3.1. KDM6A mutation is frequent in BC
We downloaded the data, including follow-up profiles and gene expression levels, of 411 and 103 BC tissues from the TCGA and ICGC databases, respectively. Among these samples, 107/411 (25.97%) and 25/103 (24.27%) harboured a KDM6A mutation (Sup. 1A–C).
Missense and truncating mutations were the main mutation types spanning the entire gene (Sup.2A-B), with the former mutation type being most frequent (Sup. 3A); single nucleotide polymorphism was more common than deletion or insertion (Sup. 3B), and C>T was the most common single nucleotide variant in BC (Sup. 3C). The number of variant bases in each sample was counted and the mutation types are shown in a box plot in different colours in Sup. 3D and 3E.
3.2. KDM6A mutation is associated with higher mutation counts
BC samples with KDM6Amutations had higher overall mutation counts than wild-type samples in the TCGA cohort (210 vs 156.5; p = 0.0383, Kruskal-Wallis equality-of-populations rank test) (Fig. 1).
3.3. KDM6A mutation is associated with reduced KDM6Aexpression levels and poor prognosis in BC
Patients with KDM6A mutations had lower KDM6A expression levels than those with wild-type KDM6A (Fig. 2A). As shown in Sup.4, patients with KDM6A mutationshad a higher metastasis stage than those with wild-type KDM6A. However, no correlation was observed between KDM6A mutation and age, histologic grade, or the tumour-node-metastasis (TNM) stage. Kaplan-Meier analysis showed that BC patients with a high KDM6Aexpression (top 15%) levels hadsignificantly longer overall survival than those with low KDM6A levels (bottom 15%) (Fig. 2B). However, univariate and multivariate Cox regression analyses showed that a low KDM6A level or presence of KDM6A mutation was not an independent prognostic factor.
3.4. KDM6A mutation is significantly correlated with tumour-infiltrating lymphocytes (TILs) in the tumour microenvironment
Tumor Immune Estimation Resource (TIMER) analysis showed that KDM6A mutation was associated with a lower number of TILs. The infiltration levels of macrophages (p = 1.44e-2), CD8+ T cells (p = 5.79e-4), neutrophils (p = 4.40e-3), and resting dendritic cells (p = 3.70e-3) in the KDM6A mutation group were lower than those in the KDM6A wild-type group (Fig.3A). The association between KDM6A mutation and different types of immune cells was significant, including dendritic cells (p = 9.70e-3, cor = –0.1272), CD8+ T cells (p = 6.0e-4, cor = –0.1689), macrophages (p = 1.03e-2, cor = –0.1262), and neutrophils (p = 5.1e-3, cor = –0.1375).
To further assess the correlation of KDM6A mutation with TILs in the BC microenvironment, the CIBERSORT algorithm was used. The composition of 22 TILs in each sample varied significantly (Fig. 3B). Moreover, naïve B cells were more enriched in the KDM6A mutation group, whereas M2 macrophages and resting mast cells were more enriched in the KDM6A wild-type group (Fig. 3C). Correlation analysis showed that memory-activated CD4+ T cells were positively correlated with CD8+ T cells, and also had a positive association with the number of M1 macrophages. In turn, resting macrophages were positively correlated with levels of activated natural killer (NK) cells. However, activated NK cells were negatively correlated with activated mast cells (Fig. 3D).
3.5. Enrichment pathway analysis of KDM6A mutation
We further explored the relationship between KDM6A mutation and the immune response. GSEA revealed that the intestinal immune network for IgA production, the chemokine signalling pathway, natural killer cell-mediated cytotoxicity, the B cell receptor signalling pathway, the T cell receptor signalling pathway, the Fc epsilon Ri signalling pathway, Fc gamma R-mediated phagocytosis, primary immunodeficiency, and the Toll-like receptor signal pathway was significantly downregulated in the KDM6A mutation group (Figure 4A–I). These results implied that KDM6A mutation downregulates signalling pathways involved in the immune system in BC.