Gender disparities in clinical outcomes of urothelial carcinoma linked to X chromosome gene KDM6A mutation

Patient inclusion

The FUSCC cohort, authorised by the Clinical Research Ethics Committee of Fudan University Shanghai Cancer Center (FUSCC), comprises 101 patients with UC. These individuals underwent radical cystectomy (RC), diagnostic biopsy or transurethral resection of bladder tumour, with genomic data sourced from the FUSCC-UC panel. In contrast, the ZSHS cohort, approved by the Clinical Research Ethics Committee of Zhongshan Hospital, Fudan University (ZSHS), consists of 132 patients. All were diagnosed with MIBC and underwent RC. The clinicopathological characteristics of these patients are detailed in table 1.

Public data sets

The Memorial Sloan Kettering Cancer Center (MSKCC) cohort18 comprises 1218 pathologically diagnosed cases with UC, with survival data available for 1146 of these cases. The Cancer Genome Atlas (TCGA) cohort5 includes 391 cases with MIBC with DNA sequencing, mRNA expression and clinical status data. The IMvigor210 cohort19 has 274 patients diagnosed with metastatic UC (mUC) treated with atezolizumab. The Urothelial Cancer - Genomic Analysis to Improve Patient Outcomes and Research (UC-GENOME) cohort20 includes 123 patients with mUC who received immunotherapy, and survival data are accessible for 109 of these individuals. Lastly, the Memorial Sloan Kettering Cancer Center - IMPACT (MSKCC-IMPACT) cohort21 encompasses 199 patients with UC treated with immunotherapy. The clinicopathological traits of patients across these datasets are summarised in online supplemental table 1. The specific inclusion criteria for these cohorts are presented in online supplemental figure 1. In total, our study enrolled 2438 patients with UC from two local and five independent public cohorts, all of whom had sufficient clinical and genomic data.

The clinical response to the immune checkpoint inhibitor (ICI) was assessed using the Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1 for the FUSCC, IMvigor210 and UC-GENOME cohorts.22 To further understand the clinical implications of KDM6A in adjuvant chemotherapy, we conducted a pooled analysis of AJCC stage II and III patients who received chemotherapy across the ZSHS, MSKCC and TCGA cohorts. For data sourcing, mutation annotation files, copy number alteration (CNA) data and associated clinicopathological data for the MSKCC, TCGA, UC-GENOME and MSKCC-IMPACT cohorts were procured from http://www.cbioportal.org.23 Meanwhile, clinical and genomic data from the IMvigor210 trial were accessed through the IMvigor210CoreBiologies R package from http://research-pub.gene.com/IMvigor210CoreBiologies.

Evaluation of KDM6A status

In the MSKCC and TCGA cohorts, KDM6A^Mut refers to both point mutations and homozygous deletions of KDM6A. For the FUSCC, IMvigor210, UC-GENOME and MSKCC-IMPACT cohorts, since no patients underwent CNA testing for KDM6A, only point mutations of KDM6A are labelled as KDM6A^Mut, as detailed in online supplemental figure 2. Considering that roughly one-third of patients displayed a KDM6A alteration (online supplemental table 1) and that in the TCGA cohort, point mutations of KDM6A significantly diminished its transcriptomic expression (online supplemental figure 3A, Mann-Whitney U test, p<0.001), the initial 33% of patients exhibiting low KDM6A expression as assessed by immunohistochemistry (IHC) assay were categorised as KDM6A^Mut in the ZSHS cohort. Representative images are available in online supplemental figure 3B.

IHC assay

IHC staining was performed on tissue microarrays (TMAs) derived from formalin-fixed, paraffin-embedded specimens. The process by which the TMA was prepared is detailed in a prior publication.24 Digital scans of all slides were obtained using the NanoZoomer-XR (Hamamatsu) and subsequently analysed with Image Pro Plus 6.0 software. Two independent pathologists, blinded to patient data, evaluated the positively stained cells and scored all samples. Cells were examined and quantified under a 200× magnification. The IHC score, derived from staining intensity and the proportion of positive cells, ranged between 0 and 300, with the final value being the average of the two pathologists’ scores.

For the ZSHS cohort, the expression status of KDM6A was evaluated using the IHC for KDM6A (Abcam, ab235989). Given the recognition of PD-L1 IHC staining by the United States Food and Drug Administration (FDA) as a supplementary diagnostic tool for assessing PD-L1 expression levels in patients deliberating immunotherapy, we used the DAKO 22-C3 and DAKO 28-8 assays for the FUSCC cohort. Patients with a combined positive score greater than 10 using the 22-C3 assay or a tumour positive score exceeding 1% with the 28-8 assay were identified as PD-L1 positive.25

Processing of genomic and transcriptomic data

For the genomic data, we used Maftools to illustrate the mutation sites of KDM6A in both male and female patients.26 A tumour mutation burden (TMB) surpassing 10 mutations per megabase was categorised as high TMB in line with FDA guidelines.27 For transcriptomic analysis, gene set enrichment analysis (GSEA) was undertaken to assess the immunogenomic characteristics.28 Gene sets representing the regulon activity of the androgen receptor (AR) and oestrogen receptor 1 (ESR1) were sourced from Shi et al.29 Within the TCGA cohort, the microenvironment cell populations-counter tool was harnessed to gauge immune infiltration in clinical specimens based on their transcriptomic data.30 Concurrently, the single sample gene set enrichment analysis (ssGSEA)31 was also employed to simulate the assessment of tertiary lymphoid structures (TLSs) and the angiogenesis process in these samples (online supplemental table 2). Furthermore, we incorporated the SOPRANO algorithm on the genomic profiles of our patient samples.32 SOPRANO calculates the trinucleotide context-corrected dN/dS both within and outside designated genomic regions. The immune dN/dS and immune escape status, as determined by SOPRANO, were intended to showcase immune-specific selection during the interplay between tumour cells and their adjacent microenvironments. All relevant data were extracted from the supplementary data accompanying the algorithm.

Statistical analysis

For survival analysis, the Kaplan-Meier approach combined with the Log-rank test was employed to affirm KDM6A^Mut as a distinct prognostic factor among male and female patients. The incidence comparison between the two groups was accomplished via the χ² test or Fisher’s exact test. For continuous variables, the Mann-Whitney U test and the Kruskal-Wallis test were executed to contrast distributions among two or more predefined subgroups. Outcomes are presented as median values alongside their IQRs. GSEA was conducted using GSEA software V.4.0.2 (Broad Institute, Massachusetts, USA).33 In this study, a p value <0.05 was deemed statistically significant. All statistical evaluations were executed using IBM SPSS Statistics V.22.0 and R V.3.6.3 (http://www.rproject.org/).

Patient and public involvement statement

This investigation incorporated 2438 patients with UC across seven independent public cohorts, all of whom had sufficient clinical and genomic data. As the research approach was retrospective in nature, patient involvement in the study’s design was deemed unnecessary. Therefore, patients did not play a role in shaping the research question, collecting and analysing data, interpreting findings or manuscript preparation. Our research results will be disseminated to the wider public via press releases, social media channels, presentations at global conferences, reports furnished to pertinent governmental bodies and scholarly associations.

KDM6A mutation is associated with superior OS in male patients with UC

KDM6A mutations were detected in 679 out of 2306 (29.4%) patients with UC from six cohorts with genomic data, with 505 out of 1768 (28.6%) occurring in men and 174 out of 538 (32.3%) in women (online supplemental figure 2). Among all patients in both ZSHS and MSKCC cohorts, KDM6A^Mut patients exhibited improved survival compared with KDM6A^WT counterparts. Specifically, in the ZSHS cohort, the median OS for KDM6A^Mut patients was 142.0 months, notably longer than the 61.0 months observed for KDM6A^WT patients (figure 1A left, Log-rank p=0.005). Similarly, in the MSKCC cohort, the median OS for KDM6A^Mut patients reached 95.6 months, slightly exceeding the 90.2 months observed for KDM6A^WT patients (figure 1B left, Log-rank p=0.028). Moreover, in the TCGA cohort, the median OS was 56.4 months for KDM6A^Mut patients and 33.0 months for KDM6A^Mut patients, though statistical significance was absent (figure 1C left, Log-rank p=0.204).

Remarkably, KDM6A mutations consistently exhibited a favourable impact on OS in male patients across three cohorts. In the ZSHS cohort, male patients with KDM6A mutations displayed a median OS of 142.0 months, contrasting with 63.0 months for wild-type individuals (figure 1A middle, Log-rank p=0.002). Within the MSKCC cohort, the median OS was 95.6 months for male patients with KDM6A mutations versus 78.1 months for those without (figure 1B middle, Log-rank p=0.010). In the TCGA cohort, male patients with KDM6A mutations displayed a median OS of 59.3 months, compared with 33.0 months for wild-type individuals (figure 1C middle, Log-rank p=0.057). However, in female patients across the same cohorts, no substantial differences in survival were observed based on KDM6A mutation status (figure 1A–C right).

KDM6A mutation and improved chemotherapy outcomes in female patients with UC

Given the limited sample size of female patients with UC undergoing chemotherapy, a pooled analysis encompassing the ZSHS, MSKCC and TCGA cohorts was conducted. Generally, patients with KDM6A mutations demonstrated a pattern of prolonged OS subsequent to chemotherapy (figure 2A). The median OS for KDM6A^Mut patients was 130.0 months compared with 61.0 months for wild-type individuals, although the difference did not reach statistical significance (Log-rank p=0.064). Remarkably, KDM6A mutations did not exhibit substantial clinical associations among male patients who received chemotherapy (figure 2B). Conversely, KDM6A^Mut women did not reach the median OS, whereas KDM6A^WT women showed an OS of 82 months (figure 2C, Log-rank p=0.026).

Gender-specific differences in immunotherapeutic response attributed to KDM6A mutation in UC

We further explored if the KDM6A alteration might explain gender differences in immunotherapy outcomes among patients with UC. In both the IMvigor210 and UC-GENOME cohorts, the combination of KDM6A mutation and gender exhibited notable stratifying potential (figure 3A,B. IMvigor210 cohort, Log-rank p=0.060; UC-GENOME cohort, Log-rank p=0.009). In the IMvigor210 cohort, male patients harbouring KDM6A mutations did not reach median OS compared with 7.1 months for female patients with the same mutations (figure 3A, Log-rank p=0.039). Additionally, the response rate was notably higher in KDM6A^Mut male patients at 45.7% compared with 0.0% in KDM6A^Mut female patients. In the UC-GENOME cohort, KDM6A^Mut male patients exhibited a median OS of 93.7 months, significantly higher than the 34.5 months for female patients (figure 3B, Log-rank p=0.004). Furthermore, the response rate was notably elevated in KDM6A^Mut male patients at 56.3% compared with 20.0% in KDM6A^Mut female patients. Moreover, within the MSKCC-IMPACT cohort, male patients with KDM6A mutations consistently demonstrated extended OS compared with their KDM6A^WT counterparts. The median OS for KDM6A^Mut male patients was not reached, while it was 14.0 months for KDM6A^WT male patients, although the difference did not reach statistical significance (figure 3C, Log-rank p=0.051). This implies that the KDM6A mutation consistently predicts positive clinical outcomes in men, whereas in women, it may indicate poor immunotherapy outcomes.

To further substantiate this observation, we examined the hypothesis in the local FUSCC cohort. Of the 27 KDM6A^Mut patients with UC, 23 were men and 4 were women (figure 3D). All female patients underwent immunotherapy but displayed notably poorer clinical outcomes compared with their male counterparts. KDM6A^Mut men did not reach the median OS, whereas KDM6A^Mut women showed an OS of 8.6 months in FUSCC cohort (online supplemental figure 4, Log-rank P = 0.053). For example, two patients, FUSCC-P011 (Male, KDM6AMut) and FUSCC-P093 (Female, KDM6AMut), both diagnosed with mUC, received Tislelizumab, an anti-PD-1 agent. After four treatment cycles, the male patient demonstrated significant remission (RECIST: PR), while the female patient showed disease progression (RECIST: PD) and experienced mortality event within two months (figure 3E).

Impact of KDM6A mutation on immune response via interaction with sex hormone receptor signalling in UC

We sought to explore the association between KDM6A^Mut and biological dysfunctions related to sexual or immunological discrepancies in UC. Intriguingly, GSEA disclosed that in KDM6A^Mut samples, the AR filtered regulon pathway was significantly upregulated (figure 4A, ES=0.528; online supplemental figure 5, Mann-Whitney U test, p=0.003), while the ESR1 filtered regulon pathway was conversely downregulated (figure 4A, ES=−0.813; online supplemental figure 5, Mann-Whitney U test, p=0.003). Notably, within a gender-specific context, KDM6A^Mut patients consistently exhibited higher AR and lower ESR1 filtered regulon activity, irrespective of their gender (figure 4B, AR: Kruskal-Wallis test, p=0.003; ESR1: Kruskal-Wallis test, p=0.003).

Regarding immunogenomic characteristics, KDM6A^Mut male patients in the IMvigor210 cohort were notably enriched for neoantigens (figure 4C, Kruskal-Wallis test, p=0.002). Concurrently, immune escape status, as defined by SOPRANO,32 was more prevalent in KDM6A^Mut male patients (figure 4C, χ² test, p=0.018), which implies that in addition to the activity of hormone receptors, the intrinsically different genomic contexts between male and female KDM6A^Mut patients might also have contributed to the different outcomes from immunotherapy. Concerning immune cell infiltration, KDM6A^Mut male patients exhibited elevated infiltration levels of T cells, cytotoxic T cells and NK cells. Transcriptomic analyses indicated an increased TLS signature in KDM6A^Mut male patients, while the angiogenesis signature was predominantly enriched in KDM6A^Mut female patients (figure 4D). Similarly, in the ZSHS cohort, IHC assays disclosed a more prevalent presence of TLS in KDM6A^Mut male patients, especially when compared with KDM6A^Mut female patients (figure 4E, incidence rate: 35.1% vs 14.3%).