CFT8634

Chemical Proteomic Profiling of Bromodomains Enables the Wide- Spectrum Evaluation of Bromodomain Inhibitors in Living Cells

▪ INTRODUCTION

Bromodomains are epigenetic “readers” of lysine acetylation (Kac) on histones and non-histone proteins.1,2 These Kac- recognition motifs, in collaboration with other epigenetic regulators, play important roles in chromatin biology and the regulation of gene expression. The bromodomains have emerged as “hot spots” for drug discovery, as their malfunctions often lead to human diseases such as cancer.3−5 The discovery of small molecular ligands that block the bromodomain of KAT2B (or PCAF) marked the start of bromodomain inhibitor development.6,7 Later endeavors led to the identification of JQ18 and I-BET7629 that inhibit the bromodomains of BET (bromo and extraterminal domain) family members (e.g., BRD4), representing a hallmark of targeting the bromodomains as a novel therapeutic strategy.

The development of bromodomain inhibitors in the past decade has provided useful tools to probe the biological significance of these epigenetic modules in the maintenance of normal cell physiology and the pathogenesis of diseases. Multiple bromodomain inhibitors are even in different stages of clinical trials for the treatment of cancer, atherosclerosis, as well as diabetes.4,10

The human genome encodes 61 bromodomains that can be divided into 8 subfamilies on the basis of their sequence and structure similarity, which exist in 46 bromodomain-containing proteins (BCPs).2 While varying in primary sequences, the bromodomains share a high structural similarity. It is therefore important to examine the selectivity of an inhibitor toward its authentic target against other bromodomains to avoid off- target effects. A variety of approaches have been applied to assess the selectivity of inhibitors using recombinant bromodomains in vitro, such as differential scanning fluorim- etry (DSF), surface plasmon resonance (SPR), biolayer interferometry (BLI), AlphaScreen, and fluorescence polar- ization (FP).11−13 However, the results obtained from the in vitro assays can be different from the real situations of how the inhibitors engage with the bromodomains in the cellular environment. The BCPs are usually multidomain proteins and function as members of various nuclear complexes. The conformation of a bromodomain can be altered by the interactions with the adjacent domains in a same BCP or with other biomolecules. In turn, the inhibitor engagement with the bromodomain can also be affected.

Fluorescence recovery after photobleaching (FRAP)14 and Nanoluc luciferase-based bioluminescence resonance energy transfer (NanoBRET)15 are two commonly used methods to investigate the interactions of bromodomain inhibitors with their targets in living cells. Although working in the cellular context, these two methods are still dependent on the recombinant bromodomains instead of the endogenous ones, as they require overexpression of the bromodomains of interest in fusion with a fluorescent protein or luciferase for the detection. In addition, the setup of FRAP or NanoBRET normally allows the test of only one bromodomain at one time. These methods are suitable for the validation of the target engagement of bromodomain inhibitors rather than to examine inhibitor selectivity. The methods that allow the assessment of bromodomain inhibitors against multiple endogenous bromo- domains in living cells are of great value for the development of bromodomain inhibitors. Toward this end, attempts to profile endogenous bromodomains by developing chemical proteo- mics approaches have been reported. For example, a series of BET bromodomain inhibitor GW841819X (a JQ1 analogue)- based photoaffinity probes were employed to examine the proteome-wide target selectivity of the inhibitor,16,17 while the high specificity of GW841819X toward the BET bromodo- mains prevents the applications of these probes in the evaluation of other bromodomains. Recently, an activity- based probe has been reported to covalently capture bromodomains.18 However, this probe could only enrich less than 10 BCPs from cell lysates in vitro, which limits its scope of application in broad-spectrum evaluation of bromodomain inhibitors. More importantly, the probe has not been shown to be suitable for use in living cells. Here we describe the development of a useful chemical proteomics approach, which combines the use of a photoaffinity probe that promiscuously captures endogenous bromodomains and the stable isotope labeling with amino acids in cell culture (SILAC)-based quantitative mass spectrometry, to profile bromodomains and evaluate their inhibitor selectivity in living cells.

RESULTS

Development of a Photoaffinity Probe to Profile Bromodomains. A wide-spectrum evaluation of bromodo- main inhibitors requires a tool to label as many as possible, if not all, cellular bromodomains. Recently, a series of tricyclic or dicyclic triazolo fused ring derivatives were developed to target a broad range of bromodomains.19,20 Structural optimization has led to the identification of bromosporine (Figures 1A and S1) as a nonselective inhibitor with nanomolar to micromolar binding affinity toward a collection of bromodomains covering all eight subfamilies.20 Considering its high target promiscuity, we reasoned that bromosporine could be a good prototype to design a chemical probe for capturing the bromodomains. In our design, bromosporine was armed by a bifunctional moiety carrying both a diazirine group (for photo-cross-linking) and a terminal alkyne (for bioorthogonal reaction). We expected that such derivatization could facilitate the covalent labeling of the bromodomains upon UV activation and subsequent con- jugation of the target proteins to fluorescent or affinity tags for visualization or enrichment, respectively. In bromosporine, the C8 carbamate and C3′ sulfonamide are two positions that the bifunctional moiety can be attached to without significantly changing the molecule skeleton or complicating the synthesis of the probe (Figure 1A). Importantly, the incorporation of the moiety should not interfere with the inhibitor−bromodomain interactions. By analyzing the reported cocrystal structures of four bromodomains, BRD4(1) (PDB code: 5IGK), BRPF1 (PDB code: 5C7N), BRD9 (PDB code: 5IGM), and TAF1L(2) (PDB code: 5IGL), in complex with bromosporine, we found that all of the bromodomains accommodated bromosporine in a similar pattern (Figure 1B), in which the dicyclic heteroaromatic core of bromosporine was buried deeply in the binding pockets and the C8 carbamate and C3′ sulfonamide were located at the pockets’ entry regions (Figure 1C−F). While the C3′ sulfonamide pointed to the open space in the BRPF1 and TAF1L(2) complexes, there could be steric clash if a larger group is introduced to this position in the BRD4(1) and BRD9 complexes. In contrast, the ethyl tail of bromosporine at the C8 carbamate was solvent accessible in all four complexes, and thereby suitable to introduce substituents. Guided by these analyses, we designed a photoaffinity probe called photo-bromosporine (photo-BS, Figure 1A) by replac- ing the ethyl of the C8 carbamate in bromosporine with the bifunctional moiety. The synthesis of photo-BS followed a 13- step route (Scheme S1) with an overall yield of 0.3%.

Figure 1. Design of photo-BS as a photoaffinity probe for bromodomain profiling. (A) Structure of bromosporine and photo- BS. (B) Superimposition of the structures of BRD4(1), BRPF1, BRD9, and TAF1L(2) in complex with bromosporine. Bromosporine molecules are depicted as sticks, and proteins are shown as ribbons in indicated colors. (C−F) Surface views of bromosporine bound to BRD4(1) (C), BRPF1 (D), BRD9 (E), and TAF1L(2) (F) to show the solvent-exposed C8 carbamate ethyl tail. Red arrows indicate potential steric hindrance for further derivatization. Proteins are colored in light gray. Bromosporine is colored in salmon. Figures were created using Pymol.

Photo-BS Labels Bromodomains in Vitro. We first examined if photo-BS can label the bromodomains. Twenty- one recombinant bromodomains in all eight subfamilies, including ASH1L, ATAD2, BAZ2B, BRD2(2), BRD3(2), BRD4(1), BRD4(2), BRD9, BRWD1(2), CECR2, CREBBP, EP300, KAT2A, KAT2B, PBRM1(1), PBRM1(5), PBRM1(6), TAF1(2), TRIM24, TRIM28, and ZMYND8, were respec- tively incubated with increasing concentrations of photo-BS. After UV irradiation, the probe-labeled proteins were conjugated to rhodamine-azide (Rho-N3) via “click chemistry”, followed by SDS-PAGE and in-gel fluorescence scanning. We observed a dose-dependent labeling of all 21 bromodomains by photo-BS (Figures 2A and S2). The labeling efficiencies of the tested bromodomains by photo-BS seemed to follow a similar trend with the binding affinities of the parent compound bromosporine toward the bromodomains;20 as the more tightly bromosporine bound to a bromodomain, the higher efficiency photo-BS labeled the bromodomain (Figure 2B). In addition, the presence of excessive bromosporine suppressed the photo- BS-mediated labeling of all tested bromodomains (Figure 3A), indicating that the labeling was bromosporine−bromodomaininteraction-dependent instead of nonspecific photo-cross- linking.

We next tested if the photo-BS-induced labeling of bromodomains could be selectively competed by the inhibitors with defined target specificities. To this end, BRD4(1), BRD4(2), CECR2, TAF1(2), and ASH1L were cross-linked with photo-BS in the presence of three bromodomain inhibitors, bromosporine,20 JQ1,8 and PFI-121 (Figure S1), respectively. As expected, bromosporine efficiently blocked the labeling of its known targets BRD4(1), BRD4(2), CECR2, and TAF1(2), while little competition was observed on ASH1L, a poor binder of bromosporine. The JQ1 and PFI-1 treatment led to selective signal reduction on both BRD4(1) and BRD4(2), which were characterized with high affinities to these two inhibitors, while the fluorescence intensities of the rest of the three bromodomains that were inert to JQ1 and PFI-1 were left unaffected (Figure 3B). Furthermore, we varied the concentration of PFI-1 and JQ1 in the photo-BS-mediated cross-linking reactions of BRD4(1) and BRD4(2), respectively. From the competition curves (Figures 3C and S3), the apparent Ki values22 of PFI-1 and JQ1 toward BRD4(1) and BRD4(2) were in line with the reported corresponding dissociation constants (Figure 3D), indicating that photo-BS could quantitatively differentiate the inhibitory activities of a compound against different bromodomains.

Photo-BS Labels Endogenous BCPs in Living Cells. To examine whether photo-BS could be used to label bromodomains in complex proteomes, we added the recombinant CECR2 and BRD4(1) into the lysates of HEK293T cells and performed photo-cross-linking reactions using photo-BS. As shown in Figure 4A, both bromodomains were robustly labeled by the probe and the cross-linking efficiencies were comparable to those using the single proteins.

Figure 2. Photo-BS labels bromodomains in vitro. (A) In-gel fluorescence and concentration dependence curves of photo-BS- induced photo-cross-linking toward representative recombinant bromodomains. For all of the photo-cross-linking experiments, the protein concentration used was 5 μg/mL. After UV irradiation, the photo-BS-labeled proteins were conjugated to Rho-N3 and visualized by in-gel fluorescence scanning. The fluorescence intensity of each band was quantified by ImageJ. Data are reported as mean ± s.d. (n = 2). (B) Comparison of bromosporine binding affinity (Kd) and photo- BS labeling efficiency (EC50) toward the bromodomains. The Kd (data acquired from ref 20) and EC50 (for data, see Figure S2) are displayed on the human bromodomain phylogenetic tree as spheres with indicated sizes and colors.

Remarkably, the labeling of CECR2 and BRD4(1) in the lysates was also selectively competed by their corresponding inhibitors, suggesting that photo-BS-based cross-linking assay could be used to evaluate bromodomain inhibitors in complex proteome samples.

Figure 3. Bromodomain inhibitors selectively compete the photo-BS-induced labeling. (A) The photo-BS-induced labeling toward 21 selected recombinant bromodomains could be competed in the presence of excessive bromosporine. (B) Bromosporine, JQ1, and PFI-1 selectively competed the photo-BS-induced labeling toward BRD4(1), BRD4(2), CECR2, TAF(2), and ASH1L. (C) In-gel fluorescence and competition curves of PFI-1 against BRD4(1) and BRD4(2). The fluorescence intensity of each band was quantified by ImageJ. All curves were normalized between 100 and 0% at the highest and lowest fluorescence intensities, respectively. Data are reported as mean ± s.d. (n = 2). (D) Summarization of the Kd, IC50, and apparent Ki values of PF1-1 and JQ1 toward BRD4(1) and BRD4(2), respectively. aKd values of PFI-1 and JQ1 toward each bromodomain were obtained from refs 21 and 8, respectively. bCalculated apparent Ki. For details of calculation, see the Supporting Information. BS: bromosporine.

Encouraged by the in vitro studies, we next sought to explore the application of photo-BS in living cells. HEK293T cells were incubated with the different concentrations of photo-BS. After UV irradiation, the cells were lysed, and the lysates were “clicked” with Rho-N3. In-gel fluorescence scanning revealed that the cross-linking was dose-dependent (Figure 4B) and saturated between 10 and 20 μM (Figure S4). We further showed that the coincubation with bromosporine greatly inhibited the labeling (Figure 4C). On the other hand, when we used photo-BS to label the lysate derived from the HEK293T cells, the fluorescence intensity of the labeled proteins kept increasing and was not saturated even at a concentration of 50 μM photo-BS (Figure S5). In addition, we noticed the different protein labeling patterns between the lysate and the live cell samples (Figure 4D), suggesting that the target engagement profiles of photo-BS in cells were significantly different from those in vitro.

We next asked if the endogenous BCPs could be labeled by photo-BS in the living cells and, more importantly, if the labeling could be selectively competed by the bromodomain inhibitors. To this end, HEK293T cells were preincubated with or without the bromodomain inhibitors bromosporine, JQ1, and GSK280123 (Figure S1), respectively. Photo-BS was then added to conduct photo-cross-linking in living cells. After lysing the cells, the photo-BS-labeled proteins were conjugated to biotin-azide, followed by streptavidin enrichment. The resulted protein mixtures were resolved by SDS-PAGE and analyzed by immunoblotting using antibodies against BRD4, BAZ2A, and KAT2A. As shown in Figure 4E, photo-BS successfully enriched all three selected BCPs from the cells. Bromosporine and JQ1 dramatically attenuated the enrichment of BRD4 that contains bromodomains targeted by these two inhibitors, while the enrichment of BAZ2A was only impeded by GSK2801, a selective BAZ2A bromodomain inhibitor. In contrast, the enrichment level of KAT2A was not significantly affected by the three tested inhibitors, as it is known that the KAT2A bromodomain is not the target of these inhibitors. This result demonstrated the ability of photo-BS in the enrichment of endogenous BCPs and its potential to evaluate the selectivity of the bromodomain inhibitors.

Photo-BS Leads to Wide-Spectrum Profiling of BCPs.We next used photo-BS, coupled with SILAC-based quantitative proteomics,24 to profile the endogenous BCPs in cells. In a forward SILAC experiment, “heavy” (grown in medium containing 13C,15N-substituted arginine and lysine) and “light” (grown in medium with arginine and lysine in natural isotope abundance forms) HEK293T cells were shone by UV light in the presence and absence of photo-BS (10 μM),respectively. After cell lysis, the “heavy” and “light” lysates were pooled. The captured proteins were biotinylated via “click chemistry”, followed by streptavidin-based affinity purification, gel separation, and in-gel trypsin digestion. The resulting peptide mixtures were analyzed by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). To ensure the data reliability, a reverse SILAC experiment, in which the “light” but not “heavy” cells were cross-linked with photo-BS, was performed in parallel. The same set of SILAC experiments was also performed in U2OS cells. We designated the proteins with SILAC ratio heavy/light (H/L) > 2 in the forward experiment and L/H > 2 in the reverse experiment as bromosporine-interacting proteins. In both cell lines tested, we identified dozens of known BCPs that lay in the regions above the thresholds (Figure 5A,B).

Figure 4. Photo-BS enriches endogenous BCPs in living cells. (A) Recombinant bromodomains could be labeled by photo-BS (5 μM) in the presence of a complex proteome, and the labeling was selectively competed by corresponding bromodomain inhibitors. (B) Concentration- dependent labeling of photo-BS in living cells. (C) Preincubation of the cells with bromosporine reduced the photo-BS-induced labeling intensity. (D) Photo-cross-linking in living cells resulted in a distinct labeling pattern compared with in cell lysate (1 mg/mL); protein bands with different intensities are indicated by red arrows. For all of the photo-cross-linking experiments, after UV irradiation, cells were lysed (for cell-based assays) and the photo-BS-labeled proteins were conjugated to Rho-N3 and visualized by in-gel fluorescence scanning. (E) Photo-cross-linking pull-down in living cells with photo-BS (2 μM) in the absence or presence of different bromodomain inhibitors (5 μM). Samples without photo-BS were prepared as a negative control. After photo-cross-linking and cell lysis, the photo-BS-labeled proteins were conjugated to biotin-N3 and enriched by streptavidin. The eluted protein mixtures were analyzed by immunoblotting against indicated antibodies. The blotting is representative of three independent experiments. BrD: bromodomain. BS: bromosporine.

We then manually picked up the BCPs for further analysis (Figure 5C,D). Specifically, 26 BCPs were detected in the samples derived from HEK293T cells. While BRPF3 was enriched by 2.9-fold in the forward sample, no signal in the reverse one was obtained. The rest of the 25 BCPs gave both forward and reverse SILAC ratios higher than 2. From the U2OS cells, 31 BCPs were detected, in which 26 of these BCPs matched the thresholds for bromosporine-interacting proteins. The proteins BRWD3, CREBBP, EP300, TRIM28, and TRIM33 were either unmet to the SILAC ratio criteria or only detected in one SILAC experiment. In total, 28 BCPs showed more than 2-fold enrichment from both SILAC experiments in HEK293T and U2OS cells. More specifically, 23 of these BCPs were shared by both cell lines. KAT2B and TRIM33 were only enriched from HEK293T, and the unique enrichment of BPTF, BRPF3, and SP100 was obtained from U2OS. These results revealed a good coverage of the photo-BS toward the endogenous BCPs (two-thirds of all 42 BCPs were enriched).

Evaluation of Target Selectivity of Bromodomain

Inhibitors in Living Cells. The wide-spectrum profile of BCPs by photo-BS promoted us to examine whether this probe can be used to evaluate the target selectivity of bromodomain inhibitors in living cells. To this end, both the “light” and “heavy” cells were photo-cross-linked with the probe, and either the “heavy” (forward experiment) or the “light” cells (reverse experiment) were preincubated with the inhibitors of interest. BCPs that showed the high SILAC ratios of inhibitor- untreated/inhibitor-treated in both forward and reverse experiments would be considered as the targets of the inhibitor (Figure 6A). The logarithmic (log2) SILAC ratios (H/L) of the identified BCPs in the forward (x axis) and reverse (y axis) experiments would be plotted in two-dimensional plots. The target BCPs of the tested inhibitors would stand out as the top- left outliers in the plots.

Figure 5. Photo-BS leads to the wide-spectrum profile of BCPs. (A and B) SILAC ratio plots for total proteins identified in HEK293T (A) and U2OS (B) cells. Forward: “heavy” cell with photo-BS (10 μM), “light” cell with DMSO as a control. Reverse: “light” cell with photo-BS (10 μM), “heavy” cell with DMSO as a control. BCPs were highlighted as red dots. See data sets S1 and S2 for SILAC ratios. (C) Heat map showing the SILAC ratios of BrD obtained from parts A and B. Proteins with SILAC ratios of photo-BS-treated/photo-BS-untreated > 2 in both forward and reverse experiments were defined as photo-BS-enriched ones. The names of BCPs enriched by photo-BS in both cell lines are labeled in red, the names of those enriched in either one cell line are labeled in black, and the names of those enriched in neither cell line are labeled in gray. See Table S1 for SILAC ratios. (D) Venn diagram of the enriched BCPs from the SILAC experiments in HEK293T (red) and U2OS (blue) cells.

We first examined the selectivity of bromosporine toward endogenous BCPs in both the HEK293T and U2OS cells. As shown in Figure 6B, five BCPs, BRD2, BRD3, BRD4, BRD7, and BRD9, were identified to bind bromosporine in the HEK293T cells. Similarly, BRD2, BRD3, BRD4, and BRD7 were also found to be targeted by bromosporine in the U2OS cells (Figure 6C). The identification of the three BET family members, BRD2, BRD3, and BRD4, as the endogenous targets of BS was in line with the previous studies that showed that the bromosporine treatment led to BET inhibition signatures of transcription in a number of leukemia cell lines.20 The engagement of bromosporine with BRD7 or BRD9 in living cells, however, has not been characterized. To validate if BRD7 and BRD9 were the targets of bromosporine, photo-BS was incubated with HEK293T and MV4;11, a leukemia cell line, in the presence or absence of bromosporine. As expected, both BRD7 and BRD9 were enriched by photo-BS from both cell lines tested. The addition of bromosporine indeed dampened the enrichment of BRD7 and BRD9 but not KAT2A, a BCP that showed a SILAC ratio around 1 (Figure 6D). These data indicate that bromosporine can engage with endogenous BRD7 and BRD9 in living cells.

Using this photo-BS-based chemical proteomics approach, we evaluated another bromodomain inhibitor, GSK685325 (Figure S1), that selectively targets BRPF1 bromodomain. Remarkably, only BRPF1 showed high SILAC ratios in both the forward and reverse experiments, while other identified BCPs all had a ratio close to 1 (Figure 6E). This is consistent with the reported high specificity of GSK6853 toward BRPF1. Together, these data suggest that our photo-BS-based proteomics method is suitable for comprehensive analysis of target profiles of bromodomain inhibitors in living cells.

▪ DISCUSSION

The evaluation of inhibitors, especially for their selectivity, is a critical step during drug discovery. The development of chemical proteomics approaches has provided powerful methods26,27 to examine the selectivity of an inhibitor against dozens or even thousands of proteins in one experiment, particularly for the assessments toward a family of structurally and functionally related proteins. Most of the reported chemical proteomics studies for inhibitor assessment have been focusing on enzymes,28−32 for example, kinases that transfer a phosphate from an adenosine triphosphate (ATP) to the substrates. Early examples include ATP-derived probes33,34 and beads immobilized with kinase inhibitors (kinobeads)35 to profile kinases in cell lysates. In a recent study, the development of sulfonyl fluoride-based probes, which cova- lently captured up to 133 endogenous kinases (∼25% of the human kinome) from a single cell line, enabled the cell-based assessment of the kinase inhibitors.36 The bromodomains have emerged as “hot spots” for inhibitor development, highlighting the significance of the methods used for the assessment of bromodomain inhibitors. Photo-BS developed in this study represents, to the best of our knowledge, the first chemical probe that captures wide-spectrum endogenous bromodo- mains and enables the evaluation of bromodomain inhibitors in living cells.

Figure 6. Target selectivity evaluation of bromodomain inhibitors in living cells. (A) Schematics illustrating the chemical proteomics method for the evaluation of bromodomain inhibitors using photo-BS. (B and C) Two-dimensional plots showing the log2 SILAC ratios of BCPs for the evaluation of bromosporine in HEK293T (B) and U2OS (C) cells. (D) Photo-cross-linking pull-down in living cells with photo-BS in the absence or presence of bromosporine to validate the engagement of bromosporine with endogenous BRD7 and BRD9. After photo-cross-linking and cell lysis, the photo-BS-labeled proteins were conjugated to biotin-N3 and enriched by streptavidin. The eluted protein mixtures were analyzed by immunoblotting against indicated antibodies. The blotting is representative of three independent experiments. BS: bromosporine. (E) A two- dimensional plot showing the log2 SILAC ratios of BCPs for the evaluation of GSK6853 in HEK293T cells. For all of the SILAC experiments, both “heavy” and “light” samples were treated with photo-BS (10 μM). In forward experiments, “heavy” cells were treated with bromosporine (10 μM) or GSK6853 (5 μM). In reverse experiments, “light” cells were treated with bromosporine (10 μM) or GSK6853 (5 μM). BCPs as top-left outliers were highlighted in red. See Figures S6, S7, and S8 for two-dimensional plots of other identified proteins. See data sets S3−S5 for SILAC ratios.

In recent years, photo-cross-linking has become a powerful approach to examine protein−protein and small molecule− protein interactions.37 Given the nature of the method that converts noncovalent interactions to covalent chemical bonds, photo-cross-linking enables capture of not only strong binding partners but also weak and transient ones that are usually considered as “nonspecific” binders. In addition, the commonly used photo-cross-linkers (e.g, benzophenone and diazirine) are nonpolar and hydrophobic. They could therefore mediate nonspecific protein interactions, which often results in ligand- independent nonspecific photo-cross-linking.38,39 Not surpris- ingly, the application of photo-BS in living cells captured many other proteins other than the BCPs (Figure 5A and B). We reasoned that most of these captured proteins should be involved in weak interactions either with the bromosporine scaffold or with the diazirine moiety. As a common practice, the competitive photo-cross-linking experiment using a relatively low concentration (10 μM) of bromosporine as the competitor was performed to distinguish the strong or “specific” protein binders from the “nonspecific” including the weak ones. Consistent with our hypothesis, among thousands of proteins captured by photo-BS, only six of them showed competition by bromosporine in either HEK293T (Figure S6) or U2OS (Figure S7) cells. Notably, five (in HEK293T) and four (in U2OS) of the six proteins, respectively, were BCPs known or validated to bind bromosporine. This result demonstrated the robustness and feasibility to use photo-BS, in combination with the competitive chemical proteomics approach, for the identi- fication of specific targets of bromodomain inhibitors, even in the presence of a high photo-cross-linking background. In addition to the BCPs, the other three proteins (PI4K2B, SDR39U1, and RPUSD2) that were specifically competed by bromosporine are potentially interesting “off-targets” of this inhibitor at the whole-proteome level, which need further validation and characterization.

While the photo-BS provided a good BCP coverage (two- thirds of the BCPs were enriched), 14 BCPs were not enriched by the probe in our SILAC experiments (Figure 5C,D). By analyzing these 14 BCPs, we noticed that the enrichment failure could be due to the limited expression levels of the BCPs. For example, in HEK293T and U2OS cells, the abundances of BRDT, which is a testis specific BET family member, and SP140, which is mainly expressed in leukemia and lymphoma cell lines, are too low to ensure a robust detection by mass spectrometry (the expression data were obtained from the Human Protein Atlas40). We believe that more BCPs would be added to the enrichment list if the photo- BS-based SILAC experiments were performed in additional cell lines. On the other hand, the low binding affinity of some BCPs toward bromosporine could also be the reason for why these proteins were undetected or did not meet the enrichment criteria in our SILAC experiments. For example, the acquired EC50 values of photo-BS-induced labeling toward ASH1L, BAZ2B, and TRIM28 bromodomains were all larger than 20 μM (Figure S2). It is therefore not surprising that 10 μM photo-BS used in the SILAC experiments failed to enrich these BCPs. One solution is to use a higher concentration of photo- BS in the SILAC experiments or develop new photoaffinity probes on the basis of other bromodomain inhibitors with target profiles different from bromosporine,CFT8634 which will be an important next step.