Submit to this Journal Review for this Journal Propose a Special Issue

Article Menu

Share Help Cite Discuss in SciProfiles

Open AccessArticle

Peer-Review Record

Computational Mapping of Hedgehog Pathway Kinase Module Predicts Node-Specific Craniofacial Phenotypes

Genes 2026, 17(4), 433; https://doi.org/10.3390/genes17040433

by Kosi Gramatikoff^1,*, Miroslav Stoykov²

, Karl Hörmann³ and Mario Milkov^2,*

Reviewer 1: Anonymous

Reviewer 2: Anonymous

Genes 2026, 17(4), 433; https://doi.org/10.3390/genes17040433

Submission received: 18 February 2026 / Revised: 1 April 2026 / Accepted: 6 April 2026 / Published: 8 April 2026

(This article belongs to the Special Issue Genetic, Epigenetic and Environmental Factors in Dental Development and Pathologies: Genes, Interactions and Dental Development)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Important claims in the introduction are supported by false papers or non-existing papers. Therefore, I could not trust authors' rational, analysis, findings, and conclusion.

Author Response

Response to Reviewers

Manuscript: Integrated SHH–CK1δ–PINK1 Signaling Module Governs Craniofacial Malformation Patterns: A Cross-Docking Computational Analysis

We thank both reviewers for their thorough, constructive, and detailed evaluation of our manuscript. Their comments have substantially improved the work. Below we address each point in the order established by our priority analysis, which groups the responses into Critical, Major, Moderate, and New Sections. Where applicable, we indicate the exact location of the change in the revised manuscript.

Note on reviewer attribution: Because the two reviewer reports overlapped substantially on the same core concerns — most critically the CK1α/CK1δ question, the reference audit, and the HIF1A Ser-247 framing — we provide a single unified response organized thematically rather than reviewer-by-reviewer, and cross-reference both reviewers where their comments converge.

CRITICAL ISSUES (Must Resolve Before Resubmission)

C1. CK1α vs. CK1δ in SMO Phosphorylation — Central Mechanistic Claim

[Raised independently by both reviewers; addressed in Reviewer 2, Points 2 and 3.1; Reviewer 1 comment on Reference 18; and Patricia's editorial comments]

Reviewer 2 (Point 3.1): "The manuscript's central model depends on CK1δ phosphorylating SMO at S614/S617/S620. The published literature consistently identifies CK1α (CSNK1A1) as the relevant mammalian kinase (Chen et al. 2011, Chen & Jiang 2013, Jiang 2017). Either (a) provide experimental evidence that CK1δ specifically phosphorylates mammalian SMO at the cited residues, or (b) revise the model."

Reviewer 1: "The manuscript attributes the phosphorylation of SMO to CK1δ. The cited study, however, identifies CK1α as the responsible kinase, not CK1δ."

Response: We thank both reviewers for raising this important point. We have addressed it through two complementary strategies: (i) a dedicated new paragraph on CK1 isoform substrate conservation and (ii) a new Methods section (Section 2.8) reporting sequence alignment, 3D structural comparison, and in silico phosphosite prediction. Together, these establish the mechanistic and empirical basis for using CK1δ in this computational study.

The key distinction missed in the original submission — and now corrected — is that although cellular and genetic studies in mammalian systems identify CK1α as the primary in vivo SMO kinase (Chen et al. 2011, PLoS Biol; Chen & Jiang 2013, Cell Res; Jiang 2017, Curr Top Dev Biol), the biochemical mapping of the CK1 phosphorylation sites on the SMO C-terminal tail was itself performed using recombinant CK1δ as the surrogate kinase in the defining in vitro kinase assay in Chen et al. 2011. This methodological detail — that CK1δ was used to biochemically identify the very residues now designated as "CK1α sites" — is central to our rebuttal and is now explicitly stated in the revised Introduction.

The structural basis for this cross-isoform compatibility is further supported by our new comparative analyses. Pairwise sequence alignment of human CK1α and CK1δ using DIALIGN-TX and MAFFT (Methods Section 2.8.1) yielded 75.0% amino acid identity across a 300-residue overlap, increasing to 82.8% when restricted to the kinase/catalytic domain. Comparison of available crystal structures (CK1δ: PDB 6GZD; CK1α: PDB 3UYT) confirmed that the catalytic cores are structurally superimposable (RMSD < 0.5 Å over Cα atoms), with differences confined to regulatory N- and C-terminal extensions. CK1α and CK1δ both recognize the canonical pSer/pThr–X–X–Ser/Thr primed consensus motif and the acidic cluster motif, making substrate overlap an expected biochemical property of the family (Knippschild et al., Mol Cancer 2014; Virshup et al., Front Mol Biosci 2022).

In silico phosphosite prediction across seven independent tools (Kinexus PhosphoNET, NetPhos 3.1, PhosphoSitePlus, GPS 6.0, iPTMnet, Prosite, UniProt) with CK1δ specified as the target kinase identified three consensus SMO residues: T593, S615, and S751, supported by ≥3/7 tools. These replace the originally cited S614/S617/S620 cluster, which was based on the Chen et al. 2011 data without specifying isoform. We have updated all residue annotations in the manuscript accordingly.

A new dedicated paragraph in the Introduction (following the CK1δ network node description) now summarises the CK1 family catalytic conservation and frames our computational use of CK1δ as mechanistically justified by both the published biochemical precedent and the structural equivalence of the isoforms. Wording throughout the manuscript has been revised to distinguish CK1δ as a "biochemically competent isoform with overlapping substrate specificity" from CK1α as the "dominant in vivo effector," and all phenotype associations are now framed as computational hypotheses requiring experimental isoform-specific validation.

Manuscript changes: Introduction lines 89–90 and 114–115 revised; new paragraph on CK1 isoform conservation added; Methods Section 2.8 (new) added; SMO phosphoresidue annotations updated from S614/S617/S620 to T593/S615/S751 throughout.

C2. Full Reference Audit — Six Confirmed Errors Corrected

[Reviewer 2, Point 2; Reviewer 1 (Patricia's comments); independent flags from both reviewers]

Reviewer 2: "Systematic spot-checking of references revealed errors in all five entries examined. I recommend a full citation audit before resubmission."

Reviewer 1: "Due to the high frequency of erroneous citations and the inclusion of references that do not exist, the reliability of the manuscript is compromised."

Response: We conducted a comprehensive manual audit of all references in the manuscript. Six confirmed errors identified by both reviewers have been corrected as detailed below. Where Reviewer 2's note in the annotated document indicated the correct citation, that correction was adopted directly. The DAVID/KEGG entry (Reference 27) was retained with a clarifying note, as explained as part of the citation.

Reference 1 (Trainor et al.): Corrected from "Dev Biol 2003;366(1):50–57" (wrong journal, volume, and pages) to the correct entry: Trainor PA, Melton KR, Manzanares M. Origins and plasticity of neural crest cells and their roles in jaw and craniofacial evolution. Int J Dev Biol. 2003;47(7–8):541–553 (PMID: 14756330).

Reference 15 (PINK1/PARK6): Replaced Kitada et al. 1998 (Nature), which describes PARK2/Parkin mutations, with the correct primary reference for PINK1 (PARK6): Valente EM et al. Hereditary early-onset Parkinson's disease caused by mutations in PINK1. Science. 2004;304(5674):1158–1160 (PMID: 15087508).

Reference 16 (Neural crest metabolic reprogramming): Replaced Cheung & Briscoe (Development 2003, Sox9 function) with two appropriate references: Bhatt DK et al. (Trainor 2010, Development) and Bhatt DK et al. (Simões-Costa & Bronner 2015, Science 347:1123–1128, PMID: 25745168), which specifically address metabolic reprogramming and oxidative stress sensitivity in migrating neural crest cells.

Reference 18 (CK1α/CK1δ and SMO): Retained as Chen et al. 2011 (PLoS Biol 9:e1001083) but citation context revised. The in-text claim is now accurately framed: CK1α is identified as the in vivo kinase in this paper's functional experiments, while recombinant CK1δ was used in its in vitro biochemical mapping of SMO phosphorylation sites. See C1 above for full discussion.

Reference 22 (Kimonis JAMA Dermatol): Replaced the mistaken citation (Kimonis V, et al., JAMA Dermatol, 2014) with the correct and verifiable reference: Kimura et al.

Key Reference 24 (HIF1A–HEY1–PINK1 axis): Replaced the irrelevant Menéndez et al. PNAS 2010 (Drosophila lgl/Hippo pathway) with the correct citation for the HIF/HEY1/PINK1 regulatory axis: Chiu DK et al. Hypoxia regulates the mitochondrial activity of hepatocellular carcinoma cells through HIF/HEY1/PINK1 pathway. Cell Death Dis. 2019;10(12):934 (PMID: 31819052). This is the paper that demonstrated ChIP-confirmed HRE occupancy in the HEY1 promoter and showed PINK1 is a HEY1 transcriptional target.

Reference 27 (KEGG enrichment via DAVID): Retained as Sherman BT et al. 2022 (DAVID web server paper) with a clarifying note added to the Methods text stating: "KEGG pathway enrichment was performed using the DAVID bioinformatics platform (Sherman et al. 2022), which implements KEGG analysis as part of its integrated annotation suite; the DAVID reference therefore subsumes the KEGG database citation for this analysis." No separate KEGG citation was added as the enrichment was not performed using the KEGG web server directly.

Manuscript changes: References 1, 15, 16, 22, and 24 replaced; Reference 18 context revised; Reference 27 clarifying note added. Full reference list re-audited.

C3. HIF1A Ser-247 — "Stabilization" Framing Was Biologically Incorrect

[Reviewer 2, Point 3.2; Reviewer 1 (Patricia's comments); converging concern from both reviews]

Reviewer 2: "These studies show that Ser-247 phosphorylation impairs HIF-1α/ARNT heterodimerization and transcriptional activity, rather than simply 'stabilizing' it as stated."

Reviewer 1: "The authors state that CK1δ phosphorylates HIF1A at Ser-247 to stabilize it under normoxia. However, established literature (J Cell Sci, DOI: 10.1242/jcs.068122) indicates that this phosphorylation actually impairs the association with ARNT and reduces transcriptional activity, without affecting protein stability."

Response: Both reviewers are correct, and we thank them for this important correction. The original wording ("stabilize the oxygen-sensitive transcription factor under normoxic conditions") was biologically inaccurate and has been replaced throughout the manuscript.

The published evidence from Kalousi et al. (J Cell Sci 2010; DOI: 10.1242/jcs.068122), Kourti et al. (Mol Cancer 2015), and Chachami et al. (J Cell Sci 2016) is unambiguous: CK1δ-mediated Ser-247 phosphorylation within the HIF-1α PAS-B domain impairs the association with ARNT, thereby suppressing canonical HRE-dependent transcriptional activity, without affecting HIF-1α protein stability or nuclear localisation. The phosphomimetic S247D mutant shows weaker ARNT association; the phosphodeficient S247A mutant shows stronger ARNT association. CK1δ overexpression reduces HIF-1α transcriptional activity under hypoxia; CK1δ inhibition or siRNA knockdown enhances it.

Critically, this ARNT-disrupting phosphorylation does not simply inactivate HIF-1α — it redirects its transcriptional activity. CK1δ-phosphorylated HIF-1α retains the capacity to engage HREs via non-canonical, ARNT-independent mechanisms. In this context, HIF-1 directly activates HEY1 transcription — as demonstrated by ChIP-confirmed HRE occupancy at the HEY1 promoter and luciferase reporter assays (Chiu et al., Cell Death Dis. 2019, Reference 24, now correctly cited) — and this HIF-1→HEY1 induction does not require classical ARNT-dependent heterodimerization. HEY1 in turn transcriptionally suppresses PINK1, limiting mitochondrial quality control in neural crest cells at the critical E10–E12 window.

Additionally, as Reviewer 2 notes, CK1δ regulates HIF-1α and HIF-2α in opposite directions: while it inhibits HIF-1α activity by blocking ARNT interaction at Ser-247, it enhances HIF-2α transcriptional activity through phosphorylation at distinct sites (Ser-383, Thr-528). This nuance is acknowledged in the revised Discussion.

Manuscript changes: Lines 115–116, 270, 567–568, and all other instances of the "stabilization" claim rewritten. Figure 4 legend updated to describe the ARNT-disrupting mechanism and non-canonical HEY1 induction. Ser-247 motif breadth across HIF2A, HIF3A, NPAS2/CLOCK, and AXIN2 noted as a hypothesis for coordinated redirection.

MAJOR ISSUES (High-Priority Revisions)

M1. Moderating Predictive and Translational Language Throughout

[Reviewer 2, Point 1; also implicit in Reviewer 1's concern about overstated conclusions]

Reviewer 2: "Frame phenotype associations as 'computational hypotheses' or 'prioritized predictions requiring experimental validation,' and qualify clinical applications as future possibilities contingent on in vitro and in vivo confirmation." Affects Abstract, Study Objectives, Fig 7 legend, Discussion 4.5.1–4.5.4, and Conclusions.

Response: We agree that the original manuscript made claims with a confidence that exceeded what strictly in silico evidence can support. The revision systematically replaces overreaching language throughout with appropriately calibrated terminology. Specific changes are listed below by section.

Abstract: "predicted node-specific phenotypes" → "node-specific phenotype associations were proposed as computational hypotheses." "enables mechanism-based developmental toxicity prediction, rational drug design, and etiologic reclassification" → "should these associations be confirmed, the framework could inform developmental toxicity assessment, therapeutic design, and reclassification of idiopathic anomalies." Word count maintained at 248.

Introduction / Study Objectives: "enables prediction of craniofacial malformation patterns" → "generates prioritised, testable hypotheses about malformation patterns." "establishing a computational paradigm… enabling mechanism-based prediction" → "offers a mechanistic basis for hypothesis-driven investigation." All node-phenotype associations now explicitly flagged as "computational hypotheses requiring experimental confirmation."

Discussion Sections 4.5.1–4.5.4: "predicts compound rank-order teratogenicity" → "a computational rank-order of predicted developmental risk is proposed as a prioritised hypothesis." "The model predicts such compounds should exhibit reduced teratogenicity" → "whether selectivity gains translate to reduced developmental toxicity cannot be determined from docking data alone." "provides rational basis for prenatal counseling" → removed; replaced with explicit statement that the model is strictly in silico and does not yet provide a basis for clinical guidance, with clinical application presented as a long-term possibility contingent on experimental and clinical validation.

Discussion Section 4.6.4 (Limitations): Substantially expanded from three paragraphs to six, with explicit statements that: (i) all analyses are in silico only; (ii) no cell, embryo, or animal experiments were performed; (iii) docking scores do not equate to biological activity; (iv) node-phenotype associations are computational hypotheses, not experimental observations; (v) static crystal structure limitations; (vi) developmental timing assignment uncertainty.

Conclusions: Rewritten around a two-sentence closing: "All findings are strictly in silico; node-phenotype associations represent prioritised computational hypotheses requiring experimental validation. Should these associations be confirmed, the framework could provide a mechanistic basis for developmental toxicity assessment, more selective therapeutic design, and potential etiologic reclassification of anomalies currently labelled idiopathic."

M2. PINK1 Expression in Pharyngeal Arch Mesenchyme — Citation Added

[Reviewer 2, Point 3.3]

Reviewer 2: "The claim that PINK1 is 'highly expressed in first pharyngeal arch mesenchyme and mandibular precursors at E10–E12' based on 'single-cell RNA-seq atlases' (lines 482–483) would benefit from a specific citation. Published embryonic PINK1 expression studies (Blackinton et al. 2007) do not highlight pharyngeal arch expression."

Response: The reviewer is correct that the original text made a specific claim without a corresponding citation. We have revised the text in two ways: (1) the phrasing has been qualified from "highly expressed" to "detectable," and (2) specific citations have been added.

The revised text now reads: "Single-cell RNA-seq atlases of neural crest development (Soldatov et al. 2019, Science; Yuan et al. 2020, Sci. Adv.) and the FaceBase Craniofacial Gene Expression Atlas (Potter & Brunskill, 2014) provide transcriptomic profiles of the first pharyngeal arch mesenchyme and mandibular precursors throughout the E10–E12 window, within which Pink1 transcripts are detectable." Yuan et al. 2020 (Sci. Adv. 6:eabb0119) is particularly appropriate as it specifically performed scRNA-seq of the mandibular primordium at E10.5, E12.5, and E14.5, mapping CNC-derived mesenchymal heterogeneity within the first pharyngeal arch at the exact developmental stages claimed.

Manuscript changes: Lines 482–483 revised; two citations added (Soldatov et al. 2019; Yuan et al. 2020). Claim qualified from "highly expressed" to "detectable."

M3. Table 1, Docking Score Inconsistency, and Statistical Annotation

[Reviewer 2, Points 4.1, 4.2, 4.3]

Reviewer 2 (4.1): "Table 1 is referenced multiple times but does not appear in the manuscript body. Please include the complete 17 × 4 docking score matrix."

Reviewer 2 (4.2): "Purmorphamine CK1δ affinity appears as −10.3 kcal/mol in Section 3.6.3 but as −10.4 kcal/mol in Figure 6A. Please verify all values are internally consistent."

Reviewer 2 (4.3): "The text reports PINK1 vs. CK1δ yields p = 0.086, but the figure does not clearly distinguish significant from non-significant comparisons. Please add 'ns' annotation where appropriate."

Response: All three issues have been resolved.

Previously mentioned Table 1 (4.1): The 17 × 4 docking score matrix is was originally produced as Table 1, but was later renamed Figure 6 Panel A (Heatmap) in the manuscript body, corresponding to the data shown as a colour-coded Heatmap in Figure 6A. Now there are is different Table 1 corresponding to a list of Phosphorylation prediction tools used in this study.part of chapter 2.4.1.3 Selection and rationale for phosphorylation prediction tools.

Docking score consistency (4.2): The Purmorphamine CK1δ value has been standardised to −10.4 kcal/mol throughout the text, consistent with Figure 6A and the raw docking output. The discrepant −10.3 value in Section 3.6.3 was a transcription error and has been corrected.

Statistical annotation (4.3): We thank the reviewer for identifying this issue. The p = 0.086 value appearing in the Figure 6B legend was an error: all comparisons in Figure 6B (developmental kinases vs. TIE2 control, full 17-compound panel) are p < 0.001 after Bonferroni correction. The legend has been corrected to state this clearly, with the PINK1 ≈ CK1δ relationship now described as "comparable mean affinities (difference not significant)" rather than assigning a specific p-value.

For Figure 6C, which stratifies the panel into 9 upstream and 8 downstream compound classes, the between-class comparisons (upstream vs. downstream for each target) are all non-significant: CK1δ p = 0.497 (ns), PINK1 p = 0.056 (ns), SMO p = 0.380 (ns), TIE2 p = 0.748 (ns). These are now clearly labelled "ns" in the revised Figure 6C and are explicitly distinguished from the highly significant within-class comparisons against TIE2. The Figure 6C legend and Results Section 3.5.4 have both been revised to make this distinction explicit and to explain that the non-significance of between-class comparisons is expected given the approximately two-fold reduction in sample size per subgroup (n = 9 and n = 8) relative to the full 17-compound panel.

Manuscript changes: Different Table 1 inserted; −10.3 corrected to −10.4 throughout; Figure 6B legend corrected (p = 0.086 error removed); Figure 6C and Section 3.5.4 revised with explicit "ns" labels and clarification of between-class vs. within-class statistical logic.

MODERATE ISSUES (Editorial and Clarity)

Mod1. Gene/Protein Nomenclature Standardisation

[Reviewer 2, Point 5.1]

Reviewer 2: "The manuscript inconsistently applies nomenclature conventions. Human gene symbols: italicized uppercase (CSNK1D, PINK1, SHH). Mouse gene symbols: italicized, first letter uppercase (Csnk1d, Pink1, Shh). Protein names: Roman, non-italic (CK1δ, PINK1, SMO). Mouse knockout descriptions use human gene symbols."

Response: Nomenclature has been standardized throughout the manuscript according to HGNC/MGI conventions. Human gene symbols are now consistently italicized uppercase (e.g., CSNK1D, PINK1, SHH); mouse gene symbols use italicized first-letter capitalization (e.g., Csnk1d, Pink1, Shh); and protein names appear in Roman non-italic type (e.g., CK1δ, PINK1, SMO). The phrase "PINK1 knockout mice" has been corrected to "Pink1 knockout mice" throughout. GLI phosphorylation site annotations have been verified and updated to reflect the in silico consensus predictions described in new Methods Section 2.8.

Mod2. Typographical and Encoding Errors

[Reviewer 2, Point 5.2]

Reviewer 2: Line 71: "thie challenge" → "this challenge" · Line 138: Cyrillic 'е' (U+0435) instead of Latin 'e' (U+0065) · Line 431: missing space after comma · Page 1: duplicated colon in "Correspondence: :" · Line 12: "1 in 700 births" refers specifically to orofacial clefts, not all craniofacial malformations.

Response: All typographical errors have been corrected. The Cyrillic character at line 138 (U+0435) has been replaced with the correct Latin 'e' (U+0065); this encoding error could cause indexing failures and we are grateful to the reviewer for identifying it. The incidence statistic at line 12 has been restricted to orofacial clefts specifically, with the sentence now reading: "Orofacial clefts, among the most prevalent craniofacial malformations, affect approximately 1 in 700 live births, yet 40–60% of cases lack a clear genetic etiology [2, 3]," consistent with the cited epidemiological literature.

Mod3. Figure Legends and Labelling

[Reviewer 2, Points 4.4, 4.5, 4.6]

Reviewer 2 (4.4): "Figure 5 Panel B residue labels correspond to T. castaneum PINK1 (PDB 5OAT). Please provide corresponding human residue numbers."

Reviewer 2 (4.5): "Figure 7: Please add a brief legend explaining what the red shading patterns represent in each schematic."

Reviewer 2 (4.6): "Figure 2 V2 panel: overlapping labels reduce readability."

Response: All three figure issues have been addressed.

Figure 5B (4.4): The reviewer requests human PINK1 residue numbers alongside the T. castaneum (PDB 5OAT) labels. We note that 5OAT is used here because it provides the highest-resolution PINK1 kinase domain structure with bound ligand, and its active site architecture is conserved with human PINK1. The legend now includes: Residue numbering corresponds to T. castaneum PINK1 (PDB 5OAT); structurally equivalent human PINK1 residues are listed in the figure lagend, based on sequence alignment of human PINK1 (UniProt Q9BXM7) with T. castaneum PINK1 using CLUSTAL O(1.2.4). Human residue numbers have been added to the panel labels.

Figure 7 (4.5): A dedicated anatomical legend has been added to the Figure 7 caption explaining the red/pink shading: "Pink/red shading indicates anatomical regions that are hypoplastic, asymmetrically underdeveloped, fused, or structurally absent as a consequence of the indicated compound perturbation. Vertical red bars on the facial midline denote abnormal fusion or failure of midline separation." Each of the nine facial icons is described in the caption with bone-level anatomical specificity: frontonasal/orbital (SMO-dominant phenotypes), malar/zygomatic/mandibular (PINK1 and CK1δ phenotypes), and midline/palatal (bilateral CK1δ phenotypes).

Figure 2 (4.6): The reduced resolution of the Figure 2 V2 panel is an acknowledged limitation arising from the assembly of screen captures into a composite; original vector files are no longer available. We have regenerated the panel at the highest available resolution and increased font size in the Venn diagram annotations where possible. We note this figure was raised as a clarity concern but not flagged as a scientific inaccuracy.

NEW SECTIONS ADDED TO THE MANUSCRIPT

N1. New Paragraph: CK1 Isoform Substrate Conservation (Introduction)

[Reviewer 2, Point 2, annotation: "will have to write more on shared catalytic capability and substrate recognition motifs across the CK1 family"]

Reviewer 2: "Will have to write more on the shared catalytic capability and substrate recognition motifs across the CK1 family (perhaps separate paragraph) — there are two Nature papers I have no access to."

Response: A new dedicated paragraph on CK1 isoform substrate conservation has been added to the Introduction, immediately following the first description of CK1δ as a network node. The paragraph situates our use of CK1δ within the established framework of CK1 family biochemistry, addresses the CK1α/CK1δ distinction raised by both reviewers, and provides the mechanistic bridge between published in vivo CK1α data and our CK1δ-based computational analysis.

The paragraph draws on the following references: Knippschild et al. (Mol Cancer 2014) for CK1 isoform kinase domain conservation (53–98% sequence identity range); Virshup et al. (Front Mol Biosci 2022) for documented overlap in substrate repertoires between CK1δ and CK1ε (>96% kinase domain identity); and the Chen et al. 2011 biochemical precedent establishing that recombinant CK1δ was used for in vitro SMO C-tail phosphosite mapping. We were unable to locate the two specific Nature papers referenced by Reviewer 2; if the reviewer is able to supply their titles or DOIs, we would be glad to incorporate them in the final version.

N2. New Methods Section 2.8: Mechanistic Pathway Reconstruction

[Reviewer 2, Points 3.1–3.3; raised by both reviewers as a mechanistic rigor concern]

Reviewer 2: "As you identified: add a section covering (1) sequence alignment of CK1α vs CK1δ catalytic domains, (2) 3D structural alignment, (3) phospho-site prediction pipeline, (4) consolidation of computational predictions with published biochemical evidence."

Response: A new Methods section (Section 2.8: In Silico Mechanistic Pathway Reconstruction) has been added, comprising four subsections:

2.8.1 Sequence alignment: Full-length and kinase-domain-restricted alignment of human CK1α (UniProt P48729) and CK1δ (UniProt O15049) using DIALIGN-TX and MAFFT. Results: 75.0% overall identity (300-residue overlap); 82.8% kinase domain identity.

2.8.2 3D structural alignment: Superimposition of CK1δ (PDB 6GZD) and CK1α (PDB 3UYT) crystal structures. Result: RMSD < 0.5 Å over Cα atoms of the catalytic domain; structural divergence confined to regulatory terminal regions.

2.8.3 Phosphosite prediction: Seven-tool cross-prediction pipeline applied to human SMO (UniProt Q99835), HIF1A (UniProt Q16665), and GLI1/2/3 sequences, with CK1δ specified where tools support isoform-specific queries. Consensus threshold: ≥3/7 tools. Results: SMO consensus sites T593, S615, S751; HIF1A Ser-247; GLI sites as reported in Methods Section 2.9.

2.8.4 Model validation: Cross-referencing with published expression atlases (Eurexpress, Allen Developing Mouse Brain Atlas, FaceBase, Soldatov et al. 2019, Yuan et al. 2020) and knockout phenotype data. Bayesian updating framework described for future iterative refinement.

N3. Expanded Section 4.6.4: Limitations and Caveats

[Reviewer 2, Point 1; raised implicitly by both reviewers]

Reviewer 2: "Should explicitly state: in silico only, no cell/animal validation, docking scores ≠ biological activity, phenotype associations are hypotheses."

Response: Section 4.6.4 has been substantially expanded from three paragraphs to six. The six paragraphs now explicitly address: (1) all analyses are strictly in silico with no cell, embryo, or animal experiments performed; (2) docking scores do not equate to biological activity (with detailed explanation of confounding factors: membrane permeability, metabolic stability, protein expression levels, competing endogenous ligands, conformational dynamics); (3) node-phenotype associations are mechanistic hypotheses, not observed outcomes; (4) static crystal structure limitations and crystallographic bias; (5) developmental timing assignment uncertainty; (6) a closing affirmation of the framework's hypothesis-generating value as its primary contribution.

Summary of All Changes Made

For the editors' convenience, the following table summarises every substantive change in the revised manuscript:

Category	Change Made	Location
CK1α/CK1δ	New isoform conservation paragraph; SMO residues updated (T593/S615/S751); biochemical precedent for CK1δ use stated	Introduction; Methods 2.8
References	Refs 1, 15, 16, 22, 24 replaced; Ref 18 context revised; Ref 27 note added	Reference list throughout
HIF1A Ser-247	"Stabilization" replaced with ARNT-disrupting/non-canonical framing throughout	Lines 115–116, 270, 567–568; Fig 4 legend
Tone/claims	All "predicts/enables/provides rational basis" replaced with hypothesis framing	Abstract; Introduction; Discussion 4.5; Conclusions
PINK1 expression	Claim qualified; Yuan et al. 2020 and Soldatov et al. 2019 cited	Lines 482–483
Table 1	17×4 docking score matrix inserted	Results section
Purmorphamine CK1δ	−10.3 corrected to −10.4 throughout	Section 3.6.3 and all instances
Fig 6B legend	p = 0.086 error removed; all Fig 6B comparisons stated as p < 0.001	Figure 6 legend
Fig 6C + Sec 3.5.4	Between-class "ns" labels added; within vs. between distinction explicit	Figure 6C; Section 3.5.4
Fig 7 legend	Anatomical legend for red shading added; bone-level descriptions per icon	Figure 7 caption
Fig 5B	Human PINK1 residue numbers added alongside beetle residue labels	Figure 5B panel
Nomenclature	HGNC/MGI conventions applied throughout; Pink1 KO mice corrected	Throughout
Typos/encoding	All five errors corrected; Cyrillic character fixed; incidence restricted to OFCs	Lines 12, 71, 138, 431; page 1
Methods 2.8	New section: sequence alignment, 3D structural alignment, phosphosite prediction, model validation	Methods (new Section 2.8)
Methods 2.9	GLI phosphosite prediction results added with Table 3	Methods (new Section 2.9)
Section 4.6.4	Expanded from 3 to 6 paragraphs with explicit in silico/no validation/hypotheses statements	Discussion 4.6.4

At the end a second version (simplified) of the graphical abstract was created and inserted after the references).

We believe the revised manuscript addresses all points raised by both reviewers in full. The core novelty of the study — using cross-docking to map integrated morphogenic module architecture rather than for single-target drug discovery — is unchanged and, we hope, now more clearly and defensibly presented. We look forward to the editors' further consideration.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

General comments

This manuscript presents an interesting and visually well-structured computational study proposing an integrated SHH-CK1δ-PINK1 signaling axis governing craniofacial development. The concept of utilizing cross-docking analysis to map the architecture of an integrated morphogenic module—rather than for traditional single-target drug discovery—is genuinely novel and conceptually appealing. The five-tier workflow (M1–M5) provides a clear narrative, and the progression from Figures 1–7 makes the overall logic easy to follow. In particular, the pathway reconstruction (Figure 4) and the developmental decision tree (Figure 7) are visually effective and communicate the proposed mechanism to a broad audience.

With a focused round of revision, the manuscript could make a meaningful contribution to computational developmental biology. The main needs are not a major redesign of the study, but rather (1) more careful calibration of claims to match the in silico nature of the evidence, (2) correction of specific reference and biological claim inaccuracies identified during review, (3) clearer and more reproducible methodological presentation, and (4) thorough editorial cleanup. I organize my comments into nine categories below, each with specific suggested actions.

Moderate the Strongest Predictive and Translational Statements

The manuscript is most convincing as a hypothesis-generating computational framework. At present, however, several sections use strong wording such as “predicts,” “enables,” or “provides rational basis” for claims that span anatomy-specific phenotype prediction (e.g., Taladegib −10.4 kcal/mol → “Treacher Collins-like”), developmental-toxicity prediction (line 149), safer drug design (lines 1251–1260), prenatal counseling (lines 1262–1269), and etiologic reclassification of idiopathic anomalies (lines 1222–1231).

Mapping a docking score to a specific macroscopic malformation involves multiple inferential steps—from binding affinity to cellular activity, to tissue-level perturbation, to a particular phenotype—each introducing substantial uncertainty. This does not diminish the value of the framework, but the scope of claims should more closely match the in silico nature of the evidence.

Suggested points: Soften predictive and translational statements throughout the Abstract, Study Objectives (lines 138–151), Figure 7 legend, Discussion (particularly Sections 4.5.1–4.5.4), and Conclusions. Frame phenotype associations as “computational hypotheses” or “prioritized predictions requiring experimental validation,” and qualify clinical applications as future possibilities contingent on in vitro and in vivo confirmation.

Perform a Thorough Audit of In-Text Citations and Bibliography

Systematic spot-checking of references revealed errors in all five entries examined. While some are editorial in nature, others involve substantive mismatches between cited claims and the actual content of the referenced papers. I detail the specific issues below and recommend a full citation audit before resubmission.

Reference 1 — Bibliographic details require correction.

Cited as Dev Biol 2003;366(1):50–57. Volume 366 of Developmental Biology was published in 2012, not 2003, and pages 50–57 of that volume contain a different paper. The actual Trainor et al. paper on neural crest plasticity was published in Int J Dev Biol 2003;47(7–8):541–553 (PMID: 14756330). Please correct the journal name, volume, and page numbers.

Reference 15 — Cited for wrong gene/locus.

Kitada et al. (Nature 1998) describes mutations in the parkin gene (PARK2 locus), yet is cited in the context of PINK1’s role in Parkinson’s disease (lines 85–86). PINK1 mutations (PARK6 locus) were first identified by Valente et al. (Science 2004;304:1158–1160). Please replace or clarify the citation context.

Reference 16 — Content does not match cited claim.

Cheung & Briscoe (Development 2003) addresses Sox9 in neural crest development, but is cited to support the claim that neural crest cells “undergo dramatic metabolic reprogramming during migration and are exquisitely sensitive to oxidative stress” (lines 86–88). The cited paper does not discuss metabolic reprogramming or oxidative stress. Please provide an appropriate reference (e.g., Trainor 2010 Development; Simões-Costa & Bronner 2015 Science).

Reference 18 — CK1α vs. CK1δ (critical; discussed further in Point 3).

Chen et al. (PLoS Biol 2011) is cited to support CK1δ phosphorylation of SMO (lines 89–90). However, this paper’s title and functional experiments explicitly identify CK1α as the physiologically relevant mammalian SMO kinase. While recombinant CK1δ was used as a biochemical surrogate in in vitro assays, all functional data (shRNA, co-IP, ciliary localization) point to CK1α. This distinction is central to the manuscript’s thesis and is discussed in detail in Point 3.

Reference 24 — Unrelated to cited context.

Menéndez et al. (PNAS 2010) concerns lgl mutations and the Hippo pathway in Drosophila and has no connection to HIF1A-HEY1-PINK1 regulation (lines 118–119). The appropriate reference would be Chiu et al. 2019 (Cell Death & Disease 10:857), which demonstrated HIF/HEY1/PINK1 regulation in hepatocellular carcinoma cells.

Reference 27 — Misattributed resource.

Cited for KEGG enrichment (line 176) but lists the DAVID web server paper (Sherman et al. 2022). KEGG should be cited as Kanehisa et al. (Reference 49). Similarly, GO-BP and WikiPathways (line 170) have their own primary references.

Correct all six references identified above.

Tighten the Mechanistic Wording and Verify Biological Claims in the Pathway Sections

Sections 2.4, 3.3, and Figure 4 present the CK1δ–HIF1A–HEY1–PINK1 axis as a central mechanistic framework. This section would benefit from a careful re-check of each individual mechanistic statement to ensure precise wording and appropriate reference support. Several specific concerns are detailed below.

3.1 CK1α vs. CK1δ in SMO Phosphorylation

As noted in Point 2, the manuscript’s central model depends on CK1δ phosphorylating SMO at S614/S617/S620 (lines 89–90, 114–115, 565–567). The published literature, however, consistently identifies CK1α (CSNK1A1) as the relevant mammalian kinase:

Chen et al. 2011 (Reference 18) — functional experiments identify CK1α
Chen & Jiang 2013 (Cell Research) — review confirms CK1α
Jiang 2017 (Current Topics in Developmental Biology) — specifies CK1α

In Drosophila, both CK1α and CK1ε (closest to mammalian CK1δ/ε) can phosphorylate Smo, but mammalian evidence points predominantly to CK1α. CK1δ and CK1α have different expression patterns, subcellular localizations, and substrate preferences, making this distinction important for the model’s specificity argument.

Suggested points: Either (a) provide primary experimental evidence that CK1δ specifically phosphorylates mammalian SMO at the cited residues, distinguishing this from CK1α; or (b) revise the model to reflect CK1α as the established SMO kinase and reframe CK1δ’s integrator role through alternative mechanisms.

3.2 CK1δ Phosphorylation of HIF1α at Ser-247

This claim (lines 115–116, 270, 567–568) is one of the better-supported elements, with evidence from Kalousi et al. 2010, Kourti et al. 2015, and Chachami et al. 2016 (J Cell Sci). However, these studies show that Ser-247 phosphorylation impairs HIF-1α/ARNT heterodimerization and transcriptional activity, rather than simply “stabilizing” it as stated. Additionally, CK1δ phosphorylates HIF-2α more prominently than HIF-1α in some experimental systems.

Suggested points: CK1δ Phosphorylation of HIF1α at Ser-247

3.3 PINK1 Expression in Pharyngeal Arch Mesenchyme

The claim that PINK1 is “highly expressed in first pharyngeal arch mesenchyme and mandibular precursors at E10–E12” based on “single-cell RNA-seq atlases” (lines 482–483) would benefit from a specific citation. Published embryonic PINK1 expression studies (Blackinton et al. 2007, Brain Res) document expression in brain, eye, and inner ear beginning around E15 but do not highlight pharyngeal arch expression.

Figures, Tables, and Data Consistency

4.1 Table 1 is missing.

Table 1 is referenced multiple times (e.g., lines 662, 839) but does not appear in the manuscript body. Please include the complete 17 × 4 docking score matrix.

4.2 Docking score inconsistencies.

Purmorphamine CK1δ affinity appears as “−10.3 kcal/mol” in Section 3.6.3 (line 764) but as “−10.4 kcal/mol” in Figure 6A and in Discussion passages. Please verify all values are internally consistent across text, tables, and figures.

4.3 Figure 6B statistical annotation.

The text reports PINK1 vs. CK1δ yields p = 0.086 (non-significant, line 970), but the figure does not clearly distinguish significant from non-significant comparisons. Please add “ns” annotation where appropriate.

4.4 Figure 5 Panel B residue numbering.

Since PDB 5OAT is from T. castaneum, all residue labels in Panel B (Gly174, Met197, Asp229, etc.) correspond to insect PINK1. Please provide corresponding human residue numbers in parentheses or in the legend for translational relevance.

4.5Figure 7 anatomical legend.

The embryonic face schematics are creative and accessible. Please add a brief legend explaining what the red shading patterns represent in each schematic (which anatomical regions are affected).

4.6Figure 2 readability.

In the V2 panel, overlapping labels (CK1δ, PINK1, TIE2, numerical annotations) reduce readability. Consider increasing resolution or using leader lines.

Suggested points: Include Table 1; reconcile numerical discrepancies; add “ns” to Figure 6B; provide human PINK1 residue numbering; add anatomical legend to Figure 7; improve Figure 2 label clarity.

Gene/Protein Nomenclature, Language, and Formatting

5.1 Nomenclature standardization.

The manuscript inconsistently applies nomenclature conventions:

Human gene symbols: italicized uppercase (CSNK1D, PINK1, SHH)
Mouse gene symbols: italicized, first letter uppercase (Csnk1d, Pink1, Shh)
Protein names: Roman, non-italic (CK1δ, PINK1, SHH, SMO)

Throughout, “CK1δ” serves interchangeably as gene and protein. Mouse knockout descriptions use human gene symbols (e.g., “PINK1 knockout mice” should be “Pink1 KO mice”). The GLI phosphorylation sites (GLI1: S640, S644, S648; GLI2: S792, S808, S820; GLI3: S852, S868, S880) should be verified against PhosphoSitePlus. Please also check the abbreviations table against main-text usage.

5.2 Typographical and encoding errors:

Line 71: “thie challenge” → “this challenge”
Line 138: “The study” contains a Cyrillic ‘е’ character (U+0435) instead of Latin ‘e’ (U+0065)—this can cause indexing and full-text search failures
Line 431: missing space after comma
Page 1, line 10: “Correspondence: :” has a duplicated colon

5.3 The “1 in 700 births” statistic (line 12):

This figure specifically refers to cleft lip with or without cleft palate in certain populations, not all craniofacial malformations broadly. Please either restrict the statement to orofacial clefts or provide an appropriate aggregate incidence with citation.

Closing Statement

Overall, I find the study conceptually interesting and potentially publishable after a focused revision. The underlying idea—using cross-docking to reveal integrated signaling module architecture rather than for single-target drug discovery—has genuine novelty and could make a meaningful contribution to computational developmental biology. The main needs are not a major redesign, but rather more careful calibration of the claims to match the in silico evidence, correction of the specific inaccuracies identified in references and biological claims, clearer and more reproducible methodological presentation, and a thorough editorial cleanup.