TOR Signaling as a Central Integrator of Embryogenic Reprogramming During 2,4-D-Induced Somatic Embryogenesis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Overall evaluation: This paper focuses on the 2,4-D-induced somatic embryogenesis (SE) in Arabidopsis thaliana and uses public transcriptome data combined with STRING high-confidence protein interaction networks for analysis. It proposes that TOR is the core integration hub for somatic embryo reprogramming, and divides it into three regulatory axes: TOR-FKBP12/RPS6A, TOR-CBP20, and TOR-TAP46. Main contributions: It constructs a global regulatory framework of TOR for 2,4-D-induced SE from a systems biology perspective, connecting the hormone - metabolism - transcription - genome homeostasis pathways. It anchors the three key downstream molecules of TOR and provides theoretical basis for in vitro plant regeneration and gene editing target selection. Main reviewers' comments are as follows:

1. The study divides the three regulatory axes of TOR, but does not conduct specialized KEGG and GO enrichment analysis for the key genes (RPS6A, CBP20, TAP46) of each axis, relying solely on network connections to determine pathway associations. The lack of enrichment data makes it difficult to quantify the proportion of biological functions of each module. It is recommended to supplement independent enrichment results for the three regulatory axes and attach statistical tables.
2. The conclusion states that TOR is the core integration hub for SE universal reprogramming, but the article lacks in vivo TOR kinase activity inhibition/activation test data. The key conclusion lacks direct experimental support. It is necessary to supplement the corresponding SE phenotype results of inhibitor treatment to improve the conclusion.
3. The STRING screening threshold in this paper is fixed at 0.9. No multi-gradient confidence control is set, making it impossible to demonstrate the rationality of the threshold selection and the robustness of the results.
4. In the analysis of cross-regulation between BR and auxin, only protein interaction relationships of ARF, BZR1, etc. are listed. The existing transcriptome is not utilized for the prediction of target gene cis-acting elements. It is impossible to verify the coordinated regulation of downstream SE key genes by BZR1-ARF from the transcriptional level. It is recommended to use PlantCARE to conduct promoter element prediction to supplement the supplementary evidence.
5. The 411 EMBL genes in this paper are only subjected to distribution statistics, and no enrichment analysis is conducted for representative EMBL genes. The depth of gene function mining is insufficient.
6. The conclusion states that the research target can be used for crop gene editing breeding, but no candidate gene list and application direction are specifically listed. The conclusion is not practical. It is recommended to combine network topology and EMBL annotation to screen candidate genes and briefly describe the editing and improvement ideas.
7. The 34 functional modules in this paper are only classified and listed in text, and no clustering algorithms such as MCODE and ClusterONE are used to quantify the basis for module division. The module division is entirely based on manual classification and lacks algorithm support, making the scientific nature insufficient. It is recommended to supplement clustering algorithm parameters and re-divide the functional modules based on topological coefficients.
8. The serial relationships of each pathway in this paper are all predicted by bioinformatics. No qPCR temporal expression verification is conducted to verify the coordinated expression pattern of genes during the induction process.
9. The conclusion only generally states that the model can be used for SE induction in multiple species. The application of soybean only has a textual description without experimental data, making it impossible to support the conclusion of cross-species applicability. Supplementary physiological and gene expression data for soybean SE induction should be provided to improve the conclusion statement.

Author Response

Dear Reviewer 1, you will find within the manuscript marked with yellow color all the comments and corrections you asked. We appreciate yor valuable comments about this manuscript.

 

 

Answers to Reviewer 1.

 

Overall evaluation: This paper focuses on the 2,4-D-induced somatic embryogenesis (SE) in Arabidopsis thaliana and uses public transcriptome data combined with STRING high-confidence protein interaction networks for analysis.

 

It proposes that TOR is the core integration hub for somatic embryo reprogramming, and divides it into three regulatory axes: TOR-FKBP12/RPS6A, TOR-CBP20, and TOR-TAP46.

 

Main contributions: It constructs a global regulatory framework of TOR for 2,4-D-induced SE from a systems biology perspective, connecting the hormone - metabolism - transcription - genome homeostasis pathways. It anchors the three key downstream molecules of TOR and provides theoretical basis for in vitro plant regeneration and gene editing target selection. Main reviewers' comments are as follows:

 

1. The study divides the three regulatory axes of TOR, but does not conduct specialized KEGG and GO enrichment analysis for the key genes (RPS6A, CBP20, TAP46) of each axis, relying solely on network connections to determine pathway associations. The lack of enrichment data makes it difficult to quantify the proportion of biological functions of each module. It is recommended to supplement independent enrichment results for the three regulatory axes and attach statistical tables.

 

Response:

 

We thank the reviewer for this valuable suggestion. To provide a quantitative functional characterization of the three TOR-associated regulatory axes, we performed independent Gene Ontology (GO) and KEGG enrichment analyses using the gene sets associated with the TOR–FKBP12–RPS6A, TOR–CBP20, and TOR–TAP46 subnetworks.

 

 

 

 

These analyses identified distinct biological signatures associated with translational regulation and ribosome biogenesis, transcriptional and developmental reprogramming, and nucleotide metabolism, DNA replication, and genome maintenance, respectively. The enrichment analyses provide additional support for the functional interpretation of the proposed TOR-centered modules and complement the network-based findings.

 

 

Changes in the Manuscript:

 

A new subsection entitled “Functional Enrichment Analysis of TOR-Associated Regulatory Axes” has been incorporated into the Results section. GO and KEGG enrichment results have been included as Supplementary Tables Reviewer_1 and Figure Reviewer_1.

 

1. The conclusion states that TOR is the core integration hub for SE universal reprogramming, but the article lacks in vivo TOR kinase activity inhibition/activation test data. The key conclusion lacks direct experimental support. It is necessary to supplement the corresponding SE phenotype results of inhibitor treatment to improve the conclusion.

 

Response:

 

We appreciate the reviewer’s comment and fully agree that pharmacological or genetic manipulation of TOR activity would provide valuable experimental validation of the proposed regulatory model. However, the primary objective of this study was to perform a systems-level reanalysis of publicly available transcriptomic datasets and construct a high-confidence protein interaction framework associated with 2,4-D-induced somatic embryogenesis. Therefore, no new biological material or experimental treatments were generated as part of this investigation.

 

To address this concern, we have carefully revised the manuscript to emphasize that the proposed TOR-centered framework represents a predictive network model based on transcriptomic and protein interaction evidence. We have moderated statements implying direct causal regulation and explicitly acknowledge that experimental validation through TOR inhibition, activation, or genetic perturbation studies will be required to confirm the proposed regulatory relationships.

 

Changes in the Manuscript:

 

The Discussion and Conclusions sections have been revised to clarify the predictive nature of the proposed model and to explicitly identify experimental validation of TOR function during somatic embryogenesis as an important future research direction.

 

 

 

 

 

 

1. The STRING screening threshold in this paper is fixed at 0.9. No multi-gradient confidence control is set, making it impossible to demonstrate the rationality of the threshold selection and the robustness of the results.

 

Response:

We thank the reviewer for this important methodological observation. To evaluate the robustness of the network architecture and the effect of confidence threshold selection, we performed additional STRING analyses using confidence scores of 0.700, 0.800, and 0.900. Although lower thresholds increased network density and the number of interactions, the major network hubs and functional organization remained largely conserved. Importantly, TOR retained a central position across all confidence levels examined. These results support the robustness of the identified regulatory framework and justify the use of the highly stringent confidence threshold (0.900) adopted in the original analysis.

 

 

Changes in the Manuscript:

A new supplementary analysis comparing network topology across multiple STRING confidence thresholds has been incorporated. The corresponding results are presented in Supplementary Tables Reviewer_1.

 

 

1. In the analysis of cross-regulation between BR and auxin, only protein interaction relationships of ARF, BZR1, etc. are listed. The existing transcriptome is not utilized for the prediction of target gene cis-acting elements. It is impossible to verify the coordinated regulation of downstream SE key genes by BZR1-ARF from the transcriptional level. It is recommended to use PlantCARE to conduct promoter element prediction to supplement the supplementary evidence.

 

Response:

 

 

We thank the reviewer for this valuable suggestion. To provide additional evidence supporting the transcriptional integration between brassinosteroid and auxin signaling pathways, we performed promoter cis-element analysis of the key regulatory genes BZR1 and ARF6 using PlantCARE.

 

The analysis revealed that both promoters contain multiple shared regulatory motifs, including G-box, MYB, MYC, W-box, STRE, MBS, and TCA-related elements, which are commonly associated with hormone-responsive transcriptional regulation, developmental processes, and stress-responsive signaling. Notably, ARF6 and BZR1 promoters exhibited similar cis-regulatory architectures enriched in MYB/MYC- and G-box-associated elements, suggesting potential convergence of upstream transcriptional regulatory mechanisms.

 

These results provide complementary evidence that auxin- and brassinosteroid-associated regulators may be coordinately integrated at the transcriptional level during somatic embryogenesis. While cis-element prediction alone cannot demonstrate direct transcription factor binding or regulatory activity, the observed promoter architecture is consistent with the functional interactions identified in the PPI network and supports the hypothesis of coordinated BR–auxin regulatory programs during embryogenic reprogramming.

 

 

 

 

 

 

 

 

 

 

 

Changes in the Manuscript:

A new supplementary section describing PlantCARE promoter analyses has been added. The identified cis-regulatory elements are presented in Supplementary Figure SX and Supplementary Table SX.

 

 

1. The 411 EMBL genes in this paper are only subjected to distribution statistics, and no enrichment analysis is conducted for representative EMBL genes. The depth of gene function mining is insufficient.

 

Response:

 

We thank the reviewer for this important observation. The objective of the EMBL gene analysis was not to identify overrepresented biological processes through a separate enrichment analysis, but rather to evaluate the representation and distribution of embryo-essential genes within the somatic embryogenesis (SE)-associated protein–protein interaction network.

 

To strengthen the biological interpretation of these results, we expanded the description of the 411 EMBL-associated genes identified in the network. EMBL genes correspond to essential loci whose loss of function causes embryo arrest or developmental failure and are involved in fundamental developmental processes such as cell division, genome maintenance, metabolic homeostasis, and organogenesis. Among the 1,927 genes included in the SE-associated network, 411 were classified as EMBL genes based on the curated dataset reported by Meinke (2020).

 

 

Changes in the Manuscript:

A dedicated subsection describing EMBL gene enrichment analyses has been incorporated into the Results section. Associated enrichment data are presented in Figure SX and Supplementary Table SX.

 

1. The conclusion states that the research target can be used for crop gene editing breeding, but no candidate gene list and application direction are specifically listed. The conclusion is not practical. It is recommended to combine network topology and EMBL annotation to screen candidate genes and briefly describe the editing and improvement ideas.

Response:

We thank the reviewer for this valuable recommendation. To increase the practical relevance of the study, we incorporated a prioritized list of candidate genes identified through the integration of network topology, functional enrichment analyses, and embryo-lethal (EMBL) annotations. Candidate genes were selected based on their central positions within the TOR-associated regulatory network, their involvement in somatic embryogenesis (SE), and their potential suitability for genetic manipulation.

The highest-priority candidates include LEC2, AGL15, SERK1, LEC1, WRI1, TOR, CBP20, BZR1, ARF5, and TAP46. These genes are associated with key biological processes such as embryogenic competence acquisition, developmental reprogramming, hormone signaling, translational regulation, and metabolic coordination. Additional candidates include DWF4, PSKR1, ARF6, DGAT1, BCCP2, NFYA1, NFYA9, NFYC2, NFYC3, and AUX1.

Based on their biological functions, several potential crop-improvement strategies can be proposed. For example, overexpression or promoter activation of LEC2, AGL15, SERK1, and LEC1 may enhance regeneration efficiency and embryogenic competence, thereby improving the recovery of edited plants in recalcitrant species. Manipulation of WRI1, DGAT1, and BCCP2 may promote lipid accumulation and support embryo development. Modification of brassinosteroid- and auxin-related regulators such as BZR1, DWF4, ARF5, ARF6, and AUX1 could improve developmental plasticity and regeneration responses. In contrast, essential genes such as TOR, CBP20, and TAP46 are associated with embryo-lethal phenotypes when severely disrupted; therefore, regulatory editing approaches (e.g., promoter engineering or CRISPR activation) may be more suitable than complete loss-of-function mutations.

These candidate genes and their potential applications have now been incorporated into the revised Discussion and Conclusions sections to provide a clearer framework for future genome-editing and crop-improvement studies

Table genes recommended editing strategy

 

Gene	Primary Function	Recommended Editing Strategy	EMBL Risk	Expected Improvement in Crops
LEC2	Master regulator of embryogenic identity	Overexpression / CRISPRa	Low–Moderate	Increased somatic embryogenesis and regeneration efficiency
AGL15	Embryogenic reprogramming regulator	Overexpression	Low	Enhanced embryogenic competence and plant regeneration
SERK1	Acquisition of embryogenic competence	Overexpression	Low	Increased frequency of embryogenic cell formation
LEC1	Embryo identity and maturation	CRISPRa / promoter activation	Moderate	Activation of embryonic developmental programs
WRI1	Lipid biosynthesis and carbon allocation	Overexpression	Low	Improved embryo development and stress resilience
TOR	Master regulator of growth and metabolism	Promoter engineering / CRISPRa	Very High	Enhanced cellular proliferation and regeneration potential
CBP20	RNA processing and miRNA-mediated regulation	Regulatory editing	High	Improved developmental reprogramming and SE responsiveness
BZR1	Brassinosteroid-responsive transcription factor	Gain-of-function / CRISPRa	Low	Enhanced growth and developmental plasticity
ARF5 (MP)	Embryo axis formation and auxin signaling	Regulatory editing	High	Improved embryonic patterning and regeneration
TAP46	TOR downstream effector controlling cell cycle	Regulatory editing	High	Enhanced cell proliferation and developmental progression
DWF4	Rate-limiting enzyme in brassinosteroid biosynthesis	Overexpression	Low	Increased regeneration and growth vigor
PSKR1	Cell proliferation and differentiation signaling	Overexpression	Low	Improved tissue culture responsiveness
ARF6	Auxin–BR signaling integration	Overexpression	Low	Enhanced hormonal responsiveness during regeneration
DGAT1	Triacylglycerol biosynthesis	Overexpression	Low	Increased energy reserves and embryo maturation
BCCP2	Fatty acid precursor biosynthesis	Overexpression	Moderate	Enhanced membrane biogenesis during embryogenesis
NFYA1	Embryogenesis-associated transcription factor	Overexpression	Low	Improved embryo initiation efficiency
NFYA9	Embryo developmental regulator	Overexpression	Low	Enhanced embryogenic competence
NFYC2	NF-Y transcriptional complex component	Overexpression	Low	Reinforcement of embryogenic gene networks
NFYC3	Embryogenesis-related transcriptional regulator	Overexpression	Low	Stabilization of embryo developmental programs
AUX1	Auxin influx carrier	Promoter activation / overexpression	Low	Improved auxin distribution and regeneration response

 

 

1. The 34 functional modules in this paper are only classified and listed in text, and no clustering algorithms such as MCODE and ClusterONE are used to quantify the basis for module division. The module division is entirely based on manual classification and lacks algorithm support, making the scientific nature insufficient. It is recommended to supplement clustering algorithm parameters and re-divide the functional modules based on topological coefficients.

 

Response:

We appreciate the reviewer’s comment regarding the identification of functional modules in the SE-associated PPI network. We agree that algorithms such as MCODE and ClusterONE are useful for detecting topological clusters; however, these approaches are mainly based on network connectivity and do not always reflect biologically meaningful functional relationships.

 

In our analysis, we evaluated topology-based clustering approaches, but the resulting groups did not consistently correspond to specific biological functions. MCODE-derived clusters, for example, mainly identified highly connected regions of the network, but several clusters contained genes from different biological processes, while functionally related genes were distributed across different clusters. Therefore, using only topological parameters could not accurately define the functional organization of the SE network.

For this reason, the 34 modules were defined through an integrated approach combining network topology with functional annotation, Gene Ontology enrichment, pathway information, and known biological roles during somatic embryogenesis. This strategy allowed the identification of biologically interpretable modules rather than clusters based exclusively on interaction density.

 

 

1. The serial relationships of each pathway in this paper are all predicted by bioinformatics. No qPCR temporal expression verification is conducted to verify the coordinated expression pattern of genes during the induction process.

 

Response:

We agree that temporal qPCR validation would provide valuable experimental confirmation of the coordinated expression patterns inferred from the transcriptomic dataset. However, the present study was designed as a systems biology reanalysis of previously published transcriptomic data and did not involve the generation of new biological samples. Consequently, experimental validation through qPCR was beyond the scope of the current investigation.

 

To address this limitation, we have explicitly acknowledged in the Discussion that the proposed regulatory relationships are based on transcriptomic and network evidence and should be considered hypothesis-generating until validated experimentally.

 

 

1. The conclusion only generally states that the model can be used for SE induction in multiple species. The application of soybean only has a textual description without experimental data, making it impossible to support the conclusion of cross-species applicability. Supplementary physiological and gene expression data for soybean SE induction should be provided to improve the conclusion statement.

 

Response:

We thank the reviewer for this observation. We agree that the present study does not provide direct experimental evidence supporting the applicability of the proposed framework in soybean or other crop species. Our intention was not to claim demonstrated cross-species conservation, but rather to suggest that the identified Arabidopsis regulatory framework may serve as a useful hypothesis-generating model for future comparative studies.

 

To avoid potential overinterpretation, we have revised the Conclusions section and moderated statements regarding cross-species applicability. The manuscript now explicitly indicates that validation in soybean and other crop species will require dedicated transcriptomic, genetic, and physiological investigations.

Changes in the Manuscript:

The Conclusions section has been revised to remove statements that could imply demonstrated cross-species applicability and now clearly identifies validation in crop species as a future research direction.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In this manuscript, the authors reanalyzed a previously published transcriptomic dataset collected at 5, 10, and 15 days after 2,4-D induction. They selected 1,927 strongly upregulated genes using an average log2 fold-change cutoff of >4 and constructed a high-confidence protein–protein interaction network with STRING. The genes were organized into 34 functional modules involving transcription, RNA processing, translation, chromatin regulation, metabolism, hormone signaling, protein degradation, DNA replication, and DNA repair. The network also contained 411 genes associated with embryo-lethal phenotypes, suggesting that many of the identified pathways are important for embryonic development. However, I have a few suggestions to strengthen this review:

 

Comments:

The authors identified 1,927 upregulated differentially expressed genes and subsequently constructed a PPI network. In addition to the network analysis shown in Figure 1, could the authors perform functional enrichment analyses, such as Gene Ontology analysis, to clarify the major biological processes, molecular functions, and cellular components represented by these genes? This would provide readers with a more direct and comprehensive overview of the biological significance of the identified DEGs.

 

In lines 139 and 201, the authors used an average log2 fold-change cutoff of >4 to select upregulated genes. This threshold appears unusually stringent, as a log2 fold-change cutoff of >1 is commonly used to define biologically meaningful differential expression. Please provide a clear justification for selecting this cutoff and discuss how it may affect the resulting network. The authors should also clarify whether statistical significance criteria, such as an adjusted p-value or false discovery rate threshold, were applied during DEG selection.

 

The PPI analysis suggests that TOR is associated with several nuclear processes, including RNA processing, transcriptional regulation, and chromatin regulation. This is intriguing because TOR has traditionally been considered predominantly cytoplasmic, although increasing evidence supports direct nuclear functions of TOR in both mammalian cells and Arabidopsis. The authors should discuss whether they propose that these interactions occur in the cytoplasm, followed by nuclear translocation of TOR-regulated proteins, or whether TOR and its associated proteins interact directly within the nucleus. Recent studies demonstrating nuclear localization and direct nuclear regulatory functions of TOR should be incorporated into the discussion, including PMID: 42135521, DOI: 10.1038/s41589-026-02188-z, and PMID: 33658715, DOI: 10.1038/s41586-021-03310-y.

 

From Section 3.3, “CBP20 Interactions with the Master Regulators of SE,” through Section 3.9, “BRs Interacting with Auxin Signaling,” the direct relationship between TOR and many of the discussed proteins and pathways is not always clear. For example, several interactions involving master regulators of somatic embryogenesis appear to be mediated primarily through CBP20 rather than through direct interactions with TOR. The authors should more clearly distinguish direct TOR interactions from indirect or extended network associations. The TOR-centered rationale of these modules should also be strengthened, and conclusions regarding TOR regulation should be moderated when they are based only on secondary network connections.

 

Figures 5, 7, 8, and 9 use dark or black backgrounds, which substantially reduce the visibility of node labels and interaction edges. Please revise these figures using a lighter background, larger labels, and improved contrast so that the network structures and individual interactions can be clearly interpreted.

 

Author Response

Dear Reviewer 2, you will find within the manuscript marked with turqoise color all the comments and corrections you asked. We appreciate yor valuable comments about this manuscript.

Reviewer_2

In this manuscript, the authors reanalyzed a previously published transcriptomic dataset collected at 5, 10, and 15 days after 2,4-D induction. They selected 1,927 strongly upregulated genes using an average log2 fold-change cutoff of >4 and constructed a high-confidence protein–protein interaction network with STRING. The genes were organized into 34 functional modules involving transcription, RNA processing, translation, chromatin regulation, metabolism, hormone signaling, protein degradation, DNA replication, and DNA repair. The network also contained 411 genes associated with embryo-lethal phenotypes, suggesting that many of the identified pathways are important for embryonic development. However, I have a few suggestions to strengthen this review:

Comments:

The authors identified 1,927 upregulated differentially expressed genes and subsequently constructed a PPI network. In addition to the network analysis shown in Figure 1, could the authors perform functional enrichment analyses, such as Gene Ontology analysis, to clarify the major biological processes, molecular functions, and cellular components represented by these genes? This would provide readers with a more direct and comprehensive overview of the biological significance of the identified DEGs.

 Response:

We thank the reviewer for this valuable suggestion. To provide a broader functional overview of the 1,927 upregulated genes used for network construction, we performed Gene Ontology (GO) enrichment analyses covering the categories Biological Process (BP), Molecular Function (MF), and Cellular Component (CC). The results revealed significant enrichment of processes associated with transcriptional regulation, RNA processing, ribosome biogenesis, chromatin organization, protein metabolism, hormone signaling, and cell cycle-related functions, consistent with the functional modules identified through the PPI network analysis.

 

We have incorporated these analyses into the revised manuscript as a new subsection in the Results section and included with figure 2. We believe these additions provide a more comprehensive functional context for the identified DEGs and strengthen the biological interpretation of the network.

Changes in manuscript:

A new GO enrichment analysis subsection has been added to the Results section, together with Figure 2

In lines 139 and 201, the authors used an average log2 fold-change cutoff of >4 to select upregulated genes. This threshold appears unusually stringent, as a log2 fold-change cutoff of >1 is commonly used to define biologically meaningful differential expression. Please provide a clear justification for selecting this cutoff and discuss how it may affect the resulting network. The authors should also clarify whether statistical significance criteria, such as an adjusted p-value or false discovery rate threshold, were applied during DEG selection.

 Response:

We appreciate the reviewer’s comment regarding the selection criteria used for identifying upregulated genes. We agree that a log2 fold-change cutoff of >1 is commonly used in differential expression studies; however, the objective of our analysis was not to identify the complete set of differentially expressed genes, but rather to reconstruct a high-confidence regulatory network associated with somatic embryogenesis (SE).

The cutoff of average log2 fold-change >4 was selected to prioritize genes showing strong and consistent transcriptional activation across the SE induction stages (5, 10, and 15 days after induction) and to minimize the inclusion of genes with modest expression changes that could increase background noise in the protein–protein interaction (PPI) network. This stringent filtering resulted in 1,927 highly responsive genes, representing robust transcriptional changes associated with 2,4-D-induced embryogenic reprogramming.

Regarding statistical significance, the original transcriptomic dataset from Wickramasuriya and Dunwell (2015) was used as the source of normalized expression values, and the present analysis was based on the reported average expression changes across induction stages. Because our objective was network reconstruction rather than differential expression discovery, gene selection was primarily based on the magnitude and consistency of transcriptional activation. We have clarified this point in the revised manuscript.

We acknowledge that this stringent threshold may exclude genes with smaller but potentially relevant expression changes; however, it also reduces false-positive associations and improves the reliability of downstream network construction. Furthermore, the STRING confidence score was set at ≥0.900 to retain only highly reliable interactions, allowing the identification of robust functional modules associated with SE.

Changes in the Manuscript:

The transcriptomic filtering criteria section has been revised to clarify that the log2 fold-change >4 threshold was applied to prioritize strongly induced SE-associated genes for high-confidence network reconstruction. We have also added an explanation regarding the use of expression magnitude and consistency as the selection criteria for network generation.

The PPI analysis suggests that TOR is associated with several nuclear processes, including RNA processing, transcriptional regulation, and chromatin regulation. This is intriguing because TOR has traditionally been considered predominantly cytoplasmic, although increasing evidence supports direct nuclear functions of TOR in both mammalian cells and Arabidopsis. The authors should discuss whether they propose that these interactions occur in the cytoplasm, followed by nuclear translocation of TOR-regulated proteins, or whether TOR and its associated proteins interact directly within the nucleus. Recent studies demonstrating nuclear localization and direct nuclear regulatory functions of TOR should be incorporated into the discussion, including PMID: 42135521, DOI: 10.1038/s41589-026-02188-z, and PMID: 33658715, DOI: 10.1038/s41586-021-03310-y.

 

 

 

 Response to Comment 3:

We thank the reviewer for highlighting the growing evidence supporting nuclear functions of TOR. We agree that the identification of TOR-associated modules enriched in RNA processing, transcriptional regulation, and chromatin organization raises important mechanistic questions regarding the cellular context of these interactions.

Changes in the Manuscript:

Traditionally, TOR signaling has been viewed primarily as a cytoplasmic pathway controlling translation, ribosome biogenesis, and cellular growth.

 

However, accumulating evidence indicates that TOR also exerts nuclear functions that directly influence transcriptional regulation. In A. thaliana, the TOR–EIN2 signaling axis provides a mechanistic link between TOR activity and nuclear gene regulation, where TOR-mediated phosphorylation controls the nuclear localization of EIN2 and thereby modulates large-scale transcriptional reprogramming associated with growth and developmental processes.

 

Notably, disruption of this pathway compromises the expression of numerous TOR-responsive genes involved in DNA replication, cell wall biosynthesis, lipid metabolism, and other developmental functions [99].

 

Furthermore, recent studies in animal systems have demonstrated that nuclear mTORC1 constitutes a functionally distinct signaling pool capable of directly regulating transcription through chromatin-associated mechanisms, providing evidence that TOR signaling can operate within the nucleus independently of its canonical lysosomal functions [100].

 

Together, these findings support the biological plausibility of the transcriptional and chromatin-associated modules identified in our network and suggest that TOR-mediated regulation during SE may involve both indirect cytoplasmic signaling cascades that converge on nuclear regulators and more direct nuclear functions associated with transcriptional control.

 

Nevertheless, the present network analysis does not establish the subcellular localization of the identified interactions; therefore, these associations should be interpreted as hypothesis-generating rather than direct evidence for nuclear TOR-containing regulatory complexes.

From Section 3.3, “CBP20 Interactions with the Master Regulators of SE,” through Section 3.9, “BRs Interacting with Auxin Signaling,” the direct relationship between TOR and many of the discussed proteins and pathways is not always clear. For example, several interactions involving master regulators of somatic embryogenesis appear to be mediated primarily through CBP20 rather than through direct interactions with TOR. The authors should more clearly distinguish direct TOR interactions from indirect or extended network associations. The TOR-centered rationale of these modules should also be strengthened, and conclusions regarding TOR regulation should be moderated when they are based only on secondary network connections.

Response:

We appreciate the reviewer’s valuable comment regarding the distinction between direct TOR interactions and extended network associations. We agree that PPI-based analyses should carefully differentiate physical interactions from indirect functional relationships. Therefore, we have revised the text to clarify that not all proteins within TOR-associated modules represent direct TOR interactors, but rather components of interconnected regulatory pathways associated with TOR signaling during somatic embryogenesis.

In the SE-associated PPI network, TOR showed direct association with CBP20, while several embryogenic master regulators (including LEC1, LEC2, FUS3, and AGL15-associated components) were identified through extended network connections mediated by CBP20 and other intermediate nodes. These associations were not interpreted as direct TOR–transcription factor interactions; instead, they indicate potential functional coupling between TOR signaling, RNA processing, transcriptional regulation, and embryogenic regulatory programs.

This interpretation is supported by previous experimental evidence demonstrating that TOR activity regulates multiple downstream processes required for developmental reprogramming. TOR inhibition or loss-of-function affects translational control, ribosome biogenesis, and growth-related pathways, including regulation of components such as RPS6A and TAP46, which are central mediators of TOR-dependent cellular responses. Furthermore, CBP20 is functionally relevant for somatic embryogenesis, as cbp20 mutants display altered expression of key embryogenic regulators such as LEC1, LEC2, and FUS3, resulting in impaired embryogenic competence. Thus, the TOR–CBP20 association identified in our network provides a biologically plausible connection between TOR signaling and the activation of embryogenic transcriptional programs.

Additionally, TOR-associated modules containing Mediator complex components, basal transcription factors, spliceosome-related proteins, and ubiquitin-mediated regulatory pathways were interpreted as functional networks rather than direct TOR targets. The presence of EMBL-associated genes within these modules further supports their relevance for embryogenic development, because disruption of these pathways frequently results in embryo arrest phenotypes.

Accordingly, we have modified the manuscript to avoid overinterpretation of secondary network connections and now describe TOR as a potential integrative regulatory node that may coordinate transcriptional, translational, metabolic, and developmental processes during SE induction rather than as a direct regulator of every component within these modules.

 

 

Changes in the manuscript:

The sections describing TOR-associated networks have been revised to distinguish direct TOR interactions from indirect functional associations. Conclusions regarding TOR regulation have been moderated, emphasizing that the identified modules represent interconnected regulatory landscapes supported by PPI topology and previous functional evidence.

Figures 5, 7, 8, and 9 use dark or black backgrounds, which substantially reduce the visibility of node labels and interaction edges. Please revise these figures using a lighter background, larger labels, and improved contrast so that the network structures and individual interactions can be clearly interpreted.

Response to Comment 5:

We thank the reviewer for this observation. Figures 5, 7, 8, and 9 have been redesigned using lighter backgrounds, larger font sizes, and improved color contrast to enhance readability. Node labels and interaction edges are now more clearly visible, allowing easier interpretation of network architecture and individual protein interactions.

 

 

Changes in manuscript:

The Discussion section has been expanded to incorporate recent literature on nuclear TOR functions; Sections 3.3–3.9 have been revised to distinguish direct and indirect TOR associations; and Figures 5, 7, 8, and 9 have been reformatted to improve clarity and readability.

 

Author Response File: Author Response.pdf