Toward the design and development of peptidomimetic inhibitors of the Ataxin-1 aggregation pathway

Abstract

Spinocerebellar ataxia type 1 is a degenerative disorder caused by polyglutamine expansions and aggregation of Ataxin-1. The interaction between Capicua (CIC) and the AXH domain of Ataxin-1 protein has been suggested as a possible driver of aggregation for the expanded Ataxin-1 protein and the subsequent onset of spinocerebellar ataxia 1. Experimental studies have demonstrated that short constructs of CIC may prevent such aggregation and suggested this as a possible candidate to inspire the rational design of peptidomimetics. In this work, molecular modeling techniques, namely the alchemical mutation and force field-based molecular dynamics, have been employed to propose a pipeline for the rational design of a CIC-inspired inhibitor of the ataxin-1 aggregation pathway. In particular, this study has shown that the alchemical mutation can estimate the affinity between AXH and CIC with good correlation with experimental data, while molecular dynamics shed light on molecular mechanisms that occur for stabilization of the interaction between the CIC-inspired construct and the AXH domain of Ataxin-1. This work lays the foundation for a rational methodology for the in silico screening and design of peptidomimetics, which can expedite and streamline experimental studies to identify strategies for inhibiting the ataxin-1 aggregation pathway.

Significance

Spinocerebellar ataxia type 1 is a neurodegenerative disorder related to the misfolding and aggregation of the Ataxin-1 protein. This disease is part of the so-called polyglutamine disease, a group of inherited pathologies related to an expanded tract of CAG nucleotides. The AXH domain of Ataxin-1 is responsible for the pathological aggregation of this protein. Defining pharmacological strategies to counter this aberrant aggregation pathway is fundamental to treating this pathology. This work aims at defining a computational methodology to rationally design a peptidomimetic able to interact with the AXH domain and prevent the aggregation pathway.

Introduction

Spinocerebellar ataxia type 1 (SCA1) is a neurodegenerative disease, characterized by misfolding of a protein called Ataxin-1 (ATX1) (1). SCA1 is a fatal autosomal dominant pathology whose symptoms are ataxia, ophthalmoparesis, and variable degrees of motor weakness resulting from selective loss of neurons in the brain stem and cerebellum and degeneration of the spinocerebellar tracts (2,3). The pathology belongs to so-called polyglutamine (PolyQ) diseases related to an expanded cytosine-guanine-adenine repeated motif into the encoding tract of the gene (4). It is firmly established that cytosine-guanine-adenine expansion is not the only actor in the pathological behavior of ATX1; sure enough, other domains of the protein are involved in aggregation and the resulting pathological condition. Indeed, the ATX1/HBP-1 (AXH) domain (SMART:SM00536) of ATX1 must be pointed to as a fundamental partner for the aggregation mechanism. In more detail, it has been highlighted that the isolated AXH domain in solution tends to aggregate into oligomeric forms (e.g., dimer, tetramer, and higher molecular species) (5). Furthermore, supporting the involvement in the amyloidogenic process of the whole ATX1, it has been shown that alteration of the AXH domain into ATX1 decreases protein aggregation (5). The AXH oligomerization mechanisms (6,7) have also been investigated through in silico modeling (8,9,10). Molecular dynamics (MD) simulations have been carried out to elucidate monomer conformation behavior in solution (8): results pointed out the overall stability of the AXH core, except for a motif of 20 residues, namely the N-terminal (Nter).

Computational investigation of the monomeric form has shown how the Nter is prone to fluctuate and that this domain has different conformations of equilibrium, divided by low energy barriers (8). Furthermore, a molecular investigation of the AXH dimer has given insight into the monomer-monomer interaction mechanism and, together with the results that emerged previously, both from in silico (8) and experimental observations (11), has suggested that the dimeric interaction acts stabilizing the Nter region, which is, in turn, involved in the dimeric interface (9,10). Later, the tetrameric form of AXH has been investigated, pointing out the importance of residues I580 and R638 for the protein-protein interaction, in agreement with experimental outcomes that observed a reduction in tetrameric percentage as a consequence of mutation of I580 into an alanine (9,10). In more detail, the work has shed light on the importance of specific residues in the mechanism of interaction, highlighting, in fact, how the mutation of I580 destabilizes the interface of the AXH interdimer while the residue R638 plays a key role in the stability of the tetramer. The AXH region plays an important role in the ATX1 function, mediating both RNA-binding activity and protein-protein interactions (12).

Regarding the polyQ-expanded ATX1 pathogenicity, many shreds of evidence suggest that beyond the onset of the pathology, there is a toxic gain-of-function mechanism instead of a loss-of-function process (13). Indeed, mice, where the expression of ATX1 was suppressed, did not show any SCA1-like behaviors or neurodegenerations (14). It has been observed that, interacting through the AXH domain, expanded ATX1 tends to form an abundant protein complex with the transcriptional repressor Capicua (CIC) (15), altering transcriptional repression activity of the CIC (16). The CIC-AXH interaction has been studied to point out the role of this complex in the onset of the pathology. What emerges is that the toxicity of polyQ-expanded ATX1 aggregates is closely linked to the interaction with CIC, whereas the mere presence of ATX1 aggregate accumulations is not sufficient for the onset of pathology (14). It is worth mentioning that loss of the ATX1-CIC interaction does not result in SCA1-like symptoms in mice (14). Thus, targeting the CIC-AXH interaction has been proposed as a possible strategy for restoring physiological behavior and mitigating the toxic function gained by polyQ-expanded ATX1 (17). Within this framework, CIC itself has emerged as an ideal template to design possible compounds or short synthetic peptides able to prevent aggregation of ATX1. While the interaction of AXH with a longer form of CIC tends to drive the complex to form multimeric aggregates (14, 15), a shorter synthetic construct of CIC (L-CICp) can bind AXH and prevent aggregation (15, 17).

Based on those findings (15,17), understanding the protein-protein interaction for the AXH/L-CICp complex and how this can prevent aggregation is fundamental to rationally designing and proposing new pharmacological strategies. So far, any study regarding a dynamic description of the AXH/L-CICp protein interaction has been carried out. Computational approaches, and in particular MD, have proven to be a powerful tool to understand the protein-protein interaction and elucidate molecular mechanisms with atomistic resolution (8,9,10,18,19,20). This work is focused on the first MD study on the AXH/L-CICp complex, aimed at shedding light on mechanisms that prevent AXH self-polymerization. We employed computational techniques such as force field-based MD and the alchemical mutation to develop a computational platform able to predict the effect on the binding affinity of L-CICp and AXH, both from a thermodynamic and a mechanic point of view. The employed methodology has proved to predict thermodynamic properties such as the change in peptide binding affinity in agreement with experimental results, while dynamic information obtained through MD gave clues on the interaction mechanism. Thus, this study can be seen as a methodological proof of concept for a computational platform to rationally design synthetic peptides able to interfere with and stop the toxic pathway gained by polyQ-expanded ATX1 and restore it to the physiological condition.

Materials and methods

System coordinates

Coordinates of the L-CICp/AXH complex were obtained from available structures deposited in the PDB (PDB: 2M41 (17)). For consistency with previous literature (15,17), numbering for AXH and L-CICp refers, respectively, to UniprotKB/Swiss-Prot entry nos. P54253 and Q96RK0.

Free-energy calculation

To study the change in the binding energy upon amino acid mutation, the alchemical mutation protocol has been employed (21,22). Python package pmx (21) has been used to build the topology files and properly preprocess the molecular structure. In detail, an alanine scan has been performed on the L-CICp sequence (VFPWHSLVPLASPQ), mutating one by one each residue and calculating the subsequent variation in the binding free energies ($\Delta G$). Both extremes and proline residues have been excluded from the mutation protocol according to the known limitations related to this methodology (21). To build the system topology, the Amber99sb^∗-ILDN, an updated version of the Amber99sb-ILDN force field with a pregenerated mutation library, has been employed. The system has been named according to the residue mutated: for example, if residue 37 was mutated from tryptophan (W) to an alanine (A), the corresponding name is CIC^W37A. Each system has been placed in a cubic box and solvated with explicit water, and the total charge of the system was neutralized by adding Na⁺ and Cl⁻ at a physiological concentration of 0.15 M. The protocol of the nonequilibrium free-energy calculation was employed (23,24,25). Indeed, physical end states were obtained from 10 ns of equilibrium ensemble, and then an alchemical transition of 100 ps between states was performed, measuring the work done by the system during the transitions. In the alchemical transition between states, nonbonded interactions were treated by applying a soft-core function (26). In the end, pmx analysis modules have been employed to calculate the $\Delta \Delta G$ values between bounded and unbounded states by the Crooks fluctuation theorem (27) and associated errors via the maximum-likelihood estimator (28). In order to compare results with experimental data, a measure reported in previous literature as Kd (17) has been converted to $\Delta \Delta G$ with the following equation:

$$\begin{array}{l}\text{ΔΔ}{G}_{A\to B}=-R\ast T\ast \ln \left(\frac{K{d}_B}{K{d}_A}\right)\hfill \end{array}\text{,}$$ (1)

where R is the Boltzmann constant 8.31 J/(mol^∗K), T is the temperature in Kelvin, and Kd_A and Kd_B are, respectively, the dissociation constant for the complex with the mutated form and the wild-type (WT) form reported from previous investigations (17).

Force field-based MD of AXH/L-CICp complex

Five systems were constructed, one containing only the AXH domain (AXH) and four others with a complex consisting of the AXH domain and a variant of the L-CICp peptide. To evaluate the mechanisms of the L-CICp peptide interaction on the dynamics of the AXH domain, and the effect of a single point mutation, the AXH/L-CICp WT was compared with the case of L-CICp mutants that experimentally result in an absolute $\Delta \Delta G$ above the thermal fluctuation energy: the mutants W37A L40A, and S39A. The conformational dynamics of the above-mentioned models were compared with those of the AXH domain alone, in the absence of the CIC peptide. All systems were placed in a dodecahedron box filled with explicit water model TIP3P (29) and neutralized with Na⁺ Cl⁻ ions added at a physiological concentration of 0.15 M. System energies were minimized through the steepest descent algorithm followed by two preliminary position restrained MD simulations, restraining α-carbon positions, first in an ensemble with constant volume and temperature and then with constant pressure and temperature. Finally, systems were simulated by removing constraints in constant pressure and temperature ensembles for 200 ns. Four replicas have been performed for each system. The engine employed for MD simulations and analysis was GROMACS 2020.3 (23). Protein topology was built applying Amber99sb-ILDN force field (30). For the temperature coupling, two groups were created, one containing the solute (AXH and, where present, CIC) and another for solvent (water molecules and ions), both coupled by a v-rescale (31) algorithm at 298 K (time constant ${\tau }_T=0.1$ ps). Pressure coupling at 1 bar (with a time constant ${\tau }_P$ = 5 ps) was done with the Parrinello-Rahman algorithm (32). Summary of the production replicas is reported in supporting material (Table S1).

Structural analysis

Trajectories were visualized and inspected using the visual MD (VMD) package (33). The flexibility of the protein was evaluated by computing the root-mean-square fluctuation of the α-carbons (Cα-RMSF) during the last 50 ns of each replica and fitting the structures on the Cαs. GROMACS analysis suite has been employed to perform the calculation. To evaluate the change in mechanical properties, the force constant has been calculated; this is a measure of the fluctuations of the mean distance of each residue from the rest of the structure (34,35). The calculation of force constants was implemented according to the formula

$$\begin{array}{l}{k}_i=\frac{3k_{B}T}{⟨{\left(d_i-⟨d_i⟩\right)}^2⟩}\hfill \end{array}\text{,}$$ (2)

where $d_i$ is the mean distance of the i-th residue from the rest of the structure, $k_B$ is the Boltzmann’s constant, $T$ is the temperature of the system, and the operator stands for the average over the simulation. The distances were defined between the Cαs of the amino acids and computed on representative snapshots extracted every 100 ps for each replica and averaged; the force constant has been calculated considering just the AXH domain. Force constant calculation has been done employing the python MDAnalysis library (36).

The per-residue protein-protein contact probability has been calculated as the probability of one atom in the selected residue to be at a distance below 2 Å from an atom of the reference molecule following a procedure employed in previous studies (9).

The probability of a residue to be involved in a secondary structure (i.e., α-helix, β-sheet, turn, or coil) has been evaluated through the STRIDE software (37), on the concatenated trajectories, as an average of the structure assigned each frame.

The probability of a specific interaction between the protein and the peptide (i.e., hydrogen bond, hydrophobic interaction) has been evaluated by considering the interaction through the PLIP software (38) on each frame and then averaging the number of the occurrences of the interactions on the total number of frames. Interactions whose probability was over 50% were selected.

Then, to evaluate the effect of L-CICp on the conformational dynamics of the AXH domain, two quantities were extracted every 50 ps from the concatenated replicas extracted from the last 50 ns of each replica. The first one is the distance from the protein center of mass (COM) of the β-hairpin formed from residues 580 to 590. The second one is the angle between the COM of the protein, the COM of the head, and the COM of the tail of the α-helix, respectively, from residues 596 to 598 and 604 to 608. Moreover, a visual inspection of the tracked coordinates has been rendered employing an in-house Tcl script (https://github.com/micmar15sr/Internal_cord).

Results

Alchemical mutation predicts the change in the experimentally measured binding affinity

Double free-energy difference (ΔΔG) has been evaluated by performing a computational alanine (A) scan on the peptide sequence. In detail, the change in binding energy (ΔΔG_Calc) has been estimated by mutating one by one the amino acids belonging to the CIC sequence i.e., F35, W37, H38, S39, L40, V41, F43, and L44 to A, and comparing the results with experimental measures (ΔΔG_Exp) reported in previous work (17). It must be pointed out that the above-mentioned method does not support mixed topology for the alchemical mutation of the proline residue (P) and both the ammino and carboxyl peptide extremities. From the alchemical mutation procedure, residues W37A, L40A, and F43A showed the highest effect in the change of binding affinity (Fig. 1), +16.5 , +7.7, and +9.6 kJ/mol, respectively. The correlation between experimental data and computational evaluation has been assessed by linear regression, showing a good agreement in terms of the Pearson correlation coefficient (R = 0.87) (Fig. 1). In addition, the average unsigned error, with respect to experimental measures, was 3.81 kJ/mol, comparable with similar studies (39). Overall, those findings are in agreement with previous experimental studies, pointing out that mutation of both residues W37 and L40 affects binding with a drastic decrease in binding affinity (15,17). Furthermore, it can also be observed that the mutation of the other residues in alanine does not alter the affinity of the peptide to its protein partner, in agreement with what has been observed in previous experimental studies (17).

Figure 1.

Linear regression fitting between changes in double free-energy difference among the alchemical mutation prediction (ΔΔGCalc) and experimental data (ΔΔGExp), within 90% of confidence interval (shaded blue). On the top left, the Pearson correlation coefficient and the associated p value are reported. To see this figure in color, go online.

AXH-CIC complex showed lower fluctuations compared with the free AXH domain

To characterize the effect of the AXH/L-CICp interaction on protein conformational dynamic and the AXH dynamics itself, four MD replicas for the five different configurations have been performed. These five systems were correspondingly named AXH, AXH-CIC^WT, AXH-CIC^W37A, AXH-CIC^S39A, and AXH-CIC^L40A. The variation of Cα-RMSF between AXH unbound and AXH.CIC^WT was evaluated to examine the effects of the peptide-protein interaction on the local flexibility of the AXH domain. The Cα-RMSF of the AXH domain has been calculated for each system and for each replica and then reported as the mean and standard deviation in Fig. 2 A. Since all the systems under investigation showed, in agreement with the previous investigation (8), the great mobility of the Ntail, data have been reported neglecting that region (for the RMSF of the whole protein, see Fig. S1). The presence of the CIC peptide, as expected, reduces the AXH conformational plasticity, in particular in the region 580–611. Furthermore, to characterize the change in mechanical profile as an effect of the protein-peptide interaction, a per-residue force constant has been evaluated (Fig. 2 B). In agreement, with the Cα-RMSF, the AXH-CIC^WT complexes showed overall greater mechanical characteristics if compared with the free AXH. Then, we compared the profile of the two systems AXH and AXH-CIC^WT in terms of variation of relative highest peaks (i.e., with values of force constant above the average) to understand which residues reduce their relative peak of stiffness in the absence of the protein-peptide interaction. It is possible to observe that in the AXH residues belonging to the α-helix (591–599) and residues 612 and 614 lose the role of stiffer elements if compared with AXH-CIC^WT.

Figure 2.

CIC interaction reduce AXH fluctuation and increase its mechanical properties. (A) RMSF of AXH protein reported as mean and standard deviation per residue. (B) AXH per-residue force constant reported as mean of the four replicas. To see this figure in color, go online.

AXH-CIC mutants dynamics reveal fundamental interactions for the protein-ligand complex

To characterize how the single point mutation (W37A, S39A, L40A) affects the interaction between CIC and AXH, the protein-ligand contact probability (CP) has been evaluated. In detail, after calculating the AXH-CIC^WT CP (Fig. 3 A), we reported the relative difference to the WT for the mutants AXH-CIC^W37A, AXH-CIC^S39A, and AXH-CIC^L40A (Fig. 3, D, B, and C, respectively). It is possible to observe that the residue Q582 is the one involved in the most probable interaction for AXH-CIC^WT (CP = 51%) and increases for the mutants AXH-CIC^S39A (CP = 0.65%) and AXH-CIC^L40A (CP = 0.68%), while it suddenly decreases for AXH-CIC^W37A (CP = 23%). Notably, the interaction with residue L686 with a relatively high probability for the WT (CP = 41%) is conserved in the L40A (CP = 42%) and W37A (CP = 43%), i.e., the two destabilizing mutants, and increases for the stabilizing mutant S39A (CP = 57%). Regarding the secondary structure (Fig. S2), it is possible to observe a tendency of a helix-to-coil transition for the residues 596–605 of AXH, AXH-CIC^W37A, and AXH-CIC^L40A, while this helix is preserved for AXH-CIC^WT and AXH-CIC^S39A. A further difference can be observed in the content of the β-sheet (residues 609–621) and the α-helix (residues 663–670): AXH-CIC^S39A and AXH-CIC^L40A complexes showed a tendency to lose the above-mentioned secondary structures, whereas the other systems preserved the ordered structural stability. Previous results are in agreement with the experimentally observed structural plasticity of the AXH domain and the ability to remodel its interaction surface to adapt to different binding partners (6, 11).

Figure 3.

AXH-CIC contact probability. Per-residue contact probability between the AXH domain and the CIC peptide for the (A) AXH in complex with CIC^WT (red) and relative difference of contact probability (black) for the AXH in complex with (B) CIC^S39A, (C) CIC^L40A, and (D) CIC^W37A. On the right side of each plot, a visual rendering of the AXH domain is reported in the “new cartoon” style. Residues have been colored according to values of contact probability. Amino acids with contact probability above 50% were rendered in van der Waals representation. For the sake of clarity, the same color scale has been employed going from 0% (white) to 50% (CIC^WT: red; CIC^S39A: magenta; CIC^L40A: green; CIC^W37A: black). To see this figure in color, go online.

Analyzing the CP of the CIC peptide (Fig. S3), it is possible to observe that the mutation at position W37 affects the contact of the residue in position 37, while in the other CIC mutants, S39 and L40, and in the WT, this contact is preserved. In the case of L40 mutants, it is possible to observe a slightly increased probability of contact for the residue in position 40, if compared with the other systems, and a decreased probability of contact for P42. Given the change in the peptide CP, the probability of it being involved in a secondary structure for the peptide in complex with the AXH has been investigated to analyze whether the peptide forms a structured interaction. CIC^WT is characterized by β-strands between residues 35 and 36 that pair up with AXH residues 580–582, an α-helix between residues 37 and 39, and another β-strand with the AXH core residues (Fig. 4 B). The mutant CIC^W37A showed an altered probability in protein-peptide secondary-like structure interaction if compared with the WT, with a substantial decrease in the probability of residues 35–36 being involved in β-sheet-type structure and an increased probability of β-pairing for residue 43 (Fig. 4 B). If compared with CIC^WT, CIC^L40A shows similar behavior in terms of the secondary structure except for an increased probability of residue 35 being involved in a β-strand. Lastly, the CIC^S39A mutant, characterized by an increased binding affinity, showed a similar fashion to CIC^WT with an increased probability of residue 43 being involved in a β-strand. These results are in agreement with previous experimental studies that showed how residues V34 and F35 drive the peptide to form a β-strand that pairs up with the AXH structure (Fig. 4 A) and that this interaction is a fundamental driver for the L-CICp/AXH complex (17). Lastly, a type-specific residue-residue interaction probability analysis has been performed to characterize the interchain noncovalent contacts most involved in the AXH-CIC interface. The profile of AXH-CIC^WT is characterized by hydrogen bonds (HBs) (i.e., 582Q-35F, 582Q-37W, 682L-42P) and hydrophobic interactions (HPs) (i.e., 575F-40L, 632F-44L, 686L-37W) (Fig. 4 A). Destabilizing mutant CICW37A lacks the characteristic interactions of WT, particularly those mediated by W37 (HB 582Q-37W and HP 686L-37W) (Fig. 4 D). CIC^L40A exhibits a disruption of the HP 575F-40L and the HP 632F-44L and a reduction in the probability of the HB 686L-42P, whereas other interactions raise their probability, such as the 632F-44L HB, the 580I-35F HB, and the 582Q-35F HB. (Fig. 4 C). CICS^39A, which increases affinity for AXH, is primarily characterized by the same interaction patterns as the WT: 582Q-35F, 582Q-37W, 682L-42P, 632F-44L, and 686L-37W. Only HP 575F-40L has a decreased probability, whereas HB 582Q-35F, HB 582Q-37W, HB 686L-42P, and HP 599F-41V are stabilized.

Figure 4.

Mutations alter major noncovalent peptide-protein interactions. On the left, there is the probability of a residue being involved in a secondary structure for (A) CIC^WT, (B) CIC^S39A (C) CIC^L40A, and (D) CIC^W37A. For the sake of clarity, the α-helixes are colored in black, the β-sheets in gray, and the unstructured regions in white. On the right, a polar plot indicating the eight most probable protein-peptide noncovalent interactions is reported. For the sake of clarity, the interaction is reported as a hydrogen bond or a hydrophobic interaction, protein residue number, residue name in the one letter code, peptide residue number, and residue name in the one letter code. To see this figure in color, go online.

CIC^WT constraints AXH motion

To evaluate the conformational stability of the AXH domain and the effects of the L-CICp, the dynamics of the domain for AXH and AXH-CIC^WT were evaluated defining two collective variables. The first variable measures the distance between the COM of AXH and the COM of the β-hairpin defined by residues 580–590 (Fig. 5 B). The second one describes the angle formed between the protein COM and the helix extremes between residues 596 and 608 (Fig. 5 A). The distance has been chosen to consider the region that is involved in the interaction between L-CICp and AXH, whose contact is altered in the case of mutation W37A. On the other hand, the angle was monitored to consider that part of the AXH domain with reduced mechanical properties and a tendency to lose secondary structure both in the absence of the peptide or in the presence of CIC^W37A. The conformational dynamic of unbound AXH is characterized by high variability in space described by the two coordinates with distance values equal to 16.41 ± 2.9 Å and an angle equal to 97.33° ± 16.98° (Fig. 5 A). Bound AXH-CIC^WT stabilizes the region of residues from 580 to 608 with distance values equal to 17.27 ± 0.61 Å and an angle equal to 96.44° ± 6.11° (Fig. 5 A). The effect of mutants on AXH conformational dynamics is shown in Fig. S4. While systems with the L40A and S39A mutations appear to show comparable dynamics to the WT case (Fig. S4, A and B), the W37A mutation shows dynamics with multiple scattered equilibrium points, with dynamics similar to the free AXH case (Fig. S4 C).

Figure 5.

Change in the conformational dynamics of beta hairpin (residues) in the presence and absence of interaction with CIC. (A) Bivariate distribution of loop (residues 580–590) distance and angle flexion of the α-helix (residues 596–598 and 604–608) with respect to AXH’s center of mass. (B) A visual inspection of used internal coordinates with reference points reported as blue balls and protein rendered as a gray new cartoon. To see this figure in color, go online.

Discussion

Experimental studies showed that the CIC is a fundamental actor driving the mechanisms of aggregation for ATX1 (15), whereas shorter forms of CIC have been shown to prevent protein aggregation (15, 17). Therefore, evaluating the above-mentioned protein-protein interaction has been suggested as a possible strategy toward protein-specific drug design (17). A previous study, given the ability of a short construct of the CIC peptide, namely L-CICp, to prevent AXH aggregation, assessed which of the 11 amino acids belonging to this peptide played a key role in the affinity of the complex (17). Starting from this experimental work, the current study has been focused on analyzing the protein-peptide interaction between the AXH domain of ATX1 and the synthetic peptide L-CICp, both from a thermodynamic and a mechanical point of view, with the final aim of providing a methodological proof of concept for a computational platform to design peptidomimetic able to prevent AXH aggregation (7,15). In this connection, computational strategies have proven to be an effective tool in elucidating the mechanism of AXH aggregation, as demonstrated by our previous studies (8,9,10). The main findings of the present research have highlighted 1) how the alchemical mutation can be an effective tool to test in silico mutation, and 2) a molecular description of how L-CICp interacts with AXH. In this article, we proved that alchemical mutation methodology can be an effective tool to predict thermodynamical properties, such as the binding affinity between the L-CICp peptide and the AXH domain, as a result of a single point mutation on the peptide. It is worth mentioning that the alchemical mutation approach, like any other model, is characterized by a number of known limitations related to insufficient sampling, force field inaccuracies, and the intrinsic experimental uncertainty of the starting conformational structure. Moreover, we considered the assumption that point mutations do not substantially disrupt the original binding conformation of the WT form. Overall, the computational predictions presented in this article are characterized by a Pearson correlation coefficient of 0.87 and an average unsigned error of 3.28 kJ/mol (Fig. 1). These results are in line with previous work in this field, showing the limit of this protocol to predict the ΔΔG within 1 kcal/mol (4.186 kJ/mol) (39). As suggested from the literature, and confirmed by our calculation, the residue W37 seems to be the most important for the AXH/L-CICp interaction (17). Moreover, an MD investigation has been performed to characterize the conformational arrangement that the peptide induces on protein and emphasize the possible mechanism of inhibition. For this reason, we compare the dynamics of the monomeric AXH domain, AXH, in complex with the WT L-CICp and AXH in complex with the L-CICp mutants characterized by an ΔΔG value above the thermal fluctuation: the destabilizing W37A, the worst in terms of binding affinity, the destabilizing L40A, and the stabilizing S39A. Previous MD studies have proven to be able to characterize the effect of mutation occurring at the AXH domain, in particular to the residues I580 and R638, and the consequential effects on the thermodynamic equilibrium between monomeric, dimeric, and tetrameric AXH aggregates (10). Our findings were able to replicate previous data on the free monomeric form in terms of RMSF (Fig. 2) and secondary-structure stability (Fig. S2). We observed a similar mechanism of stabilization induced by the peptide to that observed in AXH by the dimeric interaction. Indeed, CIC^WT in complex with AXH can stabilize the domain, with a similar effect of the dimeric form for the AXH Nter (8). Moreover, we found a central role of the residue I580 for the interaction in the protein-peptide complex as seen for the monomer-monomer interface (9). Furthermore, resemblance in terms of secondary-structure stability of AXH was observed between the dimeric AXH/AXH complex and the AXH/L-CICp WT (Fig. S2). Results from our MD simulation suggested that when mutated to alanine, the residue W37 results in a change of conformational dynamics of the AXH domain that shares an intermediate behavior between the free AXH domain and the one in complex with WT L-CICp. Indeed, the complex increases the overall rigidity of the AXH domain if compared with the unbound state, suggesting an overall effect of stabilization induced by the peptide. Furthermore, the main AXH residues involved in the interface are also the ones that have been found, from previous studies, to be crucial for the dimer interface, such as I580 and L686 (9). Additionally, investigating the CP between the L-CICp and the AXH domain reveals that the mutation of residue W37 affects the probability of residue F35 to be in contact with the AXH chain, destabilizing the probability of this residue being involved, together with the residue P36, in a β-sheet-like structure. This piece of evidence is in agreement with previous experimental studies that found that the role of residues V34 and F35 increase, by two orders of magnitude, the binding affinity for the complex (17). Moreover, investigation of the W37A, L40A, and S39A mutants has provided a hint regarding strategies to design compounds that mimic L-CICp interactions. Indeed, Fig. 4 provides a detailed description of chemical features and residues fundamental for L-CICp-AXH binding. Lastly, the MD trajectories have provided clues on the possible effect that the L-CICp WT has on AXH domain conformational dynamics. Indeed, two collective variables were defined: one that describes the motion of the β-hairpin formed between residues 580 and 590, and another that tracks the motion of the α-helix (residues 596 to 608). Analyzing the motion of the above-mentioned variables between the bound and unbound cases (i.e., AXH, AXH-CIC^WT), it has been observed how the stabilization mechanism of the AXH-CIC^WT complex involves the confinement, during the observed times of dynamics, in a specific configuration, hindering the modifications of the tertiary structure (Fig. 5 A). On the other hand, it is possible to observe how, partially in the case of loss of affinity of L-CICp with the target protein (Fig. S4) and more markedly in the case in which the peptide is removed, the protein regains conformational mobility, accessing a greater number of configurations (Fig. 5 A). Results reported and discussed here highlight that this framework of molecular modeling techniques could support and unlock the development of a computational platform to systematically test a possible peptidomimetic and fine-tune the affinity of the mimetic molecule for the AXH. As suggested from previous studies (17), this work can be a further step aimed at integrating and enabling virtual fine-tuning in order to speed up the experimental activity.

Conclusion

This article has proposed a proof of concept for a computational methodology to study peptidomimetics able to prevent AXH aggregation. The alchemical mutation technique has been employed to predict the effect of the affinity of the complex AXH/L-CICp in an alanine scan performed on the peptide sequence. Force field-based MD was used to evaluate the effects on AXH conformational stability in a complex with WT L-CICp and compared with the one bounded with the peptide, where the mutation is more destabilizing for the complex, to elucidate the molecular mechanism that stabilizes the complex. Moreover, a further effort has been devoted to comparing the dynamic of the complexes with the one of the solo AXH domain to shed light on the overall effect of the heterodimers that compete with the formation of dimeric AXH. The alchemical mutation predicts binding affinity with a good correlation to previous experimental studies. MD highlighted the effect of the peptide to stabilize the overall motion of the AXH domain and enhance the mechanical stiffness of the overall structure. Moreover, from the dynamic investigation emerges the ability of the L-CICp to interact with important residues involved in the dimerization process, such as I580, and compete with the dimer formation. In addition, a possible stabilization mechanism exerted by L-CICp emerges that can stabilize the AXH domain and limit its movement to an equilibrium conformation, while the AXH domain in the absence of interaction with the peptide shows a different equilibrium configuration. In summary, the results of this article suggest a possible combination of in silico methodologies to rationally design peptidomimetic inhibitors of the ATX1 aggregation pathway.

Author contributions

M.A.D. and G.G. designed the research, M.M. performed the computation, M.M. and G.G. analyzed the results. All the authors interpreted the data and wrote the manuscript.

Editor: Jing Chen.