KDM5D Histone Demethylase Identifies Platinum-Tolerant Head and Neck Cancer Cells Vulnerable to Mitotic Catastrophe

Abstract

Head and neck squamous cell carcinoma (HNSCC) is a major contributor to cancer incidence globally and is currently managed by surgical resection followed by adjuvant chemoradiotherapy. However, local recurrence is the major cause of mortality, indicating the emergence of drug-tolerant persister cells. A specific histone demethylase, namely lysine-specific demethylase 5D (KDM5D), is overexpressed in diverse types of cancers and involved in cancer cell cycle regulation. However, the role of KDM5D in the development of cisplatin-tolerant persister cells remains unexplored. Here, we demonstrated that KDM5D contributes to the development of persister cells. Aurora Kinase B (AURKB) disruption affected the vulnerability of persister cells in a mitotic catastrophe-dependent manner. Comprehensive in silico, in vitro, and in vivo experiments were performed. KDM5D expression was upregulated in HNSCC tumor cells, cancer stem cells, and cisplatin-resistant cells with biologically distinct signaling alterations. In an HNSCC cohort, high KDM5D expression was associated with a poor response to platinum treatment and early disease recurrence. KDM5D knockdown reduced the tolerance of persister cells to platinum agents and caused marked cell cycle deregulation, including the loss of DNA damage prevention, and abnormal mitosis-enhanced cell cycle arrest. By modulating mRNA levels of AURKB, KDM5D promoted the generation of platinum-tolerant persister cells in vitro, leading to the identification of the KDM5D/AURKB axis, which regulates cancer stemness and drug tolerance of HNSCC. Treatment with an AURKB inhibitor, namely barasertib, resulted in a lethal consequence of mitotic catastrophe in HNSCC persister cells. The cotreatment of cisplatin and barasertib suppressed tumor growth in the tumor mouse model. Thus, KDM5D might be involved in the development of persister cells, and AURKB disruption can overcome tolerance to platinum treatment in HNSCC.

Keywords: AURKB; KDM5D; cisplatin; drug-tolerant persister; head and neck cancer.

Figure 1

Differentially expression of KDM5D linked HNSCC Tumors, CSCs, and Cisplatin Resistance. (A) Individual volcano plots depicted overexpression of KDM5D as DEGs in respective phenotypes and datasets: Tumor versus Normal (GSE9844), CSCs versus Non-CSCs (GSE72384), and Cisplatin Resistant versus Sensitive (GSE102787). The position of KDM5D was marked in red dot with respective arrow. (B) Venn diagram depicts shared common DEGs between HNSCC tumors, CSCs subset, and cisplatin resistant cells, in which KDM5D belongs to one of the shared genes. (C) The heatmap of GSE72384 microarray dataset described cluster analysis of previously described top DEGs, which appropriately categorized each sample into CSCs and non-CSCs subsets. (D) Compared with all recognizable lysine demethylases belonging to the JARID or KDM family in humans, KDM5D was relatively highly expressed in HNSCC tumors and cisplatin-resistant cells. (E) The GSEA findings revealed that several crucial signaling pathways were activated or deactivated and linked to cancer stemness in HNSCC. (F) The scatter plot revealed that KDM5D expression was positively correlated with genes upregulated in the diapause-like state (r = 0.22, p = 0.0032) and negatively correlated with genes downregulated in the diapause-like state (r = −0.26, p = 0.0077). Diapause gene set scores of each sample were calculated by GSVA method. (G) A higher KDM5D expression level was associated with poorer overall survival in patients in the TCGA-HNSC dataset (p = 0.0049). NS: Not Significant, FC: Significant in Log2 Fold-Change, p: Significant in p-value, FC_P: Significant in Log2 Fold-Change and p-value, NES: Normalized Enrichment Score.

Figure 2

Overexpression of KDM5D was associated with poor platinum responses in HNSCC patients. (A) Representative images of KDM5D staining in tissue specimens of TMU-SHH HNSCC cohort in respective order: normal adjacent epithelium, well-differentiated tumor, and poorly differentiated tumors. Higher KDM5D expression was noted in HNSCC tissues than in adjacent normal epithelial tissues. (B) The highest KDM5D expression was observed in poorly differentiated squamous cell carcinoma tissues. (C) Representative images of KDM5D staining in tissue specimens of TMU-SHH HNSCC cohort according to cisplatin response and recurrence disease. (D) Among the patients with HNSCC, KDM5D expression was higher in the platinum non-responders than in the responders. (E) Among HNSCC patients who responded to platinum-based chemotherapy, higher KDM5D expression was observed in those with early disease recurrence than those with no/late-recurrence. Significance level: * p < 0.05; *** p < 0.001; **** p < 0.0001. Scale bar: 200 μm.

Figure 3

KDM5D and AURKB co-expression delineates cluster of HNSCC persister cells. (A) Representative tSNE plots of single-cell profiling in GSE103322 dataset showed eight distinct clusters of cells. (B) Array of tSNE plots portrayed expression level of interest genes such as HNSCC tumor markers (KRT6A, KRT14, CDH1), CSCs and persister marker (ALDH1A3), putative main targets of this study (KDM5D and AURKB), and diapause-related genes (CCND1, FAS, ALDH6A1). (C) Dot plot described level of expression of each gene (KDM5D, ALDH1A3, AURKB, CCND1, FAS, ALDH6A1) in eight distinct clusters. (D) Scatter plot depicted co-association of each interest gene; Pearson’s coefficient and p-value was provided in the top margin. The KDM5D expression was significantly correlated to AURKB (r = 0.75, p = 0.012) and ALDH1A3 (r = 0.38, p = 0.035). (E) Scatter plot portrayed correlation between several diapause-related genes and KDM5D, such as CCND1 (r = 0.21, p = 0.037), FAS (r = 0.24, p = 0.021), and ALDH6A1 (r = 0.39, p = 0.018). (F) tSNE plot illustrates Diapause signature scores in each individual tumor cell. The Diapause_UP module score consisted of gene scores that are overexpressed during diapause, while Diapause_DOWN module score comprised genes that are downregulated at diapause stage. Persister cells were enriched among the clusters within the circle marker (red dash line). Predominant ALDH1A3 and KDM5D expression were noted in cell clusters no. 1, 3, and 5 which was consistent with diapause state activation. (G) The clusters, which were speculated to activate diapause state (clusters no. 1 and 3) exhibited several common features of diapause state, including the activation of NRF2, glutathione, drug metabolism, glycolysis, and epithelial-mesenchymal transition pathways.

Figure 4

KDM5D promotes the generation of platinum-tolerant persister cells. (A) Brief schematic figure shows the steps to generate cisplatin-tolerant persister cells. Parental HNSCC cells were treated with a short-course of cisplatin treatment followed by a ‘drug-holiday’ or recovery stage and a final cisplatin course. At the end of this stage, the HNSCC cells were relatively viable and exhibited increased platinum tolerance. (B) KDM5D knockdown significantly reduced the expression levels of AURKB and ALDH1A3 in both SAS and FaDu persister cells, indicating that KDM5D regulates AURKB and ALDH1A3 expression. (C) KDM5D knockdown significantly reduced tumor sphere formation in both PT-SAS and PT-FaDu cells. (D) Both platinum-tolerant persister HNSCC cells exhibited cell cycle arrest, as indicated by a significant increase in the G0/G1 subpopulation and a decrease in S and G2/M subpopulations. (E,F) KDM5D silencing re-sensitized platinum-tolerant persister cells upon cisplatin treatment, indicating that KDM5D plays a crucial role in promoting platinum tolerance in HNSCCs. Significance level: * p < 0.05; ** p < 0.01, *** p < 0.001. Scale bar: 100 μm.

Figure 5

KDM5D abrogates DNA damage and cell cycle arrest upon platinum treatment. (A) The suppression of KDM5D expression increased the vulnerability of both PT-SAS and PT-FaDU cells against cisplatin, as indicated by a significant reduction in the number of colonies upon cisplatin treatment. (B) Both HNSCC cell lines (SAS and FaDu) with KDM5D overexpression exhibited increased tolerance against several incremental dosages of cisplatin. (C) The distribution of the cell cycle did not significantly change after cisplatin treatment in both PT-SAS and PT-FaDU cells. (D) KDM5D silencing through shRNA-mediated knockdown induced cell cycle arrest in PT-SAS and PT-FaDU cells, as indicated by an increase in the G1 cell subpopulation and a decrease in the S cell subpopulation. (E) The level of DNA damage upon cisplatin treatment was significantly increased in response to KDM5D silencing, as indicated by the percentage of γH2AX positive cells with >10 foci in both PT-SAS and PT-FaDu cells. (F) Cellular proliferation was significantly reduced following KDM5D silencing and become more suppressed after cisplatin treatment, as reflected by decrease in EdU-positive cell fraction in both PT-SAS and PT-FaDu cells. (G) Representative immunofluorescence images describe differential expression of DNA damage marker, γH2AX and proliferation marker, EdU following KDM5D knock-down and cisplatin treatment. Silencing of KDM5D markedly increased the expression of γH2AX but reduced that of EdU in both PT-SAS and PT-FaDu cells, suggesting a protective role of KDM5D to attenuate cisplatin-mediated DNA damage. Significance level: * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. Scale bar: 100 μm.

Figure 6

AURKB inhibition promoted mitotic catastrophe in HNSCC persister cells. (A) Treatment with AURKB inhibitor, barasertib, induced cell cycle arrest in both PT-SAS and PT-FaDU cells by markedly increasing the G1 cell subpopulation while reducing the G2 and S cell subpopulations. (B) Bar graph described dose-dependent and time-dependent significant perturbation of mitotic activity of PT-SAS and PT-FaDU cells upon barasertib treatment. (C) The percentage of mitotic defects following barasertib treatment increased in a dose-dependent manner. (D) AURKB inhibition modulated several cell cycle regulators and activated mitotic catastrophes markers, such as Chk1, Cdc25 and Cyclin B1, suggesting the inhibition of cell cycle progression despite the enhancement of mitosis. Significance level: * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Figure 7

Cisplatin and barasertib co-treatment extend tumor suppression potential in vivo. (A) Treatment with a combination of cisplatin and barasertib significantly reduced the growth of HNSCC tumors compared with monotherapy or vehicle treatment. (B) Body weight did not significantly differ between the combination treatment, monotherapy, and vehicle treatment groups. (C) Mice in the combination group exhibited longer survival than did those in the monotherapy or vehicle group, suggesting the superior efficacy of the combination treatment in HNSCC. (D) Tissue staining revealed a significant reduction in cellular proliferation, aggressiveness, and epithelial-to-mesenchymal transition potential, as indicated by decreased levels of Ki67, Vimentin, and slug, respectively. Significance level: ns: not significant; ** p < 0.01; *** p < 0.001; **** p < 0.0001. Scale bar: 200 μm.

Figure 8

Schematic illustration of KDM5D contribution to affect the clinical outcome and biological development of platinum-tolerant persister cells in HNSCC. Left panel schema shows association between high KDM5D expression and poor clinical outcome in HNSCC patients encompassing poor response to platinum treatment or early recurrence disease. Right panel schema illustrates development of platinum-tolerant persister cells characterized by high expression of KDM5D/AURKB axis which disrupts AURKB by barasertib treatment deregulated tolerance mechanism and promoted mitotic catastrophe in platinum-tolerant cells.