KMT5A-methylated SNIP1 promotes triple-negative breast cancer metastasis by activating YAP signaling

Smad nuclear-interacting protein 1 (SNIP1) is a transcription repressor related to the TGF-β signaling pathway and associates with c-MYC, a key regulator of cell proliferation and tumor development. Currently, the mechanism by which SNIP1 regulates tumorigenesis and cancer metastasis is unknown. Here, we identify that SNIP1 is a non-histone substrate of lysine methyltransferase KMT5A, which undergoes KMT5A-mediated mono-methylation to promote breast cancer cell growth, invasion and lung metastasis. Mechanistically, we show KMT5A-mediated K301 methylation of SNIP1 represents a sensing signal to release histone acetyltransferase KAT2A and promotes the interaction of c-MYC and KAT2A, and the recruitment of c-MYC/KAT2A complex to promoter of c-MYC targets. This event ultimately inhibits the Hippo kinase cascade to enhance triple-negative breast cancer (TNBC) metastasis by transcriptionally activating MARK4. Co-inhibition of KMT5A catalytic activity and YAP in TNBC xenograft-bearing animals attenuates breast cancer metastasis and increases survival. Collectively, this study presents an KMT5A methylation-dependent regulatory mechanism governing oncogenic function of SNIP1. SNIP1 methylation initiates its oncogenic functions. Here, the authors show that SNIP1 is methylated by KMT5A and this leads to downstream signalling that activates the YAP pathway, resulting in tumorigenesis and metastasis.

B reast cancer is the most prevalently diagnosed cancer type and a lethal primary malignant tumor among women across the globe 1,2 . Triple-negative breast cancers (TNBCs) account for~15% of all invasive breast cancers; they are insensitive to endocrine-targeted therapy due to lack of receptors 3 . Despite an initial response to systemic chemotherapy, increased risk of recurrence, metastasis, and mortality have been reported in patients diagnosed with TNBCs 4 . Therefore, there is an urgent need for a better understanding of the aggressive phenotype and higher metastatic potential, as well as identifying molecular targets.
Smad nuclear interacting protein 1 (SNIP1) is a nuclear protein cloned and characterized based on its ability to inhibit the TGF-β signal transduction pathway 5 . It is expressed in various cell types, including the brain, placenta, and kidney. SNIP1 is also upregulated in a wide range of tumors and associated with cell transformation and DNA damage response 6,7 . Several functions have been identified to SNIP1, suggesting its ability to interact with multiple cellular partners; notably, the majority of these functions described so far involve transcriptional regulation 8 . The N-terminal nuclear localization signal (NLS) domain of SNIP1 binds the C/H1 domain of transcriptional coactivators CBP and p300, competing for the binding of Smad4 and p65/RelA to CBP/p300 9 . SNIP1 overexpression inhibits Smad4 and p65/RelA transactivation by preventing their interaction with p300 through its NLS domain 9 . Conversely, SNIP1 also enhances the transcriptional activity of c-Myc by stabilizing it against proteasomal degradation and bridging the c-Myc/p300 complex, binding the N terminus of c-Myc through its own C terminus 7 . Therefore, SNIP1 might be an important endogenous modulator of multiple transcriptional pathways requiring transcriptional coactivators CBP/p300.
Accumulating evidence reveals that the epigenetic landscape is dysregulated in breast cancer, including TNBC, partially due to the aberrant patterns of posttranslational modifications (PTMs) [10][11][12] . Lysine methylation has evolved as a vital PTM implicated in various biological functions 13 . Further, lysine methylation of non-histone proteins regulates numerous molecular events including protein-protein interaction, protein stability, protein subcellular localization, and transcription [14][15][16] . Mechanistically, lysine methylation of non-histone proteins facilitates their interactions with specific readers for downstream signaling; besides, the crosstalk between lysine methylation and other PTMs on non-histone proteins impacts the biological functions of modified proteins 17,18 . Lysine methylation is primarily catalyzed by a family of protein methyltransferases (KMTs), including seven major families namely SUV39, SET1, STE2, RIZ, SMYD, EZ, and SUV4-20 19 . Although a significant number of KMTs have been detected in the human genome, only a few non-histone proteins including YAP, p53, STAT1, and AKT, undergo lysine methylation [20][21][22][23] . This indicates that a considerable number of lysine methylated species remain unidentified. Moreover, whether SNIP1 methylation occurs, and regulates oncogenic signaling, as well as is involved in TNBC cancer metastasis, remains to be investigated.
This work uses mass spectrometry (MS) analysis to identify KMT5A (also known as PR-Set7/9, SETD8, and SET8) as a SNIP1-interacting protein that methylates non-histone SNIP1 at K301. Consequently, KMT5A-mediated K301 methylation of SNIP1 releases KAT2A as well as promotes c-MYC/KAT2A complex formation and transcriptional activation of c-MYC targets. Subsequently, methylation of SNIP1 by KMT5A enhances key oncogenic pathway-Hippo signaling, thereby promoting TNBC metastasis. As such, our data recognize KMT5A-mediated SNIP1 methylation as an essential step for c-MYC/KAT2A complex formation and c-MYC target activation. In addition, our findings further highlight the histone methyltransferase KMT5A as a potential therapeutic target for hyperactive SNIP1-driven TNBC metastasis.

SNIP1 methylation promotes its oncogenic functions.
To identify important non-histone proteins regulated in a methylationdependent manner, a specific pan-lysine methylation antibody was used to perform a mass spectrometry (MS)-based screening of cell lysates derived from two TNBC cell lines, i.e., BT549 and MDA-MB-231 (Fig. 1a). Notably, several SNIP1-derived peptides were identified, comprising methylated modifications at three nearby evolutionarily conserved lysine residues (K301, K325, and K342) in the FHA (281-344aa) domain (Fig. 1b, c). Notably, the K301 methylation was detected in both two cell lines. Then, the SNIP1 mutant was generated by replacing the above lysine residues in the FHA domain with arginine. This lysine (K)-to-arginine (R) substitution prevents methylation but maintains a positive charge, thereby mimicking the non-methylated form of a protein. Interestingly, only the SNIP1 K301R mutant, but not the other mutants, was not methylated (Fig. 1d), suggesting Lys301 as the SNIP1 methylation site in the FHA. And LC-MS/MS revealed that K301 was mono-methylated ( Fig. 1e). Furthermore, SNIP1 monomethylation was confirmed using Kme1-specific antibody in cells treated with a global histone methylation inhibitor, 3-deazaneplanocin A ( Supplementary Fig. 1a). Moreover, the Lys301 site was highly and evolutionarily conserved in SNIP1 proteins among different species ( Supplementary Fig. 1b), suggesting the functional specificity of lysine in SNIP1. Taken together, these findings indicate that SNIP1 is methylated at Lys301.
Given that the K301 site was identified at endogenous levels through a large-scale non-biased MS approach, this paper investigated the contribution of K301 mono-methylation to SNIP1 activity and its oncogenic functions. To reveal the potential biological functions of mono-methylation within its FHA region, a methylation deficient variant of SNIP1 (K301R) was ectopically re-expressed in SNIP1 knockout MDA-MB-231 and BT549 cell lines (MDA-MB-231/sgSNIP1 and BT549/sgSNIP1) (Supplementary Fig. 1c). Furthermore, a rabbit polyclonal antibody specifically recognizing the K301 mono-methylated SNIP1 (Supplementary Fig. 1c-e) was generated through a commercial vendor. Dot blot analysis was performed to demonstrate a minimal cross-reactivity of the polyclonal antibody with un-methylated K301 containing peptide ( Supplementary Fig. 1f). Also, its specificity for SNIP1-K301me1 was verified in MDA-MB-231 cells and a clinical breast cancer specimen ( Supplementary Fig. 1g, h).
Then in contrast with the control (sgC), the knockout of SNIP1inhibited the colony formation in vitro ( Supplementary  Fig. 1i, j), and reduced tumor growth in vivo (Fig. 1f, g) with significantly downregulated Ki67 expression in primary tumor tissues (Fig. 1h, i); while re-expression of wild-type SNIP1 (SNIP1 WT ) but not the SNIP1 K301R mutant significantly restored the ability of cell colony formation ( Supplementary Fig. 1i, j), tumor growth (Fig. 1f, g) and Ki67 expression (Fig. 1h, i). Moreover, SNIP1 knockout exhibited significantly inhibited the invasiveness of breast cancer cells in vitro ( Supplementary Fig. 1k, l), tumor metastasis to the lungs and metastatic lung nodule formation ( Fig. 1j-m), and prolonged animal survival (Fig. 1n); similarly, re-expression of SNIP1 WT but not the SNIP1 K301R mutant restored the ability of cell invasion ( Supplementary  Fig. 1k, l), tumor lung metastasis (Fig. 1j-m), and shortened animal survival (Fig. 1n). Together, these findings confirm that SNIP1 methylation potentially promotes TNBC growth and metastasis both in vitro and in vivo.
SNIP1 interacted with and was methylated at Lys301 by KMT5A. To identify the physiological upstream methyltransferase (or methyltransferases) that mediates SNIP1 K301 mono-methylation, a co-immunoprecipitation (co-IP) assay and LC-MS/MS were conducted to capture the SNIP1-interacting proteins from MDA-MB-231 cells. As a result, methyltransferase KMT5A was the only lysine methyltransferase among the top SNIP1interacting proteins based on the abundance ratio above 10 and P < 0.05 (Fig. 2a). This suggests that KMT5A potentially methylates SNIP1. We indeed observed a relatively strong physical interaction between endogenous SNIP1 and KMT5A in MDA-MB-231 and BT549 cells (Fig. 2b)   also detected in the anti-Flag-KMT5A immunoprecipitates of HEK-293T cells transfected with indicated plasmids (Fig. 2c). Further, His-tagged SNIP1 was pulled down by GST-KMT5A and vice versa, indicating a direct physical interaction between SNIP1 and KMT5A (Fig. 2d, e). To determine the SNIP1 domain that binds to KMT5A, a deletion-mapping approach was adopted ( Supplementary Fig. 2a, b); then these deletion mutant protein isoforms were transduced into HEK-293T cells. As shown in Supplementary Fig. 2a The physical interaction between SNIP1 and KMT5A suggests a potential methylation regulation. Indeed, the methyl lysine (K) levels of SNIP1 were substantially impaired in both MDA-MB-231 and BT549 cells with KMT5A knockdown, as established by the anti-pan methyl antibody (Fig. 2f). Considering SNIP1 was mono-methylated at Lys301 which located in FHA domain bound to KMT5A, K301 was further predicted for KMT5A methylation site. A mutant SNIP1 was generated using arginine (R) to replace the K301 residue and two other lysine residues (K325 and K342) in the FHA domain. As shown in Fig. 2g, in vivo methylation analysis confirmed that only the replacement of K301 with R had a significant effect on SNIP1 lysine methylation by KMT5A when examined by an anti-panmethylation antibody. Furthermore, in vitro methylation analysis demonstrated that only K301R (but not K325R and K342R), showed a decrease in methylation signals (Fig. 2h). This indicates that KMT5A acts as a putative SNIP1 methyltransferase and potentially responsible for K301 methylation on SNIP1.
KMT5A mediated SNIP1 methylation activates Hippo/YAP signaling. To elucidate the functional significance of SNIP1 methylation, RNA sequencing analysis was performed to compare gene expression profiles in MDA-MB-231/sgSNIP1 cells with re-expression of sgRNA-resistant HA-SNIP1 WT or HA-SNIP1 K301R mutant. A total of 632 downregulated and 50 upregulated genes were identified (raw data accessible via Bioproject ID: PRJNA797682) by re-expressing sgRNA-resistant SNIP1 K301R mutant compared to the SNIP1 WT control group (fold change >2, P < 0.05) (Fig. 3a). KEGG pathway analysis of 632 downregulated genes identified the Hippo signaling pathway as the most significant pathway ( Fig. 3b and Supplementary data 1). Previous studies identified 57 YAP target genes in various cell types including breast cancer cells 24,25 . Among these, 40 genes, also identified as YAP-regulated downstream genes, were downregulated in MDA-MB-231/sgSNIP1 cells with re-expressing sgRNA-resistant SNIP1 K301R mutant compared to the SNIP1 WT group (Fig. 3c, d). Conversely, YAP target gene levels were significantly increased in MDA-MB-231/sgSNIP1 cells with reexpressing sgRNA-resistant SNIP1 K301 methylation mimics mutant (SNIP1 K301M ) compared to the SNIP1 WT group (Fig. 3e).
Interestingly, as shown in Fig. 3c, SNIP1 K301R also downregulated the expression of MAP/microtubule affinity-regulating kinase 4 (MARK4), which could phosphorylate the Hippo core components MST1/2 or SAV; besides, it disrupted a complex formation between MST/SAV and LATS to influence YAP/TAZ activity, which as a potent activator of YAP/TAZ activity, MARK4 inhibits YAP/TAZ phosphorylation to promote YAP/ TAZ nuclear localization, ultimately inducing transcriptional activation in breast cancer 26 . To investigate the possibility of SNIP1 methylation mediating YAP target genes by the Hippo kinase cascade MARK4/MST/SAV/YAP-mediated transcriptional activation, re-expression of SNIP1 K301R significantly downregulated MARK4, p-MST2, p-SAV, upregulated p-YAP and slightly increased p-LATS1 expression levels. Besides, it effectively altered the Hippo kinase cascade and target genes of the Hippo/YAP signaling pathway, i.e., CTGF and CYR61 27 in MDA-MB-231 and BT549 cells; excluding the typical genes including HIF-1α, VEGF, p-JAK1, JAK1, p-STAT3, STAT3, PI3KCA, p-AKT and AKT levels implicated in the major pathways of KEGG pathway analysis ( Fig. 3f-h). These data indicate that SNIP1 K301 methylation modulates the Hippo/YAP signaling pathway by regulating MARK4 expression and subsequently modifying phosphorylation of MST2, SAV, LATS1, and YAP, but not other components.
KMT5A-mediated SNIP1 K301 methylation promotes TNBC metastasis by activating YAP signaling. KMT5A-mediated SNIP1 K301 methylation positively regulates the Hippo/YAP pathway, and numerous studies have firmly established the role of Hippo/YAP during tumorigenesis and metastasis 28 . We hypothesized that SNIP1 K301 methylation modulates YAP signaling to promote TNBC metastasis in a KMT5A-dependent manner. . e LC-MS/MS spectrum of the tryptic peptide IADIPIDHPSCSK identified a monomethylated residue at K301, carrying a mass of +14.016 Da. f, g Cells derived from MDA-MB-231/sgSNIP1 cells re-overexpressed with indicated SNIP1-WT or SNIP1-K301R and selected with hygromycin (200 μg/ml) for 72 h before the collection was subjected to mouse xenograft assays by orthotopic injection. Tumor sizes were monitored and analyzed (f, g) (n = 6 mice per group). h Representative photos of Ki-67 positive staining cells in murine primary tumor sections of indicated tumor tissues. Scale bar, 100 μm (IHC). i Quantification of the number of cells positive for Ki-67 staining per field of vision in tumor sections from (h) (n = 6 for each group). FOV field of view. j-m Cells generated in (f) were subjected to mouse xenograft through tail vein injection. Lung metastases were monitored (j, l) and calculated (k, m) (n = 8 mice per group). n Kaplan-Meier survival of mice in (j) (n = 8 mice per group). Data information: In (g, i, k, m), statistical analysis was performed by a two-tailed t-test. In (n), by log-rank test. ***P < 0.001. Data are represented as mean ± SEM. Panels (d) and (f-n) show one experiment representative of three independent experiments with similar results.   First, an immunohistochemical staining analysis was performed in 100 clinical TNBC primary tumors and matched metastatic lymph nodes to gain insight into the role of KMT5A in TNBC metastasis. As a result, KMT5A was more highly expressed in matched metastatic lymph nodes than in the primary tumor ( Supplementary Fig. 3a, b). Moreover, TNBC patients with high KMT5A levels had a significantly worse prognosis than those with low KMT5A expression using Kaplan-Meier survival analysis ( Supplementary Fig. 3c). In addition, unlike the normal breast epithelium cell line MCF-10A cells, KMT5A expression  Fig. 4f). Taken together, these results demonstrate that KMT5A-dependent SNIP1 methylation promotes TNBC tumor and lung metastasis by activating MARK4 transcription and subsequent modification of the Hippo/YAP signaling pathway.

SNIP1 K301 methylation disrupts its interaction with KAT2A
and releases the inhibitory effect of SNIP1 on KAT2A HAT activity. Given the evidence that SNIP1 directly interacts with c-MYC and enhances the transcriptional activity of c-MYC 7 , we hypothesized that SNIP1 K301 methylation upregulated MARK4 by transcriptional activation of c-MYC. First, to assess whether c-MYC directly mediates MARK4 transcription, MARK4 expression was detected in c-MYC-knockdown breast cancer cells, where its expression was significantly downregulated in MDA-MB-231 and BT549 cells ( Supplementary Fig. 5a, b). Then, the JASPAR database prediction indicated two putative c-MYC binding sites at the −1500 to −1489 bp and 97 to 108 bp of the MARK4 promoter ( Supplementary Fig. 5c). Importantly, ChIP-qPCR assays also revealed that c-MYC could be recruited to the MARK4 promoter at the −1500 to −1489 bp or 97 to 108 bp site but not at the −3500 to −4000 bp control site, which was not predicted by the JASPAR database in MDA-MB-231 and BT549 cells ( Supplementary Fig. 5d, e). Furthermore, the MARK4 transcription was significantly enhanced by c-MYC overexpression in HEK-293T cells, while the single mutation (MUT1) showed a decreased signal at least to some extent. Both two mutations (MUT1 + MUT2) dramatically suppressed the transcription activity ( Supplementary Fig. 5f). These data suggest that c-MYC is directly bound to the MARK4 promoter, inducing MARK4 transcription.
A growing number of functions have been attributed to SNIP1, suggesting its ability to interact with multiple cellular partners. In addition, the ability of c-MYC to activate transcription relies partly on the recruitment of cofactor complexes comprising histone acetyltransferase KAT2A 29 . KAT2A is recruited to chromatin by the transcriptional activator c-MYC, and acts as a co-activator of gene transcription by acetylating histone H3, resulting in c-MYC binding to its target gene promoter and enhancing the transcription of its target genes 29,30 . Specific posttranscriptional modifications of non-histone proteins might influence their binding affinity for their partners. Tandem affinity purification combined with mass spectrometry suggested that c-MYC interacts with endogenous SNIP1 and KAT2A (Fig. 5a). Immunoprecipitation and western blotting assays are shown in Fig. 5b, c, where SNIP1 was further bound to KAT2A and c-MYC, suggesting that SNIP1/KAT2A/c-MYC form a complex in breast cancer cells. Further, we proposed that SNIP1 K301 methylation activates transcription by affecting its relationship with c-MYC or KAT2A. Interestingly, the findings demonstrated that the SNIP1 K301R mutant presented a more robust association with the KAT2A protein than SNIP1 WT , but not c-MYC (Fig. 5d). However, this interaction was dramatically inhibited by SNIP1 K301 methylation (Fig. 5e). In addition, GST pull-down assays revealed that His-KAT2A were pulled down by purified GST-SNIP1 but not by GST alone, and GST-SNIP1-K301M obviously weakened this combination (Fig. 5f), indicating SNIP1 methylation disrupts a physical interaction between SNIP1 and KAT2A.
SNIP1 interacts with the catalytic domain of p300, causing the HAT activity of p300 to be strongly inhibited; while signalinduced binding of BCAR4 to SNIP1 releases an inhibitory role of SNIP1 on HAT activity of p300 31 . Therefore, this work explored whether SNIP1 has a similar inhibitory effect on KAT2A HAT activity. Thus, the HAT activity of KAT2A in presence of SNIP1 WT , SNIP1 K301R , or SNIP1 K301M was determined. Similarly, it was strongly inhibited by recombinant SNIP1 WT , and even much weaker by recombinant SNIP1 K301R , but could be fully rescued by SNIP1 K301M (Fig. 5g). Furthermore, in vitro observations indicate that SNIP1 directly binds to KAT2A, suggesting a possible role of KMT5A-mediated SNIP1 K301 methylation in regulating the HAT activity of KAT2A. In the presence of BSA and empty vector (EV), KAT2A exhibited dosedependent HAT activity; this was abolished in the presence of Fig. 3 KMT5A-mediated SNIP1 K301 methylation activates Hippo/YAP signaling. a Heatmap of RNA-Seq analysis of differentially expressed genes downregulated by SNIP1-K301R (SNIP1 K301R ) (twofold change and FDR < 0.05). SNIP1 sgRNA-resistant Flag-SNIP1 WT and SNIP1 K301R mutant were reexpressed in sgSNIP1/MDA-MB-231 cells. b Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis indicates that the genes downregulated by the SNIP1 K301R mutant were predominantly associated with the Hippo signaling pathway. c Venn diagram shows YAP target genes that are regulated by SNIP1 demethylation (SNIP1 K301R ). d Heatmap representing expression levels of 55 YAP target genes determined by RNA-Seq in (a). Forty genes were found to be significantly downregulated in K301R mutant cell lines were highlighted in red. e qPCR analysis of a group of YAP target genes that were induced upon SNIP1 K301 methylation (SNIP1 K301M ) (n = 3). f, g these representative typical pathway genes that may be regulated by SNIP1 methylation were measured and verified in MDA-MB-231 (f) and BT549 (g) cells by qRT-PCR (n = 3). h Western blot was performed in MDA-MB-231/sgSNIP1 and BT549/ sgSNIP1 cells transfected with SNIP1 sgRNA-resistant Flag-SNIP1 WT and SNIP1 K301R mutant (n = 3). Data information: In (d, e-g), statistical analysis was performed by a two-tailed t-test. mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. Panels (e-h) show 1 experiment representative of three independent experiments with similar results.  (Fig. 5h). In contrast, in the presence of KMT5A WT but not inactivated kinase activity mutant KMT5A R336G , KAT2A HAT activity was largely rescued (Fig. 5h). These data suggest that the KMT5A-mediated SNIP1 K301 methylation releases its interaction with KAT2A, ultimately causing the activation of KAT2A enzyme activity. transcription start sites of genes; it also provides a specific tag for epigenetic transcription activation 32,33 . To confirm whether KAT2A regulates MARK4 expression by epigenetically activating transcription, two different lentiviruses specifically targeting KAT2A or a non-silencing control were used to deplete KAT2A in MDA-MB-231 and BT549 cells (Supplementary Fig. 6a). Interestingly, KAT2A knockdown did not affect the binding of SNIP1 and c-MYC; however, it lowered MARK4 and CTGF expression; and increased YAP phosphorylation (Supplementary Fig. 6a).
Furthermore, KAT2A knockdown sharply suppressed MARK mRNA expression ( Supplementary Fig. 6b), MARK4 transcription activity ( Supplementary Fig. 6c), and the recruitment of KAT2A to the MARK4 promoter ( Supplementary Fig. 6d, e). Also, it also decreased H3K79succ and H3K9ac levels of the MARK4 promoter after KAT2A silencing in MDA-MB-231 and BT549 cells (Supplementary Fig. 6d, e). Furthermore, SNIP1 and KAT2A cooccupied the promoter of c-MYC targeted gene MARK4 but less frequently occupied the promoter of a control gene VEGF . e Effects of SNIP1 WT or SNIP1 K301M mutant on SNIP1 association with KAT2A. His-KAT2A were co-transfected into HEK-293T cells with HA-SNIP1 constructs or empty vector control (EV) (n = 3). f GST pulldown and western blotting for SNIP1 WT or SNIP1 K301M association with KAT2A (n = 3). g In vitro HAT activity assays of KAT2A in the presence of SNIP1 WT and its corresponding mutants (SNIP1 K301R and SNIP1 K301M ) as indicated (n = 3). h In vitro HAT activity assays of KAT2A in the presence of SNIP1 and KMT5A WT or its methyltransferase activity-deficient mutant KMT5A R336G (n = 3). Data information: In (g-h), statistical analysis was performed by a two-tailed t-test. Error bars ± SEM. **P < 0.01, ***P < 0.001. Panels (b-h) show one experiment representative of three independent experiments with similar results.
SNIP1 strongly inhibits Smad/p65-dependent transcriptional activity by competing with Smad4 and p65/RelA for binding the co-activator p300 5,9 . Further, an interaction between c-MYC, SNIP1, and the KAT2A was examined to address the mechanism by which SNIP1 methylation activates MARK4 transcription. Immunoprecipitation of KAT2A followed by western blotting revealed that forced overexpression of c-MYC tampered with the interaction between KAT2A and SNIP1 (Fig. 6a). Conversely, SNIP1 overexpression also inhibited KAT2A-c-MYC interaction (Fig. 6b). These data suggest that SNIP1 might be a candidate modifier of the MARK4 transcriptional activity by competing with c-MYC for binding KAT2A. Importantly, KMT5A WT but not KMT5A R336G , which could methylate SNIP1 K301, promoted Flag-c-MYC binding to His-KAT2A; while it significantly impaired the relationship between HA-SNIP1 and His-KAT2A (Fig. 6c). Moreover, re-expression of KMT5A WT but not KMT5A R336G in MDA-MB-231/sgKMT5A and BT549/sgKMT5A significantly enhanced interaction between endogenous c-MYC and KAT2A, SNIP1 k301 methylation, MARK4 expression and the Hippo/YAP target gene CTGF expression, simultaneously, this was with a sharply decreased endogenous SNIP1 binding to KAT2A, and re-expression of KMT5A WT also reduced YAP phosphorylation (Fig. 6d).
These findings indicate that KMT5A-mediated SNIP1 K301 methylation releases KAT2A to promote the KAT2A HAT activity and c-MYC/KAT2A complex formation as well as the recruitment of c-MYC/KAT2A complex to MARK4 promoter. This ultimately inhibits the Hippo kinase cascade in a kinasedependent manner by transcriptionally activating MARK4 expression.
KMT5A catalytic activity combined with YAP signaling inhibition impedes triple-negative breast cancer progression. As previously reported, YAP/TAZ interacts with transcription factor TEAD to jointly activate target gene expression, playing critical roles in cancer biology including EMT, metastasis, drug resistance, and cancer stem cells 34,35 . Notably, verteporfin is a small molecule compound effectively inhibiting YAP/TAZ-TEAD interaction 36 .
Considering the critical role of the Hippo/YAP signaling pathway in KMT5A/SNIP1/MARK4-driven breast cancer progression and metastasis, the effects of verteporfin treatment combined with KMT5A-depletion on breast cancer metastasis were assessed. Consistent with our findings, KMT5A knockdown significantly curbed lung metastasis of MDA-MB-231 and 4T1 cell lines in vivo ( Supplementary Fig. 7a, b). In contrast with the empty vector control, re-expression of shRNA-resistant KMT5A WT but not the KMT5A R336G mutant re-established the KMT5A knockdowninhibited lung metastasis in both two models ( Supplementary  Fig. 7a, b). These data confirm that the pro-metastatic functions of KMT5A require its catalytic activity. Furthermore, the treatment of verteporfin or KMT5A depletion showed a modest inhibitory effect on MDA-MB-231 and 4T1 cells invasiveness, respectively (Fig. 7c). Strikingly, combined treatment of verteporfin with KMT5A depletion was more effective than that of single-agent treatment in both breast cancer cell lines invasion ( Supplementary Fig. 7c). To find out whether this synergy occurs in vivo, athymic nude mice injected intravenously with MDA-MB-231 cells and Balb/c mice injected orthotopically with 4T1 cells were adopted ( Supplementary Fig. 7d); 7 days after inoculation, mice transplanted with MDA-MB-231 and 4T1 with or without KMT5A depletion were treated with either verteporfin (50 mg/kg, i.p.) or vehicle daily for another 14 days; tumor metastasis was monitored at day 21 post-implantation, respectively. Notably, tumor metastasis was robustly decelerated in both implant models ( Supplementary  Fig. 7e-g) with prolonged animal survival ( Supplementary Fig. 7h, i) by the combination treatment.
To furthermore examine the potential for therapeutic targeting of KMT5A in clinical practice in the near future, we also treated MDA-MB-231 and 4T1 cells with a KMT5A-selective inhibitor, UNC0379, which has been evaluated to inhibit hepatocellular carcinoma and ovarian cancer progression and metastasis 37,38 . The result showed that relative to control, the treatment of UNC0379 or verteporfin presented a modest inhibitory effect on cell proliferation in vitro and tumor growth in vivo (Fig. 7a-c); then, combined treatment of UNC0379 and verteporfin was more effective than that of single-agent treatment in cell proliferation and tumor growth (Fig. 7a-c). Furthermore, we also determine whether these two small molecule inhibitors synergy occurs in cell invasion in vitro and lung metastasis in vivo. As expected, combined treatment with UNC0379 and verteporfin significantly inhibit cell invasion, compared with single-drug treatment (Fig. 7d); and the scheme of lung metastasis in vivo was consistent to the protocol of combined treatment with verteporfin or/and KMT5A depletion (Fig. 7e). Seven days after cell inoculation, transplantation tumors were treated with either verteporfin (50 mg/kg, i.p.) or/and UNC0379 (25 mg/kg, i.p.) daily for another 14 days; tumor metastasis was monitored at day 21 post-implantation, respectively. Consistently, tumor metastasis was robustly decelerated in both implant models (Fig. 7f-h) with prolonged animal survival (Fig. 7i, j) by the combination treatment. All these data suggest that the KMT5A/SNIP1/ MARK4 axis might comprise a clinically targetable vulnerability of TNBC driven by the aberrant Hippo/YAP signaling pathway.
KMT5A expression significantly correlated with activated Hippo/YAP signaling in triple-negative breast cancer. To further demonstrate the clinical relevance of our findings, the expression levels of KMT5A, SNIP1K301me1, and MARK4 were detected in 100 clinical TNBC species. As shown in Fig. 8a, KMT5A, SNIP1K301me1, and MARK4 were significantly coexpressed in TNBC tissues. According to the quantification of protein expression via immunohistochemical staining analysis, KMT5A significantly and positively correlated with SNIP1K301me1 and MARK4, as demonstrated using the chi-square test (Fig. 8b). Then the roles of KMT5A, SNIP1K301me1, and MARK4 in TNBC progression, we analyzed the correction between differences in the expression levels of KMT5A, SNIP1K301me1, and MARK4 and various clinicopathological parameters of individuals with TNBC ( Table 1). All these protein levels did not show a significant correlation with age, tumor size, and T staging. However, the expression of KMT5A and MARK4 were significantly correlated with Pathological grading; additionally, SNIP1K301me1 and MARK4 levels were correlated with N staging, and SNIP1 K301me1 has a certain correlation with AJCC staging ( Table 1), indicating that KMT5A-mediated methylation of SNIP1 may enhance TNBC malignant metastasis. Kaplan-Meier analysis showed that coexpression of KMT5A/SNIP1 K301me1 or KMT5A/MARK4 at high expression levels positively correlated with a poor prognosis in patients with TNBC (Fig. 8c). Moreover, we detected KMT5A, SNIP1K301me1, and MARK4 expression in normal breast specimens and tumor tissues, and these protein levels were significantly upregulated in tumor tissues (Fig. 8d); simultaneously, we also detected KMT5A and SNIP1K301me1 expression in the tissues of different tumor stage, and found that the expression of KMT5A and SNIP1K301me1 were closely related to the progress of TNBC (Fig. 8e). Eventually, KMT5A, SNIP1K301me1, and MARK4 expression were dramatically upregulated in metastatic LN compared with primary TNBC tumors (Fig. 8f). In conclusion, all these data confirm that KMT5A-mediated methylation of SNIP1 significantly correlates with activated MARK4 kinase cascade signaling in TNBC.

Discussion
Effective treatment options for breast cancer, specifically for TNBC are limited. PTMs, like methylation modifications-based mechanisms, might be the crucial nodal points for therapeutic intervention in TNBC 39,40 . To identify non-histone methylated proteins implicated in oncogenic signaling pathways, an MSbased high throughput screening was performed. Consequently, we detected numerous proteins modified by lysine methylation. The present study shows that Smad nuclear-interaction protein 1 (SNIP1), could be methylated in its FHA domain by methyltransferase KMT5A. Besides, the absence of SNIP1 K301 methylation significantly repressed cell growth, invasion, and lung metastasis. More strikingly, KMT5A-mediated K301 methylation of SNIP1 represents a sensing signal to release KAT2A for activated HAT activity, while driving an interaction of c-MYC and KAT2A. Further, it promotes the recruitment of the c-MYC/KAT2A complex to the promoter of the c-MYC target; ultimately inhibiting the Hippo kinase cascade in a kinasedependent manner by transcriptionally activating MARK4 (Fig. 8g). In addition, loss of KMT5A catalytic activity by KMT5A depletion or KMT5A-selective inhibitor combined with YAP signaling inhibition impedes breast cancer progression using a mouse xenograft model of breast cancer metastasis. Furthermore, a concomitant upregulation in KMT5A, SNIP1K301me1, and MARK4 expression depict poor prognostic factors for metastatic TNBC patients.

ARTICLE
TGF-β and NF-кB signaling pathways; it disrupts the recruitment of co-activator p300 9 . It is essential for Cyclin D1 mRNA stability and expression 42 ; augments c-Myc function 7 , which also cooperates with both c-Myc and H-Ras to induce foci formation in an in vitro transformation assay 7 , implicating the role of SNIP1 in tumorigenesis. SNIP1-TET2 interaction enhances the ability of TET2 to trans-activate c-MYC target genes, ultimately promoting anchorage-independent cell growth and cisplatin resistance of MCF-7 breast cancer cells 6 . Previous research demonstrated that SUMOylation of SNIP1 attenuates its inhibitory effect in TGF-β signaling, causing the loss of its-mediated inhibition on the expression of TGF-β target genes PAI-1 and MMP2; it eventually enhances TGF-β-regulated cell migration and invasion 43 . Herein, we found that SNIP1 K301 was innovatively mono-methylated. Further, the absence of SNIP1 K301 methylation significantly decreased growth, invasion, and lung metastasis of breast cancer cells, indicating that SNIP1 methylation regulates TNBC metastasis.
Once the methylation of SNIP1 at K301 is validated, it is vital to identify the methyltransferase involved. This paper performed mass spectrometry (MS) analysis to screen known histone methyltransferases (HMTs) and determine whether they might function as SNIP1 methyltransferases. Consequently, KMT5A was identified as a SNIP1-interacting protein methylating nonhistone SNIP1 at K301. KMT5A is the only known monomethyltransferase of histone 4 at lysine 20 (H4K20me1). In addition to histone modifications 44 , KMT5A mono-methylates non-histone protein p53 and PCNA 21,45 . Furthermore, KMT5A involvement in cell proliferation and genome stability as well as the importance of its non-histone target p53 in cell-cycle control predict its possible role in tumorigenesis 21 . KMT5A facilitates hepatocellular carcinoma growth by enhancing aerobic glycolysis 46 . Notably, it also acts as a dual epigenetic modifier on the promoters of TWIST target genes E-cadherin and N-cadherin via its H4K20 mono-methylation activity, demonstrating its metastatic potential in breast carcinoma 47 . We also demonstrated that KMT5A methylates SNIP1 K301 to activate vital oncogenic pathway-Hippo/YAP signaling, as well as its overexpression in TNBC and relationship with TNBC prognosis. Moreover, loss of KMT5A catalytic activity by KMT5A depletion or KMT5Aselective inhibitor, and combined with YAP signaling inhibition significantly impairs breast cancer progression and metastasis. In addition, elevated KMT5A and methylated SNIP1 are associated with clinical TNBC metastasis and prognosis.
c-MYC transformation domain recruits human STAGA (SPT3-TAF-KAT2A acetylase) coactivator complex and particularly requires KAT2A HAT activity, as a co-activator for c-MYC in regulating c-MYC target genes 29 . Our current research revealed that KMT5A-mediated SNIP1 K301 methylation released KAT2A, promoting c-MYC/KAT2A complex formation, and transcriptionally activating the target gene MARK4 of c-MYC; this was consistent with previous studies that a model for c-MYCdependent transcription activation on chromatin involved direct recruitment by c-MYC of KAT2A 50 . c-MYC is also a substrate of KAT2A acetyltransferase activity; its acetylation of K323 triggers a dramatic increase in c-MYC protein stability 51 . Therefore, these data suggest that a stable existence of the c-MYC/KAT2A complex promotes a continuous transcriptional activation of c-MYC target genes.
KAT2A was identified as a histone acetyltransferase, enhancing H3K9 acetylation 33 . Lately, KAT2A coupled with the -KGDH complex regulates H3K79 succinylation 52 , which plays a crucial role in transcription activation by providing a scaffold for the formation of complex transcriptional components 32 . KAT2A HAT activity for histone H3 modifications has been confirmed to be associated with the poor prognosis of many cancers. For instance, KAT2A-mediated H3K79 succinylation in the promoter of YWHAZ (encoding 14-3-3zeta expression) is essential for βcatenin stability to promote pancreatic carcinoma metastasis 53 . KAT2A recruited a nuclear receptor binding protein TIF1β by mediating H3K9 acetylation to activate NF90-maintained HIF1α stability, further enhancing nasopharyngeal carcinoma progression 33 . We revealed that SNIP1 K301 methylation released its interaction with KAT2A, and reversed the inhibition of KAT2A HAT activity by SNIP1; this was similar to the previous research, where SUMOylation of SNIP1 attenuated its inhibitory effect in TGF-β signaling as SUMO-conjugated form of SNIP1 with impaired ability to disrupt the interaction between p300 and Smads 43 . Furthermore, KAT2A increased H3K79succ and H3K9ac levels of the MARK4 promoter and co-occupied the promoter of MARK4 with SNIP1 to enhance its transcriptional activation. In addition, the EZH2/H3K27me3/DNMT1 complex promotes the YAP pathway in breast cancer 54 . As the major functional transducer of the Hippo pathway, YAP has been identified as an oncogene, which was regulated by diverse upstream signals like MARK4 26 and has been linked to many cancers such as liver cancer, breast cancer, and pancreatic cancer 55,56 . And our results identified UNC0379, a KMT5A selective inhibitor, cooperates with YAP signaling inhibition to prevent breast cancer progression and metastasis, which may help overcome the narrow therapeutic index of single treatment strategy.
In summary, we identify KMT5A-mediated SNIP1 K301 methylation as a crucial initiation event for subsequent c-MYC- Fig. 7 KMT5A catalytic activity depletion combined with YAP signaling inhibition impedes triple-negative breast cancer progression. a, b Effects of UNC0379 or/and verteporfin on MDA-MB-231 and 4T1 cells proliferation. Cells were treated with UNC0379 (5 μM), verteporfin (10 μM) or placebo (vehicle) as indicated (n = 3). c Cells generated in (a) were subjected to mouse xenograft assays by orthotopic injection in athymic nude mice, then 7 days after cell inoculation, these mice were treated with either verteporfin (50 mg/kg, i.p.) or/and UNC0379 (25 mg/kg, i.p.) daily for another 14 days, and tumor sizes were monitored and analyzed (n = 6 mice per group). d Invasion analysis of MDA-MB-231 and 4T1 cell lines treated with UNC0379 or/and verteporfin. Cells were treated with UNC0379 (5 μM), verteporfin (10 μM) or placebo (vehicle) as indicated (n = 3). e Treatment schedules for the administration of UNC0379 (25 mg/kg, intraperitoneal injection once daily) or/ and verteporfin (50 mg/kg, intraperitoneal injection once daily) to mice grafted with MDA-MB-231 cells (upper) or 4T1 cells (lower). Control mice received placebo (vehicle). Intravenous injection (i.v.). Orthotopic mammary fat pad injection (i.m.f.p.). Intraperitoneal injection (i.p.) (n = 6 mice per group). f Representative bioluminescence images in (e) on day 21. g, h Quantification of the bioluminescence activity of MDA-MB-231 (g) and 4T1 (h) tumor xenografts from UNC0379 or/and verteporfin treated and control mice in (e). i, j Kaplan-Meier survival analysis of mice with MDA-MB-231 (i) and 4T1 (j) tumor xenografts (n = 6 mice per group). Data information: In (a-c), statistical analysis was performed by a two-tailed t-test. In (d, g, h), by two-tailed Student's t-test or one-way ANOVA. In (i, j), by log-rank test. ***P < 0.001, **P < 0.01. Data are represented as mean ± SEM. Panels (a-d) and (f-j) show one experiment representative of three independent experiments with similar results.  f Representative IHC staining of KMT5A, SNIP1K301me1, and MARK4 in primary TNBC and matched metastatic lymph nodes (n = 100 paired samples). Scale bars: 50 μm. g A working model of KMT5A-promoted Hippo/YAP signaling pathway modification and triple-negative breast cancer metastasis through the SNIP1 methylation status. KMT5A methylated SNIP1 at K301, which released KAT2A for activated histone acetyltransferase (HAT) and recruited by c-MYC, leading to enhanced KAT2A anchoring onto the MARK4 promoter and MARK4 transcription, ultimately activating the Hippo/YAP signaling pathway. Data information: In (b), Statistical analysis was performed by chi-square test. In (c), by log-rank test. In (e), by two-tailed t-test. *P < 0.05, ***P < 0.001. Data are represented as mean ± SEM. Panel (d) shows one experiment representative of three independent experiments with similar results.
activating the key oncogenic pathway-Hippo signaling, which depends on SNIP1 K301 methylation releasing KAT2A for activated HAT activity and promoting the formation of c-MYC/ KAT2A complex, eventually transcriptional activating MARK4. Importantly, this study reveals that KMT5A-mediated SNIP1 K301 methylation not only serves as a key signal driving c-MYC hyper-activation and cancer metastasis, but also acts as a marker of poor prognosis of TNBC patients. These findings add to our understanding on the role of SNIP1 in cancer metastasis. Furthermore, these results provide ideas for developing therapeutic strategies for preventing cancer metastasis.  (Gibco™,10100147) and 100 U/ml penicillin-streptomycin. MCF10A cells were cultured in DMEM/F12 (Invitrogen) supplemented with 20 ng/ml epidermal growth factor, 5% horse serum, 0.5 μg/ml hydrocortisone, 10 μg/ml insulin, 100 ng/ ml cholera toxin, and 100 μg/ml penicillin-streptomycin. 4T1 cells were maintained in RPMI-1640 medium supplemented with 10% FBS and 100 U/ml penicillinstreptomycin. All cell lines were cultured at 37°C and 5% CO 2 . Mycoplasma contamination was detected with a LookOut Mycoplasma PCR Detection Kit (Sigma-Aldrich).

Methods
Patient specimens. Human tissue samples of TNBC were provided by Zhejiang Provincial People's Hospital (Hangzhou, China). The study protocol was approved by the Research Ethics Board of the Zhejiang Provincial People's Hospital. Written informed consent was obtained from all participants in the study. All the research was carried out according to the provisions of the Declaration of Helsinki of 1975. None of these patients had received any chemotherapy or radiotherapy before surgery. The study is compliant with the "Guidance of the Ministry of Science and Technology (MOST) for the Review and Approval of Human Genetic Resources" in China.
Chromatin immunoprecipitation (ChIP), re-ChIP, and quantitative RT-PCR. ChIP was performed by a SimpleChIP® Plus Enzymatic Chromatin IP Kit (Magnetic Beads, Cell Signaling Technology, 9005 S). Briefly, cells were crosslinked with 1% formaldehyde in PBS at RT for 10 min, quenched with 2.5 M glycine at RT for 5 min and washed with PBS three times. Nuclei were prepared, and chromatin was incubated with micrococcal nuclease at 37°C for 20 min, which was followed by an appropriate amount of sonication. The supernatants were immunoprecipitated by being incubated with 3 µg of anti-c-MYC, anti-KAT2A, anti-SNIP1 or nonspecific rabbit IgG at 4°C for 12-16 h. The next day, the immunocomplexes were rotationally incubated with 30 µl of ChIP-Grade Protein G Magnetic Beads for 2 h at 4°C and then were washed three times using low salt wash buffer and 1 time with high salt wash buffer at 4°C for 5 min per wash. Chromatin was eluted by ChIP elution buffer for 30 min at 65°C with gentle vortex mixing (1200 rpm) and crosslinks were reversed by treatment with 5 M NaCl and proteinase K overnight at 65°C. Samples were then incubated with RNase at 37°C for 1 h. CHIP DNA was purified and subsequently quantified by quantitative real-time PCR (qPCR). Gene expression levels were normalized to the expression level of beta-Actin. Data analysis was finally presented as percentages of the input DNA. Re-ChIP assays were performed using the Re-ChIP-IT kit (Active Motif). Briefly, the precipitated chromatin from the first ChIP reaction was eluted by 100 µl diluted Re-ChIP-IT elution buffer at RT for 30 min. Next, the eluted precipitate was desalted by the desalting column provided in the kit. The second ChIP was performed with 30 µl Protein G magnetic beads, 90 µl desalted chromatin, and 3 µg second antibody. Next, the second precipitate was washed, eluted, and reverse cross-linked as in the first ChIP. DNA was obtained by phenol and phenol/chloroform extractions, and subjected to real-time PCR evaluation. Primer sequences for PCR are available upon request.
Western blotting and immunoprecipitation (IP) assays. Total cells were lysed by incubation at 4°C for 30 min in IP lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 2 mM Na 3 VO 4 , 5 mM NaF, and 1% Triton X-100) supplemented with complete protease inhibitor cocktail (Roche). The lysates were cleared by centrifugation, separated by SDS-PAGE and analyzed by immunoblotting. To detect SNIP1 methylation in vivo, cells were lysed in IP buffer supplemented with a complete protease inhibitor cocktail, which was followed by sonication and centrifugation at 12,000 × g for 15 min at 4°C. The supernatants were immunoprecipitated by incubation with a SNIP1 antibody or an HA antibody for 12 h at 4°C. The immunoprecipitants were washed six times with IP buffer and then were boiled in 1X SDS-loading buffer for immunoblot analysis.
To detect protein interactions, cells were lysed in IP buffer, and the supernatants were immunoprecipitated with the indicated antibodies, slowly shaken on a rotating shaker at 4°C overnight and then incubated with Pierce™ Protein A/G Magnetic Beads (Invitrogen) at room temperature for 1 h with mixing. The immunoprecipitants were washed five times using wash buffer (25 mM Tris, 0.5 M NaCl, 0.05% Tween-20, pH 7.5), then were eluted, and boiled in 1X SDS loading buffer to prepare the samples for immunoblot analysis. For SNIP1associated protein detection,~4 × 10 8 MDA-MB-231 cells overexpressing Flag-SNIP1 or Flag-GFP control were harvested in IP lysis buffer supplemented with a complete protease inhibitor cocktail for assays. Equal amounts of cell lysates, including FLAG-tagged SNIP1, c-MYC or Flag-GFP control protein, were immunoprecipitated with anti-FLAG (Sigma-Aldrich) and Pierce™ Protein A/G Magnetic Beads (Invitrogen) at 4°C for the appropriate time. Beads were washed with wash buffer and the binding protein was eluted with elution buffer (0.1 M glycine, pH 2.0) at RT with mixing for 10 min. After magnetically separating the beads, the saved supernatants containing the target antigen were neutralized immediately by neutralization buffer (1 M Tris HCl, pH 8.5), then were boiled, resolved by SDS-PAGE and silver-stained. Bands were excised and subjected to LC-MS/MS sequencing and data analysis.
GST and His pulldown assay. Recombinant GST-conjugated KMT5A and histidine (His)-conjugated SNIP1 were generated by transforming Escherichia coli BL21 with pGEX-4T-1-KMT5A and pET-28a (+)-SNIP1, respectively. Five milliliters of bacterial cultures that were grown overnight at 37°C in a 150 rpm shaking incubator were inoculated into 500 ml ampicillin-or kanamycin-resistant Luria-Bertani broth (LB). Cell cultures were grown at 37°C until the log phase was reached, as indicated by an optical density of 0.8; then protein expression was induced by the addition of 0.25 mM IPTG, and cells were placed at 16°C for 16 h with continuous shaking. After induction, the cultures were centrifuged at 4°C 5000 rpm for 10 min, supernatants were discarded, and pellets were resuspended in 30 ml lysis buffer (1X PBS, 500 mM NaCl, 1% Triton X-100, 0.5 mg/ml lysozyme, and 1X EDTA-free protease inhibitor cocktail) and sonicated on ice. The lysate was centrifuged at 15000 rpm for 15 min at 4°C. Supernatants were incubated with the corresponding volume of 50% glutathione Sepharose 4B bead slurry (GE Healthcare) or Ni-NTA Magnetic Agarose Beads (Qiagen) at 4°C for 8 h with end-overend mixing following the manufacturer's instructions. The glutathione beads were washed three times using lysis buffer and then were stored at 4°C in lysis buffer, samples were eluted with elution buffer (10 mM glutathione in 50 mM Tris-HCl pH 8.0). The Ni beads were washed three times with 50 mM Tris-Cl (pH 8.0) wash buffer containing 20 mM imidazole (1 ml per wash), and samples were eluted with Tris elution buffers containing 250 mM imidazole.
In-gel digestion. For in-gel tryptic digestion, gel pieces were destained in 50 mM NH 4 HCO 3 in 50% acetonitrile (v/v) until they were clear. Gel pieces were dehydrated with 100 μl of 100% acetonitrile for 5 min, the liquid was removed, and the gel pieces were rehydrated in 10 mM dithiothreitol and incubated at 56°C for 60 min. Gel pieces were again dehydrated in 100% acetonitrile, the liquid was removed, and gel pieces were rehydrated with 55 mM iodoacetamide. Samples were incubated at room temperature in the dark for 45 min. Gel pieces were washed with 50 mM NH 4 HCO 3 and then dehydrated with 100% acetonitrile. Gel pieces were rehydrated with 10 ng/μL trypsin resuspended in 50 mM NH 4 HCO 3 on ice for 1 h. Excess liquid was removed, and gel pieces were digested with trypsin at 37°C overnight. Peptides were extracted with 50% acetonitrile/5% formic acid, which was followed by 100% acetonitrile. Peptides were dried to completion and resuspended in 2% acetonitrile/0.1% formic acid.
LC-MS/MS analysis. Tryptic peptides were dissolved in 0.1% formic acid (solvent A) and directly loaded onto a custom-made reversed-phase analytical column (15cm length, 75 μm i.d.). The gradient comprised an increase from 6 to 23% solvent B (0.1% formic acid in 98% acetonitrile) over 16 min, 23 to 35% in 8 min before climbing to 80% in 3 min and then holding at 80% for the last 3 min; this all occurred at a constant flow rate of 400 nl/min on an EASY-nLC 1000 UPLC system. The peptides were subjected to NSI source followed by tandem mass spectrometry (MS/MS) in Q Exactive HF (Thermo) coupled online to the UPLC. The electrospray voltage applied was 2.0 kV. The m/z scan range was 350-1800 for a full scan, and intact peptides were detected in the Orbitrap at a resolution of 70,000. Peptides were then selected for MS/MS with the NCE set at 28, and the fragments were detected in the Orbitrap at a resolution of 17,500. A datadependent procedure that alternated between one MS scan followed by 20 MS/MS scans with 15.0 s dynamic exclusion. The automatic gain control (AGC) was set at 5E4.
Data processing. The resulting MS/MS data were processed using MASCOT. The retrieval parameters were set as follows: the database was set to UNIPROT human (20214 sequences). Trypsin/P was specified as a cleavage enzyme allowing up to two missing cleavages. The mass error was set to ±10 ppm for precursor ions and 0.02 Da for fragmentations. Carbamidomethyl (C) and methyl (K/R) were specified as fixed modifications, and acetyl (protein N-term) and oxidation (M) were specified as variable modifications. Peptide confidence was set at high, and the peptide ion score was set at >20.
In vitro KAT2A HAT assay. The in vitro KAT2A HAT assay was performed using the HAT assay kit from the Active Motif according to the manufacturer's instruction.
Luciferase reporter assays. pGL3.0 basic with the wild-type MARK4 promoter and the promoter with mutated c-Myc binding sites were cotransfected with or without c-Myc using Lipofectamine™ 3000 Transfection Reagent (Thermo Fisher) according to the manufacturer's protocol. A pRL Renilla luciferase control reporter vector (Promega) was utilized as a negative control. A dual-luciferase assay was performed 48 h after cotransfection using the Promega E1960 Dual-Luciferase® Reporter System following the manufacturer's recommendations. membrane (Millipore) at 48 and 72 h after transfection and then concentrated. MDA-MB-231 and BT-549 cells were infected using lentivirus expressing shRNAs or shGFP control with 8 µg/ml polybrene. Infected cells were selected by treatment with 5 μg/ml puromycin 48-72 h after infection. Multiple monoclonal cultures were screened for shRNAs by western blotting and RT-PCR analysis. For transient transfection, plasmids were transfected using Lipofectamine 3000 (Invitrogen) reagent following the manufacturer's instructions.
Transwell invasion assays. A total of 5 × 10 4 cells suspended in a medium without FBS were plated on the upper chamber membranes (8 µm pore size, 5 mm diameter, Corning), which were coated with Matrigel (BD Biosciences). The insert was incubated in 500 µl of medium supplemented with 10% FBS (Invitrogen) for 16-24 h. To evaluate the cell invasion capacity, noninvasive cells were removed by softly swiping the top of the membrane using cotton swabs, and migrated or invasive cells were fixed with methanol, stained with crystal violet, and counted under a light microscope.
Histology and immunohistochemical staining. Mice lungs were collected by necropsy, and fixed in 4% paraformaldehyde, embedded in paraffin. Four-step sections (100 mm per step) were obtained from each set of lungs. The sections were stained with H&E and scanned using a Scanscope XT digital slide scanner (Aperio Technologies). Digital images of lung sections were used to analyze the metastatic burden. Immunohistochemistry (IHC) for patients' tissues was performed using anti-KMT5A (1:500), anti-SNIP1K301me1 (1:500), and anti-MARK4 (1:100) antibodies. Each sample was assigned a score according to the intensity of the staining (0 = no staining, 1 = weak staining, 2 = moderate staining, and 3 = strong staining) and the proportion of the stained cells (0 = 0%, 1 = 1-25%, 2 = 25-50%, 3 = 50-75%, and 4 = 75-100%). Negative control slides without primary antibodies were included. Core staining was scored as negative (0) when <10% of tumor cells exhibited positive expression. A score ranging from 0 to 7 was calculated by adding the staining extent score with the intensity score, resulting in a low (0-3) level or a high (4-7) level value for each specimen. The stained tissues were scored by three individuals blinded to the clinical parameters.
RNA-Seq and differentially expressed gene analysis. RNA-Seq and differentially expressed gene analysis were performed as previously described 32 . In brief, total RNA was extracted and purified with a Qiagen RNeasy Mini Kit (Valencia, CA, USA) according to the manufacturer's protocols. The RNA quality was evaluated by an Agilent 2100 Bioanalyzer before sequencing. Libraries for poly(A) + RNA were prepared according to the Illumina protocol. Libraries were sequenced on Illumina HiSeqX Ten platforms. Genes that were found to be statistically significantly differentially expressed between the two groups were identified through fold change filtering (fold change ≥ 2.0; P ≤ 0.05). The differentially expressed genes were used in a KEGG pathway analysis using DAVID software (https:// david.ncifcrf.gov/). Fisher's exact or chi-square test with FDR was used to calculate the Pvalues of KEGG pathways, respectively. A threshold of P < 0.05 and FDR < 0.25 was used to select significant items.
In vivo xenograft experiments. 4-5 weeks-old pathogen-free female BALB/c and athymic nude mice were purchased from the Slaccas (Shanghai) and maintained in a mouse-specific pathogen-free (SPF) facility for 1 week before injection to allow the mice to adjust to their new environment. And 5-8 mice per cohort were used for each independent experiment. MDA-MB-231 cells (1 × 10 5 ) or their derivatives were suspended in 100 μl of PBS and then were injected into the tail vein of nude mice, while MDA-MB-231 cells (5 × 10 5 ) or 4T1 cells (5 × 10 5 ) or their derivatives were suspended in 50 μl of PBS and mixed with matrigel (1:1), then were orthotopically injected into the mouse mammary fat pad following an established protocol (5-8 mice for each group). Mice were killed on day 25 after tail vein injection or day 28 after mammary fat pad injection. Tumor growth was determined by nonblinded weekly caliper measurements and tumor volumes were calculated using the following formula: (width × length 2 × π)/6. Metastasis in the lungs was detected by bioluminescence imaging (BLI). All lobes of the lung were harvested and the number of macroscopic metastatic nodules on the surface of the lung was counted. For in vivo drug treatment, 7 days after inoculation, tumor-bearing mice were treated with UNC0379 (25 mg/kg) or/and verteporfin (50 mg/kg) via intraperitoneal injection (i.p.) every day. Verteporfin was dissolved in DMSO (100 mg/ml), aliquoted, and stored at −80°C. The working solution was prepared in PBS freshly before use, 21 days after injection, luciferin was injected and the primary/metastatic tumors were detected by BLI with the IVIS 100 (Caliper Life Sciences, Hopkinton, MA, USA). Animal survival was eventually analyzed. Besides, independent experimental groups used for corresponding survival analysis were performed under the same conditions. The mice were monitored daily and were euthanized upon reaching the criteria according to UCSD Institutional Animal Care and Use Committee guidelines. The survival data were analyzed by log-rank test. Data were pooled from independent experiments showing similar results. All mouse studies were conducted following protocols approved by the Animal Care and Use Committee of the Chinese Academy of Medical Science.
Statistics and reproducibility. All co-IP and immunostaining experiments were repeated three times, with similar results detected each time. All statistical analyses were performed with GraphPad Prism version 8.3 (GraphPad Software Inc., San Diego, CA, USA). The significance of the data between experimental groups was determined by one-way analysis of variance (ANOVA) with Newman-Keuls posttest or unpaired two-tailed Student's t-test. For clinicopathologic analysis, a chisquare test or Fisher exact test (two-sided) was performed. Survival analysis was calculated using the log-rank test and the Kaplan-Meier method. Statistical analyzed data are expressed as the mean ± SD or mean ± SEM, as indicated. A Pvalue < 0.05 represented a statistically significant difference. All the calculated P values were derived from at least three independent experiments with three technical replicates for each experiment. No statistical method was utilized to predetermine the sample size.