A 5-Br-1-Propylisatin Derivative as a Promising BRD9 Ligand: Insights from Computational and STD NMR Investigation

TL;DR

This study identifies a new compound that binds to BRD9, a protein linked to diseases with faulty epigenetic control, using advanced techniques like NMR and virtual screening.

Contribution

The study introduces a novel isatin derivative as a potential BRD9 ligand and validates the use of STD NMR in fragment-based drug discovery.

Findings

A new isatin derivative (compound 2) was identified as a BRD9 ligand through virtual screening and confirmed via AlphaScreen assays.

STD NMR experiments confirmed the binding of compound 2 and mapped its interaction epitope on BRD9.

Compound 2 was shown to bind in the canonical BRD9 pocket, as confirmed by competitive displacement with a reference ligand.

Abstract

Bromodomain-containing protein 9 (BRD9) belongs to the non-canonical BAF chromatin remodeling complex and represents a relevant therapeutic target in pathologies featuring dysregulated epigenetic control. The absence of clinically validated inhibitors and the need for diversified chemical entities highlight the interest in identifying new scaffolds targeting this protein. In this study, Saturation Transfer Difference Nuclear Magnetic Resonance (STD NMR) was employed to assess its suitability for characterizing BRD9–ligand interactions within a fragment-based discovery framework. STD NMR conditions were first optimized using the known BRD9 ligand 1, verifying the presence of interaction signals. A pharmacophore-based virtual screening campaign was then performed using libraries of commercially available fragments, leading to the selection of a novel isatin derivative, i.e., compound 2, whose binding was demonstrated in AlphaScreen assays. STD NMR experiments provided epitope mapping consistent with the predicted binding mode, thus supporting the stability of the interaction in solution. Moreover, a competitive STD experiment demonstrated displacement of 2 by a reference ligand, confirming the binding within the canonical BRD9 pocket. Overall, this study establishes STD NMR as a reliable approach for probing BRD9–ligand interactions and for the identification and validation of BRD9-targeting scaffolds suitable for future structure-guided optimization.

Figures

Full text

1. Introduction

Bromodomain-containing protein 9 (BRD9) is a component of the non-canonical BAF (ncBAF) chromatin remodeling complex, which plays a key role in the regulation of gene expression, maintenance of cell identity, and oncogenic transcriptional programs. Owing to its involvement in the SWI/SNF complex, BRD9 has emerged as an attractive therapeutic target in cancers harboring perturbations of chromatin remodeling machinery, such as synovial sarcoma and acute myeloid leukemia [1,2]. The bromodomain of BRD9 specifically binds to acetylated lysine residues on histone tails, mediating protein–chromatin interactions that are crucial for transcriptional regulation. Therefore, the discovery of selective BRD9 binders is of great interest for the development of chemical probes and potential anticancer agents [3].

Despite significant advances in the structural characterization of BRD9 and the identification of potent modulators, the identification of new binders is of pivotal importance since no molecules are currently in the clinical stage following the failure of two clinical trials of compounds CFT8634 and FHD-609, which are no longer advancing as ongoing clinical trials. In particular, the clinical development of CFT8634 was discontinued following limited single-agent efficacy in the Phase 1/2 study despite demonstrating robust BRD9 degradation, and the FHD-609 study highlighted issues, including cardiac safety events (ClinicalTrials.gov identifiers NCT05355753 for CFT8634 and NCT04965753 for FHD-609 [2]. In this scenario, disclosing new fragment-like ligands offers a valuable opportunity for BRD9 agents exploration and optimization. Notably, in silico approaches provide a solid framework to support fragment-based drug discovery, representing an effective strategy for the identification and development of new bioactive compounds [4]. Different powerful tools have been leveraged with this aim, e.g., bioisosteric fragment replacement and pharmacophore screening [5,6,7]. For the subsequent evaluation of the bioactivity of the compounds identified in silico, traditional biophysical techniques are widely employed, despite some limitations arising from their lack of sensitivity for low-affinity interactions. Among them, Saturation Transfer Difference Nuclear Magnetic Resonance (STD NMR) represents a powerful and highly sensitive technique for studying protein–ligand interactions, capable of identifying binding events even in the μM–mM affinity range [8,9,10,11,12,13]. By selectively saturating protein resonances and monitoring the transfer of magnetization to bound ligands, STD NMR provides detailed information on the binding epitope and offers valuable insights into the molecular recognition process [14,15].

Here, we report the application of STD NMR spectroscopy to the characterization of BRD9–ligand interactions, for which, to the best of our knowledge, only one paper in the literature reported the application of this technique to BRD9 [16]. We started applying STD NMR to the known BRD9 binder 1 [17] (Figure 1) to evaluate the applicability of this spectroscopic technique to the target protein and to refine the experimental conditions, thus obtaining the epitope map of the investigated known ligand, in line with already reported outcomes. Starting from these robust premises, we performed a pharmacophore-based virtual screening coupled with molecular docking experiments, leading to the selection of a set of promising compounds [5,6]. Among them, an isatin derivative emerged as a novel fragment-like hit. After confirming its binding towards BRD9 through AlphaScreen assays, STD-NMR experiments were carried out to investigate its ability to interact with the protein. The obtained spectra not only confirmed the feasibility of applying STD NMR to BRD9 but also yielded consistent epitope maps in agreement with computational binding models. These findings establish STD NMR as a robust and informative approach for investigating BRD9–ligand interactions and provide the basis for the structure-guided design of novel BRD9 modulators.

2. Results and Discussion

2.1. STD NMR Experiments for Exploring BRD9 Binding

The first aim of this study was to determine and optimize the experimental conditions for STD NMR experiments to detect BRD9–ligand interactions. To this end, we used a compound previously identified by our group as a reference ligand, featuring a 2,4,5-trisubstituted-2,4-dihydro-3H-1,2,4-triazol-3-one scaffold (compound 1, Figure 1), and for which we demonstrated its ability to interact with the bromodomain of BRD9 [17]. In this way, the sensitivity and reliability of STD NMR in detecting ligand–protein interactions were evaluated in an already characterized robust system, thereby laying the groundwork for subsequent screening and validation studies of novel compounds. Compound 1 (Figure 1 and Figure S1) showed a promising binding to BRD9, assessed through AlphaScreen, SPR, and in vitro assays. For this reason, and for its dissociation constant (K_D_) in the medium micromolar range (i.e., K_D_~40 μM determined by SPR), 1 was selected to assess the suitability of the STD NMR technique to detect BRD9 binders and to optimize the experimental conditions. We considered this result very interesting since only one article is in the literature regarding the application of STD NMR on BRD9 [16].

Different experiments were carried out under different experimental conditions: for the first set of experiments, one sample was prepared following the already reported experimental conditions, with 40 µM of BRD9, 1.33 mM of 1 (see also Section 3) [16]. Although STD signals were identified from the resulting NMR spectra, they were considered unreliable due to precipitate formation. For this reason, a second sample was prepared for a second set of experiments, setting the experimental conditions as follows: 20 µM of BRD9, 1.33 mM of 1, 5% of DMSO, and 0 ppm and 40 ppm as irradiation frequencies. In the resulting spectra (Figure S2), STD signals emerged, leading to the assessment of the STD build-up curves and epitope map of 1 (Figure 1).

These results not only confirmed the feasibility of studying BRD9 using this technique but also contributed to refining the experimental conditions for working with this bromodomain-containing protein.

Interestingly, the obtained epitope map (Figure 1B) was in line with the binding mode previously reproduced by means of molecular docking experiments (Figure 1C,D) and pharmacophore screening (Figure S1). In more detail, H1 is one of the protons with minor STD intensity due to the proximity to structural water molecules, while the other protons, i.e., Ha, Hd and He, as evident from the front view of the binding mode (Figure 1D), featured less STD intensities since they are more exposed to the solvent.

Notably, these results are of utmost importance since they confirmed the possibility of investigating BRD9 ligands with this technique.

2.2. Pharmacophore-Based Virtual Screening Campaign

Taking advantage of our previously developed 3D structure-based pharmacophore models [5,6] and starting from the encouraging data above reported on the reference compound 1, a fragment-based virtual screening campaign was performed to identify new promising items whose predicted BRD9 binding could then be evaluated through STD NMR experiments. In detail, 4 virtual libraries of fragment compounds from Otava chemicals were accounted for in the in silico experiments (see Section 3) with the fragment-like “AHRR” model, where “A” indicates the acceptor feature, “H” the hydrophobic group, and “R” the aromatic moieties [5]. Through this in silico tool, a structure-based pharmacophore screening was performed. In particular, “structure-based” indicates that the pharmacophore model was generated directly within the protein binding site, using the information retrieved from the binding modes of co-crystallized ligands. Thus, this is very powerful when coupled with molecular docking experiments, accelerating the selection of promising compounds that featured the key molecular interactions for BRD9 recognition. Therefore, the general workflow involved the following steps (see also Section 3):

• (1)preparation of the virtual libraries;
• (2)molecular docking experiments within the BRD9 binding site;
• (3)pharmacophore screening with “pharm-fragment”, after the molecular docking, considering the “score in place” option in order to analyze the exact predicted docking pose;
• (4)selection of the compounds through qualitative and quantitative analyses, specifically considering the established interaction with fundamental amino acids, and the in silico predicted docking score and Phase Screen score values, respectively. The latter parameter reflects the degree of correspondence with the pharmacophoric features.

Finally, 10 compounds (2–11, Table S1) were selected and purchased for the subsequent binding assessment step, based on the qualitative and quantitative analysis and commercial availability. In more detail, the considered criteria for the selection of the compounds were: (a) docking score ≤ 6.0, considering a cut-off of 4.0 kcal/mol from the best value; (b) PhaseScreen score ≥ 1.5, considering a cut-off of 0.6 from the best value; the respect of at least 3 of the 4 features of the employed pharmacophore model, of which the three mandatory features included the acceptor, the hydrophobic and the aromatic feature, as minimum requirement for BRD9 recognition. The compounds were tested through AlphaScreen assays in order to prioritize only the fragment compounds able to interact with BRD9 for the subsequent step of structural characterization of the protein–ligand interaction. Among them, compound 2, an isatin-based derivative (Figure 2A), emerged as a promising BRD9 binder (Table 1) and was selected for further investigation via STD NMR experiments. The presence of an isatin-based chemotype is particularly relevant, as this versatile heterocyclic core, featuring well-established pharmacological potential, is known to occur across diverse biological systems, including marine organisms, where it often plays ecological or signaling roles [18]. Such nature inspired scaffolds frequently display a rich interaction potential and provide a privileged starting point for the identification of ligands capable of engaging challenging protein targets [19]. In this context, the obtained experimental data highlighted the importance of identifying this isatin-based compound as an unprecedented scaffold for developing novel BRD9 binders, owing to its privileged nature and its suitability to be easily optimized with different synthetic procedures.

Notably, 2 showed a promising ability in displacing the acetylated histone, featuring a binding mode in which interactions with fundamental amino acids, i.e., Asn100, Tyr99, Phe44, Tyr106, were established. In fact, 2, in agreement with the hydrophobic feature “H” of the model, exhibits a propyl group at the N1 position appropriately accommodated in the deep hydrophobic cavity of BRD9 (Figure 2B). This structural feature is particularly relevant, as the presence of bulky aliphatic substituents in the acetyl lysine mimetic region (corresponding to the “A” acceptor and “H” hydrophobic features, Figure 2B) is a key determinant in achieving BRD9 recognition and selectivity [20]. Furthermore, H-bonds with the key residues Asn100 and Tyr99 were established (Figure 2B).

Moreover, compound 2 featured a promising pharmacokinetic profile comparable to that of compound 1 (Figures S7 and S8).

2.3. Binding Assessment of 2 Through STD NMR Technique

After the encouraging preliminary results obtained from the AlphaScreen assays, compound 2 was further investigated using STD NMR. The experiments were carried out under the optimized conditions established for BRD9 (see Section 2.1 and Section 3). A series of six experiments was recorded using increasing saturation times (0.5, 0.75, 1, 2, 4, and 6 s), employing a concentration of 500 μM of 2, 20 μM of BRD9. Analysis of the resulting spectra (Figure S3) enabled the construction of the STD build-up curve (Figure 3A) and the corresponding epitope map (Figure 3B), which was entirely consistent with the predicted binding mode derived from molecular docking studies (Figure 2B), thus providing additional evidence for the stable interaction of 2 with BRD9 in solution.

In this context, a challenging aspect occurred during the analysis of the results. Notably, a chemical exchange event emerged from the NMR spectra, where duplicated signals were recorded for both aromatic and methyl protons (Figure 4). This aspect revealed the existence of an additional chemical entity, which we identified as the gem-diol species as the product of the reversible ketone hydration reaction, which is in equilibrium in the aqueous solvent in which the sample was prepared, as also clearly confirmed by the related mass spectra (Figures S4 and S5) [21,22,23].

The formula applied to determine the STD intensities (Equation (1)) was then slightly modified as follows to consider the duplicated protons (Figure 5) equally:

$$[eqn]$$

where the intensity in the off-resonance (reference) spectrum is indicated with $[eqn]$ for H and H^′^, respectively, and the intensity in the on-resonance (saturated) spectrum is indicated with $[eqn]$ for H and H^′^, respectively.

Moreover, to further strengthen the detected 2-BRD9 binding, a competitive STD NMR experiment was carried out by adding a reference BRD9 ligand under the same experimental conditions, i.e., I-BRD9 (Figure S6). Upon introduction of this binder, the STD NMR signals of 2 decreased significantly (Figure 5), indicating the displacement of 2 from the binding site. This competitive behavior confirmed the interaction of 2 with BRD9 within the same binding pocket as that of I-BRD9.

In summary, these results not only validate the molecular recognition of 2 by BRD9 but also demonstrate the capability of STD NMR to probe competitive binding events in the BRD9 system reliably. Moreover, the discovery pipeline here reported led to the identification of a novel fragment-based item able to bind BRD9, which paves the way for future development of optimized isatin-based compounds featuring promising biological activity towards BRD9.

3. Materials and Methods

3.1. Computational Details

3.1.1. BRD9 Grid Generation

Prior to performing docking calculations, the Protein Preparation Wizard workflow (Maestro, Schrödinger, New York, NY, USA) was employed using the crystal structure of the BRD9 bromodomain in complex with BI-9564, the latter used as a reference for grid box generation (PDB code: 5F1H) [24]. All hydrogen atoms were added, and bond orders were assigned.

3.1.2. Ligand Preparation

The commercially available virtual library employed was prepared using LigPrep software (version 74133, Schrödinger Suite, New York, NY, USA), accounting for the protonation state at a pH = 7.4 ± 1.0 and minimizing the structure with OPLS 2005 force field, retaining the specified chirality and generating all the possible stereoisomers and tautomers.

3.1.3. Virtual Screening

The virtual libraries were subjected to molecular docking experiments, performed using the Virtual Screening Workflow tool as implemented in Schrödinger Suite and using Glide (version 10.7) [25,26,27] software, considering three levels of precision: High-Throughput Virtual Screening scoring and sampling (HTVS), Standard Precision scoring and sampling phase (SP), Extra Precision scoring and sampling phase (XP). Precisely, the screened libraries from Otava chemicals were:

• General Fragment Library (13,059 compounds);
• Halogen Enriched Fragment Library (679 compounds);
• High Fsp^3^ (fraction of sp^3^ hybridized carbon atoms) Fragment Library (3309 compounds);
• Stereogenic Centers Fragment Library (2337 compounds).

For each library, the following scheme was applied: after HTVS step, 60% of the best poses were saved and used for the subsequent step; after SP level, 60% of the best poses were saved and used for the subsequent step; in the XP level, ten poses were generated for each input, and finally the first 80% ranked items were saved for the final output. Subsequently, each library was subjected to pharmacophore screening applying the developed 3D structure-based pharmacophore model “pharm-fragment”, i.e., “AHRR” [5], using the “Ligand and database screening” tool in Phase [28,29,30]. Compounds were selected considering the match of 4/4 as a requirement [5]. For the final selection, the output from the pharmacophore screening was further refined after a qualitative analysis considering the docking poses and the established interaction with key residues in the BRD9 binding site.

3.2. AlphaScreen Assays

AlphaScreen experiments were conducted by Reaction Biology Corporation (Malvern, PA, USA). Recombinant His-tagged BRD9, the tested small molecules, and the biotinylated H4 (1–21) peptide bearing K5/8/12/16 acetylation were dispensed into 384-well OptiPlates and allowed to equilibrate for 30 min at room temperature under gentle agitation. Bromosporine was included as a reference inhibitor for BRD9 and evaluated in a 10-point IC_50_ assay using a threefold serial dilution beginning at 10 μM. Following the initial incubation, streptavidin-coated donor beads and nickel-chelate acceptor beads were added, and the plates were kept protected from light for an additional 60 min with mild shaking. The recombinant bromodomain, test compounds, and bead suspensions were prepared as 4× stock solutions in assay buffer consisting of 50 mM HEPES-HCl (pH 7.5), 100 mM NaCl, 1 mg/mL BSA, 0.05% CHAPS, and 0.5% DMSO. Alpha signals (excitation at 680 nm; emission 520–620 nm) were recorded using an EnVision plate reader (Perkin Elmer, Waltham, MA, USA). Concentration–response curves were generated in GraphPad Prism 9 by fitting the data with a nonlinear regression model.

3.3. STD NMR Experiments

3.3.1. Sample Preparation for BRD9 STD NMR Experiments

Two samples were prepared for 1 and BRD9 in order to optimize the experimental conditions. The first sample was composed of 1.33 mM of 1 and 20 μM of protein and analysed at 278.15 K. The first buffer employed was a 50 mM PBS/D_2_O buffer (with 300 mM of NaCl, 10 mM of EDTA-d_8_, 10 mM of DTT-d_10_, 50% D_2_O) at pH 6.1 [16]. Due to precipitation problems, these experimental conditions were discarded.

The second 1 and BRD9 sample was composed of 1.33 mM of 1 and 20 μM of protein, and analysed at 278.15 K. Samples of 2 and BRD9 were composed of 500 μM of ligands and 20 μM of protein analysed at 278.15 K. All samples were prepared in 10 mM PBS/D_2_O buffer (with 95 mM NaCl, 2.7 mM KCl, and 5% of DMSO-d_6_) at pH 7.4. To perform a following competitive assay, 400 μM of I-BRD9 was added to the sample of 2 and BRD9.

3.3.2. Nuclear Magnetic Resonance

For 1 and BRD9, the build-up curves were acquired with irradiation frequency of 0 ppm/0.62 ppm (δ^0^) and 40 ppm (δ*), at 0.5 s, 0.75 s, 1.0 s, 2.0 s, 4.0 s, 6.0 s and 8.0 s saturation times, with 512 scans (considering the experiments at 0.5 s and 0.75 s), 256 scans (considering the experiments at 1.0 s and 2.0 s) and 128 scans (considering the experiments at 4.0 s, 6.0 s and 8.0 s). A relaxation delay (D1) of 8 s was applied.

For 2 and BRD9, the build-up curves were acquired with irradiation frequency of −1 ppm (δ^0^) and 40 ppm (δ*), at 0.5 s, 0.75 s, 1.0 s, 2.0 s, 4.0 s, and 6.0 s saturation times, with 1k scans (considering the experiments at 0.5 s and 0.75 s), 512 scans (considering the experiments at 1.0 s and 2.0 s) and 256 scans (considering the experiments at 4.0 s and 6.0 s). A relaxation delay (D1) of 6 s was applied. All the STD NMR experiments were performed with the pulse sequence stddiff.3 [8,31] and saturation was achieved by applying a train of 50 ms Gaussian pulses (0.40 mW) on the f2 channel, at 0.62 ppm, 0 ppm, −1 ppm (on-resonance experiments), and 40 ppm (off-resonance experiments). The broad protein signals were removed using a 25 ms spinlock (T1ρ) filter (stddiff.3) [8,31]. 1 and BRD9 sample were analyzed at a ^1^H frequency of 700 MHz on a Bruker Avance III spectrometer equipped with cryoprobe QCI 5 mm with Z-axis gradient, internal coil for ^1^H and external coil for ^31^P, ^13^C and ^15^N (^1^H and ^13^C refrigerated). 2 and BRD9 experiments were recorded at ^1^H frequency of 600 MHz on a Bruker Avance III spectrometer equipped with a cryoprobe QCI Cryo 5 mm (^1^H/^19^F ^15^N/^13^C) for ^1^H, ^15^N, ^13^C, and ^19^F with ^2^H decoupling. For each proton, they were then fitted mathematically to a mono-exponential equation:

$$[eqn]$$

from which the initial slopes ( $[eqn]$ ) were obtained. The binding epitope mapping was obtained by dividing the initial slopes by one of protons Hc for 1 and H3 for 2, to which an arbitrary value of 100% was assigned.

4. Conclusions

In conclusion, we demonstrated that STD NMR represents a highly effective and reliable tool for probing BRD9–ligand interactions within a fragment-based drug discovery framework, paving the way for further screening and structure-activity relationship studies. By first validating and optimizing the technique with the previously characterized BRD9 ligand 1, it was possible to establish experimental conditions that provided robust and reproducible STD responses, enabling a detailed epitope analysis consistent with known binding models. This validation step was essential not only for confirming the applicability of STD NMR to BRD9, a bromodomain for which limited STD data were previously reported, but also for defining experimental conditions suitable for subsequent experiments. Once validated, the technique was successfully integrated with a pharmacophore-based virtual screening campaign. This combined approach led to the identification of a novel isatin-like fragment, which displayed competitive and detectable binding to BRD9 in AlphaScreen assays. Analysis of the STD NMR experiments provided build-up curves and an epitope map coherent with the predicted binding mode, thereby confirming that the fragment engages the canonical BRD9 cavity, further corroborated by competitive STD experiments with the nanomolar I-BRD9 inhibitor.

Overall, the results underline the value of STD NMR as a sensitive, robust, and fast technique capable of detecting low-affinity interactions and obtaining structural insights during the fragment-based drug discovery. The identification of an isatin-based scaffold with experimentally validated engagement of BRD9 highlights a new potential semi-natural scaffold as a starting point for structure-guided optimization and illustrates the advantage of complementing in silico methods with advanced NMR-based screening. Future work will focus on the development of isatin-based analogues, systematic structure–activity relationship exploration, and further integration of STD NMR with other biophysical and computational techniques to accelerate the identification of selective and potent BRD9 modulators.