Discovery of Peptide-Based Tubulin Inhibitors Through Structure-Guided Design

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors describe an integrated computational-experimental approach for discovery of Peptide-Based Tubulin Inhibitors. The study is well-designed and used an appropriate set of experiments to support their claim. The experimental part is well-performed and discussed. The manuscript is overall well-organized with clear and suitable academic tone and is appropriate for publication in the journal of pharmaceutics.

 

I recommend publication. However, the following comments should be addressed before publication:

1. In the title the term selective is misleading and beyond the results presented.
2. The authors mentioned that residues within up to 6 Å from the dimer interface were selected, and those regions containing continuous stretches of interacting residues were identified as potential peptide candidates. This does not match with figure 2 where most of the identified sequences lie at the extremities of the interacting dimers.
3. The rational of peptide generation is not clear. How and why would these isolated selected peptides affect the dimer binding by lying at the dimer interface?
4. The docking protocol is not mentioned. Were theses peptides docked at the interface of the protein dimer or with α-tubulin only?
5. The calculation of the selectivity index (SI) would better emphasize the positive research results by comparing with the positive control paclitaxel.
6. In case that docking was not performed at the dimer interface, a competitive binding should be done to conclude higher affinity and dimerization prevention.
7. The fact that the capped hit peptide demonstrated exhibited no inhibitory effect on tubulin polymerization despite showing higher cytotoxicity in NCI-H1299 cancer cells compared to the original hit peptide needs to be further discussed and clarified. Lines 413-415 “These results indicate that substituting the N-terminus with an acetyl group and the C-terminus with a methylamide (NMe) in P14 enhances antiproliferative activity compared to the unmodified version.” Vs Lines 473-475 “In contrast, its terminally protected analogue, Ac-P14-NMe, exhibited no inhibitory effect on tubulin polymerization despite showing higher cytotoxicity in NCI-H1299 cancer cells (IC50 = 22.09 ± 3.50 μM).”
8. The results of the molecular dynamics studies are not supportive of the experimental findings. First the most stable peptide from the capped set maintained an RMSD value around 6-8 Å while from the uncapped original set NH2-P14-COOH exhibited the most stable RMSD profile, remaining nearly constant (~5-6 Å) over the entire simulation. In both cases the simulations reflect low degree of stability and contradict the experimental results where the capped analogue was found inactive. The authors need to review this part carefully and put it in context with the experimental results.
9. Peptides based therapeutics still have its shortcomings including stability and pharmacokinetics, the author should address this point based on similar previous studies.
10. Some minor editing such as IC50 should be subscript all throughout the manuscript (IC₅₀)

 

 

 

Author Response

Comments 1: In the title the term selective is misleading and beyond the results presented.

Response 1: We thank the reviewer for this thoughtful comment regarding the use of the term “selective” in the title.

In the context of this study, the term “selective” was used to describe the differential cytotoxic profile observed between cancerous (NCI-H1299) and non-cancerous (EA.hy926) cells. Specifically, NH2-P14-COOH demonstrated measurable cytotoxicity in tumor cells (IC₅₀ = 45.64 ± 3.20 μM) while exhibiting no significant toxicity toward endothelial cells (IC₅₀ > 100 μM), suggesting a favorable therapeutic index relative to classical microtubule-targeting agents such as paclitaxel.

We agree, however, that the term “selective” may be interpreted more broadly, potentially implying isoform specificity, binding-site selectivity, or validated mechanistic discrimination beyond the scope of the present work. Since our study includes evaluation in only one tumor and one non-tumor cell line, and does not include tubulin isoform specificity assays, or other receptors, we recognize that the terminology could be perceived as overstated.

To avoid ambiguity and ensure that the title accurately reflects the experimental evidence presented, we have revised the title to:

“Discovery of Peptide-Based Tubulin Inhibitors through Structure-Guided Design”

We believe this modification more precisely reflects the scope of the study without overstating the conclusions.

Comments 2: The authors mentioned that residues within up to 6 Å from the dimer interface were selected, and those regions containing continuous stretches of interacting residues were identified as potential peptide candidates. This does not match with figure 2 where most of the identified sequences lie at the extremities of the interacting dimers.

Response 2: We thank the reviewer for this careful observation.

Peptide selection was performed using an automated distance-based protocol that explicitly identifies residues located within ≤ 6 Å between adjacent αβ-tubulin dimers. The script computes inter-chain atomic distances between neighboring dimers in the microtubule lattice and subsequently identifies contiguous stretches of residues that satisfy the ≤ 6 Å criterion. Therefore, peptide identification was strictly algorithm-driven and not manually curated or visually selected.

It is important to clarify that, in our methodology, the term “dimer interface” refers to any region contributing to lateral or longitudinal inter-dimer contacts within the microtubule assembly. In microtubules, several stabilizing contacts occur in loop regions and peripheral structural elements. Although some extracted peptides may appear visually located at the extremities of the monomers in Figure 2, they are structurally part of validated ≤ 6 Å inter-dimer contact regions according to the automated selection protocol.

The script used for residue extraction has now been incorporated into the Supplementary Material to ensure full methodological transparency and reproducibility.

Comments 3: The rational of peptide generation is not clear. How and why would these isolated selected peptides affect the dimer binding by lying at the dimer interface?

Response 3: The rationale behind peptide generation was based on a structural interface-mimicry strategy. The αβ-tubulin dimer-dimer interactions that stabilize the microtubule lattice are mediated by defined clusters of residues forming contiguous contact regions. These regions contribute significantly to the energetic stability of lateral and longitudinal assembly.

By extracting short, contiguous stretches of residues located within ≤ 6 Å across adjacent dimers, we aimed to isolate structural fragments that naturally participate in inter-dimer recognition. The underlying hypothesis is that such fragments, when synthesized as independent peptides, may retain the ability to recognize complementary surfaces on tubulin.

If binding occurs, two principal effects may arise:

1. Competitive interference: The peptide may occupy part of the native interface surface, thereby weakening or destabilizing dimer–dimer contacts required for proper microtubule assembly.

2. Interface perturbation: Even partial or transient binding at an interfacial hotspot may locally alter contact geometry or flexibility, which can influence the dynamic equilibrium between polymerized and depolymerized states.

Importantly, peptide extraction was not assumed to guarantee activity. The generated fragments were subjected to a multistep computational filtering process (docking, molecular dynamics, MM-GBSA) to evaluate whether they could establish stable and energetically favorable interactions. Only candidates demonstrating persistent binding and favorable energetic profiles were advanced to experimental validation.

Thus, the approach does not assume that isolated fragments inherently disrupt the interface; rather, it systematically tests whether structurally derived interfacial segments can function as modulators of tubulin assembly.

We have clarified this conceptual framework in the revised manuscript to improve methodological transparency.

Comments 4: The docking protocol is not mentioned. Were theses peptides docked at the interface of the protein dimer or with α-tubulin only?

Response 4: We thank the reviewer for highlighting the need for clarification regarding the docking protocol. We have now expanded the Methods section to provide a more detailed description of the FlexPepDock procedure.

Specifically, we used FlexPepDock in refinement mode with default settings for full-atom peptide-protein refinement. Importantly, the peptide starting pose was not placed arbitrarily: it corresponds to its native interfacial location in the original lattice-derived complex. Therefore, no blind or global docking was performed. The protocol was used exclusively to locally refine peptide backbone conformation, rigid-body orientation, and side chain packing at the predefined αβ-tubulin dimer interface.

Comments 5: The calculation of the selectivity index (SI) would better emphasize the positive research results by comparing with the positive control paclitaxel.

Response 5: We thank the reviewer for this valuable suggestion. Following this recommendation, we calculated the Selectivity Index (SI) for both the tested peptides and the positive control paclitaxel, using the ratio between IC₅₀ values in non-cancerous and cancer cell lines. The comparative SI values have now been incorporated into Table 3 and discussed in the Results and Discussion sections to provide clearer context regarding the relative selectivity of the identified peptide compared to a clinically established microtubule-targeting agent.

Comments 6: In case that docking was not performed at the dimer interface, a competitive binding should be done to conclude higher affinity and dimerization prevention.

Response 6: We thank the reviewer for this important comment.

In our study, docking and subsequent molecular dynamics simulations were performed directly at the native αβ-tubulin dimer interface, as peptides were extracted from regions participating in inter-subunit contacts. Therefore, the peptides were evaluated in the exact structural context of the interface rather than at distal pharmacological pockets.

Regarding the suggestion of performing an explicit competitive binding simulation including the complementary tubulin subunit, we agree that this would represent a rigorous way to evaluate displacement. However, such a setup is technically non-trivial. The peptide and the native partner subunit occupy overlapping interfacial volumes; placing both simultaneously at the interface generates severe steric clashes that cannot be resolved by standard minimization without introducing artificial distortions. A physically meaningful assessment of competition would require defining a displacement pathway and sampling rare events involving partial interface opening or subunit rearrangement. These processes are governed by substantial kinetic barriers and typically require enhanced sampling approaches (e.g., umbrella sampling, metadynamics) and significantly larger systems including additional lattice contacts.

Such free-energy competition calculations would represent a separate computational project beyond the scope of the present work.

Instead, our strategy focuses on:
(i) evaluating the stability of peptide binding at structurally validated interfacial hotspots,
(ii) extending selected systems to 1 µs simulations to assess long-timescale behavior, and
(iii) analyzing tubulin conformational dynamics (RMSF and PCA) in the presence and absence of peptide to determine whether interface flexibility is modulated.

These analyses directly address whether interface-derived peptides can stably engage and influence interfacial regions of tubulin, which constitutes the mechanistic objective of the study.

Comments 7: The fact that the capped hit peptide demonstrated exhibited no inhibitory effect on tubulin polymerization despite showing higher cytotoxicity in NCI-H1299 cancer cells compared to the original hit peptide needs to be further discussed and clarified. Lines 413-415 “These results indicate that substituting the N-terminus with an acetyl group and the C-terminus with a methylamide (NMe) in P14 enhances antiproliferative activity compared to the unmodified version.” Vs Lines 473-475 “In contrast, its terminally protected analogue, Ac-P14-NMe, exhibited no inhibitory effect on tubulin polymerization despite showing higher cytotoxicity in NCI-H1299 cancer cells (IC50 = 22.09 ± 3.50 μM).”

Response 7: We thank the reviewer for this important observation and for highlighting the need for clarification.

We agree that the apparent discrepancy between cytotoxicity and tubulin polymerization inhibition requires further discussion. While Ac-P14-NMe exhibited higher antiproliferative activity in NCI-H1299 cells, it did not inhibit tubulin polymerization in vitro. This indicates that its cytotoxic effect is unlikely to be primarily mediated through direct modulation of tubulin assembly.

In contrast, NH₂-P14-COOH demonstrated a clear correlation between inhibition of tubulin polymerization and cytotoxic activity, supporting a tubulin-dependent mechanism of action. Therefore, we have revised the manuscript to clarify that enhanced antiproliferative activity observed for the capped analogue does not imply improved tubulin targeting.

We now explicitly discuss the possibility that terminal modification may alter physicochemical properties such as membrane permeability, intracellular stability, or off-target interactions, potentially leading to cytotoxic effects through alternative mechanisms. The Discussion section has been revised accordingly to distinguish between general cytotoxicity and mechanistically validated tubulin inhibition.

Comments 8: The results of the molecular dynamics studies are not supportive of the experimental findings. First the most stable peptide from the capped set maintained an RMSD value around 6-8 Å while from the uncapped original set NH2-P14-COOH exhibited the most stable RMSD profile, remaining nearly constant (~5-6 Å) over the entire simulation. In both cases the simulations reflect low degree of stability and contradict the experimental results where the capped analogue was found inactive. The authors need to review this part carefully and put it in context with the experimental results.

Response 8: We thank the reviewer for this careful evaluation.

We respectfully clarify that RMSD magnitude in peptide-protein systems should not be directly interpreted as a measure of binding affinity or functional stability. In short, RMSD reports structural deviation relative to the initial pose, not thermodynamic favorability nor interfacial persistence.

For short, flexible peptides bound to solvent-exposed protein interfaces, RMSD values in the 5-10 Å range are not indicative of instability per se. Instead, they reflect intrinsic peptide flexibility and partial surface adaptation during equilibration. Importantly, RMSD alone does not quantify whether the peptide remains productively engaged at key interfacial hotspots.

In the capped set, although Ac-P7-NMe maintained relatively low RMSD values (~6-8 Å), MM-GBSA analysis revealed unfavorable binding energetics in the extended simulations. Detailed inspection showed that only a localized segment remained associated with tubulin, while the rest of the peptide displayed increased flexibility and weak interfacial engagement. Thus, low RMSD in this context reflects geometrical proximity rather than energetically productive binding.

In contrast, NH2-P14-COOH not only exhibited stable RMSD behavior (~5-6 Å), but also showed the most favorable binding free energy (ΔG = –57.6 kcal/mol), persistent interaction networks, and modulation of interfacial residues. The combined structural, energetic, and dynamic analyses align with its experimentally confirmed inhibition of tubulin polymerization.

We have revised the manuscript to explicitly clarify that RMSD was interpreted together with MM-GBSA energetics, per-residue decomposition, RMSF, and interaction persistence analyses. This integrated evaluation resolves the apparent discrepancy and strengthens the correlation between computational predictions and experimental findings.

Comments 9: Peptides based therapeutics still have its shortcomings including stability and pharmacokinetics, the author should address this point based on similar previous studies.

Response 9: We thank the reviewer for this valuable comment.

We agree that peptide-based therapeutics present well-recognized limitations, including susceptibility to proteolytic degradation, limited metabolic stability, rapid systemic clearance, and, in many cases, suboptimal pharmacokinetic profiles. These aspects represent important translational challenges in the development of peptide-based drugs.

In the revised manuscript, we have expanded the Introduction and Discussion sections to explicitly acknowledge these limitations and to place our findings within the broader context of peptide drug development. In particular, we emphasize that although our study focuses on structure-guided identification and biological validation of tubulin-targeting peptides, further optimization strategies would be required to improve in vivo stability and pharmacokinetic behavior.

Such strategies may include terminal modifications, peptide stapling, backbone cyclization, incorporation of non-natural amino acids, PEGylation, or formulation approaches, which have been successfully applied in previous peptide-based therapeutic developments. 

We have clarified that the present work represents an early-stage discovery and mechanistic study, providing a structural and biological foundation upon which future pharmacokinetic optimization could be built.

Comments 10: Some minor editing such as IC50 should be subscript all throughout the manuscript (IC₅₀)

Response 10: We thank the reviewer for this observation. The formatting has been corrected throughout the manuscript to ensure that IC₅₀ is consistently presented with proper subscript notation.

Reviewer 2 Report

Comments and Suggestions for Authors

This paper is devoted to the design of peptide tubulin inhibitor. It was done by computational methods and verified by experimental techniques. The paper is well-written and quite understandable, and could be recommended for publication.

Questions to authors:

1. The authors took 3J6F structure, it was the first high resolution cryo-EM microtubule structure published in 2014 by Nogales group. Why the authors did not take the newer and better resolved structures from Eva Nogales (6DPV) or Carolina Moores (6EVZ) published in 2018, or even newer structures like 7SJ7 of human tubulin by Nogales (2022)? Would the results of the paper be changed if the authors have chosen another structure?

The choice of the tubulin structure SHOULD BE described in the paper.

2. In the Methods section, there is no information about ionic strength and pH in MD simulations.

3. 50 ns and EVEN 500 ns MD may not be enough to make sure the system is stabilized. In the papers of tubulin MD  the tubulin dynamics is often analyzed AFTER 500 ns MD. Did the authors study only the behaviour of the peptides, without addressing the possible conformational changes in tubulin?

Author Response

Comments 1: The authors took 3J6F structure, it was the first high resolution cryo-EM microtubule structure published in 2014 by Nogales group. Why the authors did not take the newer and better resolved structures from Eva Nogales (6DPV) or Carolina Moores (6EVZ) published in 2018, or even newer structures like 7SJ7 of human tubulin by Nogales (2022)? Would the results of the paper be changed if the authors have chosen another structure?

The choice of the tubulin structure SHOULD BE described in the paper.

Response 1: We thank the reviewer for this important observation.

The selection of PDB ID 3J6F was based on its representation of a complete microtubule lattice with well-defined lateral and longitudinal αβ-tubulin interfaces, which were essential for our interface-driven peptide extraction strategy. In particular, 3J6F provides a structurally consistent dimer-dimer arrangement suitable for identifying contiguous interfacial regions under a distance-based criterion (≤ 6 Å), ensuring geometric continuity and structural completeness at both lateral and longitudinal contacts.

To address the reviewer’s concern regarding potential structural model dependency, we performed a comparative in silico mapping using three additional high-resolution cryo-EM structures: 6DPV, 6EVZ, and 7SJ7. The same interface-based peptide extraction protocol was systematically applied to all structures.

This comparative analysis revealed a strong recurrence of several interface fragments across independent reconstructions. Conserved or highly similar motifs identified in multiple structures include:

• RQLFHPE
• GSQQYRAL
• GQIFRPD / QIFRPD
• EKAYHEQ
• MSMKEVDE
• SETGAGKHV

The recurrence of these motifs across distinct cryo-EM reconstructions, obtained by different groups and at different resolutions, demonstrates that the extraction protocol identifies structurally conserved interfacial regions rather than model-specific artifacts. Importantly, the spatial localization of these segments within the longitudinal and lateral interfaces remains consistent across structures.

Regarding peptide P7 (RFDGALNVDLTEFQTNLVPYP) and peptide P14 (AMFRRKAFLHW), both exhibited consistent structural correspondence across models, although with subunit-dependent variations.

For peptide P7, the fragment extracted in 3J6F corresponds to a longitudinal interface region located in α-tubulin. In 6DPV, 6EVZ, and 7SJ7, a homologous segment was identified in the adjacent β-tubulin subunit, occupying an analogous interfacial position. While sequence boundaries differ slightly between reconstructions, the core interfacial residues and their spatial positioning remain conserved, reflecting preservation of the longitudinal interface architecture rather than structural inconsistency.

A similar pattern was observed for peptide P14, the most active candidate experimentally. In 3J6F, this fragment was extracted from β-tubulin. In the alternative cryo-EM models, analogous aromatic–basic segments were identified in the complementary α-tubulin subunit, maintaining the same interfacial topology and physicochemical character. Although exact sequence identity was not reproduced under identical filtering criteria, the structural hotspot is clearly conserved across reconstructions.

Thus, for both P7 and P14, cross-structure comparison supports conservation of the relevant interfacial hotspots at the structural level. The minor variations observed reflect expected differences in map refinement, conformational state, and segmentation boundaries among cryo-EM reconstructions, rather than instability of the identified regions.

Overall, these analyses indicate that the peptide identification strategy is structurally robust and reproducible across independent high-resolution tubulin models. The selection of 3J6F is therefore fully justified within the context of the present study.

A detailed justification of the tubulin structure selection and the cross-structure validation procedure has now been incorporated into the Methods section.

Comments 2:  In the Methods section, there is no information about ionic strength and pH in MD simulations.

Response 2:   We thank the reviewer for pointing out this omission.

In the molecular dynamics simulations, only counterions (Na⁺ or Cl^-) were added to neutralize the total net charge of each system. The initial αβ-tubulin structure, including bound GTP and GDP molecules, presents a net charge of -36. Therefore, 36 Na⁺ ions were added as a baseline to achieve charge neutrality.

Subsequently, depending on the net charge of each peptide variant (NH₂-PX-COOH or Ac-PX-NMe), the number of counterions was adjusted accordingly. For positively charged peptides, fewer Na⁺ ions were required; for negatively charged systems, additional Na⁺ ions were added to maintain overall neutrality. No additional salt was included beyond neutralizing counterions.

All simulations were performed assuming physiological pH (pH 7.0). Protonation states of titratable residues were assigned accordingly during system preparation.

This information has now been explicitly incorporated into the Materials and Methods section to clarify the simulation conditions.

Comments 3: 50 ns and EVEN 500 ns MD may not be enough to make sure the system is stabilized. In the papers of tubulin MD  the tubulin dynamics is often analyzed AFTER 500 ns MD. Did the authors study only the behaviour of the peptides, without addressing the possible conformational changes in tubulin?

Response 3:  We thank the reviewer for this important and technically relevant comment.

We agree that 50 ns simulations may not be sufficient to fully assess the stabilization of large systems such as the tubulin dimer. In our workflow, the initial 50 ns simulations were intentionally used as a first-level energetic and structural filter, given the total number of systems evaluated (14 peptides x 2 terminal variants = 28 systems, each simulated including 3 conformations, resulting in 72 independent simulations). This strategy allowed us to efficiently reduce the candidate space before committing extended computational resources.

For the six peptides selected after MM-GBSA filtering, we substantially extended the simulations to 1.0 μs per system, which we consider an appropriate timescale to evaluate peptide stability, interfacial persistence, and possible induced conformational effects at the αβ-tubulin interface. These extended trajectories allowed for improved conformational sampling and more reliable energetic analysis.

Importantly, we did not restrict our analysis solely to peptide behavior. To evaluate whether peptide binding influences tubulin dynamics, we also performed 1.0 μs molecular dynamics simulations of the αβ-tubulin dimer in the absence of peptide under identical conditions. Comparative analyses were then carried out, including RMSF profiles of tubulin residues, Principal Component Analysis (PCA) of tubulin collective motions, and Comparative stability and flexibility mapping between apo and peptide-bound systems.

These additional analyses enabled us to assess whether peptide binding induces measurable conformational modulation of tubulin beyond local interfacial stabilization.

We believe that incorporating these extended simulations and comparative tubulin-focused analyses significantly strengthens the manuscript and directly addresses the reviewer’s concern. The corresponding methodological details and results have now been included in the revised version of the manuscript.