjuvant dd-MVAC and followed for a median of more than seven years, three observations stand out. First, neoadjuvant platinum-based chemotherapy substantially remodels the bladder-cancer immune microenvironment, significantly reducing intratumoral FOXP3⁺, NY-ESO-1⁺, PD-L2⁺ and CD3⁺ populations. Second, neither the tumour mutational burden nor the core mutational landscape changed meaningfully after chemotherapy, indicating that the genomic backbone of the tumour is preserved through treatment. Third — and most relevant for clinical practice — no baseline immune or genomic marker, including PD-L1, reliably distinguished platinum-refractory from platinum-sensitive patients.

Cisplatin remains the most active single cytostatic agent in urothelial cancer despite a primary response rate of only about 10–15% in cisplatin-naïve metastatic disease [19]. Its principal mechanism is the formation of platinum–DNA adducts that block transcription and induce apoptosis; resistance arises chiefly from increased DNA-damage repair (notably nucleotide-excision repair components such as ERCC2 [20]), increased glutathione conjugation, and reduced drug uptake [21,22]. Carboplatin, with comparable molecular biology but a four-fold lower potency and a different toxicity profile, has consistently produced inferior outcomes when substituted for cisplatin in combination [23]. The clinical value of cisplatin in MIBC therefore rests on combination with other active agents — MVAC, dd-MVAC and cisplatin–gemcitabine. dd-MVAC clearly showed an advantage over GC in the VESPER trial, with a cisplatin dose–response correlation [6,24,25,26]. Combination with immune checkpoint inhibitors is mechanistically attractive: platinum induces immunogenic cell death, releases tumour antigens, depletes some immunosuppressive populations and can convert "cold" tumours into inflamed ones [27,28], creating substrate for an anti-PD-(L)1 response [29]. Across tumour types, this rationale has translated into consistent perioperative benefit — CheckMate 816, KEYNOTE-671 and AEGEAN in NSCLC, KEYNOTE-522 in TNBC, MATTERHORN in gastric/GEJ cancer, and NIAGARA in MIBC [8,30,31,32,33,34,35] — yet in urothelial cancer specifically the ENERGIZE (NCT03661320) and KEYNOTE-866 (NCT03924856) trials have failed to extend the NIAGARA result, indicating that the chemo-immunotherapy interaction is more nuanced than an additive effect and thus context-dependent.

Our observation of broad lymphocytic attrition after dd-MVAC — affecting both regulatory (FOXP3⁺) and total (CD3⁺) T-cell compartments, as well as PD-L2⁺ and NY-ESO-1⁺ populations — adds to this nuance. It contrasts with the immunostimulatory framing often invoked for platinum and is consistent with the rationale for concurrent rather than purely sequential ICI–chemotherapy in the perioperative setting [8].

The stability of TMB and of the driver mutational landscape after NAC is concordant with the view that platinum resistance in UC is not predominantly acquired through gross genomic evolution during a short course of neoadjuvant therapy. Rather, determinants of cisplatin sensitivity — such as alterations in DNA-damage-response genes including ERCC2 and other nucleotide-excision-repair components [20] — are likely present from the outset, supporting pre-treatment genomic profiling as the appropriate window for biomarker-driven selection. Molecular subtype has also been associated with differential benefit from NAC [18,36], and integrating subtype with DNA-damage-response status may ultimately outperform single markers.

The absence of any association between baseline PD-L1, or other immune markers, and pCR echoes the experience of the neoadjuvant immunotherapy trials ABACUS and PURE-01, in which PD-L1 enriched for, but did not reliably predict, response [37,38]. In the contemporary context, the KEYNOTE-B15/EV-304 result [7] — a markedly higher pCR rate with EV–pembrolizumab than with cisplatin–gemcitabine — together with NIAGARA [8] and CheckMate 274 [9] is reshaping perioperative management. Our data argue that, even as non-platinum regimens move forward, a subgroup of patients still derive durable benefit from platinum: the 11 pCR patients in this cohort had no recurrences and only a single death over a median follow-up exceeding eight years, with significantly better Kaplan–Meier overall and recurrence-free survival than patients with residual disease or no cystectomy. Identifying these patients before treatment would spare them perioperative ICI exposure with its potential long-lasting toxicity and would, on a population scale, substantially reduce the burden of MIBC care.

In an exploratory re-staging analysis, eight additional patients downstaged to non-muscle-invasive disease at cystectomy (T1/Tis/Ta) had overall survival indistinguishable from the pCR group (log-rank p = 0.91) and only a non-significant numerical excess of recurrence events (2/8 vs. 0/11; p = 0.15). Across all four re-staging categories (pCR, non-MIBC downstaging, MIBC residual ≥T2, no surgery) both OS and RFS differed significantly (log-rank p = 0.018 and p = 0.023, respectively), with the pCR and downstaging curves clearly separated from the other two. Although these numbers are small for robust head-to-head inference, the finding is consistent with the wider MIBC literature treating post-NAC ≤T1 N0 as a "near-pCR" outcome and broadens the apparent platinum-sensitive subset in our cohort from 11/42 (26%) to 19/42 (45%) of operated patients.

For locally advanced and metastatic disease, platinum-based chemotherapy retains an important role in later lines after first-line ADC-based regimens [12,13,14]. Characterising which patients remain platinum-sensitive across the treatment trajectory is therefore not of historical interest only but is directly relevant to sequencing decisions today.

This study has important limitations. It is retrospective and single-centre, the sample size is modest, and the paired genetic and immune analyses were possible only in a subset, limiting statistical power and possibly introducing selection bias, as patients progressing early on PBC were not included; the immune-marker comparisons are exploratory and unadjusted for multiple testing. Immune infiltration was scored semi-quantitatively rather than by quantitative digital pathology. A small proportion of patients in any neoadjuvant cohort would achieve a pCR after TURBT alone and therefore derive no incremental benefit from chemotherapy, although in this high-risk MIBC population that proportion is expected to be very low. These constraints notwithstanding, the long and near-complete follow-up, the inclusion of all 54 consecutive eligible patients, and the availability of paired pre- and post-treatment tissue are notable strengths.

5. Conclusions

Neoadjuvant dose-dense MVAC remodels the muscle-invasive bladder cancer immune microenvironment — depleting regulatory and total T-cell populations and PD-L2 and NY-ESO-1 expression — without altering tumour mutational burden or the core genomic landscape. No pre-treatment immune or genomic marker, including PD-L1, identified platinum-refractory patients in this long-term cohort, although pathologic complete response remained strongly associated with overall and recurrence-free survival. As antibody–drug conjugate and immunotherapy regimens reshape both the perioperative and metastatic settings, better predictive biomarkers are needed to identify the patients who still benefit from platinum and to guide rational treatment sequencing across the urothelial cancer trajectory.