Article
Open access
Published: 26 June 2024

Loss of Kmt2c or Kmt2d drives brain metastasis via KDM6A-dependent upregulation of MMP3

Nature Cell Biology volume 26, pages 1165–1175 (2024)Cite this article

2903 Accesses
11 Altmetric
Metrics details

Subjects

Abstract

KMT2C and KMT2D, encoding histone H3 lysine 4 methyltransferases, are among the most commonly mutated genes in triple-negative breast cancer (TNBC). However, how these mutations may shape epigenomic and transcriptomic landscapes to promote tumorigenesis is largely unknown. Here we describe that deletion of Kmt2c or Kmt2d in non-metastatic murine models of TNBC drives metastasis, especially to the brain. Global chromatin profiling and chromatin immunoprecipitation followed by sequencing revealed altered H3K4me1, H3K27ac and H3K27me3 chromatin marks in knockout cells and demonstrated enhanced binding of the H3K27me3 lysine demethylase KDM6A, which significantly correlated with gene expression. We identified Mmp3 as being commonly upregulated via epigenetic mechanisms in both knockout models. Consistent with these findings, samples from patients with KMT2C-mutant TNBC have higher MMP3 levels. Downregulation or pharmacological inhibition of KDM6A diminished Mmp3 upregulation induced by the loss of histone–lysine N-methyltransferase 2 (KMT2) and prevented brain metastasis similar to direct downregulation of Mmp3. Taken together, we identified the KDM6A–matrix metalloproteinase 3 axis as a key mediator of KMT2C/D loss-driven metastasis in TNBC.

Histone deacetylase 11 inhibition promotes breast cancer metastasis from lymph nodes

Article Open access 13 September 2019

Dependency of NELF-E-SLUG-KAT2B epigenetic axis in breast cancer carcinogenesis

Article Open access 28 April 2023

FOXQ1 recruits the MLL complex to activate transcription of EMT and promote breast cancer metastasis

Article Open access 01 November 2022

Main

Breast cancer is the most common cancer and is a leading cause of cancer-related deaths in women worldwide¹. It is classified, based on the expression of oestrogen and progesterone receptors and human epidermal growth factor receptor 2 (HER2), into hormone receptor-positive, HER2-positive or triple-negative breast cancer (TNBC). TNBC has the highest risk of distant metastases and the worst outcome among the major subtypes². Approximately 30% of patients with metastatic TNBC have brain metastases associated with the shortest overall survival³.

Sequencing of primary and metastatic tumours identified KMT2C as being more commonly mutated in distant metastases compared with primary breast tumours⁴. Both KMT2C and KMT2D are also frequently mutated in breast cancer brain metastases⁵, implying functional relevance. Histone–lysine N-methyltransferase 2C (KMT2C) and KMT2D function as histone H3 lysine 4 (H3K4) methyltransferases, converting unmethylated H3K4 to methylated H3K4 (H3K4me1) via monomethylation of lysine 4 on the histone H3 protein subunit. H3K4me1 is enriched in active or primed enhancers and promoters and enables recruitment of histone H3K27 acetyltransferases (for example, P300) and H3K27 demethylases (for example, lysine-specific demethylase 6A (KDM6A))⁶. KDM6A and KMT2C or KMT2D are components of the epigenetic regulatory complex of proteins associated with SET1 (COMPASS)⁶, and KDM6A depends on KMT2C regardless of its catalytic activity^7,8. Similarly, binding of KMT2D to other components of the COMPASS complex is independent of its enzymatic domain⁹. Thus, KMT2C and KMT2D are considered core proteins of the COMPASS complex, and deficiency in either of them affects epigenomic landscapes. Loss of KMT2C or KMT2D induces hybrid epithelial–mesenchymal transition (EMT) states and alters EMT balance, which can promote metastasis^10,11. KMT2C deficiency in lung cancer represses DNA methyltransferase 3A expression, leading to DNA hypomethylation and a subsequent increase in metastasis¹². However, it is unknown whether loss of KMT2C and KMT2D induces similar or unique epigenetic alterations and which underlying pathways might drive metastasis in KMT2C- or KMT2D-deficient tumours.

Results

Deficiency of Kmt2c or Kmt2d enables metastasis

KMT2C and KMT2D are among the top frequently mutated genes in TNBC, with mutations detected in 8.0 and 8.7% of tumours, respectively (Extended Data Fig. 1a). There are no mutational hotspots in the SET domain mediating enzymatic activity (Extended Data Fig. 1b), but most mutations are truncating or missense, implying loss of function. The expression of KMT2C and KMT2D is significantly reduced in distant metastases compared with matched primary TNBC (Fig. 1a), suggesting a role for the loss of KMT2C/D function in metastasis. Data from the DepMap database¹³ also showed a significant inverse correlation between metastatic potential and KMT2C expression in breast cancer cell lines and a similar trend for KMT2D (Extended Data Fig. 1c).

**Fig. 1: Loss of *Kmt2c* or *Kmt2d* induces multi-organ metastasis.**

To investigate the consequences of Kmt2c and Kmt2d deficiency, we deleted these genes in the non-metastatic 168FARN and 67NR mouse mammary tumour cell lines by transfecting an mCherry-tagged single guide RNA (sgRNA) pool and a plasmid expressing Cas9 and green fluorescent protein (GFP) (Fig. 1b). Sequencing of targeted genomic regions showed 90% (Kmt2c) and 100% (Kmt2d) mutant reads in 168FARN cells (Extended Data Fig. 1d) and 64% (Kmt2c) and 98% (Kmt2d) mutant reads in 67NR cells (Extended Data Fig. 1e). Immunoblot analysis confirmed reduced protein expression in knockout (KO) derivatives of both cell lines (Fig. 1c and Extended Data Fig. 1f).

Characterization of the cells showed no differences in cell proliferation (Extended Data Fig. 1g) and a slight but significant decrease in the half-maximal inhibitory concentration for doxorubicin in 168FARN Kmt2c KO cells and for the KDM6A inhibitor GSK-J4 in 168FARN Kmt2c and Kmt2d KO cells (Extended Data Fig. 1h,i). We then transduced 168FARN cells with H2B–mCherry and injected them into the mammary fat pads (MFPs) of syngeneic immunocompetent female BALB/c mice (Fig. 1d). Loss of Kmt2c or Kmt2d had no significant effect on primary tumour growth (Fig. 1e), but we detected mCherry⁺ micro-metastatic lesions in the liver, lungs and brain of mice injected with KO but not wild-type (WT) cells (Fig. 1f,g). Although we established cultures from all primary tumours, mCherry⁺ cancer cells only grew in culture from distant organs of mice injected with Kmt2c and Kmt2d KO cells, consistent with our staining data (Fig. 1g). Immunoblot analyses of these cell cultures showed generally lower expression of KMT2C and KMT2D in metastases compared with matched primary tumours, suggesting that they were derived from KO cells. Importantly, there was no compensatory upregulation of KMT2C in Kmt2d KO cells and vice versa (Extended Data Fig. 1j,k).

To assess metastases in a more quantitative manner, we conducted intracardiac injections and performed flow cytometry from dissociated brain, liver, and lungs. mCherry⁺ cells were significantly enriched in the brains of mice injected with Kmt2c and Kmt2d KO but not WT derivatives of both 168FARN and 67NR lines, but not in liver or lung tissues (Fig. 1h,i and Extended Data Fig. 1l–n). To validate our detection method, we also used quantitative PCR to measure mCherry expression and found a significant correlation between the two assays (Extended Data Fig. 1o). Haematoxylin and eosin staining confirmed macro-metastases in brains, consistent with Fig. 1h,i (Fig. 1j). Additionally, femoral bone sections revealed macro-metastases exclusively in Kmt2c and Kmt2d KO groups (Extended Data Fig. 1p,q), demonstrating the gain of a multi-organ metastatic phenotype following Kmt2c or Kmt2d loss.

Kmt2c and Kmt2d KO-specific tumour immune microenvironment

Loss of Kmt2d sensitizes tumours to immune checkpoint inhibitors (ICIs)¹⁴. However, it is not known whether loss of Kmt2c and Kmt2d has a similar impact on the immune microenvironment and if this is responsible for the observed metastatic phenotype. Thus, we performed single-cell RNA sequencing (scRNA-seq) of primary tumours derived from MFP injection of 168FARN derivatives. Tumour epithelial cells were identified based on the expression of mCherry and the epithelial marker Krt8, whereas stromal cells were annotated using known cell type-specific markers (Fig. 2a, Extended Data Fig. 2a,b and Supplementary Table 1). Tumour cells showed clear separation of Kmt2c and Kmt2d KO from WT (Fig. 2b), and differential gene expression (DGE) analysis identified Ly6a, Bst2, Ifi27l2a and Stat1 as the top highly upregulated genes in both KO tumour cells compared with the WT (Extended Data Fig. 2c,d and Supplementary Table 2). This implied a pro-inflammatory milieu via potential activation of interferon signalling, as confirmed by pre-ranked gene set enrichment analysis (Fig. 2c). However, several immune checkpoint genes, including Cd274 (encoding programmed death ligand 1) were increased in KO compared with WT cells (Fig. 2d,e, Extended Data Fig. 2e and Supplementary Table 2). Upregulation of programmed death ligand 1 was also confirmed by immunofluorescence in both Kmt2c KO and Kmt2d KO primary tumours (Fig. 2f,g).

**Fig. 2: Distinct impact of *Kmt2c* and *Kmt2d* loss on the tumour immune microenvironment.**

We then subclustered and annotated non-epithelial cells and grouped them according to genotype (Fig. 2h and Extended Data Fig. 2f,g). Fibroblasts were classified into inflammatory and myofibroblastic cancer-associated fibroblasts, and myofibroblastic cancer-associated fibroblast signatures were higher than inflammatory cancer-associated fibroblast ones in the Kmt2c KO but not the Kmt2d KO compared with the WT (Extended Data Fig. 2h,i). Endothelial cells from Kmt2c KO and Kmt2d KO tumours were distinct from the WT and shared top upregulated genes, including Cd74, Cxcl9, H2-Ab1 and H2-Eb2 (Extended Data Fig. 2j,k). The most abundant immune cells were macrophages, with a higher ratio of M1/M2 macrophages in Kmt2c but not Kmt2d KO tumours compared with the WT (Extended Data Fig. 2l,m). T cell populations were classified into total CD4⁺, T regulatory, total CD8⁺, cytotoxic CD8⁺ and exhausted CD8⁺ (Extended Data Fig. 2n). The signature of immune-suppressive T regulatory cells was higher in Kmt2d KO tumours and showed an increasing trend in Kmt2c KO tumours compared with the WT (Fig. 2i–k). The ratio of total CD8⁺ T cells was higher in the Kmt2c KO; however, these cells had a trend of higher exhaustion than cytotoxic signatures (Fig. 2l–n). Immunofluorescence confirmed significantly more CD8⁺ T cells specifically in Kmt2c KO tumours compared with the WT (Fig. 2o,p). In line with this—CCL5, a chemoattractant for monocytes and T cells¹⁵—was only increased in Kmt2c but not Kmt2d KO cell supernatants and tumours, whereas CXCL1 and matrix metalloproteinase 3 (MMP3) were higher in all KO samples (Extended Data Fig. 2o).

We then performed intracardiac injection in immunodeficient NSG mice. Again, we detected mCherry⁺ cancer cells only in the brains of mice injected with Kmt2c or Kmt2d KO cells but not WT cells (Fig. 2q). This suggested that Kmt2c or Kmt2d loss-associated immune suppression might not be the main driver of their metastatic phenotype.

Kmt2c and Kmt2d KO changes chromatin and the transcriptome

To dissect Kmt2c or Kmt2d loss-induced cell intrinsic changes, we first focused on histones as the direct targets of KMT2C and KMT2D. Quantitative histone mass spectrometry showed similar histone modifications in both KO cells compared with the WT (Fig. 3a). H3K4me1 and trimethylation of lysine 27 on the histone H3 protein subunit (H3K27me3) were significantly decreased in the Kmt2c KO but not in the Kmt2d KO, whereas acylation of lysine 27 on the histone H3 protein subunit (H3K27ac) was significantly increased in both KO lines (Fig. 3b). Immunoblot analysis showed similar levels of H3K4me1, but decreased H3K27me3 and increased H3K27ac in both KO lines (Fig. 3c).

**Fig. 3: Altered histone and KDM6A occupancy in *Kmt2c* and *Kmt2d* KO cells.**

To explore alterations in chromatin profiles, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) for H3K4me1, H3K27ac, and H3K27me3. We found significantly gained (7,367 for Kmt2c KO and 9,824 for Kmt2d KO) or lost (4,085 for Kmt2c KO and 4,896 for Kmt2d KO) H3K4me1 peaks (Fig. 3d and Supplementary Table 3), implying locus-specific rather than global changes in H3K4me1. Intersection of gained or lost peaks between Kmt2c and Kmt2d KO showed more unique than shared peaks (Fig. 3e). H3K27ac ChIP-seq also showed significantly gained (5,656 for Kmt2c KO and 6,031 for Kmt2d KO) and lost (3,992 for Kmt2c KO and 3,866 for Kmt2d KO) peaks in both KO cell lines with again more unique than shared changes (Fig. 3f,g and Supplementary Table 3). Because KMT2C and KMT2D are H3K4 methyltransferases, changes in H3K27ac might be an indirect effect of their KO. Thus, we performed ChIP-seq for the H3K27 acetyl transferase P300 but only detected a few significantly gained peaks (401 for Kmt2c KO and 36 for Kmt2d KO). However, overall changes of P300 and H3K27ac signal intensities were significantly correlated in KO cells (Extended Data Fig. 3b). Moreover, the P300 signal intensity was significantly higher in gained H3K27ac peaks compared with unchanged peaks in both Kmt2c and Kmt2d KO cells and lower in lost H3K27ac peaks in Kmt2c KO cells (Fig. 3h). An increase in H3K27ac can be a consequence of a decrease in H3K27 methylation, thus we performed H3K27me3 ChIP-seq. There were no significantly changed peaks, probably due to the overall lower number of total called peaks in the Kmt2c and Kmt2d KO lines compared with the WT (Fig. 3i and Extended Data Fig. 3c). However, consistent with our immunoblot analyses (Fig. 3a–c), the global H3K27me3 signal was decreased in both KO lines (Fig. 3i).

The H3K27me3 demethylase KDM6A is part of the KMT2C/D COMPASS complex. Thus, we performed KDM6A ChIP-seq, which identified significantly gained (39,467 for Kmt2c KO and 5,234 for Kmt2d KO) and very few lost (341 for Kmt2c KO and 1,003 for Kmt2d KO) peaks in KO cells compared with WT ones (Fig. 3j and Supplementary Table 3). Intersection again showed more unique than shared gained KDM6A peaks between the two KO cell lines (Fig. 3k). Motif analysis for each ChIP-seq dataset showed shared motifs, including ZNF263 and HOXA1 for gained H3K4me1 and H3K27ac, respectively. The Tlx and HIC1 motif was enriched in gained KDM6A peaks, suggesting a putative role of these factors upon loss of Kmt2c or Kmt2d (Extended Data Fig. 3d). Intersection of gained H3K27ac and gained KDM6A peaks within Kmt2c and Kmt2d KO cells showed minimal overlap (Extended Data Fig. 3e), highlighting P300 as a mediator of the observed differences in H3K27ac.

We then performed RNA-seq of Kmt2c and Kmt2d KO cells and identified DGEs compared with the WT (Fig. 4a and Supplementary Table 4). The topmost significantly upregulated genes in 168FARN Kmt2c KO cells included Mmp3, Scara5 and Cerk, whereas Cfh, Pdzrn3 and Scara5 were most significantly upregulated in Kmt2d KO cells. Of note, Kdm6a expression was similar in all genotypes. Overall, DGEs were more unique than shared between both KO lines (Fig. 4b, Extended Data Fig. 4a,b and Supplementary Table 4). Consistently, gene set enrichment analysis for Hallmark pathways showed apical surface, apoptosis and coagulation enriched in the Kmt2c KO and coagulation and angiogenesis in the Kmt2d KO (Extended Data Fig. 4c). However, prediction for transcription regulators of upregulated genes using the LISA algorithm¹⁶ showed NR3C1, MED1 and SMARCA4 among the top for both KO lines (Extended Data Fig. 4d). Interestingly, in 67NR cells, significantly changed genes and enriched pathways were more similar between Kmt2c and Kmt2d KO cells (Extended Data Fig. 4e–g and Supplementary Table 4). EMT was the top enriched pathway in Kmt2c and Kmt2d KO 67NR cells, consistent with previous studies¹¹. Of note, EMT was not detected in 168FARN KO cells (Extended Data Fig. 4c), probably because the parental cells were already mesenchymal. However, we found significant likelihoods of common deregulation when we overlapped differential genes between 67NR and 168FARN cell lines, suggesting that loss of Kmt2c or Kmt2d induces similar expression changes independent of the cell line (Extended Data Fig. 4h).

**Fig. 4: KDM6A and histone remodelling correlates with transcriptomic changes upon *Kmt2c* and *Kmt2d* KO.**

To assess the impact of altered chromatin on the transcriptome, we integrated RNA-seq and ChIP-seq data using binding and expression target analysis (BETA)¹⁷. Gained H3K4me1, H3K27ac and KDM6A peaks significantly correlated with upregulated genes, whereas lost H3K4me1 and H3K27ac peaks significantly correlated with downregulated genes in Kmt2c KO and Kmt2d KO cells, respectively (Fig. 4c). Lastly, we overlapped genes with correlations of upregulation from BETAs in Kmt2c or Kmt2d KO cells. The largest gene set was shared between all three BETAs, suggesting similar target gene regulation by H3K4me1, H3K27ac and KDM6A (Fig. 4d).

Blocking KDM6A decreases brain metastasis by decreasing Mmp3

To identify common epigenetically regulated drivers of metastatic phenotype of both Kmt2c and Kmt2d KO cells, we integrated significantly gained H3K4me1 and H3K27ac peaks and genes significantly upregulated in RNA-seq and scRNA-seq of cancer cells, identifying Mmp3 as the only overlapping gene (Fig. 5a). We confirmed enriched H3K4me1 and H3K27ac signal intensities at the Mmp3 locus (Fig. 5b) and significantly upregulated Mmp3 expression in both 168FARN (Fig. 5c) and 67NR KO derivatives (Fig. 5d). Of note, MMP3 was also a top hit in our secreted protein analysis (Extended Data Fig. 2o). Importantly, MMP3 expression is significantly higher in human TNBC with KMT2C mutations compared with WT tumours, and MMP3-high tumours have significantly higher frequencies of KMT2C mutation (Fig. 5e and Extended Data Fig. 5a).

**Fig. 5: *Mmp3* drives brain metastasis of *Kmt2c* and *Kmt2d* KO cells.**

To functionally characterize Mmp3, we downregulated it using small hairpin RNA (shRNA) in 168FARN Kmt2c and Kmt2d KO and WT cells (Extended Data Fig. 5b). The cells were injected intracardially into mice and metastatic cells were quantified by flow cytometry. Mmp3 downregulation significantly reduced brain metastases in both Kmt2c and Kmt2d KO cells (Fig. 5f,g). Previous clinical trials testing MMP inhibitors showed high toxicity and low efficacy¹⁸. However, our data suggest that KDM6A might be a mediator of Kmt2c and Kmt2d loss-associated transcriptomic changes (Fig. 3c,d); thus, inhibiting KDM6A might be an indirect way of targeting MMP3. We therefore generated 168FARN Kmt2c and Kmt2d KO and WT cells expressing Kdm6a-targeting shRNAs (Extended Data Fig. 5c). Kdm6a downregulation led to a significant decrease in Mmp3 expression (Fig. 6a) and significantly fewer brain metastases of Kmt2c KO and Kmt2d KO cells (Fig. 6b,c). The frequencies of bone metastases were not affected by knockdown of Mmp3 or Kdm6a (Extended Data Fig. 5d). Similarly, downregulation of neither Mmp3 nor Kdm6a had a significant effect on primary mammary tumour growth (Extended Data Fig. 5e).

**Fig. 6: Inhibition of KDM6A reduces *Mmp3* expression and brain metastases induced by *Kmt2c* or *Kmt2d* KO.**

To explore the potential mechanisms of Mmp3 upregulation in Kmt2c and Kmt2d KO cells, we first treated cells with inhibitors of KDM6A (GSK-J4), KDM1A (ORY-1001) and P300 (A-485). GSK-J4 treatment decreased H3K27ac in Kmt2c and Kmt2d KO but not WT 168FARN cells without affecting H3K27me3 (Extended Data Fig. 5f). Similar to Kdm6a downregulation, GSK-J4 treatment significantly reduced Mmp3 expression in the 168FARN and 67NR cell lines of both KOs, but not in WT 168FARN and 67NR cells (Fig. 6d and Extended Data Fig. 5g). Inhibition of KDM1A increased H3K4me1 in Kmt2c and Kmt2d KO cells but not in WT cells, while P300 inhibition decreased H3K27ac in all cell lines (Extended Data Fig. 6a). KDM1A inhibition did not affect Mmp3 levels, whereas P300 inhibition significantly reduced Mmp3 expression, specifically in Kmt2c and Kmt2d KO cells (Fig. 6e). Quantification of reads per million in our ChIP-seq data in the Mmp3 promoter region showed a consistent decrease of H3K27me3 and an increase of P300 in Kmt2c and Kmt2d KO cells compared with WT cells, whereas KDM6A did not show consistent changes, indicating that KDM6A does not affect Mmp3 expression via promoter binding (Extended Data Fig. 6b). The MMP gene cluster locus containing MMP1, MMP3, MMP10, MMP12 and MMP13 has been shown to be coordinately regulated, putatively by distal elements, resulting in correlative expression patterns¹⁹. Accordingly, we found a significant increase of Mmp1b, Mmp10 and Mmp13 in Kmt2c KO cells and a significant increase of Mmp10 and Mmp13 in Kmt2d KO cells (Extended Data Fig. 6c). Fold changes of signal intensities within this cluster again showed increased H3K4me1, H3K27ac and P300 and decreased H3K27me3, whereas KDM6A did not show consistent changes (Extended Data Fig. 6d,e). However, inhibition of KDM6A or P300 decreased the expression of Mmp1b and Mmp10 in Kmt2c and Kmt2d KO cells, supporting a role for distal regulatory functions of KDM6A (Extended Data Fig. 6f). In line with this, KDM6A binding exclusively in promoter regions does not correlate with upregulated gene expression (Extended Data Fig. 6g). However, there was a progressively increased significance of correlations when assessing KDM6A peaks with increasing distances from the transcription start site (Extended Data Fig. 6h). Mutations in KMT2D have been shown to disrupt COMPASS complex formation⁹, suggesting a mechanism for altered KDM6A activity. However, we did not observe any differences in COMPASS complex components between WT and Kmt2c or Kmt2d KO cells in KDM6A immunoprecipitates (Extended Data Fig. 6i).

Finally, we assessed the therapeutic potential of KDM6A inhibition to prevent brain metastases. We pre-treated cells and mice with GSK-J4 for 2 d, followed by intracardiac injection and continued GSK-J4 treatment for 12 d. KDM6A inhibitor treatment was well tolerated and we did not observe any obvious alterations in vital organs such as the liver, intestine, spleen or kidneys (Extended Data Fig. 7a,b). However, we found that KDM6A inhibition significantly reduced Kmt2c KO cells in the brain, with a similar trend for Kmt2d KO cells (Fig. 6f and Extended Data Fig. 7c), whereas it did not affect the incidence of bone metastases (Extended Data Fig. 7d). We then repeated intracardiac injection of cells pre-treated with GSK-J4 for 2 d and harvested the brains 2 d later (Extended Data Fig. 7e). We detected significantly fewer Kmt2c KO cells in the brain with a similar trend observed for Kmt2d KO, suggesting an effect at early metastatic colonization (Extended Data Fig. 7f). Taken together, these results show that indirect inhibition of MMP3 via targeting KDM6A prevents metastasis of KMT2C or KMT2D mutant tumours (Fig. 4g).

Discussion

Perturbed epigenetic programs due to mutations in histone-modifying enzymes, including KMT2C, can promote metastasis^10,11. However, the underlying mechanisms are poorly understood, creating a major obstacle for therapeutic targeting of epigenetic drivers. Loss of both Kmt2c and Kmt2d can induce EMT or alter the EMT balance, which has been associated with metastasis^10,11. However, we observed metastasis-promoting effects even in the absence of EMT.

KMT2C has been shown to affect DNA damage repair²⁰, whereas KMT2D is also commonly mutated in human mismatch repair-deficient tumours²¹ with a superior response to ICI¹⁴. Additionally, KMT2D was identified as a determinant of the response to ICI. Consistently, we observed that Kmt2c or Kmt2d KO tumours have evidence of both immune activation and suppression.

Differential chromatin binding of monomethylated H3K4 in Kmt2c and Kmt2d KO cells and changes in H3K27 modifications and KDM6A binding imply involvement of other COMPASS complex components leading to broader epigenetic alterations. Indeed, in Drosophila, the KDM6A-interacting domain of Trr (KMT2C/KMT2D) was sufficient to rescue Trr-null lethality by stabilizing Utx (KDM6A)⁷. However, this might only happen in double mutants, which frequently occur in bladder or colorectal cancer but not in other cancer types²². Loss of KDM6A can increase KMT2D chromatin binding, implying a rescue function despite their distinct enzymatic function²³. In line with this, our data showed increased KDM6A binding upon Kmt2c or Kmt2d loss. This might also explain our finding of increased H3K4me1 and H3K27ac in contrast with previous publications^24,25. H3K4me1 and H3K27me3 inversely correlate in a KDM6A-dependent manner, suggesting an indirect impact of KDM6A on H3K4me1 regulation²⁶. Higher bimodal H3K4me1 signal at promoters correlates with increased gene activity²⁷, suggesting that H3K4me1 is also indirectly regulated upon Kmt2c and Kmt2d KO.

MMP3 has been shown to promote mammary tumorigenesis and has been proposed as a biomarker of cancer risk and tumour progression in patients with breast cancer²⁸. An unanswered question in our study is whether MMP3 only plays a role in metastatic seeding or if its expression is also sustained in established metastases. MMP family members are important in tumour growth and metastasis, but MMP inhibitors have not been successful in clinical trials, in part due to their toxicity²⁹. However, our study provides an alternative approach for targeting MMP3 via KDM6A inhibition to prevent brain metastases of KMT2C or KMT2D mutant tumours. As KDM6A inhibition seems to be well tolerated in mice, it could potentially be used in patients with TNBC who have KMT2C or KMT2D mutant tumours, in combination with currently used therapies. Alternative strategies are combinations including P300 and bromodomain and extra-terminal protein inhibitors, due to gained H3K27ac in Kmt2c/Kmt2d KO cells. Bromodomain and extra-terminal protein inhibitors have been shown to be more effective in Kmt2c KO cancer cells¹¹.

A remaining question is why KDM6A chromatin binding is enhanced upon Kmt2c or Kmt2d KO. Less is known about the redundancy of the COMPASS complexes and whether loss of either can rescue or even increase the activity of the other. It is likely that other COMPASS components beyond KDM6A also play a role in Kmt2c or Kmt2d loss-induced epigenetic and phenotypic changes. Our multi-dimensional profiling data provide a rich resource for further investigations into these unexplored areas.

Methods

Ethical compliance

This research complies with all of the ethical and safety rules. Animal work was performed following protocol 11-023, approved by the Dana-Farber Cancer Institute Animal Care and Use Committee.

Mammary tumour cell lines

168FARN and 67NR murine mammary tumour cell lines were obtained from the Karmanos Cancer Institute. HEK293T (CRL-3216) cells were obtained from the American Type Culture Collection and cultured following the provider’s recommendations. Cell lines were regularly tested for Mycoplasma.

Animal models

For MFP injection and intracardiac seeding, female BALB/c or NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were purchased from The Jackson Laboratory at 5–6 weeks of age. Animal experiments were performed according to protocol 11-023 approved by the Dana-Farber Cancer Institute Animal Care and Use Committee. The maximum tumour size limit was 2 cm in any dimension and was not exceeded in this study. Mice were housed five to a cage with ad libitum access to food and water under an ambient temperature of 20 °C with a humidity of 40–50% and a 12 h light/12 h dark cycle.

Cell line models

Single-guide RNA sequences (Supplementary Table 5) against exon 1 of murine Kmt2c or Kmt2d were chosen using University of California Santa Cruz Genome Browser CRISPR target tracks with a Doench/Fusi 2016 efficiency of >55. Three guide sequences each were then cloned into an sgRNA scaffold mCherry-tagged vector using BsmBI. Target cells were co-transfected with three sgRNA vectors and Cas9–GFP vector in a 1:1:1:3 ratio using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s protocol. After 2 d, mCherry/GFP double-positive cells were sorted via flow cytometry and cultured for 8 d until transient expression of the fluorescence marker was lost. The process was repeated two additional times. Finally, cells were sorted for mCherry/GFP double negativity to select out cells with random stable integration of either of the vectors. To stably integrate H2B–mCherry or shRNA vectors for knockdown of Mmp3 or Kdm6a, HEK293T cells were transfected with packaging plasmids (pMD2.G and psPAX2) and respective vectors in a 1:2:2 ratio using Lipofectamine 3000 (Invitrogen). After 2 d, the supernatant was harvested and filtered through 0.22 µm filters, then 200 µl was used to infect target cells with the addition of polybrene (4 µg ml⁻¹) for 2 d. Cells were then either selected using puromycin (2 µg ml⁻¹) or sorted for mCherry positivity. For knockdown of Mmp3 and Kdm6a, three different SMARTvector Lentiviral shRNA constructs incorporating the mEF1α promoter but no fluorescence marker (Horizon Discovery) were used.

Mouse tumorigenesis and metastasis assays

For MFP or intracardiac injections, mice were anaesthetized by constant isoflurane application. Instruments and injection sites were sterilized, and for MFP injection a 1 cm incision was made between the midline and the fourth nipple. Using insulin syringes, 200,000 cells in 50 µl phosphate-buffered saline (PBS) were injected into the MFP and the incision was closed with surgical clips. Post-operational analgesia was ensured with topical ropivacaine application. For intracardiac injection, 50,000 cells in 50 µl PBS were injected directly into the left ventricle. Successful injection was validated by controlling blood flow into the syringe before and after the procedure. For KDM6A inhibition experiments, mice were treated with GSK-J4 HCl (Selleckchem; 10 mg kg⁻¹ in PBS) via intraperitoneal injection every other day, starting 2 d before intracardiac injection and lasting until the experimental endpoint.

Cell proliferation and viability assays

Cells were cultured at 37 °C under 5% CO₂ and humidified conditions. For proliferation or dose curve analyses, 500 cells were plated in 96-well plates and treated with the indicated drugs in dimethyl sulfoxide (to a final concentration of 0.1%) or left untreated. At 5 d after treatment, the start cell count was measured with a Celigo Image Cytometer (Nexcelom Bioscience) using the mCherry signal of the cells. For proliferation analyses, the cell count was analysed every day for 5 d.

Immunoblot analyses

Cells were collected and proteins were isolated using isolation buffer (1 mM ethylene glycol tetraacetic acid, 1 mM ethylenediaminetetraacetic acid (EDTA), 150 mM NaCl, 20 mM Tris and 1% Triton-X) for 30 min on ice after passing through a QIAshredder (QIAGEN). Histones were isolated using the EpiQuik Total Histone Extraction Kit (EpigenTek). The protein concentration was measured using Pierce 660 nm Protein Assay Reagent (Thermo Fisher Scientific) and equal amounts were loaded onto 3–8% gradient Tris-Acetate or 10% Tricine (histone extracts) gels. Proteins were transferred onto polyvinylidene difluoride membranes using wet blot devices and transfer buffer containing either 20% (histone extracts) or 10% (other extracts) methanol for 1 h (histone extracts) or 4 h (other extracts) at 90 V. Membranes were incubated with the primary antibodies (Supplementary Table 5) KMT2C (a gift from A. Shilatifard; 1:1,000), KMT2D (orb18454; Biorbyt; 1:1,000), tubulin (T6199; Millipore; 1:20,000), total histone H3 (39736; Active Motif; 1:1,000), H3K4me1 (710795; Invitrogen; 1:1,000), H3K27me3 (39155; Active Motif; 1:1,000), H3K27ac (ab4729; Abcam; 1:1,000), PAX-interacting protein 1 (ABE1877; Sigma–Aldrich; 1:1,000), RBBP5 (13171S; Cell Signaling Technology; 1:1,000) or WDR5 (13105S; Cell Signaling Technology; 1:1,000) in 5% milk overnight and with the secondary antibodies anti-mouse HRP (62-6520; Invitrogen; 1:10,000) or anti-rabbit HRP (65-6120; Invitrogen; 1:10,000) in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 h. Signals were developed with Clarity Western ECL Substrate (Bio-Rad) on a ChemiDoc MP device (Bio-Rad). If needed, antibodies were stripped off from membranes using Restore Western Blot Stripping Buffer (Thermo Fisher Scientific).

Combined immunoprecipitation and immunoblot analyses

Approximately 15 million cells were collected and lysed in lysis buffer (300 mM NaCl, 50 mM Tris (pH 7.5), 1% IGEPAL CA-630 and 0.1% sodium deoxycholate) for 5 min in a cooled water bath sonicator and passed five times through a 30 G syringe. Next, 1.5 mg protein lysate was diluted up to 1 ml with low IP buffer from the Nuclear Complex Co-IP Kit (Active Motif), then 2.5 µg antibody (KDM6A; 33510S; Cell Signaling Technology) was added before incubation on a rotator at 4 °C. Following this, 50 µl Dynabeads Protein G was added for 2 h on a rotator at 4 °C. The beads were washed four times with washing buffer (150 mM NaCl and 50 mM Tris (pH 7.5)) and the proteins were eluted with 40 µl 4× LDS buffer at 95 °C for 5 min.

Immunofluorescence staining

Tissue sections were deparaffinized and antigen retrieval was performed using Target Retrieval Solution, pH 6 (Agilent) for 40 min with a non-pressure food steamer. Slides were incubated with the primary antibodies mCherry (43590S; Cell Signaling Technology), CD8 (98941S; Cell Signaling Technology) and PDL-1 (64988S; Cell Signaling Technology) in 5% normal goat serum in TBST overnight and the secondary antibodies goat anti-rabbit Alexa Fluor 647 (A-21245; Invitrogen) and goat anti-rabbit Alexa Fluor 555 (A-21428; Invitrogen) in 5% normal goat serum in TBST for 2 h. Endogenous fluorescence was quenched using a TrueVIEW Autofluorescence Quenching Kit (Vector Laboratories) for 5 min. Images were taken with a Nikon ECLIPSE Ti2-E fluorescence microscope. Quantification was performed with automated scripts using Fiji (ImageJ) from three to five different fields of view per section. Values within each replicate were averaged to calculate the final value per replicate.

Flow cytometry

Immediately after collection, tissues were smashed with micro pestles and digested for 10 min (brain and bone marrow) or 1 h (tumour, liver and lungs) using digestion media (2% wt/vol collagenase IV, 2% wt/vol hyaluronidase and 2% wt/vol bovine serum albumin in Dulbecco’s modified Eagle medium) at 37 °C on a shaker. Solutions were filtered through a mesh, washed with PBS and frozen in 10% dimethyl sulfoxide/foetal bovine serum at −80 °C or directly used for flow cytometry. For this, cells were passed through a 70 or 100 μm cell strainer, incubated with 4′,6-diamidino-2-phenylindol (1:20,000) and analysed using an LSRFortessa (BD Biosciences). Gating strategies can be found in Supplementary Fig. 1.

Next-generation sequencing amplicon

Cells were washed with PBS and DNA was isolated using DNeasy Blood & Tissue Kits (QIAGEN). Regions of interest were amplified via PCR using primers for Kmt2c or Kmt2d (Supplementary Table 5) and run on an agarose gel electrophoresis. PCR products were purified from agarose gel using the Monarch DNA Gel Extraction Kit (New England Biolabs) and sequenced with GENEWIZ (Azenta).

Quantitative real-time PCR

RNA was isolated using the Monarch Total RNA Miniprep Kit (New England Biolabs). Then, 2 μg RNA per reaction was used for reverse transcription with the PrimeScript RT Reagent Kit (Takara Bio). Real-time PCR was performed using TB Green Premix Ex Taq II (Takara Bio) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Values were calculated using the ∆∆Ct method normalized to actin expression. Primer sequences for β-actin, Mmp3, Mmp1b, Mmp10, Mmp13 and Kdm6a are indicated in Supplementary Table 5.

Cytokine array

Cells were washed with PBS and cultured in foetal bovine serum-free Opti-MEM (Gibco) for 24 h before the supernatant was collected. Snap-frozen tumour tissue was mechanically homogenized and lysed with 1% Triton-X in PBS and the supernatant was collected. For each experimental condition, the protein concentration in the supernatant from five biological replicates was measured using Pierce 660 nm Protein Assay Reagent (Thermo Fisher Scientific) and the cytokine concentration was determined with a Proteome Profiler Mouse XL Cytokine Array (R&D Systems) on a ChemiDoc MP (Bio-Rad) device.

Histone mass spectrometry

Cells were washed twice with cold PBS scraped off the plates, collected via centrifugation and snap frozen at −80 °C. Histones were isolated via acidic extraction followed by trichloroacetic acid precipitation, as previously described³¹. Briefly, 10 µg histone extract of each sample was propionylated, desalted and digested with trypsin overnight. The generated peptides were propionylated, desalted and reconstituted for mass spectrometry analysis. A reference mixture of isotopically labelled synthetic peptides for histones H3 and H4 was added to each sample. Peptides were separated on a C18 column (EASY-nLC 1000; Thermo Fisher Scientific) and analysed by mass spectrometry using the parallel reaction monitoring method (Q Exactive Plus Orbitrap; Thermo Fisher Scientific). Chromatographic peak areas of endogenous (light; L) and synthetic standard (heavy; H) peptides were extracted in Skyline and the ratios of light to heavy peak areas (L:H) were calculated. Ratios were normalized to respective ratios of a typically unmodified region of H3 (41–49) or H4 (68–78) and were log₂ transformed. Kmt2c KO and Kmt2d KO cells were further normalized to the WT for each histone mark.

Single-cell RNA-seq preparation and sequencing

Samples from tumour tissues were prepared as described for flow cytometry. To deplete dead cells and debris, Percoll purification was performed. First, layers of 50 and 40% Percoll solution (90% Percoll and 10 mM HEPES in PBS) were prepared in 15 ml conical centrifugation tubes. Cells were pelleted, resuspended in 20% Percoll solution and layered on top of the former layers. Samples were centrifuged for 30 min at 2,000g and 4 °C without a break. Then, white layer interphases between 20 and 40% layers were collected, washed with PBS, filtered through a 100 μm strainer and pelleted. Pellets were resuspended in 0.04% UltraPure BSA (Sigma–Aldrich) in PBS and immediately processed for library preparations. Approximately 26,000 single cells were loaded onto a 10x Genomics Chromium instrument (10x Genomics) according to the manufacturer’s recommendations. The scRNA-seq libraries were generated using a Chromium Next GEM Single Cell 5’ HT v2 Kit (10x Genomics). Quality controls for amplified complementary DNA libraries and final sequencing libraries were performed using a Bioanalyzer High Sensitivity DNA Kit (Agilent). Equimolar ratios of libraries were sequenced on an Illumina NovaSeq 6000 (Illumina) targeting 40 million 150-base pair (bp) read pairs per library at the Dana-Farber Cancer Institute Molecular Biology Core Facilities.

Single-cell RNA-seq analyses

Raw data were processed using Cell Ranger (10x Genomics) with the bcl2fastq function to obtain fastq files for each sample. Files were further processed using Cell Ranger Count 7.0.1 and Cell Ranger Aggr version 7.0.1 (10x Genomics) to obtain counts and aggregate samples into groups. Countmatrix files were then processed in R studio using the Seurat package. Low-quality reads were removed according to the following criteria: nFeature_RNA > 500 & nFeature_RNA < 7500 & percent.mt < 15. Normalization was performed using LogNormalize followed by SCTransform for regression of the mitochondrial percentage and cell cycle (S.Score and G2M.Score; Supplementary Table 1). Clustering was performed using a resolution of 0.2 with the top 30 principal components. Next, clusters were manually annotated using gene.module.scores with specific marker genes (Supplementary Table 1). For further characterization, the clusters tumour cells and non-tumour cells were first subset. Non-tumour cells were then subset into the clusters fibroblast, endothelial cells, macrophage and T cell. Within each subset, cells were again annotated using gene.module.scores, and scores were plotted according to each genotype.

RNA-seq preparation and sequencing

Cells were washed with PBS and collected via trypsinization. RNA was isolated using the RNeasy Mini Kit (QIAGEN). Libraries were prepared using KAPA mRNA HyperPrep (Roche) strand-specific sample preparation kits from 100 ng purified total RNA, according to the manufacturer’s protocol, on a Biomek i7 (Beckman Coulter). The finished double-stranded DNA libraries were quantified by Qubit fluorometer (Thermo Fisher Scientific) and 4200 TapeStation (Agilent). Uniquely dual-indexed libraries were pooled in an equimolar ratio and shallowly sequenced on an Illumina MiSeq (Illumina) to further evaluate library quality and pool balance. The final pool was sequenced on an Illumina NovaSeq 6000 (Illumina) targeting 40 million 150-bp read pairs per library at the Dana-Farber Cancer Institute Molecular Biology Core Facilities.

RNA-seq analyses

RNA-seq data were analysed using the VIPER³² pipeline. Briefly, reads were aligned using STAR to the mm9 mouse genome. Genes with no counts in all samples were excluded and the remaining counts were normalized via log₂-transformed trimmed mean of M values transformation to counts per million (log₂[TMM–CPM + 1] from edgeR)³³ for further processing. To plot the heatmaps, batch effects among different biological replicates were removed using the removeBatchEffect function of the LIMMA³⁴ package, considering each collection time as an individual batch, and z scores normalized to the average of all WT samples were plotted. For differential expression analyses, DESeq2 (ref. ³⁵) was performed using genotypes as factor levels using non-normalized counts.

ChIP-seq library preparation and sequencing

Cells were washed with cold PBS and fixed with fixing buffer (1% paraformaldehyde, 0.05 M HEPES (pH 7.5), 0.1 M NaCl and 1 μM EDTA (pH 8.0)) for 10 min at room temperature on a shaker. Fixation was stopped by adding 1/10 vol/vol 1.25 M glycine for 5 min. Cells were scraped off the plate, pelleted via centrifugation and stored at −80 °C. Then, cell membranes were lysed using lysis buffer (0.25% Triton-X, 0.5% IGEPAL CA-630, 10% glycerol, 0.5 μM EDTA, 0.14 M NaCl and 0.05 M HEPES (pH 8.0)) for 10 min at 4 °C and nuclei preparations were collected via centrifugation. Nuclei were washed with washing buffer (0.5 μM EDTA, 0.2 M NaCl and 0.01 M Tris-HCl) for 10 min at 4 °C, washed again and subsequently resuspended in shearing buffer (0.01 M Tris-HCl, 0.5 μM EDTA and 0.1% sodium dodecyl sulfate (SDS)). The solution was transferred into AFA Fiber tubes and sonicated with a Covaris E220 ultrasonicator (Covaris; peak incident power = 150 W, duty cycles = 5%; cycles per burst = 200) for 15 min at 4 °C. Then, debris were pelleted and the supernatant was transferred to new reaction tubes before the addition of Triton-X (1%), NaCl (0.14 M) and pre-washed Dynabeads Protein G (Thermo Fisher Scientific) and incubation for 1 h at 4 °C on a rotator. The supernatant was collected and the antibodies H3K4me1 (ab8895; Abcam; 2.5 μg per ChIP), H3K27me3 (9733S; Cell Signaling Technology; 2.5 μg per ChIP), H3K27ac (C15410196; Diagenode; 2.5 μg per ChIP), KDM6A (33510S; Cell Signaling Technology; 2.5 μg per ChIP) or P300 (ab275378; Abcam; 2.5 μg per ChIP) were added before incubation at 4 °C on a rotator overnight. The next day, pre-washed Dynabeads Protein G were added before incubation at 4 °C on a rotator for 2 h. The supernatant was removed and the beads were washed with low salt wash buffer (2 mM EDTA, 1% Triton-X, 0.1% SDS, 0.15 M NaCl and 0.02 M Tris-HCl (pH 8.0)), high salt wash buffer (2 mM EDTA, 1% Triton-X, 0.1% SDS, 0.5 M NaCl and 0.02 M Tris-HCl (pH 8.0)) and LiCl wash buffer (1 mM EDTA, 1% sodium deoxycholate, 1% IGEPAL CA-630, 0.25 M LiCl and 0.01 M Tris-HCl (pH 8.0)) at 4 °C for 5 min each. Beads were washed twice with TE buffer, and elution buffer (1% SDS and 100 mM NaHCO₃) was added and incubated for 30 min while being vortexed every 5 min at room temperature. The beads were removed and the supernatant was incubated at 65 °C overnight. Then, 0.2 mg ml^–1 RNase A was added before incubation for 30 min at 37 °C. Subsequently, 0.2 mg ml^–1 proteinase K was added before further incubation for 1 h at 55 °C. DNA was extracted by adding 1:1 vol/vol phenol/chloroform (pH 8.0) and the upper phase was used for DNA precipitation by adding isopropanol (42.5%), glycogen (0.007%) and sodium perchlorate (0.28 M), then pelleted using centrifugation. The DNA pellet was washed with 70% ethanol and resuspended in low TE buffer. ChIP-seq libraries were prepared using xGen DNA Library Prep reagents (Integrated DNA Technologies) on a Biomek i7 (Beckman Coulter) liquid-handling platform from approximately 1 ng DNA with 14 cycles of PCR amplification, according to the manufacturer’s protocol. Finished sequencing libraries were quantified by Qubit fluorometer (Thermo Fisher Scientific) and 2200 TapeStation (Agilent). Library pooling and indexing was evaluated with shallow sequencing on an Illumina MiSeq (Illumina). Subsequently, libraries were sequenced on an Illumina NovaSeq 6000 (Illumina) targeting 40 million 150-bp read pairs by the Molecular Biology Core Facilities at the Dana-Farber Cancer Institute.

ChIP-seq analyses

ChIP-seq data were analysed using the CoBRA³⁶ pipeline. In brief, reads were aligned to the mm9 genome using the BWA-MEM³⁷ aligner and peaks were called using MACS2 (ref. ³⁸) (false discovery rate < 0.01; broad peak mode for H3K4me1, H3K27me3, KDM6A and P300; default for H3K27ac). Reads in peaks were then counted using SAMtools³⁹ and used for differential analysis with DESeq2 (ref. ³⁵) (P value < 0.05; log₂[fold change] > 0.5) normalized to sequencing depth. Heatmaps were generated using individual BigWig files and differential peak files for each comparison were generated using computeMatrix and plotHeatmap excluding blacklisted regions. The intersection of differential peaks was identified using Bedtools⁴⁰ with at least a 1-bp overlap. Motif analysis for each differential peakset was performed using HOMER⁴¹ (-size 200 -p 10), and de novo motifs were considered. For motif heatmaps, only motifs with a −log₁₀[P value] > 25 in at least one sample were plotted. Correlation of differential peaks and gene expression was performed using BETA¹⁷. Each individual differential gained or lost peakset was analysed together with the DESeq2 output from RNA-seq of the same cells but different biological replicates using default parameters. For quantification of P300 signal intensities within H3K27ac peaks, multiBigwigSummary was used to bin the P300 signal of each biological replicate into specific peak regions. The bin count of replicates was averaged and log₂ transformed. Density plots for the correlation of changed P300 and H3K27ac signals was generated with ggplot. DESeq2 outputs from the CoBRA pipeline were used to extract all of the calculated fold changes within identified peak regions for H3K27ac or P300 in Kmt2c or Kmt2d KO cells compared with WT cells. Then, multiBigwigSummary⁴² was used with a bin size of 100 to calculate the fold changes of H3K27ac or P300 in a similar region. Quantification of read counts for H3K27me3, P300 and KDM6A within the Mmp3 promoter (transcription start site ±1 kilobases) was performed using featureCounts⁴³ with respective bam files. For comparisons of signal intensities across the Mmp3 cluster, the locus signal of biological replicates was first averaged using bigwigAverage⁴². Then, average values were compared using bigwigCompare⁴² in a 1,000-bp bin with the log2FC mode. The overall signal intensity was then calculated from average values of all of the bins.

Analyses of public data

Mutation frequencies, locations at protein sequences and gene expression were identified using the cBioPortal platform (https://www.cbioportal.org/) selecting for patients with TNBC (oestrogen receptor, progesterone receptor and HER2 status negative) from the Molecular Taxonomy of Breast Cancer International Consortium cohort. This cohort consists of 320 patients, with 299 samples having mutation information and 320 samples having gene expression information. Gene expression for matched primary and secondary tumours was extracted from ref. ³⁰ with selection for basal subtype. Plots for the correlation of metastatic potential scores and gene expression for cell lines were generated using data from the DepMap database (https://depmap.org/portal/).

Statistics and reproducibility

Experiments and sample size were designed and determined according to similar experiments in previous studies. No statistical method was used to predetermine the required sample size. All repeat experiments are included in the analyses as biological replicates. Only animals with technical failures of cardiac injection were excluded. Metastasis quantification was performed after randomization of the samples. The investigators were not blinded to allocation during the experiments and outcome assessment. Data were first tested for normal distribution using the Shapiro–Wilk test. If data for matched comparisons did not pass the test, non-parametric tests were used. For single comparisons, a non-paired (unless otherwise noted), two-tailed t-test or Mann–Whitney U-test was used. For multiple comparisons, one-way analysis of variance corrected for multiple comparisons or a Kruskal–Wallis test corrected for multiple comparison was used. Unless otherwise noted, multiple comparison was performed for Kmt2c KO and Kmt2d KO versus the WT. All tests were performed with a 95% confidence interval. P values are indicated for each experiment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All of the data needed to evaluate the conclusions in the paper are present in the paper and/or its Supplementary Information. RNA-seq, scRNA-seq and ChIP-seq data that support the findings of this study have been deposited in the Gene Expression Omnibus under the accession code GSE237392. The human TNBC data were derived from the Molecular Taxonomy of Breast Cancer International Consortium cohort. The dataset derived from this resource that supports the findings of this study is available at https://www.cbioportal.org/study/summary?id=brca_metabric. Mass spectrometry data have been deposited in ProteomeXchange with the primary accession code PXD052075. The mm9 mouse genome dataset is available at https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000001635.18/. Source data are provided with this paper. All of the other data supporting the findings of this study are available from the corresponding author upon reasonable request.

References

Sung, H. et al. Global Cancer Statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. 71, 209–249 (2021).
Article PubMed Google Scholar
Garrido-Castro, A. C., Lin, N. U. & Polyak, K. Insights into molecular classifications of triple-negative breast cancer: improving patient selection for treatment. Cancer Discov. 9, 176–198 (2019).
Article CAS PubMed PubMed Central Google Scholar
Darlix, A. et al. Impact of breast cancer molecular subtypes on the incidence, kinetics and prognosis of central nervous system metastases in a large multicentre real-life cohort. Br. J. Cancer 121, 991–1000 (2019).
Article CAS PubMed PubMed Central Google Scholar
Bertucci, F. et al. Genomic characterization of metastatic breast cancers. Nature 569, 560–564 (2019).
Article CAS PubMed Google Scholar
Morgan, A. J., Giannoudis, A. & Palmieri, C. The genomic landscape of breast cancer brain metastases: a systematic review. Lancet Oncol. 22, e7–e17 (2021).
Article CAS PubMed Google Scholar
Sze, C. C. & Shilatifard, A. MLL3/MLL4/COMPASS family on epigenetic regulation of enhancer function and cancer. Cold Spring Harb. Perspect. Med. 6, a026427 (2016).
Article PubMed PubMed Central Google Scholar
Rickels, R. et al. A small UTX stabilization domain of Trr is conserved within mammalian MLL3-4/COMPASS and is sufficient to rescue loss of viability in null animals. Genes Dev. 34, 1493–1502 (2020).
Article CAS PubMed PubMed Central Google Scholar
Shilatifard, A. The COMPASS family of histone H3K4 methylases: mechanisms of regulation in development and disease pathogenesis. Annu. Rev. Biochem. 81, 65–95 (2012).
Article CAS PubMed PubMed Central Google Scholar
Cocciadiferro, D. et al. Dissecting KMT2D missense mutations in Kabuki syndrome patients. Hum. Mol. Genet. 27, 3651–3668 (2018).
Article CAS PubMed PubMed Central Google Scholar
Zhang, Y. et al. Genome-wide CRISPR screen identifies PRC2 and KMT2D-COMPASS as regulators of distinct EMT trajectories that contribute differentially to metastasis. Nat. Cell Biol. 24, 554–564 (2022).
Article CAS PubMed PubMed Central Google Scholar
Cui, J. et al. MLL3 loss drives metastasis by promoting a hybrid epithelial–mesenchymal transition state. Nat. Cell Biol. 25, 145–158 (2023).
Article CAS PubMed PubMed Central Google Scholar
Na, F. et al. KMT2C deficiency promotes small cell lung cancer metastasis through DNMT3A-mediated epigenetic reprogramming. Nat. Cancer 3, 753–767 (2022).
Article CAS PubMed PubMed Central Google Scholar
Jin, X. et al. A metastasis map of human cancer cell lines. Nature 588, 331–336 (2020).
Article CAS PubMed PubMed Central Google Scholar
Wang, G. et al. CRISPR-GEMM pooled mutagenic screening identifies KMT2D as a major modulator of immune checkpoint blockade. Cancer Discov. 10, 1912–1933 (2020).
Article CAS PubMed PubMed Central Google Scholar
Appay, V. & Rowland-Jones, S. L. RANTES: a versatile and controversial chemokine. Trends Immunol. 22, 83–87 (2001).
Article CAS PubMed Google Scholar
Qin, Q. et al. Lisa: inferring transcriptional regulators through integrative modeling of public chromatin accessibility and ChIP-seq data. Genome Biol. 21, 32 (2020).
Article PubMed PubMed Central Google Scholar
Wang, S. et al. Target analysis by integration of transcriptome and ChIP-seq data with BETA. Nat. Protoc. 8, 2502–2515 (2013).
Article CAS PubMed PubMed Central Google Scholar
Vandenbroucke, R. E. & Libert, C. Is there new hope for therapeutic matrix metalloproteinase inhibition? Nat. Rev. Drug Discov. 13, 904–927 (2014).
Article CAS PubMed Google Scholar
Llinas-Arias, P. et al. Chromatin insulation orchestrates matrix metalloproteinase gene cluster expression reprogramming in aggressive breast cancer tumors. Mol. Cancer 22, 190 (2023).
Article CAS PubMed PubMed Central Google Scholar
Chang, A. et al. Recruitment of KMT2C/MLL3 to DNA damage sites mediates DNA damage responses and regulates PARP inhibitor sensitivity in cancer. Cancer Res. 81, 3358–3373 (2021).
Article CAS PubMed PubMed Central Google Scholar
Papke, D. J. Jr. et al. Validation of a targeted next-generation sequencing approach to detect mismatch repair deficiency in colorectal adenocarcinoma. Mod. Pathol. 31, 1882–1890 (2018).
Article CAS PubMed Google Scholar
Zehir, A. et al. Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients. Nat. Med. 23, 703–713 (2017).
Article CAS PubMed PubMed Central Google Scholar
Andricovich, J. et al. Loss of KDM6A activates super-enhancers to induce gender-specific squamous-like pancreatic cancer and confers sensitivity to BET inhibitors. Cancer Cell 33, 512–526.e8 (2018).
Article CAS PubMed PubMed Central Google Scholar
Chen, R. et al. Kmt2c mutations enhance HSC self-renewal capacity and convey a selective advantage after chemotherapy. Cell Rep. 34, 108751 (2021).
Article CAS PubMed PubMed Central Google Scholar
Gala, K. et al. KMT2C mediates the estrogen dependence of breast cancer through regulation of ERα enhancer function. Oncogene 37, 4692–4710 (2018).
Article CAS PubMed PubMed Central Google Scholar
Duplaquet, L. et al. KDM6A epigenetically regulates subtype plasticity in small cell lung cancer. Nat. Cell Biol. 25, 1346–1358 (2023).
Article CAS PubMed Google Scholar
Bae, S. & Lesch, B. J. H3K4me1 distribution predicts transcription state and poising at promoters. Front. Cell Dev. Biol. 8, 289 (2020).
Article PubMed PubMed Central Google Scholar
Duffy, M. J., Maguire, T. M., Hill, A., McDermott, E. & O’Higgins, N. Metalloproteinases: role in breast carcinogenesis, invasion and metastasis. Breast Cancer Res. 2, 252–257 (2000).
Article CAS PubMed PubMed Central Google Scholar
Fields, G. B. The rebirth of matrix metalloproteinase inhibitors: moving beyond the dogma. Cells 8, 984 (2019).
Article CAS PubMed PubMed Central Google Scholar
Siegel, M. B. et al. Integrated RNA and DNA sequencing reveals early drivers of metastatic breast cancer. J. Clin. Invest. 128, 1371–1383 (2018).
Article PubMed PubMed Central Google Scholar
Dele-Oni, D. O. et al. Proteomic profiling dataset of chemical perturbations in multiple biological backgrounds. Sci. Data 8, 226 (2021).
Article PubMed PubMed Central Google Scholar
Cornwell, M. et al. VIPER: visualization pipeline for RNA-seq, a Snakemake workflow for efficient and complete RNA-seq analysis. BMC Bioinformatics 19, 135 (2018).
Article PubMed PubMed Central Google Scholar
Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140 (2010).
Article CAS PubMed Google Scholar
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015).
Article PubMed PubMed Central Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Qiu, X. et al. CoBRA: containerized bioinformatics workflow for reproducible ChIP/ATAC-seq analysis. Genom. Proteom. Bioinform. 19, 652–661 (2021).
Article Google Scholar
Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Preprint at https://doi.org/10.48550/arXiv.1303.3997 (2013).
Feng, J., Liu, T., Qin, B., Zhang, Y. & Liu, X. S. Identifying ChIP-seq enrichment using MACS. Nat. Protoc. 7, 1728–1740 (2012).
Article CAS PubMed Google Scholar
Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 25, 2078–2079 (2009).
Article PubMed PubMed Central Google Scholar
Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
Article CAS PubMed PubMed Central Google Scholar
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Article CAS PubMed PubMed Central Google Scholar
Ramirez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res. 44, W160–W165 (2016).
Article CAS PubMed PubMed Central Google Scholar
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923–930 (2014).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

We thank members of our laboratories for critical reading of this manuscript and useful discussions. We thank A. Shilatifard (Northwestern University) for providing KMT2C antibodies and S. Spisak (Dana-Farber Cancer Institute) for providing sgRNA/mCherry and Cas9/GFP plasmids. We thank the Dana-Farber Cancer Institute Molecular Biology and Flow Cytometry Core Facilities, Dana-Farber/Harvard Cancer Center Rodent Histopathology Core facility, Dana-Farber Cancer Institute Animal Resource Facilities and Translational Immunogenomics Laboratory for outstanding services. This research was supported by the National Cancer Institute (P01CA250959 to K.P. and H.W.L. and R35 CA197623 to K.P.), Ludwig Center at Harvard (to K.P.), Saverin Breast Cancer Research Fund (to K.P.), Canadian Institutes of Health Research (to M.-A.G.) and EMBO (to M.S.). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Author information

Authors and Affiliations

Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, USA
Marco Seehawer, Zheqi Li, Jun Nishida, Pierre Foidart, Ernesto Rojas-Jimenez, Marie-Anne Goyette, Pengze Yan & Kornelia Polyak
Department of Medicine, Brigham and Women’s Hospital, Boston, MA, USA
Marco Seehawer, Zheqi Li, Jun Nishida, Pierre Foidart, Ernesto Rojas-Jimenez, Marie-Anne Goyette, Pengze Yan & Kornelia Polyak
Department of Medicine, Harvard Medical School, Boston, MA, USA
Marco Seehawer, Zheqi Li, Jun Nishida, Pierre Foidart, Ernesto Rojas-Jimenez, Marie-Anne Goyette, Pengze Yan & Kornelia Polyak
The Eli and Edythe L. Broad Institute, Cambridge, MA, USA
Andrew H. Reiter, Shaunak Raval & Malvina Papanastasiou
Center for Functional Cancer Epigenetics, Dana-Farber Cancer Institute, Boston, MA, USA
Miguel Munoz Gomez, Paloma Cejas & Henry W. Long

Authors

Marco Seehawer
View author publications
You can also search for this author in PubMed Google Scholar
Zheqi Li
View author publications
You can also search for this author in PubMed Google Scholar
Jun Nishida
View author publications
You can also search for this author in PubMed Google Scholar
Pierre Foidart
View author publications
You can also search for this author in PubMed Google Scholar
Andrew H. Reiter
View author publications
You can also search for this author in PubMed Google Scholar
Ernesto Rojas-Jimenez
View author publications
You can also search for this author in PubMed Google Scholar
Marie-Anne Goyette
View author publications
You can also search for this author in PubMed Google Scholar
Pengze Yan
View author publications
You can also search for this author in PubMed Google Scholar
Shaunak Raval
View author publications
You can also search for this author in PubMed Google Scholar
Miguel Munoz Gomez
View author publications
You can also search for this author in PubMed Google Scholar
Paloma Cejas
View author publications
You can also search for this author in PubMed Google Scholar
Henry W. Long
View author publications
You can also search for this author in PubMed Google Scholar
Malvina Papanastasiou
View author publications
You can also search for this author in PubMed Google Scholar
Kornelia Polyak
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

M.S. and K.P. conceived of the study and wrote the original draft of the manuscript. M.S. developed the methodology. M.S., Z.L., A.H.R., M.P. and P.C. performed the formal analysis. M.S., Z.L., J.N., P.F., A.H.R., E.R.-J., M.-A.G., P.Y., S.R., P.C. and M.M.G. performed the investigation. H.W.L., M.P. and K.P. provided resources. H.W.L. and K.P. supervised the study and acquired funding. All authors contributed to designing the study, writing the manuscript and reviewing and editing the manuscript.

Corresponding author

Correspondence to Kornelia Polyak.

Ethics declarations

Competing interests

K.P. serves on the Scientific Advisory Board of and holds equity options in IDEAYA Biosciences and Scorpion Therapeutics. K.P. and H.W.L. receive sponsored research funding through the Dana-Farber Cancer Institute from Novartis. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Cell Biology thanks Yibin Kang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Kmt2c or Kmt2d loss does not influence therapeutic responses but enables metastasis.

a, Top 15 most commonly mutated genes from TNBC breast cancer from METABRIC cohort (n = 299). b, Lolliplot plot of KMT2C or KMT2D mutation count. Generated with cBioPortal from data depicted in Ext. Data Fig. 1a. c, Correlation plot of log2 TPM+1 values for KMT2C and KMT2D expression with metastatic potential of breast cancer cell lines from DepMap (n = 23, simple linear regression, 95% confidence bands). d and e, Amplicon NGS sequencing for 168FARN (d) and 67NR (e) cells for Kmt2c KO or Kmt2d KO with percentage of deletions or insertion compared to WT sequence. f, Western blot for KMT2C or KMT2D from 67NR WT, Kmt2c KO or Kmt2d KO cells (n = 2 biological replicates). g, Growth curve analysis of 168FARN or 67NR WT, Kmt2c KO or Kmt2d KO cells (n = 3 biological replicates per group, non-linear regression with F test testing for growth rate, mean ± s.d., two-tailed). h,i Growth curve analysis for 168FARN (h) or 67NR (i) WT, Kmt2c KO or Kmt2d KO cells treated with the indicated drugs for 5 days normalized to matched DMSO. Significant differences were only seen for doxorubicin in 168FARN cells (1.8 nM for WT and 1.1 nM for Kmt2c KO) and GSK-J4 in 168FARN cells (0.45 µM for WT, 0.24 µM for Kmt2c KO and 0.22 µM for Kmt2d KO) (n = 6 biological replicates per group for 168FARN, n = 4 biological replicates per group (2 biological replicates for Doxorubicin) for 67NR, non-linear regression with F test testing for IC50, mean ± s.d., two-tailed). j,k Western blot analysis for KMT2C, KMT2D and tubulin in indicated cell lines derived from primary tumours or distant organs from mice injected with Kmt2c KO (j) or Kmt2d KO (k) cells indicated in Fig. 1e. Cell line from Kmt2d KO (j) or Kmt2c KO (k) tumour was used as a control (n = 1 biological replicate). l,m Quantification of mCherry positive cells from lung or liver tissue of mice 12 days (l) or 13 days (m) after intracardiac injection with 168FARN (l) or 67NR (m) KO or WT cells (n = 5 (168FARN liver and lungs Kmt2d KO and 67NR liver and lungs Kmt2c KO), 6 (168FARN liver and lung WT and Kmt2c KO), 7 (67NR liver and lung WT) or 8 (67NR liver and lung Kmt2d KO) mice per group, Kruskal Wallis test with Dunn’s multiple comparison, mean ± s.d.). n, Representative scatter plot from flow cytometry for mCherry using brain tissue from mice depicted in Fig. 1h. o, Correlation plot of mCherry positive cells and relative mCherry mRNA expression from brain of mice 11 days after intracardiac injection of Kmt2c or Kmt2d KO cells (n = 10 biological replicates, simple linear regression with Pearson correlation, two-tailed). p, H&E staining from femur sections of mice from experiment depicted in Fig. 1h and i (scale bar = 500 µm). q, Quantification of mice with or without bone metastases from experiments depicted in Fig. 1h and i (n = 5-8 mice per group). PT, primary tumour, BoM, bone metastases, LiM, liver metastases, BM, brain metastases, LuM, lung metastases.

Source data

Extended Data Fig. 2 Differential effects of Kmt2c KO or Kmt2d KO on distinct cell populations in tumour microenvironment.

a, UMAP from scRNA-seq with all cells coloured by genotype (n = 3 biological replicates per genotype). b, UMAP from scRNA-seq including all cells coloured by intensities of gene module scores for tumour (left) and non-tumour (right) gene signatures. c, Volcano plot showing significantly downregulated (blue) or upregulated (red) genes in Kmt2c KO cells compared to WT. DEGs are calculated using findMarkers (Wilcoxon Rank Sum test) from Seurat from subclustered tumour cells only and genes with log2 FC between −0.3 and 0.3 are not considered (grey). d, Volcano plot showing significantly downregulated (blue) or upregulated (red) genes in Kmt2d KO cells compared to WT. DEGs are calculated using findMarkers (Wilcoxon Rank Sum test) from Seurat from subclustered tumour cells only and genes with log2 FC between −0.3 and 0.3 are not considered (grey). e, Heatmap showing average expression of individual genes used to generate immune-checkpoint signature used in Fig. 2d. f, UMAP of all subclustered non-tumour cells coloured by annotations according to gene signatures. g, Dot plot showing expression levels of genes used to determine gene signatures for each annotation. h, UMAP of subclustered CAF populations coloured by intensities of gene signatures scores for iCAF (left) of myCAF (right). i, UMAP for subclustered CAFs coloured by genotype (left) and quantification of scores depicted in Ext. Data Fig. 2h for each genotype. j, UMAP of subclustered endothelial cells colored by different genotypes. k, Expression level of top 6 differentially upregulated genes in endothelial cells in Kmt2c KO, Kmt2d KO or WT tumours. l, UMAP of subclustered macrophage populations coloured by intensities for gene module scores for M1 (left) or M2 (right) macrophages. m, UMAP for subclustered macrophages coloured by genotype (left) and quantification of scores depicted in Ext. Data Fig. 2l for each genotype. n, UMAP for subclustered T cells coloured by intensities of gene signature scores for CD4, Treg, CD8, cytotoxic CD8 or exhausted CD8 cells. o, Heatmap showing log10 FC intensities from cytokine array using supernatant from cultured cells or tumour lysates from Kmt2c KO (blue) or Kmt2d KO (yellow) cells compared to WT cells (values are from 5 pooled samples per condition).

Source data

Extended Data Fig. 3 Distinct histone modifier changes upon Kmt2c or Kmt2d KO.

a, Heatmap of total P300 ChIP-seq signal in WT, Kmt2c, and Kmt2d KO cells. Peaks are centred at TSS in a ± 3kb window (n = 2 biological replicates per group). b, Density plot showing log2FC in P300 and H3K27ac signal intensities in Kmt2c KO (left) and Kmt2d KO (right) cells compared to WT (linear regression with Pearson correlation, data is calculated from 3 biological replicates for H3K27ac and 2 biological replicates for P300 per group, two-tailed, p = 2.2 × 10⁻¹⁶). c, Quantification of total called peaks for H3K27me3 in WT, Kmt2c KO and Kmt2d KO cells (n = 2 biological replicates per group). d, Motif analysis using HOMER for significantly gained or lost peaks from H3K4me1, H3K27ac and KDM6A ChIP-seq in Kmt2c KO (blue) and Kmt2d KO (yellow) cells compared to WT cells. Data are presented as -log10 p-value. Only motifs with -log10 p-values > 25 in at least one sample are shown. e, Venn diagram showing intersections of significantly gained H3K27ac and KDM6A peaks for Kmt2c KO (blue) and Kmt2d (yellow) compared to WT cells.

Source data

Extended Data Fig. 4 Kmt2c and Kmt2d KO cell have unique but also shared transcriptomic landscapes.

a, Volcano plot showing significantly upregulated (red) and downregulated (blue) genes in Kmt2c KO cells compared to WT cells. Genes with adjusted p-values > 0.1 or log2 FC between −0.6 and 0.6 are not considered (grey). b, Volcano plot showing significantly upregulated (red) and downregulated (blue) genes in Kmt2d KO cells compared to WT cells. Genes with adjusted p-values > 0.1 or log2 FC between −0.6 and 0.6 are not considered (grey). c, GSEA analysis showing normalized enrichment scores for Hallmark gene sets of Kmt2c KO (left) or Kmt2d KO (right) cells compared to WT cells. Only pathways with p-values < 0.05 are shown. d, Transcription regulator prediction using LISA algorithm with upregulated DGE shown in Extended Data Fig. 3d,e. Shown are top 25 candidates for each KO cell line as -log10 (p-value). e, Heatmap (top) for RNA-seq data from WT (black), Kmt2c KO (blue) and Kmt2d KO (yellow) 67NR cells. Values are shown as z-scores normalized to average of WT cells. Only genes with adjusted p-values < 0.1 or log2 FC between −0.6 and 0.6 from DEG analysis are shown (n = 3 biological samples per group). Venn diagram (bottom) showing overlap of significantly upregulated (left) or downregulated (right) genes between Kmt2c KO (blue) and Kmt2d KO (yellow) cells compared to WT 67NR cells. f, Volcano plot showing significantly upregulated (red) and downregulated (blue) genes in Kmt2c KO (top) cells compared to WT 67NR cells. Genes with adjusted p-values > 0.1 or log2 FC between −0.6 and 0.6 are not considered (grey). Volcano plot showing significantly upregulated (red) and downregulated (blue) genes in Kmt2d KO (bottom) cells compared to WT 67NR cells. Genes with adjusted p-values > 0.1 or log2 FC between −0.6 and 0.6 are not considered (grey). g, GSEA analysis showing normalized enrichment scores for Hallmark gene sets of Kmt2c KO (top) or Kmt2d KO (bottom) cells compared to WT 67NR cells. Only pathways with p-values < 0.05 are shown. h, Venn diagram showing overlap of significantly upregulated (top) or downregulated (bottom) genes for Kmt2c KO (blue) or Kmt2d KO (yellow) between 168FARN and 67NR cells. P-values are calculated as exact hypergeometric probability of overlaps of each gene set normalized to total mouse transcriptome.

Source data

Extended Data Fig. 5 Mmp3 and Kdm6a in Kmt2c and Kmt2d KO-related tumorigenesis and metastases.

a, Relative mutation frequency of KMT2C (left) or KMT2D (right) in MMP3 high (top 50%) or MMP3 low (bottom 50%) TNBC from METABRIC cohort (n = 299, Fisher’s exact test) b, Relative expression of Mmp3 in control and sh_Mmp3 knockdown derivatives of WT, Kmt2c KO and Kmt2d KO cells. Values were calculated using ddCt method normalized to actin expression (n = 3 biological replicates per group, two-tailed unpaired t-test, means ± s.d.). c, Relative Kdm6a mRNA expression in sh_ctrl and sh_Kdm6a derivatives of Kmt2c KO, Kmt2d KO and WT 168FARN cells measured by qRT-PCR (n = 4 biological replicates per group, two-tailed unpaired t-test, mean ± s.d.). d, Frequency of mice with bone metastases after intracardiac injection of WT, Kmt2c KO or Kmt2d KO cell derivatives with control shRNA or with shRNA against Mmp3 or Kdm6a, respectively (cohorts are the same as depicted in Figure 4f and i). e, Volumes of primary tumours after MFP injection of sh_Mmp3, sh_Kdm6a or sh_ctrl derivatives of WT (left), Kmt2c KO (middle) or Kmt2d KO (right) cells into BALB/c mice (n = 4 (WT sh_Mmp3), 5 (WT sh_ctrl and sh_Kdm6a, Kmt2c KO sh_sh_Kdm6a and Kmt2d KO sh_ctrl), 6 (Kmt2d KO sh_Kdm6a and sh_Mmp3) or 7 (Kmt2c KO sh_ctrl and sh_Mmp3) mice per group, non-linear regression with F test testing different growth rates, mean ± s.d., two-tailed). f, Western blot for H3K27me3, H3K27ac and total H3 in WT, Kmt2c KO and Kmt2d KO with or without GSK-J4 treatment for 2 days. (n = 2 biological replicates). g, Relative expression of Mmp3 in DMSO or GSK-J4-treated WT, Kmt2c KO and Kmt2d KO 67NR cells 3 days after treatment start (n = 4 biological replicates per group, two-tailed Mann-Whitney test for WT and Kmt2c KO, unpaired t-test for Kmt2d KO, mean ± s.d.).

Source data

Extended Data Fig. 6 Regulation of Mmp3 by KDM6A and P300 in Kmt2c and Kmt2d KO cells.

a, Representative western blot for H3K4me1, H3K27ac and total H3 proteins (n = 2 biological replicates) in WT, Kmt2c or Kmt2d KO cell treated with ORY-1001 (500 nM, KDM1A inhibitor) or A-485 (1µM, P300 inhibitor). b, Quantification of reads using featureCounts of H3K27me3, P300 and KDM6A ChIPseq data within ± 1kb from the Mmp3 promoter in WT, Kmt2c or Kmt2d KO cells (n = 2 biological replicates each). c, Bubble plot showing log2 fold change vs WT and -log10 adjusted p-value for Mmp1b, Mmp10 and Mmp13 expression in Kmt2c KO (top) or Kmt2d KO (bottom) from DEseq2 output using same data as Extended Data Fig 4a (n = 3 biological replicates each, Wald test). d, Gene tracks of log2FC of Kmt2c KO (blue) or Kmt2d KO (yellow) vs WT for H3K4me1, H3K27ac, P300, H3K27me3 and KDM6A across the Mmp3 containing Mmp cluster locus. Scores were averaged in 1 kb windows to calculate fold changes (data are averaged from n = 3 biological replicates for H3K4me1 and H3K27ac or n = 2 biological replicates for H3K27me3, P300 and KDM6A). e, Quantification of average signal intensities for gene track signals in e in 1 kb bins. f, Relative mRNA expression of Mmp1b, Mmp10 and Mmp13 in WT, Kmt2c or Kmt2d KO cells after control, GSK-J4 or A-485 treatment for 2 days. Values are plotted as z-scores after log₁₀ transformation for each gene (n = 4 biological replicates each, One-Way ANOVA with Dunnetts multiple comparison compared to control treatment within each genotype and gene, Kruskal-Wallis test with Dunns multiple comparison compared to control for Kmt2d KO cells and Mmp13). g, Correlation plots from BETA analysis using significantly gained KDM6A peaks within 3 kb of known TSS and differential gene expression in Kmt2c (left) or Kmt2d KO (right) cells compared to WT (data from n = 2 biological replicates for KDM6A ChIP-seq and n = 3 biological replicates for RNA-seq). h, Bar plot showing p-values for correlations of significantly gained KDM6A peaks within depicted distances from TSS with gene expression from BETA analysis (data from n = 2 biological replicates for KDM6A ChIP-seq and n = 3 biological replicates for RNA-seq). i, Immunoblot for KDM6A, PTIP, RBBP5 or WDR5 on control IgG or KDM6A immunoprecipitates (IP) using lysates of Kmt2c KO, Kmt2d KO or WT cells.

Source data

Extended Data Fig. 7 KDM6A inhibition is tolerable and inhibits brain metastases.

a, Schematic outline of experimental setup. b, Representative H&E-stained slides of liver, intestine, spleen and kidney from mice 12 days after intracardiac injection with WT cells treated with DMSO or GSK-J4. (scale bars = 100 µm) c, Representative H&E-stained slide from brain tissues of mice described in Fig. 4m. (Scale bars = 100 µm, representative data from 5 untreated and 4 treated mice from 1 experiment) d, Ratio of mice with detectable bone metastases from the mouse cohort described in Fig. 4m (n = 6 – 7 biological replicates per group). e, Schematic outline of experimental setup for extended data figure 7f. f, Quantification of mCherry positive cells normalized to live cells in brain tissue lysates from mice. Mice were injected via intracardiac injection with WT, Kmt2c KO or Kmt2d KO cells which were treated with or without GSK-J4 for 2 days and brains were collected 2 days later (n = 4 (WT GSK-J4), 5 (WT DMSO and Kmt2d KO GSK-J4), 6 (Kmt2c KO DMSO) or 7 (Kmt2d KO DMSO and Kmt2c KO GSK-J4) mice per group, two-tailed Mann-Whitney test, mean and SD). Panels a and e created with BioRender.com.

Source data

Supplementary information

Supplementary Information

Supplementary Fig, 1.

Reporting Summary

Peer Review File

Supplementary Table 1

Gene signatures used to define clusters in scRNA-seq analyses.

Supplementary Table 2

Differentially expressed genes from scRNA-seq for Kmt2c KO versus WT or Kmt2d KO versus WT 168FARN cancer cells using the findMarkers function of Seurat.

Supplementary Table 3

DESeq2 results using read counts in peaks from H3K4me1, H3K27ac and KDM6A ChIP-seq for Kmt2c KO versus WT or Kmt2d KO versus WT 168FARN cells.

Supplementary Table 4

DESeq2 results from RNA-seq for Kmt2c KO versus WT or Kmt2d KO versus WT 168FARN or 67NR cells.

Supplementary Table 5

List of the antibodies and DNA oligonucleotides used in this study.

Source data

Source Data Fig. 1

Statistical source data.

Source Data Fig. 2

Statistical source data.

Source Data Fig. 3

Statistical source data.

Source Data Fig. 5

Statistical source data.

Source Data Fig. 6

Statistical source data.

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 2

Statistical source data.

Source Data Extended Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 4

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 6

Statistical source data.