Firstly, an amine slide was prepared. The glass slides were immersed in the silane solution for at least 1 h with constant stirring. Secondly, the amine slides were functionalized with carboxylic acid using succinic anhydride. Thirdly, the carboxylic-functionalized slides were modified by N-hydroxysuccinimide (NHS), which were called NHS slides. Finally, the NHS slides were reacted with avidin to obtain avidin slides.

The Liberty CEM Peptide Synthesizer was introduced to synthesize a range of peptides. The peptides were purified via semi-preparative HPLC and their characteristics were identified using the Applied Biosystems 4800 Plus MALDI-TOF/TOF (matrix-assisted laser desorption/ionization–tandem time of flight) analyzer.

Histidine (6×)-tagged BRD4A-c002 (1) and BRD4A-c011 (2) bromodomain expression constructs were used in this study. All constructs were transformed into competent Escherichia coli BL21 cells. Protein expression was induced with 0.1 mM IPTG overnight at room temperature. Thereafter, purification of the protein was performed with Ni-NTA agarose beads (Qiagen, Hilden, Germany) according to the manufacturer’s protocols. The protein concentration was measured using the BCA Protein Assay Kit (#23225). The proteins were labeled with Cy3-NHS ester. The purity of the protein was determined with 12% SDS-PAGE gel, followed by both in-gel fluorescence scanning and Coomassie staining.

For the peptide immobilization experiment, the concentrations of peptides in the spotting buffer (phosphate-buffered saline/dimethyl sulfoxide, PBS/DMSO = 1:1) were prepared in 384 wells before the spotting process. Proteins were prepared in PBS buffer before the spotting process. The spotting process was performed using an ESI SMA arrayer (Ontario, Canada), followed by incubation in a humidity chamber with different periods. After the incubation process, the PBST buffer was used to wash the slices, rinsed with deionized water, and dried with nitrogen. The slides were finally scanned with a microarray scanner.

The Malvern Panalytical Ltd. microcalorimeter was used to determine the binding affinity of BRD4 to different peptides. There are several steps to this assay, including protein and peptide preparation, instrument washing, sample loading, titration, and data analysis.

All animal experiments were approved by the University of Macau Animal Ethics Committee (under the protocol UMARE-015-2019). The generation of Fgfr2-S252W mice was described in a previous work (19). Transgenic mice were continually checked to accelerate the growth of mammary tumors. Tumor samples were cut using sharp blades into small pieces of about 0.1–0.3 cm in diameter and processed three ways, as follows: 1) two larger pieces were fixed in 10% formalin and processed for histology and immunohistochemistry (IHC) staining with antibodies; 2) three pieces were immediately frozen in liquid nitrogen and used for subsequent DNA, RNA, and protein isolation; and 3) all the remaining pieces were placed in a freezing medium [10% DMSO/25% serum/65% Dulbecco’s modified Eagle’s medium (DMEM)], slowly frozen to −80°C, and transferred into liquid nitrogen a few days later. Our data indicated that the tumor tissues were well preserved under this condition and can be used for initiating tissue culture and xenograft tumors in nude mice at a later time. Primary FGFR2 tumor cells were derived from the Fgfr2-S252W mouse model using a standard procedure and maintained in F-medium.

Unique sgBRD4 sequences were individually cloned into the lentil-CRISPR v2 vector (Addgene plasmid #52961) with a puromycin resistance marker, and are listed in Table S3 . For BRD4 lentivirus production, 9µg DNA(plasmid: pCMVR8.2(addgene plasmid #12263): pMD2-VSVG =4: 3: 2) was used for transfection per 10-cm dish, and the virus was filtered with a 0.45-µm filter 48 h later. Then, Fgfr2-S252W mammary tumor cells were infected with the virus 24 h later. Infected cells were selected after 3 days with 4 µg/ml puromycin (Invitrogen, Carlsbad, CA, USA). After selection, the cells were switched to F-medium for the subsequent experiment.

A tumor slice culture method was performed in tumors developed from the Fgfr2-S252W mouse model, as previously described. In brief, the tumors were collected and extracted by punch biopsy. Thick tissue slices (250 μm) were obtained with a vibratome (Leica VT1200 S). Then, the tissues were placed on inserts with a culture medium. After 4–6 days, the tissues were stained with MTT and the slice viability was measured.

Total proteins of tumors or cells were extracted using RIPA buffer with phosphatase and a protease inhibitor. Following a previously described protocol, immunoblotting was carried out using ChemiDocTM with corresponding antibodies. The antibodies used for IHC and Western blot are listed in Supplementary Table S2 .

The total RNAs of tumors or cells were extracted using TRIzol^® Reagent, then reversed to complementary DNA (cDNA) (kit #205313; Qiagen). Then, the transcriptional levels of the target genes were examined with specific primers using the QuantStudio™ 7 Flex Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). Gene-specific data were normalized to 18S expression. The primers are listed in Table S3 .

All values were presented as the mean ± SEM of individual samples. The samples were analyzed using unpaired two‐tailed t‐tests. A p < 0.05 was considered statistically significant. All analyses were conducted in GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA).

Firstly, the fully protected peptide sequences Biotin-GGIRRYQK(ivDde)STELL (H3K56) and Biotin-GGKQTARK(ivDde)STGGK (H3K9) were synthesized via conventional solid-phase chemistry using Fmoc-based synthesis ( Figure 2 ). In both sequences, the ivDde group of lysine was then selectively removed on resin using 4% hydrazine in dimethylformamide (DMF). Subsequent modifications of the free lysine residue were carried out using 10% of the appropriate anhydride in toluene. The final biotin-containing peptides were then cleaved from the resin with concomitant removal of the protecting groups under optimized trifluoroacetic acid (TFA) cleavage conditions. In total, 12 modified peptides of H3 were synthesized in this study. Applied Biosystems 4800 Plus MALDI-TOF/TOF analysis was used to confirm the identity of the final peptides ( Supplementary Figure S1 and Table S1 ).

The peptides were then immobilized on a glass surface via the use of a highly chemoselective avidin–biotin ligation reaction. In this concept, an avidin-functionalized glass slide was quickly produced through some sequential chemical modification steps ( Figure 3A ). The peptides were then site-specifically immobilized onto the functionalized surface in 12 distinct subgrids. BRD4 (1) and BRD4 (2) were purified using Ni-NTA agarose beads (Qiagen) for purification (20). The proteins were labeled with Cy3-NHS and characterized using SDS-PAGE and Coomassie gel ( Figure 3B ). Each subgrid was then incubated with fluorescent-labeled BRD4 (1) and BRD4 (2) for half an hour. Then, the slide was washed copiously using 0.5% PBST buffer and scanned using a microarray scanner ( Figure 3C ). Based on the fluorescent intensity, we concluded that BRD4 could bind to various H3-modified peptides.

The profiling experiment revealed several potent peptide binders for H3K56, with two peptide sequences, namely, GGIRRYQKacSTELL and GGIRRYQKprSTELL, showing particular affinity to both BRD4 (1) and BRD4 (2) ( Figures 3C, D ). A similar binding profile was observed with H3K9, but with fluorescence intensity lower than that for H3K56 ( Supplementary Figure S2 ). In order to compare these results with better accuracy, the quantitative binding affinities of the H3K56 peptides in solution were measured by isothermal titration calorimetry (ITC). The ITC data for the two binding peptides, KAc- and KPr-modified H3K56, are shown in Figure 4 . The dissociation constants (K _d) for the acetylated peptide H3K56Ac were found to be 23.0 and 49.2 µM for binding to BRD4 (1) and BRD4 (2), respectively, which were similar to values previously reported in the literature (21). The propionylated sequence H3K56Pr showed an approximately 10-fold increase in binding affinity to BRD4 (1) and BRD4 (2) (K _d = 44.8 and 21.9 µM, respectively) compared with the previously reported K _d values for propionylated lysine peptides derived from histone H3 (13). Altogether, our peptide array experiments and ITC studies showed that BRD4 binds not only to N-acetylated lysine residues, as previously reported, but also to propionylated lysine residues. These results together demonstrated that peptide array technology is a powerful tool for studying peptide–protein interactions.

Recently, we have found that the activation of FGFR2 induces TNBC with multiple signaling, such as cancer stem cells (CSCs), epithelial–mesenchymal transition (EMT), and the tumor microenvironment (TME) (19). Thus, we examined whether BRD4 is involved in regulating the histone acetyltransferase (HAT) activity inactivation in the Fgfr2-S252W TNBC mouse model. We found that BRD4 was highly expressed in mouse tumors by multiple methods, such as RT-PCR ( Figure 5A ) and immunoblot ( Figure 5C ). Subsequently, we used specific H3K9ac and H3k56ac antibodies to examine whether BRD4 regulates the acetylation of the lysine residues in H3 in Fgfr2-S252W tumors. Our data showed that BRD4 was positively correlated with acetylated H3K9, and this acetylated residue was associated with open chromatin and active c-Myc ( Figures 5B, D ). The knockout of BRD4 resulted in the suppression of H3K9ac acetylation activity ( Figure 5D ). Taking all these data together, we believe that FGFR2 activates the MAPK signaling pathway and regulates BRD4 and c-Myc expressions ( Figure 5C ). These data suggest that BRD4 is a potential target gene for cancer treatment. Afterward, we investigated whether BRD4 degradation is a targeted strategy for Fgfr-S252W mammary tumors. We obtained an Fgfr2-S252W cell line from the mouse model. The cells with Fgfr2-S252W were not very sensitive to the BRD4 inhibitor (BRD4i) JQ1 when compared to their response to the fibroblast growth receptor (FGFR) inhibitor (FGFRi) BGJ398 under the cell culture conditions ( Figure 5E ). Therefore, we combined FGFRi and BRD4i; the data showed that low concentrations of FGFRi (1 μM) and BRD4i (1 μM) had synergistic effects. However, Fgfr2-S252W tumor cells still showed resistance to BRD4i. Previous studies have shown that an elevated level of programmed death-ligand 1 (PD-L1) confers resistance to FGFRi. Thus, we next examined whether PD-L1 would induce BED4i resistance. We examined the levels of PD-L1 in various treatment groups ( Figure 5F ). We found that PD-L1 was significantly decreased only when FGFRi (BGJ398) was combined with a high concentration of JQ1 (5 μM). These findings suggest that PD-L1 promotes resistance to targeted therapy. Thus, we examined the expressions of inflammatory genes such as IL-1α, IL-1β, and IL-6. The data indicated that these pro-inflammatory mediators could induce the expressions of chemokines and the recruitment of macrophages (F4/80, CD206) to produce C–C motif chemokine ligand 2 (CCL2) or CCL5 ( Figure 5G ). Fgfr2 activation created an immunosuppressive environment and enhanced tumor progression. Thus, combination strategies using inhibitors targeting both FGF/FGFR and the bromodomain and extraterminal (BET) proteins with immune checkpoint blockade (ICB) may provide a useful approach for cancer treatment.

It is important to establish quick and reliable models to assess targeted therapy or the combinations of immunotherapy for clinical therapeutics. To this end, we introduced the tumor slice culture platform to overcome PD-L1-mediated resistance due to this platform being able to maintain the activity of immune cells under the tissue culture condition (19). Firstly, equal-sized slices of mammary tumors were prepared from Fgfr2-S252W mice ( Figure 6A ) using a Biopunch (#15111-50). Thereafter, the slices were placed into a tissue slice culture insert with F-medium and treated with various drugs or in combination with ICB targeting tumor cells. The slices were then stained with MTT for 4–6 days after the culture. The data revealed that targeted therapy combined with ICB significantly increased the killing efficacy ( Figure 6B ). The data were also confirmed by colorimetric detection ( Figure 6C ). In addition, we performed IHC staining against BRD4, c-myc, IFN-γ, Ki67, and PD-L1 to explore the mechanisms underlying the synergistic actions between FGFR/BRD4 inhibitors and immunotherapy. As shown in Figure 6D , the levels of BRD4, c-Myc, PD-L1, and Ki67 significantly decreased, while the level of INF-γ markedly increased in the anti-programmed cell death 1 (PD-1)/PD-L1 group compared with the controls. Therefore, the tumor slice culture system can be used to quickly evaluate the efficacy and provide a versatile platform for immuno-oncology and drug discovery.

In summary, we have used a peptide array to demonstrate that BRD4 can recognize not only acetylated lysine residues of H3 but also propionylated lysine residues. We have successfully developed a fast and selective strategy to profile the binding affinity of lysine-modified peptides and proved that the immobilization method could serve as a powerful tool for mapping protein/substrate specificity. Furthermore, we validated this finding in the TNBC mouse model. The knockout study of BRD4 in Fgfr2-S252W mouse tumor cells suggested that BRD4 is a key regulator of H3K9 acetylation in TNBC. This acetylation is associated with the activation of c-Myc and drives tumor formation. Furthermore, we found that the combination of FGFR2i and BRD4i with ICB could significantly increase the sensitivity of cancer cells to immunotherapy.