Constrained Peptides with Fine‐Tuned Flexibility Inhibit NF‐Y Transcription Factor Assembly

Abstract Protein complex formation depends on the interplay between preorganization and flexibility of the binding epitopes involved. The design of epitope mimetics typically focuses on stabilizing a particular bioactive conformation, often without considering conformational dynamics, which limits the potential of peptidomimetics against challenging targets such as transcription factors. We developed a peptide‐derived inhibitor of the NF‐Y transcription factor by first constraining the conformation of an epitope through hydrocarbon stapling and then fine‐tuning its flexibility. In the initial set of constrained peptides, a single non‐interacting α‐methyl group was observed to have a detrimental effect on complex stability. Biophysical characterization revealed how this methyl group affects the conformation of the peptide in its bound state. Adaption of the methylation pattern resulted in a peptide that inhibits transcription factor assembly and subsequent recruitment to the target DNA.


Introduction
Thea ssembly of proteins into multimeric complexes is central to many biological processes.The underlying proteinprotein interactions (PPIs) involve am ultitude of individual amino acid contacts and require the involved proteins to adopt ad efined, but partially flexible,t hree-dimensional structure.F or selective and efficient protein assembly,t he interplay between structural preorganization and flexibility is crucial, but it is often not clear how these parameters precisely influence complex stability.T oi nvestigate this interplay,i solated peptide motifs serve as valuable model systems.I nt his respect, a-helices have drawn considerable attention since they represent ah ighly abundant secondary structure element in PPI interfaces. [1] Short and isolated helical interaction motifs predominantly exist as flexible random coils when free in solution. Organization upon complex formation is associated with considerable entropic penalties and often results in low binding affinity.P reorganization of helical binding motifs can minimize these penalties and consequently lead to increased complex stability. [1a] Te rtiary structures in naturally folded proteins are stabilized by non-covalent intramolecular interactions,the hydrophobic effect, and disulfide bridges.T oc ompensate for the lack of such structural constraints in small helices,p reorganization has been artificially achieved through backbone rigidification [2] or macrocyclization strategies,i ncluding the formation of hydrogen-bond surrogates [3] and inter-side-chain crosslinks. [1a, 4] Thel atter approach is often referred to as peptide stapling and can be implemented through avariety of crosslinking strategies,such as lactam formation, [5] 1,3-dipolar cycloaddition, [4b, 6] thiolr eactive ligation, [7] and CÀCb ond formation. [8] So-called hydrocarbon stapled peptides [9] combine two constraints:1 )backbone derivatization through amino acid a-methylation [10] and 2) the crosslinking of two alkene-bearing side chains through ring-closing metathesis. [8a] While the crosslink length has been extensively studied, [9] the precise implications of the methyl group on peptide helicity and target binding are open questions. [11] Theh ydrocarbonstapling approach has been used to stabilize a-helical interaction motifs and has given rise to various bioactive PPI inhibitors of challenging protein targets,s ome of which were the first-reported inhibitors for these targets. [12] Among PPIs,h uman transcription factor complexes represent particularly attractive therapeutic targets since many of them have implications in the onset and progression of certain forms of cancer. [13] As often observed for PPI interfaces,t he identification of potent inhibitors of transcription-factor assembly is complicated by the large size of the involved interfaces and the lack of pronounced binding pockets.E ven though constrained helical interaction motifs generally show at endency to inhibit PPIs, [1a] only very few have been reported to directly target transcription factors with sufficient affinity to enable the inhibition of complex assembly. [12a, 14] Thenuclear transcription factor Y(NF-Y) [15] is at rimeric complex (NF-YA/B/C) that binds to ap articular DNAsequence (CCAATbox, Figure 1A), thereby activating genes involved in regulation of the cell cycle and DNA repair. [16] NF-Y is considered apotential therapeutic target, [17] but its direct and selective inhibition has proven to be challenging. [18] Herein, we report the first structure-based design of astapled peptide that inhibits the assembly of NF-Y. We altered the a-methylation pattern of the involved nonnatural amino acids and were thus able to fine-tune the flexibility of the peptide.T his alteration results in increased affinity towards subunits of the transcription factor, thereby inhibiting its functional assembly as well as subsequent recruitment to the target DNA.

Results and Discussion
The 16-mer NF-YAS equence is Crucial for NF-YB/C Binding NF-Y subunits Band Cform astable heterodimer,which by itself does not possess relevant affinity to DNAt arget sequences (CCAATb ox DNA). Only upon availability and binding of the third subunit (NF-YA) is the transcriptionally active trimeric complex formed and DNAbinding occurs with affinity values in the low nanomolar range. [19] In this study,we aimed to inhibit the interaction between the NF-YB/C heterodimer (light/dark grey) and NF-YA(green) to prevent transcription-factor assembly and DNA( blue) binding (Figure 1A). Thepreviously reported crystal structure of the NF-Ytrimer in complex with DNA(PDB ID:4awl) reveals a29mer NF-YAsequence as the B/C binding motif (PBM, V267-R295). [20] Aiming for the structural characterization of PBM bound to NF-YB/C (B:a a5 1-143;C :a a2 7-120) in the absence of DNA, crystallization conditions were screened to provide crystals diffracting to 2.0 (space group P3 2 21). The resulting crystal structure ( Figure 1B,F igure S2, PDB ID: 6qmp) shows the NF-YB/C dimer in the expected histone-like fold, which superimposes closely with the corresponding domains in the DNA-bound NF-Y trimer (RMSD of 169 C a : 0.85 ,PDB ID:4awl, Figure S3). ForPBM, we observe welldefined electron density for all amino acids except G289 and R295, both located in the C-terminal part of the peptide ( Figure S2). Thec entral part of PBM adopts an a-helical conformation (Y272-A287), which is flanked by extended peptide stretches ( Figure 1B). Except for the C-terminal part (K290-E294), PBM superimposes closely with NF-YAinthe DNA-bound complex (RMSD for V267-E288:0 .52 ,F igure S4).
To identify the crucial interaction motif within PBM, six truncated sequences (1-6, Figure 1C)and PBM were synthesized and fluorescently labeled to determine dissociation constants with the NF-YB/C dimer using af luorescence polarization (FP) assay ( Figure 1C,F igure S5). PBM shows sub-micromolar affinity for NF-YB/C (K d = 0.7 AE 0.2 mm)and removal of seven (1)orten (2)C-terminal amino acids results in only moderate affinity losses (K d (1) = 1.3 AE 0.1 mm, K d (2) = 2.0 AE 0.4 mm). Further removal of three C-terminal amino acids finally reduces binding considerably (K d (3) = 9.5 AE 0.4 mm). N-terminal truncations were investigated using peptide 2 as as tarting point. Removal of the first three amino acids causes only am inor affinity reduction (K d (5) = 2.9 AE 0.1 mm). Further N-terminal truncation, however, abrogates NF-YB/C binding almost completely (K d (6) % 45 mm), which determines the central a-helix including as hort Nterminal segment as the core motif required for NF-YB/C binding.

Design of NF-YA-Derived Hydrocarbon-Stapled Peptides
Having identified the core binding motif of NF-YA, we aimed for as tabilization of the central a-helix to increase its affinity for the B/C dimer using the hydrocarbon-stapling approach. Thet wo required non-natural a-methylated and  Table S2) of PBM (V267-R295, green cartoon representation)incomplex with the NF-YB/Cdimer (grey,surface representation). Selected PBM side chains are shown as ball-and-stick representation.F or an overview of entire structure see Figure S2. C) The sequences of PBM (V267-R295) and truncated peptides 1-6,along with their dissociation constants(K d )asdetermined by FP (errors account for 1s,m easurements performed in triplicate;f or binding curves see Figure S5). olefin-bearing amino acids were introduced during solidphase peptide synthesis and subsequently crosslinked through ring-closing olefin metathesis ( Figure S6). [21] Since hydrocarbon-stapled peptides with i,i + 7a nd i,i + 4c rosslinks ( Figure 2A)h ave been shown to have the highest degree of helicity, [9b] we decided to focus on these two architectures.
Within the NF-YAh elix ( Figure 1B), we identified five solvent-exposed amino acids (Y272, H273, L276, Q280, A283; Figure 2B and Figure S7), which were used for crosslink incorporation. This resulted in four stapled helices,t wo with i,i + 7(macrocycle A and C)and two with i,i + 4(macrocycle B and D)c rosslinks (Figure 2A,B), all of which were introduced in two different lengths of the binding motif (16mer 5 and 19-mer 2,F igure 2B). Theresulting eight peptides were synthesized, and fluorescently labeled versions were used to determine dissociation constants with NF-YB/C using FP ( Figure 2B, Figure S5). Compared to the linear peptides 2 and 5,only the 19-mer peptide harboring macrocycle C (2-C) exhibits increased affinity for NF-YB/C (K d (2-C) = 0.9 AE 0.1 mm).
To compare the binding of unlabeled 2 and 2-C,w e performed isothermal titration calorimetry (ITC;F igure 2C, Table S3, Figure S8/S9), which shows aslightly increased NF-YB/C affinity of stapled peptide 2-C (K d (2-C) = 1.08 AE 0.06 mm)w hen compared to linear peptide 2 (K d (2) = 1.37 AE 0.03 mm). ITC measurements also confirm the expected 1:1 binding stoichiometry for peptide and NF-YB/C.T oi nvestigate the binding mode,w ec o-crystalized 2-C in complex with NF-YB/C,which provided crystals that diffract to 1.8 (space group P2 1 2 1 2 1 ). Ther esulting crystal structure of NF-YB/C bound to 2-C ( Figure 2D,P DB ID:6 qms) shows ap rotein dimer overlaying closely with the one bound to PBM (RMSD = 0.74 ,F igure S10). Except for the C-terminal amino acid L285, we observe well-defined, continuous 2 F o ÀF c electron density for stapled peptide 2-C that also includes the hydrocarbon crosslink ( Figure S11). Superimposition of 2-C and corresponding residues in PBM shows ag ood overlap,t hus confirming an analogous binding mode (RMSD = 0.59 ,F igure S12).

Amino Acid a-Methylation Determines Binding Affinity
When comparing affinities of the two peptide lengths (16mer vs.19-mer) in our panel ( Figure 2B), we observed in four cases the expected trend:Shorter peptides (16-mer: 5, 5-A, 5-B, 5-C)e xhibit lower affinity than their longer counterparts (19-mer: 2, 2-A, 2-B, 2-C). Formacrocycle D though, the 16mer peptide (K d (5-D) = 3.1 AE 0.1 mm)s hows higher affinity than its 19-mer analogue (K d (2-D) = 7.1 AE 0.3 mm). Notably and in contract to the other crosslinks,macrocycle D replaces the most N-terminal amino acid (Y272) in the a-helix of the parent peptide PBM ( Figure 1C). Importantly,i na ll macrocycles,t he two non-natural and crosslinked amino acids (X) harbor an a-methyl group,which influences the local peptide conformation, yet the precise implications for a-helicity and binding affinity are unclear. [11] Fort hat reason, we were interested in whether the a-methylation in macrocycle D causes the loss of affinity upon peptide elongation.   Figure S5). C) RepresentativeITC curves of 2 and 2-C (measurements were performed in triplicate;for full data see Table S3 and Figure S8 Figure S5). Thehighest affinity peptide 2-D N was co-crystalized with NF-YB/C,y ielding crystals diffracting up to 2.5 (space group P2 1 2 1 2 1 ). Theresulting crystal structure (PDB ID:6 qmq) shows the NF-YB/C dimer in the expected fold ( Figure S13). Peptide 2-D N ,w hich includes the hydrocarbon crosslink, is clearly defined by the 2 F o ÀF c electron density ( Figure 3B)a nd establishes protein contacts (Figure 3C)also observed for 2-C and PBM. Theoverlay of 2-D N , 2-C,a nd PBM in their bound state ( Figure 3D)i ndicates an almost identical peptide conformation.
To obtain more insight into the characteristics of complex formation, ITC experiments for 2-D and 2-D N with NF-YB/C were performed ( Figure S16, 17). Theb inding stoichiometry of 2-D N (N = 0.76) and 2-D (N = 1.26) differs from the expected value (N = 1), however, the binding affinity values are in line with FP data and reveal a15-fold higher affinity for peptide 2-D N (K d = 0.37 AE 0.01 mm)t han for 2-D (K d = 5.7 AE 0.5 mm). Taken together this provides thermodynamic binding data for at otal of four peptides ( Figure 3E and Table S3). Binding of unmodified peptide 2 is associated with ah igh entropic penalty (ÀTDS = 13 kcal mol À1 ), which is compensated by ac onsiderable binding enthalpy (DH = À21.0 kcal mol À1 ). Compared to peptide 2,both peptides with afully amethylated macrocycle (2-C and 2-D)e xhibit ar educed entropic penalty while the contributions by binding enthalpy are decreasing. Notably, 2-D N binding provides at hermodynamic profile similar to linear peptide 2,a nd it shows the highest entropic penalty (ÀTDS = 16.1 kcal mol À1 )a sw ell as the largest binding enthalpy (DH = À24.8 kcal mol À1 )i no ur panel. To assess whether differences in binding originate from preorganization of the unbound peptides,wedetermined the a-helicity of 2, 2-C, 2-D,a nd 2-D N using circular dichroism (CD) spectroscopy ( Figure 3F and Table S4). Linear peptide 2 shows avery small helical content (13 %, green), while i,i + 7 stapled peptide 2-C exhibits high a-helicity (81 %, blue). For the two i,i + 4stapled peptides 2-D (grey) and 2-D N (orange), we observe very similar CD spectra indicating moderate helicity (47 %, Figure 3F). This suggests that the altered amethylation pattern in these two peptides does not change their overall folding propensity.

a-Methylation Affects the Conformation of the Bound Peptide
It is surprising that peptide 2-D N binds NF-YB/C with a15-fold higher affinity than 2-D although the only difference is the absence of asingle methyl group that is not involved in direct contacts with the protein ( Figure 3C). This is partic-  Table S3 and Figure S8, S9, S16, S17). F) CD spectra in buffer (pH 7.4, c(peptide) = 75 mm)a nd calculated helical content (for complete secondary-structure distribution, see Table S4).
ularly interesting considering their similar CD spectra. For this reason, we aimed for am ore thorough investigation of their unbound states using 1 HNMR spectroscopy in aqueous solution. Comparing Ha chemical shifts with random coil references again provides similar trends for the two peptides: Amino acids H273 to K277 show ac lear deviation from random coil, thus indicating the presence of ad efined secondary structure in the center of the free peptide,w hile the remaining N-terminal (V267-Q271) and C-terminal amino acids (R278-L285) appear to be relatively flexible ( Figure S18). Overall, this is in line with their behavior in the CD measurements (47 %helicity for 2-D and 2-D N ).
We hypothesized that the differences in NF-YB/C-binding originate from peptide characteristics in the bound state. To obtain more details on the bound conformation of 2-D and 2-D N ,t ransferred-NOESY (tr-NOESY) experiments were performed. tr-NOESY provides structural information about aligand in its bound state while analyzing resonances of the free ligand. [22] NMR experiments were performed with an approximately 50-fold excess of peptide over protein at ac oncentration that facilitates rapid exchange between free and bound peptides.NOEs indicative of the free state develop slowly.T hus,N OEs observed in the transferred NOE spectrum at short mixing times are indicative of the bound conformation of the peptide.C ompared to their unbound state,b oth NF-YB/C-bound peptides show af ew additional NOE correlation peaks and ac onsiderable increase in the relative intensity of an umber of NOE peaks,w hich is indicative for folded structures (Table S5). Using the tr-NOEs of the bound peptides as input data for restrained simulated annealing calculations,e nsembles of conformers of NF-YB/ C-bound 2-D and 2-D N were obtained (Figures 4B and C). When superimposed with its crystal structure (orange) in complex with NF-YB/C,t he NMR-based conformers of bound 2-D N (light orange) show av ery good overlay (Figure 4B). Notably and in line with the TOCSY experiments, the N-terminus of 2-D (grey,F igure 4C)c onsiderably deviates from the bound form 2-D N (orange) while the rest of the peptide overlays.T aken together,the results of the NMR and ITC experiments clearly show that the a-methyl group at position 272 has astrong effect on peptide binding.This effect appears to be mainly caused by interference with the bound conformation of the peptide,s ince CD and NMR indicate very similar conformations for 2-D and 2-D N in their unbound state.P resumably,t he bioactive conformation of the peptide observed in the crystal structures of PBM, 2-C,a nd 2-D N is perturbed by an a-methyl group at position 272.

Inhibition of DNA Binding through Disruption of NF-Y Trimer Formation
Having obtained high-affinity peptide 2-D N ,w et ested its ability to interfere with the formation of the trimeric NF-YA/ B/C complex. Fort hat purpose,c ompetition pull-down experiments were designed in which ab iotinylated NF-YA fragment (biotin-PBM) was used to bind and immobilize the NF-YB/C dimer.T he pull-down of NF-YB/C was detected through SDS polyacrylamide gel electrophoresis (PAGE; Figure 5A). Applying increasing concentrations of 2-D N (c = 16-400 mm)results in aconcentration-dependent inhibition of complex formation between biotin-PBM and NF-YB/C (c = 80 mm). As expected, the low-affinity peptide 2-D exhibits only very weak inhibitory activity.
Given the ability of 2-D N to disrupt the NF-Y trimer, we were interested in whether DNAbinding of the NF-Y trimer is also affected by this PPI inhibitor ( Figure 1A). To monitor the binding state of an NF-Y target DNAc ontaining the CCAAT-box, af luorescence polarization (FP)-based assay was designed:D ouble-stranded target DNAh arboring atetramethylrhodamine (TAMRA) label was incubated with the NF-YB/C dimer and an NF-YAf ragment (aa 262-332), comprising the NF-YB/C binding site and the DNAb inding motif.F Pm easurements confirm the requirement of intact NF-Y timer for DNAbinding (88 %bound DNA, Figure 5B).
In the presence of peptide 2-D N (c = 10 mm), we observe ac onsiderable loss in the fraction of bound DNA( 32 % bound DNA). In contrast when using peptide 2-D,n o significant change in the DNAb inding state is observed (87 %b ound DNA). Finally,t hese findings were confirmed utilizing FP-based titration experiments involving preformed NF-Y trimer,which was titrated against the TAMRA-labeled DNA(K d = 2.2 AE 0.4 nm, Figure 5C). In the presence of 2-D N , the affinity between the NF-Y trimer and the DNAdeceases 3.5-fold (K d-app = 7.7 AE 0.9 nm), while peptide 2-D does not affect DNAb inding by NF-Y (K d-app = 2.4 AE 0.3 nm).

Conclusion
In this study,w es how that peptide binding motifs can respond sensitively to slight changes in their structural preorganization. Fine-tuning of their conformational properties can thus enable the design of inhibitors of challenging protein targets such as transcription factors.I nitially,w e identified a14-mer a-helix with ashort extended N-terminal stretch as the protein fragment crucial for NF-YB/C binding. To stabilize the a-helix, aset of modified hydrocarbon-stapled peptides was designed, thereby introducing an inter-residue crosslink and an a-methyl group at each of the two crosslinked amino acids.This set of modified peptides did not show ameaningful increase in binding affinity,thus indicating that solely focusing on the preorganization of the unbound peptide is insufficient. We observed an unusual binding behavior for one derivative (2-D), where peptide shortening resulted in increased binding affinity.K nowing that amino acid amethylation can restrict the conformational freedom of the peptide backbone,w ei nvestigated how variation of the amethyl pattern influences binding affinity of 2-D.O nly at position 272 did substitution of the methyl group by hydrogen affect binding,r esulting in ap eptide (2-D N )w ith 15-fold higher affinity.
It is surprising that such as mall variation in ap eptide region without direct protein contact results in aconsiderable affinity increase,i np articular in relation to the size of the ligand (MW(2-D N ) = 2351 gmol À1 ). To identify the cause of this difference,weinvestigated in detail both the free and the bound state of each peptide (2-D and 2-D N ). CD and NMR experiments indicated very similar conformational behavior in solution suggesting therefore differences in the bound state. Forthe higher-affinity peptide (2-D N ), we obtained X-ray and NMR structures in the NF-YB/C-bound form clearly showing contacts of the a-helix and the extended N-terminal stretch with the protein. For 2-D on the other hand, the NMR structure indicates that the additional a-methyl group at position 272 induces an elongation of the a-helical conformation towards the N-terminus even when bound to the protein, resulting in loss of direct protein contacts.This observation is confirmed by ITC measurements,w hich show reduced binding enthalpy for 2-D compared with 2-D N .T herefore,w e reason that specific release of conformational constraint by removal of the a-methyl group allows adaption of the correct binding mode.T aken together,o ur study indicates that the stabilization of binding motifs composed of multiple secon-  Figure S20). B) Fraction of bound DNA as determined by FP using TAMRA-labeled DNA (c = 5nm)w ith NF-YA(262-332) (c = 25 nm)a nd NF-YB/C (c = 4nm)i nthe absence and presence of peptide (c = 10 mm,triplicate measurements, errors account for 1s). ns: p > 0.05, ** p < 0.01, *** p < 0.001. C) FP titration using TAMRAlabeled DNA (c = 5nm)and varying concentrations of NF-Y trimer (c = 9.7 10 À11À 3.7 10 À7 m)i nthe absence or presence of peptide (c = 10 mm). dary-structure elements requires sections with relatively high conformational freedom to facilitate tertiary-structure-like folds.G iven these intrinsic flexibility features and the complexity of the involved interaction areas,i ti si mportant to note that afully rational affinity maturation of constrained binding epitopes is challenging.C onsequently,t he combination of structure-based design with screening approaches applying focused peptide libraries [23] could represent an appealing strategy.