High five! Methyl probes at five ring positions of phenylalanine explore the hydrophobic core dynamics of zinc finger miniproteins

The elucidation of internal dynamics in proteins is essential for the understanding of their stability and functionality. Breaking the symmetry of the degenerate rotation of the phenyl side chain provides additional structural information and allows a detailed description of the dynamics. Based on this concept, we propose a combination of synthetic and computational methods, to study the rotational mobility of the Phe ring in a sensitive zinc finger motif. The systematic methyl hopping around the phenylalanine ring yields o-, m-, p-tolyl and xylyl side chains that provide a vast array of additional NOE contacts, allowing the precise determination of the orientation of the aromatic ring. MD simulations and metadynamics complement these findings and facilitate the generation of free energy profiles for each derivative. Previous studies used a wide temperature window in combination with NMR spectroscopy to elucidate the side chain mobility of stable proteins. The zinc finger moiety exhibits a limited thermodynamic stability in a temperature range of only 40 K, making this approach impractical for this compound class. Therefore, we have developed a method that can be applied even to thermolabile systems and facilitates the detailed investigation of protein dynamics.


Introduction
Side chain rotations make protein structures dynamic so they can fulll their functional duties like molecular recognition, catalysis, and signal transmission to name a few examples. Different spectroscopic and computational methods identify mobilities on different timescales with the aim of correlating structural dynamics with functional properties. 1,2 The application of the model free quantication of signal coalescence for the determination of aromatic side chain mobility in BPTI by Wüthrich et al. 3,4 was an early hallmark in the characterization of protein dynamics. While the general idea of studying dynamics by applying temperature gradients is still used today, its application is limited to only the most stable proteins and hampered by the number of aromatic NMR signals. 5 A further development is the use of relaxation dispersion experiments which provided excellent agreement with simulations 6 and allows the study of near degenerate spin systems. [7][8][9][10] Here, we propose a method using phenylalanine derivatives with modi-ed side chains in combination with molecular dynamics simulation, to characterize the hydrophobic core dynamics in miniproteins with limited thermodynamic stability.
Miniproteins are the smallest cooperative functional polypeptide fold. 11 One prominent example is zinc nger peptides that contain a b-turn, antiparallel b-sheet, a-helix, and loop, all tightly packed around a hydrophobic core. 12,13 In the early 90's, investigations focused on illuminating structural features and motions proting from the straightforward synthesis of standalone hairpin folds and the accessibility for theoretical studies. [14][15][16][17] One structural element that is essential for a stable fold is the hydrophobic core, which consists of several aliphatic amino acids, oen accompanied by aromatic side chains. The importance of this structural feature for the stability of zinc nger peptides has been under investigations multiple times throughout the literature using various spectroscopic methods and mutation experiments. [18][19][20][21] It could be shown that especially the central phenylalanine on the position 10 or 13, depending on the size of the zinc binding site, is of great importance for a stable tertiary fold. 1 H NMR spectroscopy identies exceptionally large signal dispersion for the central aromatic amino acid F13 of the hydrophobic core and is clearly separated from the rest with a strongly shielded z-proton, due to edge-to-face interaction with His ( Fig. 1). It was noted that even though the central F13 is packed tightly in the hydrophobic core, an AA 0 BB 0 X spin system resulting from fast conformational averaging was observed for this side chain. 22 Relaxation measurements concluded on restricted rotational motion around c 2 whereas mutagenesis experiments highlighted the intolerance of the hydrophobic core against non-aromatic substitutions. 18,21,23 This highly conserved F13 was substituted by Lachenmann et al. with a cyclohexyl amino acid with the aim to differentiate p-stacking from other interactions within the hydrophobic core. 21 They discovered that the side chain volume is of importance rather than aromaticity. Building on this, further substitutions on other positions have been performed, to obtain more stable zinc nger derivatives and further elucidate the structure and functions. They range from single side mutations to precisely tuned DNA-interactions 24,25 to multiple non-standard residues, which improve stability while retaining the fold. 26 Lombardo et al. were even able to completely swap the a-helix of the zinc nger peptide with an oligourea helix and preserve metal binding properties. 27 Beyond side chain substitutions, other changes were studied in zinc ngers too, although care must be taken in the region of the b-hairpin turn to avoid the loss of the affinity for zinc in its tetrahedral coordination sphere. 26 Cys-X 4 -Cys type zinc ngers, such as the third nger of YY1 or the second nger of Sp1 studied here are known to tolerate signicant changes in the b-turn region while adopting a stable and strong fold. 28,29 We chose the third zinc nger of Yin Yang1, since the system has been well studied and provides an ideal model for miniproteins. 28,30 Here, we introduce strongly b-turn stabilizing Hot]Tap 31 which has been incorporated into several b-turn motifs 32,33 and displays superior conformational restriction of hairpin turns, in comparison to BTD. 34 Aer the substitution of the natural bturn, we observed an increased yield, crude purity, and solubility for the zinc nger derivative, while retaining the fold of the natural zinc ner peptide and the BTD derivative. As expected for a well-folded structure, the inuence of the stronger restriction by Hot]Tap is nearly indistinguishable from BTD which lacks preferred lactam puckering and b-branching. With this stable zinc nger fold at hand, we then stepwise increased the steric demand of the aromatic side chain at position Phe 13 . The combined effects reduce the ipping frequency of the Phe 13 aromatic side chain and characterize the oen underestimated contributions of dynamic modes to the fold stability of the miniprotein. Fig. 1 shows the systematic methyl-hopping on the aromatic ring of Phe 13 with the aim to identify structures with less time-averaged NOEs resulting from the symmetry-broken side chain. This additional structural information about the rotational exibility is then used to analyze the entropic Fig. 1 Tertiary structure of the third zinc finger of Yin Yang 1 protein (3YY1), based on the crystal structure 30 showing the side chains involved in zinc binding and the four phenylalanines in a color code. The solution structure exhibits the same bba fold 15 with fast aromatic flips, causing an AA 0 BB 0 C spin system in the 1 H NMR of all four Phe. The central aromatic residue Phe 13 , shown in blue in the HSQC expansion above right, stands out because of its large chemical shift dispersion. The zinc finger variants (ZFVs) in the inset table differ by their methylation pattern of Phe 13 . The amino acid code identifies the position of the methyl groups from the Greek letters. Hot]Tap stands for the hydroxythreonine annulated at thiaproline according to lit. 31 The arginine residues on positions 20 and 24 were substituted with lysine to obtain higher solubility. Every zinc finger variant is listed in this table with the corresponding mutations and fold stability, qualitatively determined by temperature dependent 1 H NMR spectroscopy (ESI †).
contribution of phenyl groups to the overall fold stability. The classical Cys 2 -His 2 zinc nger fold allows the aromatic residue to rotate freely around its c 2 dihedral. Because of this fast exchange only ambiguous information is available about the orientation of the ring. The tolyl and xylyl amino acids of Fig. 1 deliver a characteristic set of NOE contacts, to precisely determine the orientation (c 1 ) and the rotational mobility (c 2 ) of the aromatic side chain. The additional structural information facilitated the calculations of NMR structures to further visualize the bba motif and aid in the understanding of geometrical requirements. Nevertheless, NMR structure ensembles are still limited in quantifying the internal motions of peptides. To overcome this challenge researchers are using molecular dynamics simulation to further investigate those uctuations. 35 For freely accessible side chains, those motions are on a fast timescale and thus can be easily addressed. Aer the formation of a well-dened hydrophobic core, rotations can slow down immensely and thus are out of reach for standard MD. To address this challenge, enhanced sampling methods like replica exchange MD, 36 steered MD, 37 Gaussian accelerated MD 38 and metadynamics 39 have been developed. Metadynamics is a widely used method for the reduction of dimensions in which one (or a few) specic collective variable (reaction coordinates) is selected and a positive Gaussian potential is employed at a certain timestep. [40][41][42] This bias increases sampling and furthermore allows the creation of a free energy surface along the sampled coordinate. These combined advantages, for us, make metadynamics the method of choice to investigate the rotation of modied phenylalanine residues inside the hydrophobic core.

Results and discussion
We started our investigation by synthesizing the native third zinc nger of YY1, which we denote in the manuscript as 3YY1, using Fmoc solid phase peptide synthesis, as previously reported by Viles et al. 28 Aer purication with semipreparative RP-HPLC and addition of zinc to an aqueous solution, the peptide exhibits the, for a Cys 2 -His 2 zinc nger common, bba fold. The substitution of Glu 7 and Gly 8 with Hot]Tap, a dipeptide b-turn mimic, developed in our group 31 results in the same fold with only small changes in chemical shi values. This mimic shows strong resemblance to the well-established BTD 43 yet is superior in the rigidication of the b-turn, as BTD lacks preferred lactam puckering and b-branching. 31 The hydroxy groups are liberated from the acetonide during acidic cleavage from the resin. 34 Aer successful synthesis we observed higher purity, both crude and puried (ESI †), higher yield and better Fig. 2 The aromatic region of the phenylalanine residues has been magnified and color coded in the NOESY spectra of 3YY1 (A) and ZFV-1 (B). All phenylalanine moieties show two signals for Hd and H3; while the Hz from Phe 2 , Phe 6 and Phe 17 falls into a very narrow chemical shift range. Only Phe 13 splits into 3 separate signals. Besides small changes in chemical shift values for Phe 6 , no major changes in the aromatic region could be observed upon the introduction of Hot]Tap. Below, the molecular representation of ZFV-1 with the Phe moieties color coded to the correspondent NOESY spectra and Hot]Tap is explicitly shown.
solubility. NMR analysis of the apo peptide reveals that the Hot]Tap derivative exhibits a pre-folded species with b-turn indications. 2D-NMR structural analysis aer the addition of zinc reveals the same TOCSY/NOE pattern as observed in the parent peptide, which indicates the bba fold for both peptides which is most obvious from the identity of the aromatic chemical shis of the four phenylalanines well-distributed over the entire zinc nger fold (Fig. 2) of 3YY1. All four show AA'BB'C spin systems caused by fast rotational averaging. Due to signal overlap, only d and 3 protons are clearly distinguishable. Only Phe 13 , which is part of the hydrophobic core, situated between both secondary structure elements, shows three clearly resolved signals even at lower temperatures.
With the aim of obtaining more information about the rotational behavior, we performed molecular dynamics simulations to accurately model an atomistic representation of zinc nger proteins and to supplement the experimental investigations. Researchers in the past few decades have employed MD simulations in various studies to examine various challenges, like folding/unfolding or DNA-binding mechanisms. [44][45][46][47] We simulated each zinc nger peptide for 1.5 ms and compared the exibility of the c 2 -dihedral between the phenylalanine residues.
As depicted in Fig. 3, out of all phenylalanine amino acids, Phe 6 has the most exible aromatic side chain. Even though it shows a face-to-face packing with His, 26 it appears to perform numerous turns around its c 2 -dihedral in the 1.5 ms MD simulation. Hot]Tap rigidies the b-turn, thus restricting the rotational mobility of Phe 6 . This impact can also be observed in 1 H NMR, as the H 26bProR -proton experiences a strong upeld shi of 0.5 ppm, due to a stronger aromatic ring anisotropy. Phe 17 which is situated at the beginning of the a-helix rotates freely. Furthermore, no immediate long-range effect on Phe 13 and Phe 2 is visible as both c 2 -dihedral mobilities remain similar in both zinc nger peptides. In contradiction to the TOCSY/ NOESY-pattern, Phe 13 is only able to perform a single 180 turn during the simulation for 3YY1 but is unable to rotate in ZFV-1. The high energy barrier makes this occurrence statistically unlikely to be sampled during a classical MD simulation in 1.5 ms. This difference in experimental results and simulated outcome is one major challenge of MD simulations since the timespan sampled by standard methods lies in the ns to ms range. In contrast, the timescale in 1 H NMR spectroscopy is typically much longer. Trying to resolve this discrepancy, enhanced sampling techniques have been developed to further improve the timescale visited during a simulation. We chose well-tempered (wt) metadynamics to investigate the c 2 exibility of Phe 13 since it is an established method for the computational study of dynamic systems. [48][49][50][51] The exibility during the wtmetadynamics simulation increases dramatically and multiple turns can be observed for Phe 13 in both peptides. Additionally, since only one local parameter was biased, no severe change in the exibility of the other Phe could be observed and shown in Fig. 3 for Phe 2 , Phe 6 and Phe 17 . One remarkable advantage of wt-metadynamics is the ability to readily produce the free energy surface alongside a biased parameter. We compared the free energy surface between 3YY1 and ZFV-1 and depicted them in Fig. 4. Even though the b-turn is, by no means, in the vicinity of the hydrophobic core, its inuence is considerable. 3YY1 has an energy barrier of 17 kJ mol À1 , while the incorporation of Hot]Tap results in an increase to 31 kJ mol À1 . This large impact can be explained by the change in rigidity in the b-sheet and difference in the turn-type. The Glu-Gly motif favored a type-II turn, while the replacement with Hot]Tap forces the peptide to adopt a type-II 0 turn. This feature has also been observed in the literature for the BTD derivative. 28,29 Besides the rigidifying effect in proximity, Hot]Tap also exerts a longrange interaction by stiffening the b-sheet itself. This reduction in the exibility of the secondary structure raises the energy barrier for the rotation of Phe 13 around the c 2 -dihedral.
Aer establishing the "baseline" for the exibility of Phe 13 inside the hydrophobic core and the inuence of b-turn modi-cation, we wanted to gain further insight into the dynamics of this central amino acid. Wüthrich et al. have established methods for the analysis of slow ring-ips, using elaborate temperature gradients. 3 Since the thermodynamic stability of zinc nger miniproteins is limited to around 330 K (ESI †), this method is not applicable here. We therefore sought to replace the aromatic moiety with synthetic Phe-analogs, to further gain insight into its orientation. In the past few decades, studies have focused on replacing this conserved position with nonaromatic amino acids to assess its importance for structural integrity and thermodynamic stability. 16,18,20,23 Since the library of natural aromatic amino acids is limited and has either been investigated (Tyr) or didn't fold (Trp), we focused on expanding this set of building blocks. As mentioned in the Introduction, one example of an unnatural amino acid was the implementation of cyclohexyl alanine in the zinc nger motif. 21 While this side chain has a similar hydrophobicity, the complex spin system with its many spins impedes a direct identication of the side chain orientation. Therefore, we employed the wellestablished Negishi cross coupling strategy of aryl-iodides in combination with N-Boc-3-iodo-L-alanine methyl ester to quickly generate unnatural building blocks with spin systems analogous to phenylalanine, as depicted in the ESI. † 52, 53 We were able to obtain the modied Phe amino acids of Fig. 1 in moderate to good yields, which were deprotected by hydrolysis followed by Fmoc protection. These seven unnatural building blocks were introduced into the zinc nger sequence by standard Fmoc-based solid-phase peptide synthesis. To evaluate the orientation of the side chain and rotational dynamics, each aromatic position was systematically modied. Even though, Fmz exhibits structural similarities to Tyr (which also has a modication on the z-position) we wanted to include this building block to evaluate the inuence of methylation on each aromatic position. As expected, ZFV-4 was able to adapt a stable bba fold though we observed that a higher pH value was needed to facilitate complete folding. NMR structure analysis showed a structure in close analogy to the tyrosine derivative, with the methyl group facing outwards of the hydrophobic core towards the solvent. This disposition leaves a small cle inside the hydrophobic core, which is compensated for by Leu 19 advancing inside the core. Furthermore, ZFV-4 exhibits the same dynamics as Phe and Tyr, namely an AA 0 BB 0 spin system. For clarity, only the lowest NMR structure for each variant is depicted in Fig. 5, while an ensemble of 10 conformers is represented in the ESI. † Generally, all derivatives were able to adopt a zinc nger type fold, as typical chemical shi values remained the same and TOCSY/ NOESY patterns differentiated only slightly.
Because of the steric requirements of the hydrophobic core, the introduction of a methyl group in either the delta or the epsilon position should favor one specic orientation. Furthermore, breaking the rotational symmetry of the aromatic ring was expected to yield more NOE contacts. This was most successful for ZFV-2, as a dened set of 5 distinct TOCSY/NOE signals (4 in the aromatic region plus one signal in the aliphatic region) could be observed (Fig. 5). Analysis revealed that the methyl group is orientated towards the solvent, outside of the hydrophobic core. Facing inside the hydrophobic core is disfavored since competition with Leu19 would distort the hydrophobic core, resulting in thermodynamic instability. Besides the main NOE signals, additional signals arise at low intensity. This signal set corresponds to the disfavored orientation inside the hydrophobic core, as indicated by few key NOE contacts (ESI †). By evaluating the relative intensity of both conformers which are in fast chemical exchange, we can estimate that the methyl group is orientated inside the core to only $5%, based on the NOE signal volume proportion.
In contrast to ZFV-4, methylation on the epsilon position allows for edge to face interaction with His 22 , as visible in the NMR structure (Fig. 5). Therefore, ZFV-2 shows a remarkable thermodynamic stability, exhibiting a good signal to noise ratio even at 320 K. To further rigidify the rotational exibility of Phe 13 , the Fi3 building block was introduced, which increases the steric bulk at the epsilon position signicantly. This zinc nger derivative (ZFV-5) depicts a near identical fold in comparison to ZFV-2. Furthermore, no second signal set could be observed for the disfavored orientation inside the hydrophobic core. However, a decrease in thermal stability could be detected. Aer evaluating the zeta and epsilon positions, we introduced the Fmd-building block. Despite the structural similarity to Fm3 and Fmz, ZFV-3 displays a different behavior. Though the peptide clearly retains the classical bba fold, a decrease in structural stability, as the sample degrades aer several days and loses its S/N ratio at higher temperature with no reversibility, could be observed. Additionally, aromatic signals, while clearly visible, are broadened in comparison to those of other derivatives at lower temperature, which hints at a deceleration of the rotation around c 2 on the edge of slow chemical exchange. 2D NMR analysis revealed that the conformer ratio has shied to around 70/30. This behavior could also be observed in the NMR structure ensemble. In the 10 lowest energy conformers, several examples of the methyl group facing inside the zinc nger core could be observed. The results indicate that the hydrophobic core opposes certain geometrical restrictions. Both Fmd and Fm3 contain the same steric bulk, yet ZFV-3 shows more rotational exibility, signifying a better t for the methyl group inside the hydrophobic core than in ZFV-2. The increased mobility broadens the aromatic signals. Unfortunately, the limited range of thermal stability makes large temperature gradients inaccessible to gain a detailed description for this zinc nger variant. Following the same approach, the steric demand on the delta position with Fid was increased. Aer a straightforward synthesis, we rst analyzed the apo peptide, to verify the successful incorporation of the building block. This was followed by the addition of zinc in aqueous solution and resulted in a zinc nger like fold. Surprisingly, no aromatic signals for Phe 13 and for residues 14-18 which, in the zinc nger structure, are in proximity to Phe 13 , could be observed. Therefore, structural analysis of this zinc nger derivative was incomplete. One explanation for this behavior may be found in the rotation around c 2 , which is considerably slowed down and thus leads to the broadening of the aromatic signals. Due to increased steric demand, a rotation around c 2 would require the secondary structural features to change conformation to allow passing of the isopropyl residue. If this conformational change is slow enough, broadening would make the signals from residues 14-18 undetectable. Aer the evaluation of single position modied phenylalanine derivatives, we investigated the inuence of multiple modications. This way, competition experiments could further aid in understanding of the rotational dynamics inside the hydrophobic core. As we have demonstrated, modication on the zeta position provides no benet during our investigation of the rotational behavior. Therefore, only double methylation, which would break the symmetry, at the delta and epsilon positions was evaluated. The Fxo building block, which features syn-type methylation, was successfully introduced into the zinc nger moiety, and exhibits a very stable bba fold (ZFV-7). Strong resemblance is present to ZFV-2 and only one rotamer, facing outside of the hydrophobic core, could be identied. In contrast to this, ZFV-8, which contains methyl groups at opposite positions, still adopt the same tertiary fold, but a displacement of the aromatic ring, similar to Tyr and ZFV-4, could be observed. Consequently, the most shielded aromatic signal is now the epsilon proton, because of edge-to-face stacking against the imidazole residue of His 22 . Furthermore, all aromatic signals for Fxp 13 are broadened at 290 K hinting at underlying rotational mobility, yet no second signal set for the other orientation was observed.
Motivated by the experimental results for these synthetic zinc ngers, we performed MD (ESI †) and wt-metadynamics simulations of the stable zinc nger variants. The results are summarized in Fig. 6. As expected, all zinc nger peptides exhibit the same mobility for Phe 6 and Phe 17 , since they are remote to the point of mutation and are not inuenced by long range effects. Phe 2 rotational mobility around c 2 is only slightly hindered in the substituted zinc nger peptides; one reason could be the slight displacement of Leu 19 , which changes the steric environment of Phe 2 . Wt-metadynamics revealed that aer the introduction of a methyl group in the epsilon position, an increase in the energy barrier to 40 kJ mol À1 could be detected. In contrast to expectations, the behavior changes in ZFV-5, which shows a signicantly lower barrier, similar to ZFV-1. Methylation on the d-position demonstrates only a slight increase of the energy barrier to 34 kJ mol À1 but the direction of rotation is strongly inuenced.
Fmd is only able to rotate towards the a-helix because a 70 kJ mol À1 energy barrier prevents its rotation towards the bsheet. As already evident from NMR analysis, double syn-like methylation in ZFV-7 generates a robust fold that bears great resemblance to ZFV-2. This characteristic is also reproduced in the wt-metadynamics simulation, as the progression of Fxo mimics a combination of Fm3 and Fmd with an energy barrier of 40 kJ mol À1 but also preferred rotational orientation. Temperature-dependent broadening of aromatic signals for Fxp could be observed in the NMR spectra, indicating slow chemical exchange. The energy prole obtained during the simulation conrmed this observation, as the double methylation at opposite positions lowers the energy barrier to 25 kJ mol À1 , where again a preferential direction of rotation towards the a-helix could be noted. The direct conversion of these energy barriers in ring ip rates using the Eyring equation 54 overestimates rates, since wt-metadynamics seems to underestimate the energy barrier for the rotation in the zinc nger hydrophobic core. This systematic deviation is avoided by comparing the relative rotation rates of the side chains. Relative to the MD ip rate of 10 6 for 3YY1 we estimate the factors of ZFV-1 1 Â 10 4 , ZFV-2 1 Â 10 2 , ZFV-3 1 Â 10 3 , ZFV-5 5 Â 10 2 , ZFV-7 1 Â 10 2 and ZFV-8 1 Â 10 5 . Upon the introduction of Fm3, 13 the ring ip rate is reduced by a magnitude of 10 4 . The modication of the delta position only reduces the ip rate by 10 3 . The scaled rates highlight the potential that these building blocks have for the characterization of side chain dynamics.

Conclusion and outlook
Aer the rst investigations on zinc nger structures in 1989, 15 it was soon detected that the conserved aromatic residue Fig. 6 Free energy profile of the biased dihedral for synthetic zinc finger variants with a Phe 13 substituted hydrophobic core. Snapshots obtained at different timeframes during the wt-metadynamics simulation of ZFV-3, which shows the stepwise rotation inside the hydrophobic core for Fmd. 13 exhibits exible dynamics. In the following years, the studies have focused on the elucidation of these dynamics by NMR relaxation methods or complete swapping of the aromatic residue in the hydrophobic core. [20][21][22] We extended this search by the systematic variation of the X 13 position in 3YY1, using unnatural phenylalanine amino acids. These building blocks showed different methylation/isopropylation patterns, which resulted in the splitting of the AA 0 BB 0 C spin system and allowed detailed investigation of the rotational dynamics. The systematic variation of the position of the substituents identies zinc nger variants with fast ring ips caused by decreased barriers from ground state elevation over selective line broadening in 1 H NMR from the intermediate exchange of the aromatic ring at position Phe 13 . The approach presented here is applicable to temperature sensitive proteins like zinc ngers. For each of the nine modied zinc nger derivatives, NMR structure ensembles were generated, using two-dimensional NMR spectroscopy, and displayed local and global changes of the zinc nger structure depending on the modication pattern. To overcome the discrepancy between experimental and theoretical results, welltempered metadynamics simulation, biasing the c 2 -dihedral, was employed. The analysis of the dynamics demonstrated that the rotation of X 13 was independent of other residues and a small energy barrier for 3YY1 and ZFV-1 could be observed, allowing for a fast rotation. The synthetic phenylalanine derivatives resulted in different rotational behavior. Epsilon modi-cation locked the ring in one position, while delta substitutions allowed for a rotation inside the hydrophobic core. By expanding the steric bulk to isopropyl, successful incorporation induced mobility, which distorts the secondary features considerably and resulted in the broadening of the signals beyond detection. Simultaneous modication on delta and epsilon positions on the same site again resulted in locked orientation while opposite methylation allows the ring to rotate freely due to a low energy barrier. This behavior leads to selective broadening of Fxp 13 signals in NMR spectra at room temperature and is amplied at lower temperatures. To describe the mobility of the synthetic phenylalanine derivatives in more detail, wt-metadynamics simulations were carried out, the results of which agree with the observations from NMR spectroscopy in many areas. While epsilon methylation slightly increases the rotational barrier, delta methylation generates a preferred rotational direction, as the energy barrier for the rotation towards the b-hairpin is highly restricted. Double methylated derivatives (Fxo and Fxp) exhibit an energy prole similar to that of Fmd and Fm3. Of particular interest is Fxp, since wt-metadynamics show a decrease in the energy barrier, compared to ZFV-1. However, despite the low energy barrier and signal broadening, no additional set of NOE contacts for the second orientation could be detected in the strongly restricted temperature range of ZFV-8.
In summary, we developed methods to use modied aryl amino acids as sensors to probe side chain dynamics on the positions of phenylalanine in temperature sensitive miniproteins. Experimental and computational methods were performed using a sensitive zinc nger fold as a model peptide. By applying systematic methyl-hopping around the aromatic ring, the rotational symmetry of the c 2 -dihedral is broken, thus generating new insightful contacts in NMR spectroscopy. Based on these results, we can conclude that the hydrophobic core of the 3YY1 zinc nger can accept single methylation at every position of the Phe 13 residue. However, stability and rotational mobility strongly correlate with the modied position. The most stable structures were studied by wt-metadynamics simulations, which are in good agreement with the experimental outcomes. The quantication of the restricted mobility of a buried Phe is applicable to other synthetically accessible proteins containing Phe and Tyr, especially when expanded to a molecular library approach. Further research could focus on the symmetric substitution of phenylalanine residues, which would result in an equilateral increase in the rotational barrier. This in turn would facilitate the elucidation of the rotational behavior by temperature gradients for stable miniproteins. We hope that the methodology of combining systematic methyl hopping and wt-metadynamics becomes a useful tool for the study of aromatic side chain dynamics in the hydrophobic core of miniproteins.

Author contributions
AG contributed the design and nancial support of the research project. PH performed synthesis, NMR, and conducted the modeling studies. Both authors participated in the data interpretation, wrote and edited the manuscript.

Conflicts of interest
We declare no conict of interest.