Inhibition of Aurora‐A/N‐Myc Protein–Protein Interaction Using Peptidomimetics: Understanding the Role of Peptide Cyclization

Abstract Using N‐Myc61‐89 as a starting template we showcase the systematic use of truncation and maleimide constraining to develop peptidomimetic inhibitors of the N‐Myc/Aurora‐A protein–protein interaction (PPI); a potential anticancer drug discovery target. The most promising of these – N‐Myc73‐94‐N85C/G89C‐mal – is shown to favour a more Aurora‐A compliant binding ensemble in comparison to the linear wild‐type sequence as observed through fluorescence anisotropy competition assays, circular dichroism (CD) and nuclear magnetic resonance (NMR) experiments. Further in silico investigation of this peptide in its Aurora‐A bound state, by molecular dynamics (MD) simulations, imply (i) the bound conformation is more stable as a consequence of the constraint, which likely suppresses dissociation and (ii) the constraint may make further stabilizing interactions with the Aurora‐A surface. Taken together this work unveils the first orthosteric N‐Myc/Aurora‐A inhibitor and provides useful insights on the biophysical properties and thus design of constrained peptides, an attractive therapeutic modality.


Introduction
The Aurora kinases play essential roles in mitosis and have received attention as targets for drug-development. [1]Aurora-A represents a promising target for development of anticancer therapeutics; however whilst potent and selective active site inhibitors have been identified, these have yet to be approved for therapeutic use.Such compounds might be expected to have a narrow therapeutic window because Aurora-A has many roles, including a critical function in bipolar mitotic spindle assembly. [1,2]Aurora-A is an incomplete kinase; its localization and activation are regulated through protein-protein interactions (PPIs).Targeting these PPI interfaces may represent an alternative approach to ATP-competitive inhibitors.One candidate PPI is the N-Myc/Aurora-A interaction. [3]MYCN (the gene that encodes for the N-Myc protein) was discovered through its association with neuroblastoma, a cancer of the nervous system that affects children. [4,5]Elevated levels of N-Myc have also been observed in several other cancers. [6]In 2016, Richards et al. published an X-ray co-crystal structure of Aurora-A in complex with an intrinsically disordered region of N-Myc (residues 28-89); [7] although residues 28-60 were not observed in the structure, residues 61-89 of N-Myc were found to undergo a disorder to order transition to interact in an α-helical conformation with the cleft between the Aurora-A N-and C-lobes formed by the αB-/αC-helices, the A-loop, and the αG-helix (Figure 1a).The formation of this complex has been proposed to explain why Aurora-A stabilizes N-Myc against degradation, amplifying the effects of N-Myc overexpression.
Certain ATP-competitive Aurora-A inhibitors have been shown to perturb the N-Myc/Aurora-A interaction due to inhibitor-induced conformational changes in Aurora-A that disrupt the PPI interface; competition co-precipitation experiments found that MLN8054, but not CCT137690, disrupted the N-Myc/Aurora-A interaction. [7]Similarly, Weiss and co-workers developed a series of conformation-disrupting Aurora-A inhibitors based on diaminopyrimidine and pyrazolopyrimidine scaffolds, such as CD532. [8]Although CD532 was shown to be a potent N-Myc/Aurora-A inhibitor that reduced N-Myc protein levels in xenographs, concomitant inhibition of kinase activity might still perturb other cellular essential functions. [9]An alternative approach has been to use Aurora-A degraders/ PROTACS: [10] The first-in-class Aurora-A/N-Myc degrader, HLB-0532259 was developed from a novel Aurora-A-binding ligand that engages the N-Myc/Aurora-A complex.HLB-0532259 was shown to promote degradation of Aurora-A and N-Myc with superior potency and excellent selectivity in comparison to established allosteric Aurora-A inhibitors. [11]Despite these advances, it would be desirable to have orthosteric inhibitors of N-Myc/Aurora-A interaction available as tools to dissect out the mechanistic role of this interaction and to serve as alternative leads for drug discovery.
The larger PPI surface and the N-Myc helical epitope make a peptidomimetic approach attractive.Peptides have emerged as promising alternatives to small molecules given the advantages they present in terms of potent and selective target affinity; however, they suffer from poor (proteolytic) stability and suboptimal cell permeability, motivating the development of constrained peptides. [12]Here we describe the first competitive inhibitor of the N-Myc/Aurora-A interaction.Using a peptidomimetic approach we show that a truncated and constrained peptide has enhanced inhibitory potency in biophysical experiments when compared to its unconstrained analogue.This represents a starting point for the development of reagents that could be used to explore inhibition of N-Myc/Aurora-A interaction for cancer chemotherapy.Moreover, using fluorescence anisotropy, circular dichroism, NMR structural analyses of the unbound peptides and molecular dynamics simulations for the Aurora-A bound peptides, we show that introduction of the constraint confers enhanced potency through a combination of subtle effects.Whilst the constraint does not pre-organize the peptide in an α-helical conformation, it likely: (i) restricts the accessible conformational landscape of the peptide, raising it in energy and pre-disposing the unbound peptide to Aurora-A recognition; (ii) stabilizes the bound conformation to suppress dissociation; (iii) introduces stabilizing contacts between constraint and protein.

N-Myc 61-89 peptide truncation and sequence variation
Fluorescence anisotropy titrations previously established that N-Myc 61-89 binds to Aurora-A with low μM affinity. [7]Using either Trp > Ala or Glu > Lys variants, three residues in N-Myc 61-89 were found to have a strong effect on binding affinity (PDB: 5G1X; Figure 1a): tryptophan residues Trp77 N-Myc , Trp88 N-Myc and Glu73 N-Myc , which presumably participates in a salt bridge with Aurora-A via Lys143 AurA .Based on prior studies Leu61 N-Myc was considered to make an energetic contribution to binding, however, despite a number of non-covalent contacts (e. g. π-π stacking between Phe67 N-Myc and His176 AurA , hydrogen bonds between Ser64 N-Myc and His280 AurA , Ser64 N-Myc and Arg286 AurA , Arg65 N-Myc and Arg180 AurA , Glu73 N-Myc and Ala172 AurA ) we considered this region ripe for removal.Thus, we started these studies by probing the effects on binding of removing residues from the 61-75 region (Table 1, Figure S1).A short N-Myc variant spanning the helical region (N-Myc 73-89 ) was prepared and its IC 50 value determined through competition against a fluorescein labelled N-Myc 61-89 sequence (Table 1).Despite maintaining all three confirmed hot-spot residues (Glu73 N-Myc , Trp77 N-Myc, and Trp88 N-Myc ), removal of residues 61-72 decreased the inhibitory potency of N-Myc (N-Myc 73-89 IC 50 = 179 � 19 μM verses N-Myc 61- 89 IC 50 = 42 � 4 μM).A slightly shorter variant (N-Myc 76-89 ) also exhibited similarly diminished potency (IC 50 = 215 � 11 μM).To further investigate the relevance of the Glu73 N-Myc /Lys143 AurA salt bridge for N-Myc/Aurora-A binding, additional variants with Glu73 > phosphoserine (pSer) and Glu73 > Ser substitutions were assessed (N-Myc 73-89-E73pS and N-Myc 73-89-E73S ; Table 1).The effects of both Glu73 N-Myc substitutions on potency were subtle, suggesting that this side chain does not make a major contribution to binding.We then assessed the effects of sequence elongation at the C-terminus, by preparing N-Myc 73-94 and N-Myc 73-97 , with C-terminal elongations of five and eight residues, respectively (Table 1).Overall, the results indicate that the five residues beyond Gly89 N-Myc contribute little to binding, as only further elongation of the sequence increased potency (N-Myc 73-97 , IC 50 = 79 � 4 μM).

Analyses of constrained peptides by fluorescence anisotropy competition and circular dichroism
The effects of introducing a chemical constraint within the helical region of N-Myc 61-89 were next explored.To identify suitable positions for introduction of a constraint, we adopted a systematic approach. [13,14]In this approach, cysteine pairs were introduced in the helical region of N-Myc 61-89 at all non-hot-spot residues at i and i + 4 spacings, to give a series of 7 variants (see ESI, Table S1).Each variant was then constrained using dibromomaleimide [15,16] to afford the maleimide (mal) constrained variants (Figure 1b), or allowed to oxidize to yield the disulfide bridged variant (ox), or reduced to its free-thiol form (red).The sequences of the resultant 21 variants along with their measured IC 50 , MRE 222 values and % helicity estimates (determined using circular dichroism spectroscopy) are given in the supporting information (Table S1 and Figures S2-S4).In all cases, the maleimide constraint was found to have a limited or negative impact on the inhibitory potency of the peptide.However, we found that some positions were more tolerant to modification than others, with constrained peptides N-Myc 61-89- L82C/E86C-mal , N-Myc 61-89-E80C/E84C-mal , and N-Myc 61-89-T79C/L83C-mal showing only minor loss in potency when compared to the native peptide.
Although we found that extending of the N-Myc helical region by an additional five residues at the C-terminus has a limited effect in terms of potency (N-Myc 73-94 ; IC 50 = 199 � 13 μM, Table 1), this sequence offered additional positions for the introduction of a constraint in comparison to N-Myc 61-89 (note: N-Myc 73-97 had higher potency, however our motivation was to identify the shortest sequence into which a constraint could be productively incorporated).More specifically, constrained variants involving the replacement of the i and i + 4 pairs: Asn85 N-Myc /Gly89 N-Myc , and Glu86 N-Myc /Ser90 N-Myc , could be explored.These pairs of residues were replaced with cysteines, and the peptides generated were constrained (mal), oxidized (ox), or reduced (red).The sequences and key CD spectral data for these 6 new variants are shown in Table 2. Using competition FA, both new maleimide constrained variants showed improved IC 50 potencies compared to the unmodified peptide (N-Myc 73-94 IC 50 = 199 � 13 μM; Table 2, Figure S5).Indeed, maleimide constrained N-Myc 73-94-N85C/G89C-mal (IC 50 = 49 � 5 μM) was found to exhibit 4-fold improvement (ΔΔG ~3.5 kJ mol À 1 relative to N-Myc 73-94 ) in potency to disrupt N-Myc/Aurora-A compared to the WT peptide.The inhibitory potency exhibited by the constrained N-Myc 73-94-N85C/G89C-mal variant was found to be in the same range as that observed for the longer WT sequence (N-Myc 61-89 , IC 50 = 42 � 4 μM; Table 1), despite lacking the N-terminal turn region and being ~25 % shorter.The second variant, N-Myc 73-94-E86C/S90C-mal (IC 50 = 111 � 18 μM) was also found to be 2fold more potent than the control peptide.For the most potent

Peptide
Sequence [a] IC 50 % Helicity [d] N sequence (N-Myc 73-94-N85C/G89C-mal ), a number of features should be noted: (i) Gly89 N-Myc is replaced with cysteine to introduce the mal constraint and glycine is known to have low helix propensity; [17,18] Ser90 N-Myc has the potential to act as a helixcapping residue [19] before Pro91 N-Myc which likely interrupts the helical structure. [17,20]Finally, based on the Aurora-A/N-Myc crystal structure, a mal constraint introduced at Asn85 N-Myc and Gly89 N-Myc is unlikely to be completely solvent-exposed and may form direct contacts with the surface of Aurora-A surface.
Overall, these results highlight a crucial role in selecting an appropriate sequence length to allow identification of optimal sites for constraining peptide ligands.Next, we assessed if the effects observed on peptide potency upon constraining could be correlated with changes in peptide conformation (Figure 2).The control N-Myc 73-94 and N-Myc 73-94-E86C/S90C variants showed similar CD spectra, characterized by the presence of a minima in mean residue ellipticity (MRE) at a wavelength of ~200 nm, slightly shifted to ~205 nm for N-Myc 73-94-E86C/S90C-ox (Figure S5 and S6).Qualitatively, the variants exhibited only limited differences in helicity, with all adopting predominantly random coil conformations in solution (estimated % helicity ~17 to 21; Table 2).In contrast, N-Myc 73-94-N85C/G89C-mal exhibited a different CD spectrum, indicating that the mal constraint in this position destabilizes the random coil conformation of the peptide and biases it in favor of another defined conformation (Figure 2c).Both the shift in minima to higher wavelength (from ~200 nm to ~205 nm) and the increase in positive signal at 193 nm are indicative of increased α-helix propensity (% helicity = 21 � 1; Table 2).However, the emergence of a negative signal at 230 nm was unexpected, as it has been proposed to be more characteristic of "turn" like structures (although a positive signal at ~215 nm, also characteristic of turn structures is absent). [21]The signal at 230 nm may arise as a consequence of the interaction between the chromophores of the mal group and the peptide backbone.In this regard, abnormal CD spectra have been associated with the effects of other conformational constraints. [22]Overall, the data indicates that the conformational ensemble has reduced random coil character.[e] 5 mol equivalents of TCEP were added to the samples to ensure that no oxidized species were present.2.
Analyses of the potency vs. helicity data for N-Myc 73-94 based peptides (Figure 2d, Table 2) revealed that the mal constrained analogue for the N-Myc 73-94-E86C/S90C series is the most potent, despite little difference in helicity relative to the other variants.For N-Myc 73-94-N85C/G89C variants, a clearer correlation is observed with both the free-thiol (red) and disulfide bridged (ox) analogues displaying similar potency and helicity, and the mal variant exhibiting an increase in both potency and helicity (Figure 2c).Overall, the full dataset for all peptides studied indicates a complex relationship between constraint position, peptide conformation and inhibitory potency (Tables 2, S1 and Figures 3d, S7).

Comparison of N-Myc 73-94-N85C/G89C-mal and N-Myc 73-94 by NMR analyses
Given the higher potency and unique CD spectrum observed for N-Myc 73-94-N85C/G89C-mal when compared to the linear WT sequence, we further investigated the solution structure of the unbound peptide by NMR spectroscopy.Samples of N-Myc 73-94 and N-Myc 73-94-N85C/G89C-mal were prepared, and their spectra analyzed at 5 °C.The NMR spectra for N-Myc 73-94 at this temperature showed well resolved 1 H resonances (Figure 3a), whilst constrained N-Myc 73-94-N85C/G89C-mal exhibited broadened resonances of lower intensity (Figures S8-18 for full spectra at 5 °C).Although TOCSY spectra for both peptides had some similarity (Figure S18) many N-Myc 73-94-N85C/G89C-mal 1 H signals were not observable (Cys res 85 and 89) precluding full assignment (see Figures S16-18).We hypothesize such behavior arises as a consequence of the maleimide constraint restricting the rate of interconversion between different conformers such that they match the NMR timescale, broadening the resonances to the point of coalescence.Such behavior has previously been observed in molten globules and partially folded proteins. [23,24]o confirm this hypothesis, we carried out 1 H variable temperature (VT) NMR experiments on both samples (Figure 3a-b, Figures S19-22).Upon increasing the temperature, we observed changes in the unconstrained peptide (N-Myc 73-94 ); an increase in temperature of only 5 °C was sufficient to induce broadening of the 1 H resonances (Figures S19-20).Different behavior was found for the constrained peptide (N-Myc 73-94-N85C/G89C-mal ): upon gradually increasing the temperature from 5 to 25 °C, we observed a clear shift to higher signal intensities accompanied by resonance narrowing and improved resolution, particularly in the amide NH and Hα regions (Figure S21-22).Reducing the temperature to 0 or À 3 °C restored the exchange/broadening effect, supporting the notion that slow exchange between conformers occurs.
For N-Myc 73-94-N85C/G89C-mal at 25 °C, as for the wild type peptide at a 5 °C, the data suggest there may be secondary structure propensity despite the increase in temperature.When compared, both peptides showed a similar trend in the relative magnitude of Cα secondary shifts at each residue (Figure 3d, Table S3).The most notable divergence between the peptides is observed in the constrained N-Myc region corresponding to residues 85-89, where relatively high Cα secondary shifts were observed for Glu86 N-Myc /Leu87 N-Myc .This indicates the introduction of the chemical constraint may exert a helix stabilizing effect in a region where the WT peptide shows low helical propensity (i.e., Δδ~δ RC ).Both maleimide derivatized residues, Cys85 N-Myc and Cys85 N-Myc , exhibited two out-of-trend negative Δδ Cα values, suggesting that their secondary chemical shifts are either highly affected by the presence of the constraint, or misrepresented due to the lack of an appropriate δ RC reference for the modified amino-acid. 1 H secondary shifts were also calculated for both peptides (Figures S30-31); regions of continuous up-field (i.e. negative) shifts indicate helicity. [25,27]In agreement with Cα secondary shifts, a continuous series of upfield Δδ Hα secondary shifts were observed between Glu73 N-Myc and Gly89 N-Myc for the linear control peptide at 5 °C, of relatively small magnitude (Δδ Hα ~0.2), indicating weak helix propensity.The calculated Δδ Hα values were indicative of two regions: one at the N-terminus of the peptide displaying higher Δδ Hα values around Val78 N-Myc , and a second towards the C-terminus, where consistently lower Δδ Hα are observed.For the constrained peptide at 25 °C we found a similar general trend, with Δδ Hα values in the same range but slightly lower in magnitude.

Molecular dynamics (MD) analyses on N-Myc 73-94-N85C/G89C-mal and N-Myc 73-94
Finally, to investigate how the constraint might affect the secondary structure of the peptide in the presence of Aurora-A, we employed MD analysis (Figure 4, Figures S32-S37). [14]We based our analyses on the reported X-ray structure of the N-Myc/Aurora-A complex (PDB : 5G1X), [7] where the sequence of N-Myc was extended at the C-terminus to include either the natural residues 89-94 or the maleimide constrained G85C/ N89C fragment.Results from this analysis were broadly consistent with the structural insights inferred using both CD spectra and NMR techniques, with N-Myc 73-94 showing two well defined regions in its minimum energy structure: a helical segment spanning the N-terminus -residues 73-86 -and a small transient turn region -residues 88-94 -that seems to wrap around the Aurora-A αG-helix (Figure 4a, Figure S30).MD simulations for N-Myc 73-94-N85C/G89C-mal revealed a similar and consistent energy minimum Aurora-A bound structure to that observed for the linear peptide, indicating that both peptides could access similar conformations in their Aurora-A bound states (Figure 4a and Figures S32-33).We also observed that the maleimide group sits orientated directly towards Aurora-A in proximity to Gln335 AurA , which seems to be conformationally trapped between the maleimide and residues Trp88 N-Myc and Glu84 N-Myc in N-Myc.Analyses of the time averaged structure of both peptides bound to Aurora-A (Figure 4b-c) revealed that for N-Myc 73-94 fraying of the helix occurs towards the Cterminus, whilst in comparison, the average structure observed for constrained N-Myc 73-94-N85C/G89C-mal remains closer to the energy minimum state, with the maleimide constraint potentially interacting via a π-amide interaction with Gln335 AurA and preventing the C-terminus from unfolding and undocking from the protein surface.In addition, results from these models also suggest that such an arrangement, where the C-terminus is kept in proximity to Aurora-A due to the maleimide constraint, enables a transient/dynamic close electrostatic interaction between Glu94 N-Myc and Arg343 AurA , which does not take place in the WT peptide due to helix unwinding.

Conclusions
In summary, here we show for the first time the systematic application of maleimide-constraining to identify peptidomimetic leads for inhibition of the N-Myc/Aurora-A interaction.Using N-Myc 61-89 as a starting point and through rational evaluation of the effects of sequence truncation, elongation, and maleimide constraining we identified N-Myc 73-94-N85C/G89C-mal as a minimal constrained peptide with enhanced inhibitory potency when compared to its linear parent peptide.Correlation between potency (determined by competition FA) and conformational analyses (by CD) indicated complex structure/ activity relationships for the interaction of all the different N-Myc variants with Aurora-A, suggesting that α-helicity of the peptides does not solely determine inhibitory potency.Indeed, the most potent lead, N-Myc 73-94-N85C/G89C-mal , was found to exhibit a unique CD spectrum; this may represent a maleimidestabilized turn-like secondary structure although this could also arise from exciton coupling with the maleimide group, however, the data is consistent with reduced random coil character.Overall, our results suggest that the constraint restricts the accessible conformational landscape of the peptide to an ensemble which is more compatible with recognition of Aurora-A.This in combination with enhanced helicity at the C-terminus, in the Aurora-A bound complex and additional non-covalent contacts between constraint and protein as observed in MD simulations provide a plausible explanation for the enhanced potency of N-Myc 73-94-N85C/G89C-mal .Overall, and excitingly, our work reports the first steps towards the rational structure/ folding-guided development of constrained peptide inhibitors of the N-Myc/Aurora-A PPI as potential anticancer therapeutics.We will report on further efforts towards this goal in due course.
Peptide maleimide constraining and oxidation: Peptide constraining was performed in acetonitrile/phosphate buffer (ratio 2 : 1; peptide concentration 4 mg mL À 1 ; phosphate buffer: 20 mM phosphate, 150 mM NaCl, pH = 7.8.TCEP (solution in buffer, 2 equiv.)was added, and the peptide was agitated for 30 min to reduce any disulfide bonds.Thereafter, dibromomaleimide (solution in acetonitrile, 2 equiv.)was added and the peptide was agitated for 2 h.The reaction was monitored by LC-MS.Upon completion of the reaction, peptides were purified by preparative HPLC and freeze-dried.Oxidation of the peptides to their disulfide bridged variants was accomplished by direct air oxidation of the pure free sulfhydryl materials in buffer at room temperature overnight.
Fluorescence anisotropy (FA) assays: All assays were performed using 384-well plates (Greiner Bio-one, UK).Aurora-A 122-403-C290A/C393A protein was produced as described previously. [30]All samples were prepared in 25 mM Tris, 150 mM NaCl, 5 mM MgCl 2 , pH 7.5, and tested in triplicate using an EnVisionTM 2103 MultiLabel plate reader (Perkin Elmer; Waltham, MA, USA).The parameters were set as follows: Excitation wavelength = 480 nm (30 nm bandwidth) and emission wavelength = 535 nm (30 nm bandwidth).Measured data were processed and analysed as previously described. [31]Specifically, the perpendicular intensity (P) and parallel intensity (S) were subtracted by the control values and used for calculations of intensity and anisotropy using the following Equations 1, 2, 3, 4, and 5:

Figure 1 .
Figure 1.Key features of the Aurora-A/N-Myc complex.(a) Crystal structure of the Aurora-A C290A/C393A catalytic domain (forest-green) in complex with N-Myc 61-89 (cyan; PDB: 5G1X).Magnified view of the Aurora-A/N-Myc PPI interface with A-loop and P + 1 pocket shown (limon).Key residues shown as sticks, with pThr288 AurA residue (red) and hot-spot residues (deep-pink) highlighted.Charge-reinforced or polar contacts between side chains are shown by dashed yellow lines.(b) Scheme illustrating the use of reversible chemistry for the synthesis of maleimide constrained peptides.

Figure 4 .
Figure 4. MD studies of N-Myc 73-94 and constrained N-Myc 73-94-N85/G89C-mal in the presence of Aurora-A: (a) Overlay of the minimum energy structures observed for N-Myc 73-94 (in grey) and N-Myc 73-94-N85/G89C-mal (in cyan) bound to Aurora-A (the maleimide constraint and key interacting residues are highlighted in purple); (b) Average structures of unmodified N-Myc 73-94 (shown in grey) and N-Myc 73-94-N85/G89C-mal (shown in cyan) in the presence of Aurora-A (the maleimide constraint is shown highlighted in purple); (c) Axial view of the averaged structure of N-Myc 73-94-N85/G89C-mal (in cyan) in the presence of Aurora-A showing the arrangement of the maleimide group and key residues (W88 and E84) around Gln335 AurA .

Table 1 .
Effects of sequence truncation, elongation, and single point variations in N-Myc 61-89 .

Table 2 .
Effects of sequence modification, oxidation and maleimide constraint in N-Myc 73-94 .One letter code for amino acids.Highlighted in grey: free-SH cysteines; in gold: disulfide bridged cysteines À SÀ SÀ and in magenta: maleimide constrained cysteines.[b] IC 50 values given as the mean value and corresponding standard deviation (SD) determined from triplicate competition FA assays against fluorescein labelled WT N-Myc 61-89 FAM (50 nM) in the presence of Aurora-A 122-403-C290A/C393A (15 μM) (n = 3).All assays were performed in 25 mM Tris, 150 mM NaCl, 5 mM MgCl 2 , pH 7.5 and left to equilibrate for 2 hours at room temperature before measuring.[c] Mean residue ellipticity and [d] estimated % Helicity as measured in 25 mM Tris, 150 mM NaCl, 5 mM MgCl 2 , pH 7.5 at 5 °C by CD spectroscopy and calculated using equations 6 and 7 (n = 2).