Ligandability assessment of the C‐terminal Rel‐homology domain of NFAT1

Transcription factors are generally considered challenging, if not “undruggable”, targets but they promise new therapeutic options due to their fundamental involvement in many diseases. In this study, we aim to assess the ligandability of the C‐terminal Rel‐homology domain of nuclear factor of activated T cells 1 (NFAT1), a TF implicated in T‐cell regulation. Using a combination of experimental and computational approaches, we demonstrate that small molecule fragments can indeed bind to this protein domain. The newly identified binder is the first small molecule binder to NFAT1 validated with biophysical methods and an elucidated binding mode by X‐ray crystallography. The reported eutomer/distomer pair provides a strong basis for potential exploration of higher potency binders on the path toward degrader or glue modalities.


| INTRODUCTION
Transcription factors (TFs) play a critical role in cellular processes and represent a highly relevant target class for drug discovery, given their association with a wide range of diseases, including cancers and developmental disorders. [1,2]TFs directly regulate gene transcription by binding to specific DNA sequences, serving as a crucial link between cell signaling and gene regulation. [3]Often functioning as (hetero-) dimers, two TFs cooperatively bind to maintain their sequence-specific binding properties. [4]With over 1500 proteins, TFs represent a substantial target class, posing considerable challenges for drug development and, in some cases, being deemed "undruggable." [5]Targeting the protein-DNA interaction directly has proven to be challenging due to either specificity issues on the DNA site or highly charged DNA binding interfaces on the protein site. [6]Successful strategies for targeting TFs have included inhibiting nuclear hormone receptor ligand binding domains and developing protein-protein interaction inhibitors. [7]The latter is exemplified by the discovery and clinical development of inhibitors of the interaction of the TFs p53 and Nrf2 with their specific E3 ligases mouse double minute 2 homolog and Keap1, [8,9] which leads to increased expression levels and induction of the respective pathways.Other successful targeting approaches also include the recruitment of TFs as (neo-) substrates to E3 ligases.This is exemplified by the discovery of glue molecules for E3 ubiquitin ligase cereblon (CRBN), which enhance the interaction with the TFs IKZF1 and IKZF3, [10] and molecules that enhance the β-catenin::β-transducin repeat-containing protein interaction. [11]A more modular approach involves degradation by proteolysis-targeting chimeras (PROTACs), which has recently been applied to the TF signal transducer and activator of transcription 3. [12] In this approach, a nonfunctional binder can serve as a protein of interest (POI) binder, providing a different angle on targeting multidomain TFs.
The nuclear factor of activated T cells (NFAT) proteins constitute a family of calcineurin-regulated TFs. [13]NFAT1 plays a key role in the activation and regulation of T cells. [14]In activated T cells, NFAT1 controls cytokine expression in cooperation with other TFs. [15]wever, upon suboptimal T-cell activation, NFAT1 homodimers mediate the transcriptional activation of anergy-inducing genes, thereby promoting T-cell tolerance. [15]Due to its central function in T-cell regulation, NFAT1 constitutes an attractive target for immunemodulating therapies, such as cancer immunotherapy and immunosuppressive therapies. [16,17]However, direct targeting of NFAT1 with biophysically validated small molecule binders has not yet been reported.NFAT1 is a multidomain protein comprising the NFAT homology region (NHR), the Rel-homology domain (RHD) domain, and the C-terminal domain (Figure 1a).In this study, we set out to assess the ligandability of the C-terminal domain of the RHD (RHD-C) which has shown to be essential for NFAT1 homodimer formation. [18]e evaluation of small molecule ligandability before initiating a ligand discovery campaign can substantially conserve resources and facilitate targeted investment into specific proteins, domains, and mutants.Conventional methods for assessing ligandability, such as SiteMap, [19] depend on static structures as an input and strive to pinpoint binding hot spots through the analysis of surface/pocket morphology and computation of interaction potentials for probes, typically employing a grid-based approach.Incorporating protein mobility through molecular dynamics (MD) simulations may enable the identification of cryptic pockets that are hidden in static crystal structures and are induced and therefore only visible once a matching fragment is bound. [20]Specifically, cosolvent MD methodologies have proven to be beneficial in the detection of transient pockets and assessment of ligandability since the nature of fragments bound in simulations as well as their interaction energies provide insights into chemical motifs favored in target binding. [21,37]The efficacy of combining in silico and experimental fragment-based approaches has been exemplified by the discovery of chemical matter binding to cryptic pockets on the E3 ubiquitin ligase VHL and its associated protein-protein interfaces. [22] this study, we present the results of the workflow described above.First, we crystallized the NFAT1 RHD-C domain in a non-DNA bound, apo form.Then, using the apo structure, we applied MD methodology to probe protein ligandability.Finally, we applied canonical fragment-based screening (FBS) approaches to identify small molecule binding to the NFAT1 RHD-C domain.coli.The construct was designed based on a previously solved structure on NFAT1 in complex with human immunodeficiency virus long terminal repeat kB element (Figure 1b). [23]The protein sample was subjected to a broad crystallization screening, which yielded plate-shaped crystals in several conditions.The initial apo structure was solved at 1.74 Å resolution and contained two NFAT1  molecules in the asymmetric unit (ASU) (Figure 1c, Supporting Information S1: Table 1). The strctural integrity of the truncated protein was confirmed by the superposition of a monomer to the respective domain of the template structure (protein data bank [PDB] ID: 1P7H, C a rmsd monomerA value of 0.35 Å).Interestingly, the two molecules in the ASU form a similar NFAT dimer as observed in the DNA complex structure (C a rmsd dimer value of 6.1 Å) (Figure 1d).These results encouraged further assessment of the protein's ligandability by small molecules using computational methods, based on the newly generated X-ray structure.

| Cosolvent MD simulations
The dimeric crystal structure indicated two different side-chain rotamers of Tyr659 in chain A/B (Figure 2a), which gave rise to a small hydrophobic cavity formed by Val610, Tyr659, Ile661, and Arg667 in one monomer (Figure 2b).We performed cosolvent MD simulations of the NFAT1 RHD-C domain using a ternary mixture of propane, benzene, and water to probe ligandability of this region alongside the whole surface of the individual protein monomers.Upon trajectory analysis, we identified another region on the NFAT1 RHD-C domain where probe molecule binding was much more likely compared with the small hydrophobic cavity around Tyr-659 (Figure 2c).In this region, we observed the binding of propane and in particular benzene molecules independent of the Tyr-659 rotamer state.Calculated grid-free energies for benzene molecules reached −10 kJ/mol in this region, which was found to be a unique feature of this region on the surface of the NFAT1 RHD-C domain.Based on these in silico results, we concluded that the region between the loop regions incl.Ser-605 and Ser-633 may in fact be amenable to ligand binding.

| FBS of the RHDC domain of NFAT1
Based on the expectation of challenging ligandability, we selected protein-observed nuclear magnetic resonance (NMR) using isotopically labeled protein as the FBS method. [24]The very sensitive method was previously successfully applied to identify fragments binding to TFs (e.g., NRF2). [25]We expressed and purified uniformly 15  with respect to the spectral quality using 1 H- 15 N heteronuclear single quantum coherence ( 15 N-HSQC), and the spectral quality at 30 μM protein concentration was sufficient to assess the binding of compounds by chemical shift perturbation (CSP) (Supporting Information S1: Figure 1).To identify binders, we screened a proprietary fragment library of 3082 compounds in cocktails of five fragments at 500 μM each.In the screening campaign, eight hit pools were identified.All fragments of the hit pools were measured individually, and based on the observed CSPs, only four of these remained valid after deconvolution (Figure 3a).The remaining hits were titrated to obtain their K D with 15 N SOFAST HMQC experiments where only compound 1 resulted in a saturable NMR K D of 729±69 µM.The mean K D was obtained from curves of selected cross peaks ± standard deviations (Figure 3b, Supporting Information S1: Figure 1).To further increase confidence in the hit, we purchased the (S)-2 and (R)-3 enantiomers and repeated the CSP experiment at a single concentration of 250 µM (Figure 3d,) and we unambiguously assigned the eutomer 2 and distomer 3 of the racemic mixture.

| Crystal structure NFAT1 RHD-C domain in complex with eutomer 2
For establishing a soakable system, conditions from the initial crystal screening were reevaluated.A condition out of a Morpheus Screen (Molecular Dimensions) was prioritized, as crystals could be directly flash-frozen from the respective soaking condition.After soaking with the racemic mixture (1) at a concentration of 5 mM over 24 h, we collected a data set in the same space group P2 1 with slightly different cell constants at a resolution of 1.55 Å (Supporting Information S1: Table 1).The protein conformation is very similar to the apo crystals (C a rmsd dimer value of 0.2 Å) and the compound is bound to both NFAT molecules in the ASU.As suggested by the NMR experiments, only the S-conformer is bound (Supporting Information S1: Figure 1).The binding mode to chain A is shown in  The superposition of grid-free energies for benzene binding with the experimentally generated fragment structure (Figure 4c) shows very good overlap in terms of identified pocket region as well as required chemical features for binding.While the hydrophobic pharmacophore could be predicted in silico using benzene probes, the adjacent hydrogen bonding pharmacophores were identified through experimental probing.
We hypothesize that the hydrogen bond donor features may have been accessible computationally as well if more elaborate fragment probes than simple benzene/propane probes had been employed in the first place.Based on the assumption that only eutomer 2 contributes to the binding affinity, the identified fragment hit exhibits a promising ligand efficiency of 0.40.

2. 1 |
Crystal structure of the NFAT1 RHD-C domain To validate the utility of the RHD-C domain for fragment-based applications, we expressed and purified NFAT1 575-678 from Escherichia F I G U R E 1 (a) Schematic representation of the domain architecture of nuclear factor of activated T cells 1 (NFAT1).(b) NFAT1 in complex with HIV-1 LTR kB element (PDB ID: 1P7H), NFAT1 shown in wheat, bound DNA in cyan.(c) Apo crystal structure of the NFAT1 C-terminal domain of the Rel-homology domain (RHD-C) domain, dimer shown in gray.(d) Superposition of the NFAT1 RHD-C structure (PDB ID 8R07) with the DNA-bound NFAT1 (PDB ID: 1P7H), color coding according to (b).HIV-1, human immunodeficiency virus; LTR, long terminal repeat; PDB, protein data bank.
200 ns cosolvent MD simulations were performed for both monomers in 10 replicates, which lead to a total trajectory length of 4 µs for analysis after pooling.Despite the microsecond trajectory length, we found Tyr-659 rotamer flips on the border of sampling times.We observed closure of the small cavity close to Tyr-659 in 10 out of 10 individual trajectories when starting from the open state.On the contrary, we observed pocket opening only in three out of 10 trajectories when starting from the closed state.After pooling trajectories of the replicates, we found the open state 33% and closed state 3% populated when starting from the open and closed state, respectively.The binding of benzene and propane probe molecules around Tyr-659 was only observed in the generally lower populated open pocket state.This led to the conclusion that this surface region is only poorly ligandable even in the presence of explicit benzene and propane probes.
N labeled RHD-C domain protein of NFAT1 (residues 575-678).The protein was analyzed F I G U R E 2 (a) Crystal structure of the apo nuclear factor of activated T cells 1 (NFAT1) C-terminal domain of the Rel-homology domain (RHD-C) domain in cartoon representation with two different rotamers of Tyr-659 found in the two monomers shown as sticks.(b) One rotamer of Tyr-659 opens a small hydrophobic cavity (protein surface color by hydrophobicity: lipophilic (red)-white-hydrophilic (blue)) which allows binding of small molecule probes in cosolvent simulations.Regions with a grid-free energy of binding for benzene are highlighted in yellow (contoured at −3 kJ/mol).(c) Alternative binding site region identified in cosolvent simulations between loops Ser-605 and Ser-633 (Tyr-659 is shown for reference on the left).Grid-free energies of binding for benzene are contoured in yellow at −8.5 kJ/mol on the whole protein surface indicating only a single binding hot spot at this intensity.

Figure 4a .
Figure 4a.The S-conformer 2 sticks in a hydrophobic pocket formed by the side chains of Phe 603, Val 628, and Leu 638.The pocket is flanked by two loops (residues 630-635 and 603-605).The 4-amino substituent forms two direct hydrogen bonds to the backbone carbonyl oxygen atoms of Lys 630 (2.8 Å) and Ser 633 (2.8 Å).The dihydropyran ring oxygen forms a water-mediated hydrogen bond via water 187 to the backbone carboxyl oxygen (3.2 and 3.0 Å, respectively).The binding site is located distal to the NFAT1 dimerization interface and the RHD-N domain in a likely nonfunctional site (Figure4b).
3 | CONCLUSIONIn this study, we describe the computational and experimental ligandability assessment of the RHD-C domain of the TF NFAT1.Recent advancements in the degrader and glue field targeting TFs F I G U R E 3 (a) Summary of the fragment-based screening campaign and molecular structure of the racemic hit.(b) 15 N HSQC NMR-based K D determination of the racemic mixture (1).Cross peaks monitored in the titration are marked by the numbers and arrows.Shown in blue is the apoprotein and in red are the chemical shift perturbations (CSPs) at the highest concentration used of 1500 µM.The average K D of the three peaks was 729 µM (individual K D s were obtained as follows, #1: 649 µM, #2: 819 µM and #3: 721 µM).(c) 15 N HSQC NMR of nuclear factor of activated T cells 1 (NFAT1) in the presence of 250 µM eutomer 2. (d) 15 N HSQC NMR of NFAT1 in the presence of 250 µM distomer 3. 15 N HSQC, 15 N heteronuclear single quantum coherence.suggest that small molecule tractable protein domains could render TFs more "druggable" than expected by traditional small molecule inhibitor approaches.Our analysis of the newly generated RHD-C domain crystal structure, combined with microsecond timescale cosolvent MD simulations, revealed the presence of two small hydrophobic cavities.One cavity is dependent on an also experimentally observed Tyr-659 rotamer transition, while the other one appears more ligandable by hydrophobic fragments throughout the MD trajectory.Subsequent experimental probing of the protein with fragments yielded a very low hit rate, with only one fragment displaying saturable NMR K d .CSP experiments with separated enantiomers 2 and 3 suggested a very specific binding event, which was confirmed by the first small molecule-bound NFAT structure, reported in this study.The clear separation between eutomer 2 and distomer 3 indicates that a combination of at least three pharmacophores (hydrophobic interactions, direct hydrogen bonding of the primary amine, and watermediated hydrogen bonding) is necessary for binding.The finding explains the unusually low hit rate observed in the FBS campaign.
In summary, our research has led to the discovery of a fragment binder to NFAT1 and we have shed light on its binding mode, thereby demonstrating general ligandability of the transcription regulator NFAT1.In light of the attractiveness of NFAT1 as a potential therapeutic target, our initial demonstration of NFAT1 ligandability might foster drug development approaches to tackle this to date elusive target.As NFAT1 functions both as a proinflammatory and anergy-promoting factor, targeting NFAT1 might be advantageous in the context of autoimmune diseases or graft-versus-host disease as well as in cancer immunotherapy of tumors characterized by poor T-cell infiltration due to anergy.The use of this hit as an anchor fragment for the development of molecular glues or PROTACs would require additional elaboration of the fragment.In our view, fragment elaboration via growing is limited by the small size of the identified pocket but could potentially pave the way to fully harness the potential of the described molecular interactions of the identified fragment in the future.

4 | EXPERIMENTAL 4 . 1 |
Protein purificationThe construct for expression of the NFAT1 RHD-C (residues 575-678, UniProt ID: Q13469) was obtained by gene synthesis (GeneArt, Thermo Fisher) in donor vector (pDONR-221) and transferred by recombinant cloning into pDEST17 vector.The construct included an N-terminal 6xHis-tag followed by a tobacco etch virus (TEV) cleavage site and was used to transform E. colistrain BL21(DE3).Protein was expressed using standard procedures in terrific-broth medium cellF I G U R E 4 (a)Binding mode of eutomer 2 as observed in the crystal structure in complex with nuclear factor of activated T cells 1 (NFAT1) 575-678 .The ligand is shown as a stick model; color-coded by atom type with carbon shown in green.(b) Superposition of the ligand-bound NFAT1 C-terminal domain of the Rel-homology domain (RHD-C) structure (PDB ID: 8R3F, ligand shown as spheres) with the DNA-bound NFAT1 (PDB ID: 1P7H).The ligand is color-coded by atom type with carbon shown in green and the proteins in gray (PDB ID: 8R3F) and wheat (pdb ID: 1P7H), respectively.(c) Overlay of bound fragment from crystallographic screen and grid-free energies for benzene binding (contour level: −8.5 kJ/mol) indicating a clear overlap in terms of binding site as well as chemical nature of bound aromatic fragment.PDB, protein data bank.