13Cβ‐Valine and 13Cγ‐Leucine Methine Labeling To Probe Protein Ligand Interaction

Precise information regarding the interaction between proteins and ligands at molecular resolution is crucial for effectively guiding the optimization process from initial hits to lead compounds in early stages of drug development. In this study, we introduce a novel aliphatic side chain isotope‐labeling scheme to directly probe interactions between ligands and aliphatic sidechains using NMR techniques. To demonstrate the applicability of this method, we selected a set of Brd4‐BD1 binders and analyzed 1H chemical shift perturbation resulting from CH‐π interaction of Hβ‐Val and Hγ‐Leu as CH donors with corresponding ligand aromatic moieties as π acceptors.


Introduction
NMR spectroscopy has developed into an important source of information for analyzing protein binding pockets and interaction surfaces at atomic resolution, complementing and extending information obtained by X-ray crystallography.Experimental NMR data encompasses a wide range of binding affinities, spanning from strong to weak, with K D values varying from low nanomolar (nM) to millimolar (mM).[3] In contrast to X-ray crystallography, NMR can probe the binding surface of a protein closely resembling physiological conditions. The chemical shift of a protein NMR signal is a sensitive indicator for the chemical environment of the nucleus observed.Consequently, alterations in electron density due to ligand interaction lead to chemical shift perturbation (CSP).CSP studies above a certain protein size are challenged by the limited resolution and sensitivity of NMR spectroscopy, which has been addressed in the past by evolution in NMR experimental techniques and progress in hardware development. [6]23][24][25] The majority of these strategies primarily emphasize methyl labeling due to the advantageous spectroscopic characteristics of the À CH 3 threefold symmetry leading to favorable relaxation properties and maximum signal intensities.In contrast to these commonly used isotope patterns, our study explores the possibilities of the methine 13 CÀ H system located at the branching points of the isopropyl valine and the isobutyl leucine side chains, respectively.
The binding affinity of a ligand towards a protein target arises from various interactions such as Van der Waals forces, electrostatic interactions and hydrogen bonds. [26]Among these interactions, the XHÀ O (X=N, O) hydrogen bond is widely prevalent and plays a critical role in molecular recognition, serving as a main stabilizer of active protein conformations. [27] In addition to lone pair carrying heteroatoms, aromatic π-systems are enabled to take on the role of the hydrogen acceptor.32][33] However, in most hit-to-lead structure optimization programs, the consideration of CH-π bonds is often overlooked, primarily due to the scarcity of experimental methods that can directly quantify these weak interactions.
NMR spectroscopy has been used to detect methyl-π interactions via J-couplings in the past. [34] Specifically, we investigated the geometric parameters between aromatic ligands and labeled tryptophan residues of the corresponding target protein (PI by NMR). [35]In another application, we could assign the CSP values to proS and proR hydrogen -ligand interactions in corresponding methylene isoleucine labelled protein samples. [36]We have now pursued this concept further to the aliphatic residues leucine and valine.This extension is motivated by two key factors: Firstly, minimizing the exposure of nonpolar regions to water is the main thermodynamic driving force for a protein to fold into a conformation with a binding pocket enriched in hydrophobic residues. Secondly, therapeutically active compounds frequently feature aromatic systems.Recent statistical analysis of physicochemical and structural properties reveal that 76 % of drugs approved within the last decade feature one or more heteroaromatic ring structures. [39]

Results and Discussion
A literature-based route to access 13 C-pyruvate 3 was modified and adapted to our purpose (scheme 1). [42]Methylation of the ylide 1 was accomplished using 13 C-iodomethane as an economic source of carbon-13.Ozonolysis yielded the tert-butyl pyruvate 3, which was subsequently converted to the two isomers of the dimethylhydrazone derivative 4 (= major isomer). [20]Subsequent methylation of this hydrazone was described previously, [43] but following this literature reported procedure using lithium diisopropyl amide (LDA) as a base did not result in the desired product in our hands.Instead, we optimized this key step and identified reaction conditions applying sodium bis(hexamethylsilyl)amide (NaHMDS) to abstract the proton in α-position. [36]Reduced reaction times were essential to efficiently perform the conversion of compound 4 to 5 (= major isomer).The remaining steps have been conducted according to protocols we developed in earlier work, but now applied on different isotopologues. [20]We additionally established an alternative route to access precursor 7 starting from [2-13 C] acetone 8 (scheme 1).Condensation with hydantoin in presence of D 2 O yielded compound 9.The target compound 7 was obtained after basic hydrolysis of the hydantoin ring. [44]aOD was applied in this step to avoid methyl protonation and the α-position was subsequently protonated by treatment with 1 M NaOH at room temperature.This sequence resulted in a methyl deuteration grade of > 97 %.This second synthetic approach features a different source of carbon-13 (acetone instead of methyl iodide) and is distinguished by a very effective two-step reaction sequence.[3-13 C, 4,4'-2 H 6 ] Ketoisovaleric acid 7 was added to the 15 NH 4 Cl supplemented minimal medium of a BL21(DE3) E. coli based overexpression system for Brd4-BD1.
We chose the binding domain one of the bromodomaincontaining protein 4 (Brd4-BD1) as our model system and correlated chemical shift perturbations with the presence of Scheme 1. Reagents and conditions: i. 13  CH-π interactions. An in-house collection of proteinligand co-crystal structures provides extensive structural data to evaluate our NMR based approach. [47]Our novel precursor enables the introduction of a 13 CÀ H spin system at the βposition of valine and the γ-position of leucine (Scheme 1).The 1 HÀ 13 CÀ HSQC of the resulting protein sample revealed welldefined signals, which arise from 13 CH methine spin pairs of five valine-and thirteen leucine residues in the Brd4-BD1 sequence (Figure 1).While the valine signals have been assigned using a combination of NMR experiments recorded on a uniformly 13 C labeled protein sample (see experimental part), correlations of methine resonances to the leucine residues are currently not available.In order to explicitly identify the signals of two leucine side chains, which are located in the binding site (Leu92 and Leu94; see Figure 1C), corresponding L92A and L94A mutants were overexpressed and studied by 2D NMR (supporting information).Thereby, two peaks (Figure 1D) were assigned to these two residues.Since our precursor labeling scheme does not feature a chiral carbon center, stereoselective synthesis is not required to further reduce protein spectra complexity, as e. g. in the case of single methyl group labeling.However, in the case of high molecular weight proteins it might be beneficial to separate valine from leucine labeling and reduce signal overlap.In this case, leucine isotope labeling can be inhibited by adding unlabeled α-ketoisocaproate to the overexpression medium as described in earlier work. [22]pplying a fully 13 C labeled version of precursor 7 can have the additional advantage of signal assignment using triple resonance NMR experiments.It is of special value that both synthetic routes depicted in scheme 1 may lead to this all- 13 C precursor 7 if corresponding isotope sources are used.On the one hand, the stable ylide 1 can be synthesized from commercially available [1,2-13 C 2 ] bromoacetate and be transformed to uniformly 13 C α-ketoisovalerate by using 13 CH 3 I as a reagent in the subsequent reaction steps.On the other hand, the all- 13 C isotopologue of 7 is also accessible using [ 13 C 3 ] Figure 1.A) Brd4-BD1 sequence featuring 13 leucines (purple) and 5 valines (green).B) Overlay of 1 H-13 C-HSQC spectra of selectively Leu and Val 13 CH labeled Brd4-BD1 (red) onto the 1 H-13 C-HSQC of 13 C-uniformly labeled Brd4-BD1 ( 13 CH and 13  acetone and [1,2-13 C 2 ] hydantoin, with the latter being synthesized from commercially available [ 13 C 2 ] glycine as described in literature. [21]en different ligands (compounds 10-19, see supporting information for corresponding structures) were added to the selectively Val/Leu labeled protein samples and the resulting CSPs recorded by acquisition of 1 H-13 C-HMQC SOFAST spectra.The protein-ligand affinities had been analyzed in previous studies and the ligands were synthesized following literature procedures. [35,47]The ligands contain strong (nM) and weak (μM) binders and corresponding structural details are given in the supporting information.For each of these compounds, the binding mode to Brd4-BD1 is known and some of them are available in the PDB database (6XV7, 6XUZ and 6XV3).
In the case of high affinity binders (e. g. the triazolopyrazine/5-azabenzimidazole derivative 14; K D = 38 nM), overlay of the 1 H- 13 C-SOFAST-HMQC of the protein in the apo-state and in presence of the ligand displays significant chemical shift perturbation of two leucine and one valine methine signal (Figure 2A).This finding corresponds to the reported binding mode, indicating two leucines (Leu92 and 94) and one valine (Val87) side chains within the binding site (PDB 6XV3).Taking a closer look at the binding mode, the distance between the Leu92 methine proton and the 5-azabenzimidazole ring system in ligand 14 displays a CH-π interaction, which is detected by the large CSP observed (Figure 3 top).In the case of the Val87 methine proton, the structural data again corroborates the large CSP observed.Worth to notice here is the already moderate upfield position of the Val87 methine resonance, which is due to the proximal aromatic side chain of Tyr97 (Figure 3 bottom).An additional upfield shift upon ligand interaction is due to the positioning of the triazolo-pyrazine system of the ligand, which leads to additional shielding of the methine group.
In case of weaker binders (e. g. compound 17; K D 8,9 μM), the absence of the large bicyclic π-system impedes a significant interaction with Leu92 and a CSP is consequently missing in the corresponding 1 H- 13 C-SOFAST-HMQC spectrum (Figure 2B).Spectra of Brd4-BD1 in presence of ligands 10-19 confirm the finding that a Leu92 CSP is only observed in cases where the ligand structure features a rigid bicyclic aromatic system as corresponding CH-π acceptor (supporting information).Hence, our method provides an effective and direct tool to answer the question if certain ligands serve as appropriate acceptors in noncovalent CH-π leucine/valine interaction.
The extent of chemical shift perturbations caused by CH-π interactions is determined by the proximity between the proton donor and the acceptor aryl rings, as well as their spatial orientation.The effect of an aromatic ring current on a proton signal CSP can be described by the point-dipole model introduced by Pople (Figure 4A). [48]The corresponding structural parameters in the Pople equation for Leu92 and Leu94 (r and θ,  see table SI1 in the supporting information) have been extracted from x-ray data.
The correlation of experimental CSP values for Leu92 and Leu94 with predictions made using the Pople equation for different ligands is shown in Figure 4B.Overall, the positive correlation of Δσ derived from structural data and the observed CSP suggests that the observation of methine proton signals has the capability to serve as an additional tool for evaluating binding modes within protein-ligand complexes.Interestingly, the slightly smaller slope of 0.84 for linear regression is comparable to the value (0.92) for aromatic CH-π interactions. [35]ople's approximation describes the secondary magnetic field induced by the electric ring current via a magnetic dipole positioned at the ring center.The implementation of this model offers high simplicity and decreased calculation efforts.1]

Conclusions
We developed a method to introduce 13 C-1 H spin systems into the methine positions of valines and leucines in presence of deuterated methyl groups.A corresponding isotopologue of the metabolic precursor α-ketoisovaleric acid 7 was synthetized in two different approaches using the economic isotope sources 13 CH 3 I/CD 3 I and [2- 13 C] acetone/D 2 O, respectively.The compound was applied in the E.coli-based overexpression of selectively labeled Brd4-BD1 resulting in well-resolved 13 C-HSQC NMR spectra.CSPs were observed upon addition of ten different ligand compounds with known binding modes.In this study, we could show, that the magnitudes of the observed CSPs correlate with the geometrical arrangement of the interacting moieties via Pople's equation and can be exploited as a valuable data source to characterize protein ligand interaction.Methine-based CSP experiments can thus complement methods focusing on methyl resonances, especially in the case of small to medium-sized proteins and serve as an additional guiding principle in ligand optimization processes.

Experimental Section
General synthetic procedures: Unless otherwise stated, all reagents and reactants were purchased from commercial suppliers and used without further purification.All solvents were distilled before use.Iodomethane- 13 C and Iodomethane-d 3 were purchased from Sigma Aldrich®.Oxygen-and moisture sensitive reactions were carried out under an argon atmosphere and yields refer to pure compounds.The reactions were monitored via thin layer chromatography (TLC) on silica gel 60 with fluorescent indicator UV254.Visualization of the compounds was carried out using an UV-lamp (254 nm) and by application of specific reagents: H 3 PMo 12 O 40 (10 % in ethanol) with subsequent heating using a heat-gun.Flash column chromatography was performed on silica gel 60 (0.040-0.063 mm) from Merck.Freeze-drying was done with liquid nitrogen and subsequent lyophilization by inducing high vacuum on an oil pump. 1
[3-13 C; 4,4,4-2 H 3 ] 2-Keto-3-(methyl-2 H 3 )-butanoic acid (7).A solution of 143 mg (0.47 mmol) compound 6 in a mixture of 10 mL diethyl ether (Et 2 O) and 10 mL dichloromethane (DCM) was cooled to 0 °C.Gaseous HCl was purged into this solution by adding concentrated HCl to concentrated sulfuric acid dropwise.The resulting gas was bubbled through the solution for 15 minutes at a moderate pace.After that, the reaction vessel was tightly closed and stirred at room temperature for 1 hour.The solution was then cooled again to 0 °C, and this process was repeated three times.Stirring was continued overnight after the final cycle.The following day, the solvent was removed under reduced pressure, resulting in the formation of 58 mg (99 %) of [3-13 C; 4,4,4-2 H 3 ] 2-keto-3-(methyl-2 H 3 )-butanoic acid 7 as a clear, colorless viscous liquid. 1
[3-13 C; 4,4,4-2 H 3 ] 2-Keto-3-(methyl-2 H 3 )-butanoic acid (7).108 mg of the hydantoin derivative 9 were stirred at 100 °C in 4 mL of aqueous NaOD 20 % (w/w) under an argon atmosphere for 5 hrs.After addition of 10 ml of water, 10 ml of ethyl acetate were added, and the phases separated.The aqueous phase was washed one more time with 10 ml ethyl acetate and the combined organic phases discarded.The aqueous phase was acidified by addition of 1 ml of HCl conc.at 0 °C.The resulting solution was extracted five times with 10 ml ethyl acetate and the combined organic phases carefully evaporated via rotary evaporation > 200 mbar.Then, 4 mL of a 1 M aqueous solution of NaOH were added to the flask.The mixture was stirred for 3 hours at room temperature.After that, 10 mL of water and 10 mL of ethyl acetate were added, and the two layers were separated.The organic layer was discarded, and the remaining aqueous layer was acidified with HCl conc at 0 °C.This mixture was then extracted five times with 20 mL portions of ethyl acetate.After that, the combined organic phases were dried over magnesium sulfate.Evaporation of the solvents in vacuo at > 200 mbar resulted in a pale-yellow oil of α-ketoacid 7 in a yield of 65 mg (72 %).Proton NMR spectroscopy indicates a methyl deuteration grade of ~97 %. 1 H NMR (400 MHz, CDCl 3 , ppm): δ = 3.44 (d, 1 J C,H = 131.1 Hz, 1H; 13 CH), 1.23 (s, 0.21H, residual protonation of the two methyl groups; CH 3 ). 13C NMR (100 MHz, CDCl 3 , ppm): δ = 34.84 ( 13 C).
Expression and purification of Brd4-BD1.Protein overexpression was performed as described previously. [35]Recombinant human Brd4-BD1 was expressed in E. coli BL21(DE3), which contains a Nterminal TEV-cleavable His6-Tag (plasmid was kindly provided by Boehringer Ingelheim GmbH & Co KG).Uniformly 15 N and selective Leu/Val 13 C/ 2 H labeled Brd4-BD1 (wild-type, L92A, L94A) was expressed following the expression protocol for efficient isotopic labeling of recombinant proteins using a fourfold cell concentration in M9 minimal medium, supplemented with 1 g/L 15 NH 4 Cl, 3 g/L Dglucose and 100 mg/L [3-13 C; 4,4,4-2 H 3 ] 2-keto-3-(methyl-2 H 3 )-butanoic acid 7. Uniformly 15 N and 13 C-labeled Brd4-BD1 for triple resonance signal assignment was expressed following the same protocol, M9 minimal medium was supplemented with 1 g/L 15 NH 4 Cl and 3 g/L 13 C 6 -D-glucose.Cells were grown until an OD 600nm of 0.7 and protein expression was induced with 0.4 mM IPTG final concentration.Cells were harvested by centrifugation, lysed by sonication and the lysates were subsequently centrifuged.Proteins were purified from the supernatant by Ni 2 + affinity chromatography.The purified protein was treated with TEV protease and again loaded onto a Ni 2 + column to bind the cleaved His6-Tag and the His6-Tagged TEV protease.The flow-through containing Brd4-BD1 was concentrated.Purity was analyzed by SDS-PAGE.NMR samples of Brd4-BD1 were prepared in 10 mM sodium phosphate buffer containing 0.1 mM protein, 100 mM sodium chloride, 10 % D 2 O and 1 mM DTT.

Figure 3 .
Figure 3. Orientation of ligand 14 to Leu92 (top) and Val87 (bottom) according to PDB entry 6XV3.Leu92 is in proximity to the 5-azabenzimidazole of the ligand, whereas Val87 is oriented close to the triazolo-pyrazine system, as well as the Tyr97 side-chain.
H and 13 C NMR spectroscopic data were recorded on a Bruker AVANCE-DRX 400 MHz spectrometer.NMR solvent signals were calibrated to 7.27 ppm (CDCl 3 ) and 4.79 ppm (D 2 O).Chemical shifts (δ) are given in ppm (s = singlet, d = doublet, dd = doublet of doublets, m = multiplet) and reported relative to the residual solvent peaks.Coupling constants (J) are given in Hertz (Hz).High resolution mass spectrometry experiments were performed using electrospray ionization (ESI, 3 keV, in the positive or negative ion mode) or electron ionization (EI, 70 eV).

[ 3 -
13 C] tert-Butyl pyruvate (3): A constant stream of O 3 from an ozone generator was purged through the solution of the ylid 2 in DCM (resulting from the preceding reaction) at À 78 °C until the

Figure 4 .
Figure 4. Pople equation illustrating the change in the isotropic nuclear shielding constant (Δσ) of a proton depending on the relative orientation of a proton to an aromatic π-system (A).Parameters: Δσ…isotropic nuclear shielding constant [ppm]; n…number of circulating electrons (= 6); e… elementary charge [Franklin] (= 4.803207×10 À 10 ); a…radius of the aromatic ring [cm] (= 1.39×10 À 8 for 6-membered ring and 1.18×10 À 8 for 5-membered ring); m…electron mass [g] (= 9,1094×10 À 28 ); c…speed of light [cm s À 1 ] (= 2.998×10 10 ); Θ…angle between the ring normal through the aromatic center and the proton to ring center vector [rad] (see SI for corresponding values); r…distance from the proton to the ring center [cm] (see SI for corresponding values).(B) Correlation between the changes in chemical shift obtained from the Pople equation and the experimentally observed CSP values for Leu92 and Leu94 residues of Brd4-BD1 in presence of different ligands.