Regulatory Mechanisms of Free Umami Amino Acid Accumulation in Fresh Waxy Kernels: Insights from Transcriptome and Metabolomics Analyses

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The reviewed study compares two sweet-waxy corn varieties (Q3 and H402) at 22 days after pollination, combining targeted LC–MS/MS amino-acid profiling, untargeted metabolomics, and RNA-seq to uncover genes and regulatory networks (O2, NKD1, bZIP family, snoRNA U3, rRNA 28S) potentially responsible for elevated glutamate/aspartate and broader free-amino-acid enrichment in Q3 kernels. The topic is relevant to food science, as the flavor (umami) and nutritional quality of fresh corn are legitimate targets for food-quality research. However, the manuscript contains various methodological issues, unsupported inferences, and reporting problems that make its central claims currently unreliable. Therefore, several revisions are required before this manuscript can be considered further.

Major concerns

1. There is insufficient analytical validation for the targeted quantification of amino acids. For instance, no internal standard (stable isotope or structural analogue) is reported for LC–MS/MS quantification; no recoveries, LOD/LOQ, matrix effect assessment, or calibration linearity and R² values are provided (Methods 2.2.1–2.2.2). Without this QC, reported fold-changes (e.g., Glu 1.52×, Asp 3.32×) could reflect matrix or ionization variability rather than actual biological differences. (See Methods 2.2.1–2.2.2.)
2. The HILIC column gradient appears inconsistent with column chemistry. The ACQUITY UPLC BEH Amide column is HILIC-type; typical HILIC runs start high in organic (acetonitrile). The reported gradient (0 min: 100% A aqueous) is reversed for HILIC, which is likely to cause poor retention/peak shape. Authors must confirm the chromatographic method and show example chromatograms and retention times for standards (See Methods 2.2.2.).
3. There is a loose criterion for non-targeted metabolite identification. Metabolite ID used METLIN and in-house library with mass accuracy “within 100 ppm” (Methods 2.3). A 100 ppm tolerance for high-resolution QTof is far too permissive (modern practice: ≤5–10 ppm). This raises a high risk of misannotation. Authors must reprocess IDs with strict mass accuracy, MS/MS spectral matching to standards or curated libraries, and report identification confidence levels.
4. The biological replication and pooling have ambiguous true variability. Samples are pools of kernels (each replicate pools kernels from six ears); reported n=3 per group. Pooling can mask ear-to-ear variance and reduce the ability to assess biological variability. Authors must justify the pooling strategy, and ideally present data from independent ears (or increase biological replicates) or at least report variance among ears before pooling. (See Methods 2.1 & Results Fig.1 legend.)
5. The transcriptome → regulatory inference pipeline contains major problems. TF prediction/annotation claims to use TFDB + HMMER referencing TF models “for animals and humans” (Methods 2.4, lines ~201–203). This is inappropriate for plant TF annotation — plant-specific resources (PlantTFDB, iTAK) and plant HMM profiles must be used. Using animal/human models will misclassify and miss many maize TFs, skewing the central claim about bZIPs/O2.
6. The use of STRING to infer TF → target regulatory networks is inappropriate. STRING is a protein–protein interaction/database for general PPI associations; it is not a primary tool for TF→cis-regulatory target prediction. Authors use STRING for TF–target network (Results 3.6, Fig.8). This approach cannot support direct regulatory claims. TF-target inference should combine co-expression (e.g., WGCNA), promoter motif enrichment (with plant motif databases), and ideally experimental validation (ChIP/EMSA or reporter assays).
7. The promoter motif analysis lacks key details and statistical testing. Table 1 lists counts of “bZIP binding motifs” per promoter but does not state how promoter regions were defined (length upstream), what motif PWMs were used, what significance threshold, or whether counts are enriched relative to a genomic background. Raw counts alone do not establish functional binding or enrichment. Provide methods (motif database, threshold, promoter coordinates) and statistical enrichment tests.
8. There are contradictory statements between the abstract and the results about snoRNA/rRNA expression. In this regard, the abstract states “snoRNA U3 and rRNA28S exhibited higher and lower expression in Q3, respectively” (Abstract lines 26–29). The results (3.5) report snoRNA U3-2 and rRNA 28S were significantly lower in Q3-T versus H402-T (lines 326–329). This direct contradiction must be resolved. Such inconsistencies undermine confidence in data handling and manuscript editing.
9. There is a relevant issue regarding correlation ≠ causation, as an overstatement of regulatory claims has been detected. Central claims that O2 and other bZIPs “may act as central regulators” or “directly bind promoters” (Abstract, Conclusions, Discussion) are speculative based on co-expression, motif counts, and correlation. No experimental evidence of binding (ChIP-qPCR, EMSA) or functional perturbation (mutant/overexpression/knockdown, transient assays) is provided. Authors must substantially temper their conclusions or provide experimental validation. (See Results 3.6, Discussion, Conclusions.)
10. There is also an issue related to the use of rRNA and snoRNA expression to infer translational efficiency, as it is a weak method. The argument that lower snoRNA U3 / rRNA 28S leads to reduced translation and thus accumulation of free amino acids is plausible but unproven. No polysome profiling, ribosome-footprinting, or protein synthesis rate measurement is shown. Inferring changes in translation from the expression of a few rRNA/snoRNA loci is insufficient. Authors should either perform direct translation measurements or present this as an untested hypothesis. (See Results 3.5 and Discussion.)
11. Many comparisons use t-tests (Fig.1 legend, qPCR panels). For multiple genes/metabolites, adjustments for multiple testing (FDR, Bonferroni) are required. For group comparisons with more than two groups (Q3-T, Q3-N, H402-T, H402-N), ANOVA with post-hoc tests is a more appropriate alternative to pairwise t-tests. The manuscript requires a comprehensive statistics section that describes the tests, assumptions, correction methods, and exact p/FDR values.
12. The Data Availability Statement claims “original contributions included in article/Supplementary Material”, but supplementary tables/complete metabolite lists, raw LC-MS chromatograms, MS/MS spectra, raw reads (RNA-seq FASTQ), count matrices, and DEG lists are not completely included. For omics studies, public deposition (PRIDE/MetaboLights for metabolomics; SRA/GEO for RNA-seq) with accession numbers is required. Provide raw data and complete supplementary files.

Specific comments:

1. Lines 80–88: The literature on O2 and endosperm regulators is relevant, but the Introduction should more clearly state what is novel in this study (e.g., kernel-type separated analysis in F2 ears + integrated omics). Emphasize limitations of correlative studies.
2. Lines 106–113: Samples are pooled from six ears — justify pooling and describe whether each biological replicate represents an independent ear pool. Report whether ears were from independent plants and randomized across blocks. If possible, reanalyze without pooling or increase the replicate number to capture biological variance.
3. Lines 165–169: Re-identify metabolites with stricter mass tolerance (≤5–10 ppm), require MS/MS spectral matching to standards or reference spectra with similarity scores, and report identification confidence (Schymanski levels). Replace “within 100 ppm” with appropriate criteria.
4. Lines 173–179: Report RNA-seq QC metrics per sample (raw reads, percent mapped, duplication rates, rRNA contamination), and show PCA/hierarchical clustering to demonstrate replicate consistency.
5. Lines 187–193: FPKM is described, but downstream DESeq2 requires raw counts. Clarify pipeline: present command lines and versions (HISAT2/featureCounts/RSEM/DESeq2). Share the raw count matrix in the Supplementary.
6. Lines 199–203: Replace animal/human TF models with plant-specific TF annotation (PlantTFDB, iTAK). Recompute TF lists and re-interpret TF counts.
7. Lines 214–225 & Fig.6–8: qPCR is presented, but details are incomplete: report primer efficiencies, melt curves, and whether tubulin is validated as a stable reference across kernel types. Use two reference genes if possible. Show raw Ct values in Supplementary. Also, use appropriate statistical tests (ANOVA when comparing >2 groups).
8. Table 1: Define promoter region length used (e.g., −2 kb upstream), PWM source, motif score threshold, and perform enrichment vs background to show motifs are non-random. Provide position weight matrices and example motif alignments in the Supplementary.

Comments on the Quality of English Language

The manuscript is generally understandable but contains typographical errors and inconsistent gene/transcript labels (e.g., “Naked endospern1” in Abstract vs NKD1 elsewhere), contradictory statements (snoRNA/rRNA), and inconsistent statistical descriptions (t-test vs DESeq2 thresholds). Language requires careful professional editing to ensure clarity and consistency. 

Author Response

Comment1：The reviewed study compares two sweet-waxy corn varieties (Q3 and H402) at 22 days after pollination, combining targeted LC–MS/MS amino-acid profiling, untargeted metabolomics, and RNA-seq to uncover genes and regulatory networks (O2, NKD1, bZIP family, snoRNA U3, rRNA 28S) potentially responsible for elevated glutamate/aspartate and broader free-amino-acid enrichment in Q3 kernels. The topic is relevant to food science, as the flavor (umami) and nutritional quality of fresh corn are legitimate targets for food-quality research. However, the manuscript contains various methodological issues, unsupported inferences, and reporting problems that make its central claims currently unreliable. Therefore, several revisions are required before this manuscript can be considered further.

Response1：Thank you for reviewing our manuscript and for the constructive comments, which greatly helped us to improve the manuscript. We have heavily revised our experiments. The manuscript was carefully revised and point-by-point response was listed below. We hope that your comments have been addressed accurately. The revised manuscript was marked with red color and the responses were presented in blue text.

Comment2：There is insufficient analytical validation for the targeted quantification of amino acids. For instance, no internal standard (stable isotope or structural analogue) is reported for LC–MS/MS quantification; no recoveries, LOD/LOQ, matrix effect assessment, or calibration linearity and R² values are provided (Methods 2.2.1–2.2.2). Without this QC, reported fold-changes (e.g., Glu 1.52×, Asp 3.32×) could reflect matrix or ionization variability rather than actual biological differences. (See Methods 2.2.1–2.2.2.)

Response 2：Thanks for your pointing out the methodological issues in our manuscript. As you pointed, the targeted quantification of amino is insufficient, and it is indeed the case. Since external standard method and internal standard method are commonly used quantitative analysis methods in liquid chromatography-mass spectrometry instruments. For the relatively quantitative method of the external standard was commonly used due to it is simple and low-cost in some studies. Further, this method is based on the national standard (GB/T 30987-2020): Determination of free amino acids in plant. We have added the detailed detection parameters in the Supplementary file1 for your reference, such as calibration linearity and R² values. In other aspect, we used this method to determine amino acids for three times, and the intra group variability of the obtained values is very small, indicating that the data are repeatable, and there is a significant difference between Q3 and H402. Statistically, it can be able to reflect the significant difference between groups. The fold-changes of Glu (1.52×) and Asp (3.32×) were relatively reliable. Additionally, several published studies have utilized this method to detect free amino acids (Ref1 and Ref2) as follows:

Ref1: Lou H, Yang Y, Zheng S, Ma Z, Chen W, Yu C, Song L, Wu J. Identification of key genes contributing to amino acid biosynthesis in Torreya grandis using transcriptome and metabolome analysis. Food Chem. 2022; 379:132078. doi: 10.1016/j.foodchem.2022.132078.

Ref2: Kong F, Lu S. Soil inorganic amendments produce safe rice by reducing the transfer of Cd and increasing key amino acids in brown rice. J Environ Sci (China). 2024; 136:121-132. doi: 10.1016/j.jes.2022.09.042.

Comment3：The HILIC column gradient appears inconsistent with column chemistry. The ACQUITY UPLC BEH Amide column is HILIC-type; typical HILIC runs start high in organic (acetonitrile). The reported gradient (0 min: 100% A aqueous) is reversed for HILIC, which is likely to cause poor retention/peak shape. Authors must confirm the chromatographic method and show example chromatograms and retention times for standards (See Methods 2.2.2.).

Response 3：Thank you very much for your suggestion. The BEH amide column contains amide bonded phase and full porous ethylidene bridge hybrid particles (BEH) packing. HILIC columns are usually used for polar alkaline retention. The BEH amide column can be used for acid and alkaline retention. Since the mobile phase starts with 100% B phase, while the HILIC column cannot contain more than 40% water, otherwise the theoretical tray will collapse. The gradient (0 min: 100% A aqueous) is not exist in Table1, and the chromatogram and retention time of the standard sample have been shown in Table 2 in the method section2.2.3.

Comment4：There is a loose criterion for non-targeted metabolite identification. Metabolite ID used METLIN and in-house library with mass accuracy “within 100 ppm” (Methods 2.3). A 100 ppm tolerance for high-resolution QTof is far too permissive (modern practice: ≤5–10 ppm). This raises a high risk of mis-annotation. Authors must reprocess IDs with strict mass accuracy, MS/MS spectral matching to standards or curated libraries, and report identification confidence levels.

Response 4：We are very sorry for presenting the wrong description of methods in the original version. Since non-targeted metabolomics can detect a large number of metabolite signals at the same time, but its sensitivity and qualitative and quantitative accuracy are poor, which is used for broad metabolite profiling. Since the original description was too simple and there were even some errors in method description, we have corrected them in details and added data result evaluation in Supplementary file2. Non-target qualitative analysis provides substances with a score of more than 0.5. Detailed matching rules as follows: MZ error of level 1 and level 2 matching is 25 ppm, RT error of level 1 and level 2 matching is 6 s, allowable error of parent ion Q1 search is 25 ppm, allowable error of MS2 search is 50 ppm, and the lowest level 2 score is 0.3. The final score consists of three parts: fragment score, forward search score, and reverse search score. Its weight is listed as follows: Proportion of fragment score=0.1, proportion of forward search score=0.3, proportion of reverse search score=0.6. The reference also listed below.

Ref: Shen X, Wang R, Xiong X, Yin Y, Cai Y, Ma Z, Liu N, Zhu ZJ. Metabolic reaction network-based recursive metabolite annotation for untargeted metabolomics. Nat Commun. 2019 Apr 3;10(1):1516. doi: 10.1038/s41467-019-09550-x. 

Comment5：The biological replication and pooling have ambiguous true variability. Samples are pools of kernels (each replicate pools kernels from six ears); reported n=3 per group. Pooling can mask ear-to-ear variance and reduce the ability to assess biological variability. Authors must justify the pooling strategy, and ideally present data from independent ears (or increase biological replicates) or at least report variance among ears before pooling. (See Methods 2.1 & Results Fig.1 legend.)

Response 5: Thank you for your valuable suggestions. We are terribly sorry for not describing it clearly. Maize cobs from six ears are as one biological replication, n=3, means n=3 biological replication, not 3 ears per group (Line112-117).

Comment6：The transcriptome → regulatory inference pipeline contains major problems. TF prediction/annotation claims to use TFDB + HMMER referencing TF models “for animals and humans” (Methods 2.4, lines ~201–203). This is inappropriate for plant TF annotation — plant-specific resources (PlantTFDB, iTAK) and plant HMM profiles must be used. Using animal/human models will misclassify and miss many maize TFs, skewing the central claim about bZIPs/O2.

Response 6: We are sorry to make a severe mistake of description in method of TF annotation. We have communicated with the sequencing company and finally confirmed that the TF annotation was performed by the plant model- PlantRegMap/PlantTFDB v5.0, not the animals and humans. We have carefully revised the methods of transcription factors prediction and annotation (Line330-331).

Comment7：The use of STRING to infer TF → target regulatory networks is inappropriate. STRING is a protein–protein interaction/database for general PPI associations; it is not a primary tool for TF→cis-regulatory target prediction. Authors use STRING for TF–target network (Results 3.6, Fig.8). This approach cannot support direct regulatory claims. TF-target inference should combine co-expression (e.g., WGCNA), promoter motif enrichment (with plant motif databases), and ideally experimental validation (ChIP/EMSA or reporter assays).

Response 7: Thanks for your advice. Indeed, STRING is a protein–protein interaction/database for general PPI associations, the network of TFs just showed their interactions, not the directly regulatory claims. While, what we want to express is that the network exhibited these reported TFs with other unreported bZIPs interaction, hinted bZIPs complex regulatory network.

Comment8：The promoter motif analysis lacks key details and statistical testing. Table 1 lists counts of “bZIP binding motifs” per promoter but does not state how promoter regions were defined (length upstream), what motif PWMs were used, what significance threshold, or whether counts are enriched relative to a genomic background. Raw counts alone do not establish functional binding or enrichment. Provide methods (motif database, threshold, promoter coordinates) and statistical enrichment tests.

Response 8: Thank you very much for your pointing out the issues about the methods description lacking on bZIPs binding motifs. We have revised it in Methods 2.6.

Comment9：There are contradictory statements between the abstract and the results about snoRNA/rRNA expression. In this regard, the abstract states “snoRNA U3 and rRNA28S exhibited higher and lower expression in Q3, respectively” (Abstract lines 26–29). The results (3.5) report snoRNA U3-2 and rRNA 28S were significantly lower in Q3-T versus H402-T (lines 326–329). This direct contradiction must be resolved. Such inconsistencies undermine confidence in data handling and manuscript editing.

Response 9: Thank you very much for pointing out this contradiction. We are sorry that the misleading description of the results has caused your misleading, and we have corrected it (line28-29) for details.

Comment10：There is a relevant issue regarding correlation ≠ causation, as an overstatement of regulatory claims has been detected. Central claims that O2 and other bZIPs “may act as central regulators” or “directly bind promoters” (Abstract, Conclusions, Discussion) are speculative based on co-expression, motif counts, and correlation. No experimental evidence of binding (ChIP-qPCR, EMSA) or functional perturbation (mutant/overexpression/knockdown, transient assays) is provided. Authors must substantially temper their conclusions or provide experimental validation. (See Results 3.6, Discussion, Conclusions.)

Response 10: What you said is quite right, based on co expression, motif counts, and correlation, we speculated on the possible core regulatory role of O2 and other bZIPs gene family members. We found various studies showed the regulatory roles of bZIPs in other plants on amino acids synthesis and accumulation in the discussion part. Therefore, we speculate that these unreported bZIPs in maize might combine with amino acid synthesis and protein translation related genes thereby affecting the synthesis and accumulation of free amino acids. In future study, we plan to validate these unreported bZIPs biofunction via performing (CHIP-qPCR, EMSA) or functional dehydration (mute/overexpression/known, transient assessments). This article is mainly based on a new pathway for the synthesis of free amino acids, especially umami amino acids synthesis, that is, to put forward a clue, a new possibility and a guiding role.

Comment11：There is also an issue related to the use of rRNA and snoRNA expression to infer translational efficiency, as it is a weak method. The argument that lower snoRNA U3 / rRNA 28S leads to reduced translation and thus accumulation of free amino acids is plausible but unproven. No polysome profiling, ribosome-foot printing, or protein synthesis rate measurement is shown. Inferring changes in translation from the expression of a few rRNA/snoRNA loci is insufficient. Authors should either perform direct translation measurements or present this as an untested hypothesis. (See Results 3.5 and Discussion.)

Response 11: That’s right. We all speculate on the possible regulatory pathway based on the results of omics, expression and correlation analysis. If we want direct evidence, we need to do a lot of work, such as polysome profiling, ribosome-foot printing, or protein synthesis rate measurement you mentioned. As a matter of fact, the focus of this experiment is to propose a possible new regulation pathway, which will lay the foundation for the subsequent improvement of the quality of corn and the creation of better corn varieties, which has a certain role in promoting the food processing to obtain high-quality ingredients.

Comment12：Many comparisons use t-tests (Fig.1 legend, qPCR panels). For multiple genes/metabolites, adjustments for multiple testing (FDR, Bonferroni) are required. For group comparisons with more than two groups (Q3-T, Q3-N, H402-T, H402-N), ANOVA with post-hoc tests is a more appropriate alternative to pairwise t-tests. The manuscript requires a comprehensive statistics section that describes the tests, assumptions, correction methods, and exact p/FDR values.

Response 12: Our analysis results are involved the comparison between the two groups:group A vs group B. A vs B is suitable for t-test, while metabonomic and transcriptomics do not involve A vs B vs C (ANOVA algorithm is used for multiple group comparison). The p value calculated by hypothesis testing is used for metabolome, while the p value filtered by deseq2 algorithm is used for transcriptome comparison, and the FDR value, namely Padj, is obtained by BH correction algorithm. Similarly, qPCR also involves the comparison between the two groups, so t-test is used. The above mentioned are all conventional statistical test methods.

Comment 13：The Data Availability Statement claims “original contributions included in article/Supplementary Material”, but supplementary tables/complete metabolite lists, raw LC-MS chromatograms, MS/MS spectra, raw reads (RNA-seq FASTQ), count matrices, and DEG lists are not completely included. For omics studies, public deposition (PRIDE/MetaboLights for metabolomics; SRA/GEO for RNA-seq) with accession numbers is required. Provide raw data and complete supplementary files.

Response13: Thank you very much for your feedback. We have provided the DEG and metabolite lists information in the Table S4 and Table S2, respectively. As to raw LC-MS chromatograms, raw reads (RNA-seq FASTQ), count matrices, public deposition, it takes a long time to upload the original sequence data to the database for network speed, so the raw data you mentioned will be uploaded soon after acceptance of this article.

Specific comments:

Comment1：Lines 80–88: The literature on O2 and endosperm regulators is relevant, but the Introduction should more clearly state what is novel in this study (e.g., kernel-type separated analysis in F2 ears + integrated omics). Emphasize limitations of correlative studies.

Response 1: Thank you very much for your suggestion. We have strengthened the preciseness of description and the logic of language expression (line87-93).

Comment2：Lines 106–113: Samples are pooled from six ears — justify pooling and describe whether each biological replicate represents an independent ear pool. Report whether ears were from independent plants and randomized across blocks. If possible, reanalyze without pooling or increase the replicate number to capture biological variance.

Response 2: Yes, we obtained corn ears from different independent plants. Generally, one corn plant can only produce one effective ear. There are indeed ambiguities in the method description. We have made modifies in Method 2.1. Actually, we take six ears from different independent corn plants in each variety, one group include six independent ears, a total of three groups, that is, three biological repeats were used in our experiment. The kernels of each group were collected from the middle part of fresh corn cob (Line112-117).

Comment3：Lines 165–169: Re-identify metabolites with stricter mass tolerance (≤5–10 ppm), require MS/MS spectral matching to standards or reference spectra with similarity scores, and report identification confidence (Schymanski levels). Replace “within 100 ppm” with appropriate criteria.

Response 3: Thank you for pointing out the problem. There is indeed a serious error in the method description. We have confirmed this problem, and the detailed method has been carefully described in the manuscript.

Comment4：Lines 173–179: Report RNA-seq QC metrics per sample (raw reads, percent mapped, duplication rates, rRNA contamination), and show PCA/hierarchical clustering to demonstrate replicate consistency.

Response 4：We have provided the information in the supplementary Table S3.

Comment5：Lines 187–193: FPKM is described, but downstream DESeq2 requires raw counts. Clarify pipeline: present command lines and versions (HISAT2/featureCounts/RSEM/DESeq2). Share the raw count matrix in the Supplementary.

Response 5：OK, FPKM is first described and finally we use TPM for genes expression matrix, we have provided this data in the supplementary table 4 (Table S4).

Comment6：Lines 199–203: Replace animal/human TF models with plant-specific TF annotation (PlantTFDB, iTAK). Recompute TF lists and re-interpret TF counts.

Response 6：OK, this is indeed a serious wrong description in method. The wrong description has been modified (Line333-334).

Comment7：Lines 214–225 & Fig.6–8: qPCR is presented, but details are incomplete: report primer efficiencies, melt curves, and whether tubulin is validated as a stable reference across kernel types. Use two reference genes if possible. Show raw Ct values in Supplementary. Also, use appropriate statistical tests (ANOVA when comparing >2 groups).

Response 7：OK, first we found the tubulin gene is stable and highly expressed in different tissues from RNA-seq data. The raw Ct values have been provided in the supplementary Table S8.

Comment8：Table 1: Define promoter region length used (e.g., −2 kb upstream), PWM source, motif score threshold, and perform enrichment vs background to show motifs are non-random. Provide position weight matrices and example motif alignments in the Supplementary.

Response 8：The question you raised is very timely, and we have added descriptive details in the Method 2.6 (Line342-354). We use PlantPAN to analyze the promoter regions, the reference list below:

Ref1: Chi-Nga Chow, Chien-Wen Yang, Nai-Yun Wu, Hung-Teng Wang, Kuan-Chieh Tseng, Yu-Hsuan Chiu, Tzong-Yi Lee, Wen-Chi Chang, PlantPAN 4.0: updated database for identifying conserved non-coding sequences and exploring dynamic transcriptional regulation in plant promoters, Nucleic Acids Research. 2024.

Ref2: Chi-Nga Chow, Tzong-Yi Lee, Yu-Cheng Hung, Guan-Zhen Li, Kuan-Chieh Tseng, Ya-Hsin Liu, Po-Li Kuo, Han-Qin Zheng, and Wen-Chi Chang "PlantPAN3.0: a new and updated resource for reconstructing transcriptional regulatory networks from ChIP-seq experiments in plants", Nucleic Acids Res. 2019.

Comment9：The manuscript is generally understandable but contains typographical errors and inconsistent gene/transcript labels (e.g., “Naked endospern1” in Abstract vs NKD1 elsewhere), contradictory statements (snoRNA/rRNA), and inconsistent statistical descriptions (t-test vs DESeq2 thresholds). Language requires careful professional editing to ensure clarity and consistency. 

Response 9: Thank you very much. The issues you pointed out have been corrected, such as “Naked endospern1” to “Naked endosperm1” (Line26), and also including methodological and grammatical logical problems. On the other hand, t-test is generally used to evaluate whether there is a significant difference between the two-sample means. It is suitable for the comparison of independent samples, paired samples and single samples, and also includes the comparison of the mean difference between the experimental group and the control group. While the BH method used in DESeq2, with the analysis of high-dimensional data, such as a large number of gene expression data in genomics, BH method is also applied to these fields. In high-dimensional data, the number of hypothesis tests is extremely large, so how to effectively control the false discovery rate has become an important problem. The effectiveness of BH method makes it a common tool in these fields. Therefore, the thresholds are different.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript by Zhao et al. reports a transcriptomic and metabolomics study using mass spectrometry to compare the transcriptomes between sweet waxy corn ‘Qianjiangnuo No.3’ (Q3) and ‘Huayunhuanuo 402’ (H402), aiming to understand the genetic regulatory networks governing free amino acid metabolism in maize kernels. Although the topic of the investigation is worthy of investigation, the quality of the metabolomics analysis reporting is very scant, and the manuscript cannot be considered in its current form, and so I raccomented resubmission only after major revions   In detail, the major concerns:

• Non-targeted metabolomics: It is not clear why the authors performed untargeted metabolomics. Indeed, non-targeted metabolomics is used for broad metabolite profiling, while the extract method used is more specific for polar acid metabolites, such as amino acids, so it is normal that the higher number of detected metabolites (22%) belongs to this class. Moreover, this is the only result reported for this analysis, which is insufficient because the real analysis of non-targeted metabolomics data involves using chemometrics, such as PCA or PLS-DA, to visualize the metabolites that ultimately cause the clustering of two classes, and in this specific case of the two maize kernels. Thus, the sentence in lines 260-262 is misleading; the number of detected metabolites depends on the extraction methods or mass spectrometry parameters, and therefore, it is insufficient to claim that this is evidence of the central role of amino acid metabolism in both kernel types. This could be written only if chemometric analysis highlights amino acids as important metabolites in the clustering of the two maize kernels.  Therefore, the report of non-targeted metabolomics data in this manuscript should be allowed only after chemometric analysis, and the software used should be reported in the methods section.  Nonetheless, more information should be given in the analysis. Did the author use an internal standard to assess the quality of the analysis? Did they use QC samples? Finally, the reported data for non-targeted metabolomics did not follow the minimum reporting standards for non-targeted metabolomics data. So, I suggest following the instructions reported by Alseekh, Saleh, et al.  Nature Methods 18.7 (2021): 747-756, and report the identification level, MS/MS data, and so on in a supplementary table.
• Targeted metabolomics: the quantification methods reported in section 2.2 should be improved; no data about the method used for quantification were reported. I suppose the authors used the MRM method, considering that they reported using a QTRAP 6500. Therefore, all compound-dependent parameters should be reported in a table. Moreover, it is unclear how and if the quantitative method was validated; therefore, the authors should report the R, LOD, and LOQ, along with day-to-day variation and between-day variation for each calibration curve. Moreover, it is not clear how they assessed the matrix effect, considering that no Internal standard is reported.  Finally, no information is reported about the statistical analysis, nor is it clear which software was used. and which test was conducted, and how the reported graph was generated. This information should be reported.

  One minor concern point is in the introduction, in lines 90-93, please make this sentence clearer, considering that the corn variety Q3 was still registered, this manuscript did not report its identification, but the amino acid analysis, so probably the authors should report the relevant manuscript where this variety was described.   All these major points, and so in some point this involve manuscript re-writing of the manuscirp should be addressed prior to be considered in Food.

Author Response

Comment1：The manuscript by Zhao et al. reports a transcriptomic and metabolomics study using mass spectrometry to compare the transcriptomes between sweet waxy corn ‘Qianjiangnuo No.3’ (Q3) and ‘Huayunhuanuo 402’ (H402), aiming to understand the genetic regulatory networks governing free amino acid metabolism in maize kernels. Although the topic of the investigation is worthy of investigation, the quality of the metabolomics analysis reporting is very scant, and the manuscript cannot be considered in its current form, and so I recemented resubmission only after major revisions.

Response 1：Thank you for reviewing our manuscript and for the constructive comments, which greatly helped us to improve the manuscript. We have heavily revised our experiments. The manuscript was carefully revised and point-by-point response was listed below. We hope that your comments have been addressed accurately. The revised manuscript was marked with red color and the responses were presented in blue text.

Comment2：In detail, the major concerns: Non-targeted metabolomics: It is not clear why the authors performed untargeted metabolomics. Indeed, non-targeted metabolomics is used for broad metabolite profiling, while the extract method used is more specific for polar acid metabolites, such as amino acids, so it is normal that the higher number of detected metabolites (22%) belongs to this class. Moreover, this is the only result reported for this analysis, which is insufficient because the real analysis of non-targeted metabolomics data involves using chemometrics, such as PCA or PLS-DA, to visualize the metabolites that ultimately cause the clustering of two classes, and in this specific case of the two maize kernels.

Response 2: The reason for non-targeted metabolism detection is that, we found that there were significant differences in free amino acids between two kinds of fresh sweet waxy corn grains (F2 generation) through targeted detection of free amino acids, and these detected F2 corn kernels include sweet and waxy grains. Therefore, we want to know whether there were differences between isolated sweet and waxy kernels. We discovered the differential changes in targeted amino acids in isolated kernels among different types of kernels in two kinds corn through VIP screening, which could partially explain some prediction.

Comment3：Thus, the sentence in lines 260-262 is misleading; the number of detected metabolites depends on the extraction methods or mass spectrometry parameters, and therefore, it is insufficient to claim that this is evidence of the central role of amino acid metabolism in both kernel types. This could be written only if chemometric analysis highlights amino acids as important metabolites in the clustering of the two maize kernels. Therefore, the report of non-targeted metabolomics data in this manuscript should be allowed only after chemometric analysis, and the software used should be reported in the methods section.

Response 3: Yes, the misleading caused by the statement description issue has been corrected (Line 403) and we used Variable Importance in Projection (VIP), R software and MetDNA algorithm for chemometric analysis in non-targeted metabolism (Line245-268).

Comment4：Nonetheless, more information should be given in the analysis. Did the author use an internal standard to assess the quality of the analysis? Did they use QC samples? Finally, the reported data for non-targeted metabolomics did not follow the minimum reporting standards for non-targeted metabolomics data. So, I suggest following the instructions reported by Alseekh, Saleh, et al.  Nature Methods 18.7 (2021): 747-756, and report the identification level, MS/MS data, and so on in a supplementary table.

Response 4: We have provided the above data in the Supplementary file2 and modifies in Method 2.3.

Comment5：Targeted metabolomics: the quantification methods reported in section 2.2 should be improved; no data about the method used for quantification were reported. I suppose the authors used the MRM method, considering that they reported using a QTRAP 6500. Therefore, all compound-dependent parameters should be reported in a table. Moreover, it is unclear how and if the quantitative method was validated; therefore, the authors should report the R, LOD, and LOQ, along with day-to-day variation and between-day variation for each calibration curve. Moreover, it is not clear how they assessed the matrix effect, considering that no Internal standard is reported. Finally, no information is reported about the statistical analysis, nor is it clear which software was used. and which test was conducted, and how the reported graph was generated. This information should be reported.

Response 5：Thank you for your raising questions and suggestions. Relevant information has been added to the method. The method for the determination of free amino acids in plants is liquid chromatography tandem mass spectrometry (LC-MS/MS) (The third method) according to national standard-GB/T 30987-2020; The scanning mode is the multiple response monitoring (MRM) mode. Analysis software: multi-quant TM 3.0.3 software; The detection limit and the limit of quantitation need different concentrations of standard solutions to explore, and S/N is also calculated according to the ratio of response and noise; The information of the daughter ion, parent ion, CE (collision energy), DP (de-clustering potential) and residence time of all amino acid compounds are reflected in the manuscript (Line161-194).

 Comment6：One minor concern point is in the introduction, in lines 90-93, please make this sentence clearer, considering that the corn variety Q3 was still registered, this manuscript did not report its identification, but the amino acid analysis, so probably the authors should report the relevant manuscript where this variety was described. All these major points, and so in some point this involve manuscript re-writing of the manuscript should be addressed prior to be considered in Food.

Response 6: This variety Q3 is a new fresh sweet waxy corn variety cultivated through breeding selection in our team. The information of amino acids is not involved in the variety identification. We will provide the relevant information of the variety (Line94-98).

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors addressed my comments and suggestions adequately, so the manuscript improved in quality and content, and it can be further processed.

Author Response

Dear reviewer，thank you again for your appreciation and constructive comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors addressed most of my criticisms; however, I still have some questions and suggestions for the methods section:

1. The authors report in Supplementary File 2 that the internal standard is not detected in the blank sample. This shows there is no carryover. However, they never mention which internal standard they used. The internal standard is a compound that is not present in the matrix. It is useful for non-targeted metabolomics to verify the quality of extraction or LC-MS/MS analysis. Usually, it is a deuterated or 13C-marked metabolite. Please name the compound in the proper section or in supplementary file 2 if an internal standard is used.
2. I have some doubts about section 2.2.4. For a quantitative method, the authors should calculate the accuracy of a QC sample for both between-day and within-day variation. They should not only focus on qualitative determination. Usually, variation is calculated by injecting two QC samples with known concentrations five times on the same day or the same week. Then, accuracy and precision are evaluated. This is a fundamental step for method validation.
3. Please check the quality of the english especially in the new section added after the previous revision.

Author Response

Dear reviewer, thanks for your advice, we have revised them and highlighted revisions to the manuscript.  A cover letter file was provided to respond to your comments and explainations point by point.

Author Response File: Author Response.pdf