PML1 is the most abundant isoform in estrogen receptor-positive (ER+) breast tumors
To better understand the expression patterns of PML isoforms in breast cancer patients, we interrogated RNA-seq datasets from normal breast tissues (GTEx) and breast tumors (TCGA). Our results demonstrate that the total PML transcript expression is significantly elevated across all breast cancer subtypes compared to normal tissues (Fig. S1). PML1 mRNA is the predominant isoform in normal breast tissues and ER + breast tumors. Moreover, PML1 abundance shifts dramatically from ~ 38% in normal tissues (Fig. 1A) to ~ 67% in ER + tumors (Fig. 1B), while that of PML2 mRNA is expressed at a lower level than PML1 in both normal (~ 30%) and malignant breast tissues (~ 20%). PML4, which encodes an extensively studied tumor suppressor, is expressed at a much lower level (~ 7%). Moreover, higher PML1 mRNA levels are associated with poor prognosis of ER + breast cancer patients (Fig. 1C), but there was no correlation between the expression of other PML isoforms and prognosis (Fig. S2A). Furthermore, the total PML protein abundance is elevated in ER + breast tumors (Fig. 1D). We also observed a trend in which higher PML protein abundance correlates with poor prognosis (Fig. S2B). PML1 and PML4 proteins share the first 620 amino acids, with PML4 containing a 13 a.a. unique C-terminus and PML1 possessing an additional 262 a.a (Fig. 1E). To better understand the role of PML1 in breast cancer, we generated a PML1-specific antibody. We confirmed that PML1 and PML4 proteins migrate around 130 kDa and 100 kDa, respectively (Fig. 1F) and that PML1 is the predominant isoform in ER+/HER2- breast cancer cell lines, including MCF-7, T47D, and ZR-75-1 cells (Fig. 1G). These findings suggest that PML1 is the most abundant isoform in breast cancer, and its high expression may be a potential biomarker for poor prognosis for ER + breast cancer.
PML1 promotes cancer phenotypes and fulvestrant resistance
Our previous study demonstrated that the ectopic overexpression of PML4 inhibits the proliferation, migration, and invasion of MCF-7 cells [18]. We expand our studies by investigating the effects of PML on another ER + breast cancer cell line, ZR-75-1. Our results showed that the knockdown of PML reduces the proliferation (Fig. 2A), colony formation (Fig. 2D), and invasion (Fig. 2F) of MCF-7 and ZR-75-1 cells, while PML1 overexpression has the opposite effect (Fig. B, E, and G). Furthermore, MCF-7-HA-PML1 cells, which express virally transduced HA-PML1, exhibit a significant increase in the IC50 (4.499e-008M) for fulvestrant, compared to control cells (1.046e-010M) (Fig. 2H), indicating that higher PML1 expression promotes fulvestrant resistance. This result is consistent with a recent clinical study indicating that the PML gene is amplified in 14% of ER + MBC [25](Fig. S3). Moreover, exogenous PML1 rescues the proliferation of PML knockdown cells (Fig. 2C), but PML4 does not (Fig. S4A). Additionally, PML2 inhibits the proliferation and breast cancer cell stemness (Fig. S4B), indicating that PML2 and PML1 have the opposite effects on breast cancer cells. These results suggest that PML isoforms play distinct roles in breast cancer development and progression and that PML1 may play a role in fulvestrant resistance.
PML1 binds and positively regulates stemness gene promoters and promotes breast cancer stem-like cell (BCSC) populations
The observations that PML1 promotes fulvestrant resistance and invasion of breast cancer cells prompted us to investigate PML1’s role in cancer cell stemness. Gene Set Enrichment Analysis (GSEA) revealed that affected genes in PML knockdown microarray gene expression study are enriched for genes upregulated in the Mammary_Stem_Cell_Up signature [31](Fig. 3A), suggesting PML’s role in CSC regulation. Analyses of PML ChIP-seq data in MCF-7 cells revealed that PML binds to more than half of the BCSC-associated gene promoters (Table S3). Knockdown of PML1 significantly reduced the expression of a subset of BCSC-related genes (Fig. 3B), while overexpression of PML1 increased their expression (Fig. 3C). Moreover, PML1 knockdown reduced the frequency of BCSCs in extreme limiting dilution assays (ELDAs) (Fig. 3D) and tertiary tumorsphere-formation assays (Fig. 3F, S5), while PML1 overexpression had the opposite effect (Fig. 3E, 3G, and S5). FACS analyses further showed that PML knockdown reduced the ALDHhigh cell population, while overexpression of PML1 increased it (Fig. 3H-I). These results suggest that PML1 promotes the stemness of breast cancer cells.
ChIP-seq analyses reveal crosstalk between PML1, ER, and Myc-bound promoters
Previous reports have shown that the Myc transcription factor regulates the expression of a subset of stemness genes [32] and Myc interacts with PML4 [33]. Analyses of ChIP-seq data for PML, Myc, and ER revealed that most PML-binding sites (~ 77%) are in promoter regions, which account for 23% of protein-coding gene promoters (Fig. 5A and 5E). In contrast, less than 14% of the ER-binding sites are in promoter regions, while ~ 80% are in intergenic regions or introns (Fig. 5B). Interestingly, most Myc-binding sites are in intergenic regions or introns (Fig. 5C). Focusing on PML-bound promoters (Fig. 5D), we found that PML and ER bind 1,387 common promoters (Fig. 5D-E), which accounts for ~ 70% of ER- and ~ 18% of PML1-bound promoters, respectively (Fig. 5E, S6A). The top-ranked consensus sequence among PML1 and ER commonly bound promoters is an estrogen-response element (ERE) half-site, -AGGTCA- (Fig. S6B). Myc binds ~ 94% of PML-bound promoters in MCF-7 cells (Fig. 5E). In fact, microarray analyses [34] suggest that affected genes in PML knockdown cells are enriched in Myc-targeted genes (Fig. S6C). Furthermore, ChIP-seq analyses suggest that PML1, Myc, and ER bind several BCSC-related gene promoters (Table S3), including JAG1, KLF4, MYC, SNAI1, and YAP1 (Fig. S7). Using ChIP-qPCR, we confirmed that PML binds these promoters but not NANOG (Fig. 5F) and that PML1, not PML4, binds these promoters (Fig. 5G). These analyses suggest that PML, Myc, and ER regulate gene expression in BCSCs by binding to common promoters. Moreover, microarray gene expression analyses [34] indicate that PML target genes are enriched in estradiol-responsive genes [35](Fig. 5H). Proximity ligation assays (PLA) showed that endogenous PML and ER interact (Fig. S8). Furthermore, Coimmunoprecipitation demonstrated endogenous and exogenous PML1 and ER interact (Fig. 5I-J), and the recruitment of PML1 to BCSC-related gene promoters is induced upon E2 treatment (Fig. 5K), indicating a potential role of PML1 in E2-induced ER-target gene expression. Furthermore, the knockdown of ESR1 significantly reduces the expression of stemness-related genes, phenocopying the effects of PML1 knockdown (Fig. 5L). However, the loss of PML1 had little or no effect on the ER binding to these promoters (Fig. 5M), suggesting PML1 regulates ER target gene expression without affecting ER binding to the promoters.
PML1 interacts with WDR5, a core subunit of the histone H3 lysine 4 methyltransferase (H3K4 HMTs) complexes
To investigate the underlying mechanism of how PML1 promotes ER and Myc transcriptional activity, we utilized an in-silico approach to screen for PML1-interacting proteins, which identified several putative PML-interacting proteins involved in histone modification (Fig. 6A), including proteins involved in histone H3K4 methylation (Fig. 6B), such as WDR5 [36]. WDR5 is a core subunit of all four MLL1-4 histone methyltransferase complexes. These complexes catalyze the methylation of histone H3 lysine 4 (H3K4), with MLL1/2-containing complexes responsible for H3K4me3 and MLL3/4-containing complexes catalyzing H3K4me1 [37]. Interestingly, in silico analyses also suggest an association of PML with MLL1. Using co-immunoprecipitation (Fig. 6C), GST pulldown assays (Fig. 6D), and PLA (Fig. S8), we showed that PML1 and WDR5 physically interact. Furthermore, we analyzed the ChIP-seq database to examine the H3K4me3 status of PML1-, ER-, and Myc-bound promoters. Our analysis revealed that H3K4me3 marks ~ 88% of PML-bound promoters, and ~ 90% of PML, Myc, and ER commonly-bound promoters are enriched in H3K4me3 (Fig. 6E). Specifically, several BCSC-related genes described above are enriched with the H3K4me3 mark (Fig. 6F-G).
To interrogate the role of PML1 in regulating global H3K4me3 across gene promoters, we performed PML knockdown followed by ChIP-seq, which showed that PML1 regulates H3K4me3 levels at numerous gene promoters (Fig. 7A), including gene loci associated with BCSCs (Fig. 7B). Additionally, the H3K4me3 patterns we observed on these promoters align well with publicly accessible data (Figure S7). Importantly, ChIP-qPCR confirmed that the loss of PML1 significantly reduced the H3K4me3 mark on BCSC-related gene promoters (Fig. 7C). Because PML1 and PML4 contain the WDR5-interacting domain, we examined whether PML1 and PML4 can restore the H3K4me3 mark in PML knockdown cells, and our data demonstrated that PML1, not PML4, re-establishes the H3K4me3 mark in PML knockdown cells (Fig. 7D). Furthermore, the loss of PML1 significantly reduced the associations of WDR5 (Fig. 7E), MLL1 (Fig. 7F), and MLL2 (Fig. 7G) with stemness gene promoters. We further investigated whether WDR5 is required for PML associations with these promoters and found that knockdown of WDR5 markedly reduces the expression of the BCSC-related genes (Fig. 7H) and the H3K4me3 mark (Fig. 7I) but has little or no effect on PML1 associations with these promoters (Fig. 7J). These data suggest that PML1 promotes ER and Myc transcriptional activity through its interaction with WDR5 and the subsequent enrichment of the H3K4me3 mark at target gene promoters.
Inactivation of WDR5 enhances the effectiveness of fulvestrant in PML1-overexpressing cells
The data presented above suggests that WDR5 and PML may act together to modulate the expression of stem cell-associated genes and stemness in breast cancer cells. Our results demonstrate that the knockdown of WDR5 leads to a significant decrease in BCSCs population (Fig. 8A and S9) and inhibition of MCF-7 cell proliferation (Fig. 8B). We also found that the knockdown of WDR5 significantly enhances the anti-proliferation activity of fulvestrant against PML1-overexpressing cells, reducing the IC50 from µM to nM (Fig. 8C). We next investigated the effects of pharmacological inhibitors of WDR5, OICR-9429 and compound 16 (C16), on stemness-related gene expression, cell proliferation, and the anticancer activity of fulvestrant. Both inhibitors disrupt the interaction between WDR5 and MLL1 by targeting their interacting sites [38, 39]. Our results demonstrated that both inhibitors effectively reduced the population of BCSCs (Fig. 8D), inhibited the expression of stemness-related genes (Fig. 8E), and suppressed the proliferation of both control and PML1-overexpressing cells (Fig. 8F-G). Furthermore, both inhibitors enhanced the anti-growth activity of fulvestrant (Fig. 8H). These results suggest that the PML1:WDR5 association has functional significance in regulating breast cancer stemness and fulvestrant resistance.