Progression of prostate cancer reprograms MYC-mediated lipid metabolism via lysine methyltransferase 2A

Abstract

Background: The activities of MYC, the androgen receptor, and its associated pioneer factors demonstrate substantial reprogramming between early and advanced prostate cancer. Although previous studies have shown a shift in cellular metabolic requirements associated with prostate cancer progression, the epigenetic regulation of these processes is incompletely described. Here, we have integrated chromatin immunoprecipitation sequencing (ChIP-seq) and whole-transcriptome sequencing to identify novel regulators of metabolism in advanced prostate tumors characterized by elevated MYC activity.

Results: Using ChIP-seq against MYC, HOXB13, and AR in LNCaP cells, we observed redistribution of co-bound sites suggestive of differential KMT2A activity as a function of MYC expression. In a cohort of 177 laser-capture microdissected foci of prostate tumors, KMT2A expression was positively correlated with MYC activity, AR activity, and HOXB13 expression, but decreased with tumor grade severity. However, KMT2A expression was negatively correlated with these factors in 25 LuCaP patient-derived xenograft models of advanced prostate cancer and 99 laser-capture microdissected foci of metastatic castration-resistant prostate cancer. Stratified by KMT2A expression, ChIP-seq against AR and HOXB13 in 15 LuCaP patient-derived xenografts showed an inverse association with sites involving genes implicated in lipid metabolism, including the arachidonic acid metabolic enzyme PLA2G4F. LuCaP patient-derived xenograft models grown as organoids recapitulated the inverse association between KMT2A expression and fluorine-18 labeled arachidonic acid uptake in vitro.

Conclusions: Our study demonstrates that the epigenetic activity of transcription factor oncogenes exhibits a shift during prostate cancer progression with distinctive phenotypic effects on metabolism. These epigenetically driven changes in lipid metabolism may serve as novel targets for the development of novel imaging agents and therapeutics.

Keywords: Lipid metabolism; Prostate cancer; Transcriptional regulation.

Fig. 1

Identification of transcription factors associated with MYC redistribution at AR- and HOXB13-occupied sites. A Ranked order depiction of genes whose expression correlate with MYC in an institutional cohort of 177 laser-capture microdissected prostate cancer tumor foci. B Schematic of the strategy used to identify upstream transcription regulators based on co-occupied peaks in LNCaP ChIP-seq data. anti-MYC [22], anti-AR, and anti-HOXB13 ChIP-seq data [23] used for analyses. C, D Bar plots showing the total peak number of MYC co-occupied sites, characterized as “core” or “redistributed” peaks, for anti-AR (C) and anti-HOXB13 (D) ChIP-seq. E Venn diagraph showing the overlap between transcription regulators identified by MYC/AR and MYC/HOXB13 co-occupied genes and nominated Ingenuity Upstream Regulator Analysis

Fig. 2

Decreasing association of KMT2A expression with prostate cancer drivers during primary disease progression. A–C Pearson correlation of the log₂ CPM expression level for KMT2A with a 54-gene ssGSEA MYC activity score (A), a 266-gene ssGSEA AR activity score (B), or the log₂ CPM expression level for HOXB13 (C) in a cohort of 177 laser capture microdissected foci of human prostate tumors. The error bars represent the 95% confidence bands for linear regression. For each correlation, the foci are subdivided into Gp3 (left), Gp4 (middle, including intraductal tumors), and Gp5 (right)

Fig. 3

Inverse association of KMT2A expression with prostate cancer drivers in metastatic disease. A–F Pearson correlation of the log₂ CPM expression level for KMT2A with the 54-gene ssGSEA MYC activity score (A, B), a 266-gene ssGSEA AR activity score (C, E), or the log₂ CPM expression level for HOXB13 (D, F) in the LuCaP series of patient-derived xenografts (N = 25) (A, C, D) or the West Coast Dream Team-Prostate Cancer Foundation metastatic CRPC cohort (N = 99) (B, E, F). The error bars represent the 95% confidence bands for linear regression

Fig. 4

Inverse relationship between PLA2G4F and KMT2A expression in metastatic prostate cancer. A Expression of KMT2A across the LuCaP PDX models in our study displayed as a heatmap, indicating which models are classified as KMT2A high or low. B Heatmap of anti-AR and anti-HOXB13 ChIP-seq [29] peak signals (aggregated across samples as indicated) around TSS regions (± 3 kb) in LuCaP PDX's with differential KMT2A expression. Each row is a peak ranked by KMT2A low to high. The location of PLA2G4F on heatmap is marked by a black bar. C Functional enrichment from Ingenuity Upstream Pathway Analysis of anti-AR and anti-HOXB13 ChIP-seq differential peaks overlapping genes for pathways with P < 0.1. Red indicates that PLA2G4F is present in the pathway gene set. D Integrative genome viewer tracks of the PLA2G4F gene locus from anti-AR and anti-HOXB13 ChIP-seq. LNCaP DNase-hypersensitivity regions shown in blue. E–H Spearman correlation of the log₂ CPM expression level of PLA2G4F with log₂ CPM expression level of KMT2A (E, G) or the 54-gene ssGSEA MYC activity score (F, H) in the LuCaP (E, F) or the West Coast Dream Team-Prostate Cancer Foundation metastatic CRPC cohort (G, H). Samples with log₂ CPM PLA2G4F < 0 were excluded. The error bars represent the 95% confidence bands for linear regression

Fig. 5

Arachidonic acid (ArA) metabolism is associated with KMT2A expression. A Time course of [¹⁸F]ArA uptake in LuCaP PDX organoids (N = 15) classified as either KMT2A-high (red) or KMT2A-low (blue). Individual organoids are given by light shaded lines, with each line representing the median uptake of 2–5 replicates. Dark shaded lines represent the median uptake for KMT2A-high or KMT2A-low PDX organoids. Results are expressed as median uptake per 1 million cells. B. Percent [¹⁸F]ArA uptake in LuCaPs based on KMT2A expression at 120 min. Line at median. Each organoid (representing 2–5 replicate experiments) plotted as open circles (N = 9 high, N = 6 low). P = 0.026 by Mann–Whitney U test. C, D Spearman correlation of the log₂ CPM expression level of PLA2G4F with the percent [¹⁸F]ArA uptake in LuCaPs at 120 min (C) or the rate of uptake over 2 h, measured by the slope of linear regression (D). Cases are colored by KMT2A expression status. The error bars represent the 95% confidence bands for linear regression of each scatter plot