High Expression of PgRMC1 Correlates with Poor Neoadjuvant Chemotherapy Response and Alters Chemosensitivity in Breast Cancer Cells

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this manuscript the authors have addressed a clinically relevant question whether PgRMC1 expression contributes to chemotherapy resistance and could it help identify breast cancer patients less likely to respond to neoadjuvant chemotherapy. The authors have applied clinical observations  together with in vitro functional experiments leading to an interesting translational angle in this study. The authors have shown a interesting association between high PgRMC1 expression and poor response to NAC. The observation that modulation of PgRMC1 alters chemosensitivity in breast cancer cell lines further supports its biological relevance. However, several aspects of the study design and interpretation limit the strength of the conclusions and addressing the points below would strengthen the manuscript.

1. The authors analyzed two separate cohorts, in which the NAC cohort is quite small (n=27), and PgRMC1 IHC was performed on surgical specimens after chemotherapy rather than on pre-treatment biopsies. This raises some concern that the observed expression levels might have been influenced by treatment itself. The authors should explain why exactly were pre treatment samples not usable? Were there technical issues with tumor cellularity in the core biopsies? Also, please provide a clearer breakdown of the RECIST response grades (exact numbers of patients in grade 1 vs. 2/3) and their distribution across subtypes.

Also the PgRMC1 immunohistochemical analysis in the NAC cohort was performed using post-treatment surgical specimens. Because these tissues represent residual disease after chemotherapy exposure, high PgRMC1 expression could be because of the selection of resistant clones or the treatment-induced changes itself.

Therefore, the data support an association between PgRMC1 expression and poor pathological response, they but it does not establish PgRMC1 as a true predictive marker before treatment. The authors should either provide a an explanation or revise the wording in the manuscript accordingly. If they have access to the pretreatment biopsy material, even in a subset of patients, they can analysis that.

2. The mechanistic studies could be strengthened. The in vitro experiments suggest that PgRMC1 contributes to chemoresistance and may act through ERK signaling. These findings are interesting, but the mechanistic link remains somewhat descriptive. For example, it would be informative to determine whether inhibition of ERK signaling can reverse the chemoresistant phenotype induced by PgRMC1 overexpression. Similarly, rescue experiments following PgRMC1 knockdown would help establish causality.

3. Different cell lines were used for gain and loss of function experiments. Although the rationale appears to be based on endogenous PgRMC1 expression levels, this is not clearly explained. The authors should include a brief justification for the choice of cell lines, supported by baseline expression data to improve the transparency.

Minor Comments

1. Please provide additional information regarding the normal breast tissue controls, including whether these were paired adjacent tissues and how many samples were examined.

2. The authors should report exact p-values and confidence intervals where possible would improve statistical transparency.

3. Quantification of western blot results from independent biological replicates would strengthen the signaling data.

 

Author Response

Major Comments:
Comments 1: The authors analyzed two separate cohorts, in which the NAC cohort is quite small (n=27), and PgRMC1 IHC was performed on surgical specimens after chemotherapy rather than on pre-treatment biopsies. This raises some concern that the observed expression levels might have been influenced by treatment itself. The authors should explain why exactly were pre-treatment samples not usable? Were there technical issues with tumor cellularity in the core biopsies? Also, please provide a clearer breakdown of the RECIST response grades (exact numbers of patients in grade 1 vs. 2/3) and their distribution across subtypes. Because these tissues represent residual disease after chemotherapy exposure, high PgRMC1 expression could be because of the selection of resistant clones or the treatment-induced changes itself; the authors should provide an explanation or revise the wording accordingly.

Response 1: We greatly appreciate this critical and insightful comment. We fully agree that evaluating pre-treatment core needle biopsies (CNBs) would be ideal. However, a considerable number of pre-treatment CNBs contained only a scant amount of carcinoma cells, which would have further reduced our already small sample size and precluded robust and uniform immunohistochemical (IHC) evaluation. We therefore used post-NAC operatively resected surgical specimens, which retain sufficient residual tumor cells for reliable evaluation. This technical limitation is now explicitly stated in the revised Discussion (Limitation section, lines 457–465).
Regarding the concern that the observed expression might be influenced by treatment, PgRMC1 expression has been reported to be unaffected by anticancer drug treatment [15] (Discussion, lines 407–409). The high PgRMC1 expression observed in the post-NAC specimens is therefore likely to reflect the intrinsic biological nature of the residual tumor rather than a therapy-induced change. Nevertheless, we fully agree that our data demonstrate an association and do not establish PgRMC1 as a validated pre-treatment predictive marker. Accordingly, we have softened our claims throughout the manuscript, referring to PgRMC1 as a “putative biomarker” that “may contribute” to chemoresistance (e.g., Abstract, lines 40–42; Discussion, lines 377–380; Conclusions, lines 522–528).
We also thank the reviewer for requesting a clearer breakdown of the response grades and their distribution across subtypes. First, therapeutic response was assessed pathologically according to the histological therapeutic response criteria of the Japanese Breast Cancer Society (JBCS), not RECIST 1.1 (please also see our responses to Reviewer #2, comments 3 and 5); the JBCS grade definitions are now provided in the new Supplementary Table S1. The NAC cohort comprised 44 patients (Section 2.2, line 103; Figure 1; Section 3.2, line 249). Of these, 13 could not be evaluated by PgRMC1 IHC because they achieved a complete (n = 8) or a near-complete (n = 5) pathological response, leaving no or only scant residual invasive carcinoma (Section 3.2, lines 255–257). PgRMC1 IHC was therefore evaluable in 31 cases, which were compared as Grade 1 (poor response; n = 20) versus Grade 2 (favorable response; n = 11). No Grade 0 cases were observed. The distribution of the 31 evaluable cases across intrinsic subtypes is summarized in the new Supplementary Table S2. The actual distribution across subtypes was as follows: Luminal A (6 in JBCS Grade 1, 4 in JBCS Grade 2); Luminal B (7 in JBCS Grade 1, 4 in JBCS Grade 2); HER2 (0 cases); and TNBC (7 in JBCS Grade 1, 3 in JBCS Grade 2).
No HER2-positive cases were evaluable, as all HER2-positive patients achieved a complete or near-complete pathological response (Section 3.2, lines 265–266). Higher PgRMC1 Allred scores were significantly associated with a poorer pathological response (Grade 1) than with a favorable response (Grade 2) (p = 0.0471, Mann–Whitney U test; Figure 4E, line 283). We have revised the CONSORT diagram (Figure 1, lines 110–125), Figure 4, and Supplementary Tables to reflect this accounting.

Comments 2: The mechanistic studies could be strengthened. The in vitro experiments suggest that PgRMC1 contributes to chemoresistance and may act through ERK signaling, but the mechanistic link remains somewhat descriptive. For example, it would be informative to determine whether inhibition of ERK signaling can reverse the chemoresistant phenotype induced by PgRMC1 overexpression. Similarly, rescue experiments following PgRMC1 knockdown would help establish causality.

Response 2: We thank the reviewer for this excellent and critical suggestion. To strengthen the evidence for a functional role of PgRMC1, we now emphasize our reciprocal loss-of-function findings at the signaling level: esiRNA-mediated knockdown of PgRMC1 reduced ERK1/2 phosphorylation (Figure 5B, lines 293–298), reciprocally mirroring the increase in pERK1/2 induced by PgRMC1 overexpression (Figure 5A). Combined with the gain-of-function chemoresistance data (Figure 6), these reciprocal results support a functional contribution of PgRMC1 to ERK signaling and to the chemoresistant phenotype. We fully agree that definitive establishment of causality — specifically, rescue of chemosensitivity following PgRMC1 knockdown and reversal of the phenotype by pharmacological ERK inhibition — represents an important next step, which we have now added as a future research direction in the Discussion (lines 439–442).

Comments 3: Different cell lines were used for gain and loss of function experiments. Although the rationale appears to be based on endogenous PgRMC1 expression levels, this is not clearly explained. The authors should include a brief justification for the choice of cell lines, supported by baseline expression data to improve transparency.

Response 3: We apologize for the lack of clarity regarding our cell line selection, which was based on the endogenous baseline PgRMC1 protein levels shown in Figure 2B. For forced ectopic overexpression, we used MCF7 and MDA-MB-468 cells, which exhibit relatively low endogenous PgRMC1 protein expression. For siRNA-mediated knockdown, we used MDA-MB-231 and SKBR3 cells, which show robust endogenous PgRMC1 expression. We now state this rationale explicitly and reference the baseline expression data (Figure 2B) in the revised Results (Section 3.3, lines 285–292).

Minor Comments:
Comments 4: Please provide additional information regarding the normal breast tissue controls, including whether these were paired adjacent tissues and how many samples were examined.

Response 4: We appreciate this request for clarification. For the immunohistochemical analysis, we did not use separate, independent control tissues. Every examined section contained non-neoplastic elements — normal mammary ductal/lobular epithelium and adipose tissue — adjacent to the carcinoma, and PgRMC1 immunoreactivity in the cancer cells was compared with that of the surrounding non-neoplastic tissue within the same section, which served as an internal (endogenous) control (Section 3.1, lines 216–220). For the mRNA analysis in the initial cohort (n = 112), paired adjacent non-cancerous tissue obtained from the same surgically resected specimens was used for comparison with the corresponding tumor tissue (Section 2.2, lines 100–101). We have clarified these points in the revised Materials and Methods.

Comments 5: The authors should report exact p-values and confidence intervals where possible to improve statistical transparency.

Response 5: We agree and have revised the text and figure legends to report exact p-values wherever applicable. For the group comparisons in Figure 4 and Figure 6, non-parametric tests (Mann–Whitney U) and two-way ANOVA were used, respectively, for which hazard ratios are not applicable; the exact p-values and the corresponding test statistics are now provided (Figure 3 legend, lines 243–248; Figure 4E legend, line 283; Figure 6 legend, lines 328–332). For the KM plotter survival analyses, hazard ratios with 95% confidence intervals are displayed within the figure panels (Figure 8; e.g., HR = 1.12 [1.01–1.47] for all patients; HR = 1.56 [1.09–2.23] for luminal B).

Comments 6: Quantification of western blot results from independent biological replicates would strengthen the signaling data.

Response 6: We sincerely thank the reviewer for this valuable suggestion. We agree that densitometric quantification from independent biological replicates would further strengthen the signaling data. Although we confirmed the overall trend in repeated experiments, some variability was observed among experiments, and we were unable to perform additional experiments within the revision period to allow robust statistical validation. Therefore, we have added densitometric values from the representative immunoblot images to Figure 5 (lines 309–310). Densitometric analysis was performed using ImageJ software (version 1.54g; Section 2.6, lines 188–189). We have also revised the text to avoid overinterpretation of these data.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

I would like to congratulate the authors on their article. The introduction is concise and adequately contextualises PgRMC1 biology. However, there are a few things missing. There is no mention of the established NAC response predictors and some of the literature mentioned in the discussions, like the study of Willibald et al would be better suited in the introductory part.

The two patient cohorts are difficult to compare directly. The 112-patient cohort uses frozen tissue and qPCR to assess mRNA, whereas the 27-patient NAC cohort uses post-treatment FFPE tissue and IHC to assess protein. PgRMC1 IHC was performed on post-NAC surgical specimens rather than pre-treatment biopsies, which limits its value as a predictive biomarker for NAC response. This should be better explained. 

The study does not define a pre-specified cut-off for high versus low PgRMC1 expression, and the grouping of RECIST grade 2 and 3 responses is not clearly justified.

In Methods there is a numbering error : Section 2.3 appears twice, once for "Immunohistochemistry" and again for "Real-time qPCR analyses," which should be Section 2.4.  

RECIST 1.1 is typically used for radiological response assessment in solid tumors. Its application for pathological grading of post-NAC surgical specimens requires clarification.

The culture conditions include charcoal-treated FBS supplemented with 1 nM E2 + 10 nM progesterone. Please provide the rationale for these specific hormone concentrations.

Table 1 has inconsistent lymph node data. Positive (n=66) + Negative (n=34) = 100, not 112. The remaining 12 patients are unaccounted for.

Reference 18 (cited for IHC methodology) is a neuroscience paper on Alzheimer's disease/colivelin and has no apparent relevance to IHC breast cancer methodology.

Reference 20 (cited for qPCR) is an Alzheimer's/JAK-STAT hippocampal neuron paper.

The results are generally logically ordered from descriptive (tissue expression) to correlative (clinical cohort) to functional (in vitro) to prognostic (KM database).

Numbering error - The text in Section 3.5 references the luminal B survival plot as "Figure 7C" but it is Figure 8C in the manuscript.

The paragraph beginning "We further examined the effects of forced PgRMC1 expression on the gene expression profiles of MCF7 cells..." appears twice with near-identical content.

The primary conclusion — that elevated PgRMC1 expression is associated with poor NAC response — is supported by the data, but the strength of that support is moderate given the small cohort.

The claim that "PgRMC1 contributes to chemoresistance" from in vitro data (overexpression/knockdown in cell lines) may be overstated. In vitro forced overexpression under supraphysiological hormone conditions does not necessarily recapitulate the tumor microenvironment. The wording should be softened to "may contribute."

The proposed clinical utility of PgRMC1 as a predictive biomarker in pre-treatment CNBs is a reasonable hypothesis, but the study did not test this directly (IHC was performed on post-NAC specimens, not pre-treatment biopsies). This should maybe represent as a future research direction rather than a current finding.

Figure 5 - There is a discrepancy between the figure legend, figure header, and the text that must be resolved.

 

Comments on the Quality of English Language

The English could be improved to more clearly express the results. Also, there are several grammatical errors. 

Author Response

Comments 1: There is no mention of the established NAC response predictors, and some of the literature mentioned in the Discussion, such as the study by Willibald et al., would be better suited in the introductory part.

Response 1: We greatly appreciate these insightful comments and agree with the reviewer. We have revised the Introduction to briefly mention the established clinicopathological predictors of NAC response (Section 1, lines 54–57). In addition, we have moved the discussion of the study by Willibald et al. (regarding PgRMC1 and anthracycline-based neoadjuvant therapy) from the Discussion to the Introduction (Section 1, lines 66–69) to better contextualize our study.

Comments 2: The two patient cohorts are difficult to compare directly. The 112-patient cohort uses frozen tissue and qPCR to assess mRNA, whereas the NAC cohort uses post-treatment FFPE tissue and IHC to assess protein. PgRMC1 IHC was performed on post-NAC surgical specimens rather than pre-treatment biopsies, which limits its value as a predictive biomarker for NAC response. This should be better explained.

Response 2: We fully acknowledge this limitation and have clarified it in the revised Discussion (Limitation section, lines 457–488). Ideally, PgRMC1 protein expression should have been evaluated using pretreatment CNB specimens to determine its predictive value for NAC response. However, many pretreatment CNB specimens contained only scant carcinoma cells, which would have further reduced the number of evaluable cases. We therefore used post-NAC surgically resected specimens to ensure sufficient residual tumor tissue for immunohistochemical evaluation.
We also initially attempted a preliminary analysis of PGRMC1 mRNA expression using frozen tumor tissues obtained after NAC. However, in cases with favorable NAC responses, PGRMC1 mRNA levels — as well as the expression of many other genes — were extremely low, presumably because of the marked reduction in residual tumor cells. Thus, bulk mRNA analysis of post-NAC tissues was strongly affected by tumor cellularity and was considered unsuitable for evaluating PgRMC1 expression in residual tumor cells. For this reason, we focused on immunohistochemical analysis of PgRMC1 protein in the NAC cohort (Discussion, Limitation section, lines 466–477).
Accordingly, we have revised the manuscript to state that high PgRMC1 expression was associated with a poor pathological response to NAC, rather than presenting it as a validated predictive biomarker. We have also standardized the terminology throughout the manuscript by using “PgRMC1” for the protein and “PGRMC1” (italicized) for the gene or mRNA.

Comments 3: The study does not define a pre-specified cut-off for high versus low PgRMC1 expression, and the grouping of RECIST grade 2 and 3 responses is not clearly justified.

Response 3: We sincerely thank the reviewer for this important comment. We did not define a pre-specified IHC cut-off to classify cases into PgRMC1-high and PgRMC1-low groups. Instead, PgRMC1 protein expression was evaluated using the Allred scoring system, and Allred scores were compared with clinicopathological factors, including pathological therapeutic response to NAC, WHO grade, and molecular subtype. We have clarified in the Methods and Discussion that an IHC-based cut-off was not established (Discussion, Limitation, lines 502–509), and we acknowledge that determining a reproducible Allred score–based cut-off will require validation in larger, independent clinical cohorts.
Regarding the survival analyses using publicly available databases, patients were stratified according to PGRMC1 mRNA expression using the cut-off settings provided by each database, as described in the Materials and Methods (Section 2.7, lines 190–200). These mRNA-based cut-offs are distinct from the Allred-score IHC assessment.
Regarding the response grouping, therapeutic response was evaluated pathologically according to the JBCS histological therapeutic response criteria (Supplementary Table S1), not RECIST 1.1. Of the 44 patients in the NAC cohort, 13 could not be evaluated by PgRMC1 IHC because they achieved a complete (n = 8) or near-complete (n = 5) pathological response, leaving no or only scant residual invasive carcinoma; no Grade 0 cases were observed. The immunohistochemical comparison was therefore performed between Grade 1 (poor response; n = 20) and Grade 2 (favorable response; n = 11) (Section 3.2, lines 254–263; Figure 4E). The subtype distribution is provided in the new Supplementary Table S2. We have revised the manuscript and Figure 4 accordingly.

Comments 4: In Methods there is a numbering error: Section 2.3 appears twice, once for “Immunohistochemistry” and again for “Real-time qPCR analyses,” which should be Section 2.4.

Response 4: We sincerely apologize for this oversight. The numbering has been corrected, and “Real-time qPCR analyses” is now correctly labeled as Section 2.4 (line 145).

Comments 5: RECIST 1.1 is typically used for radiological response assessment in solid tumors. Its application for pathological grading of post-NAC surgical specimens requires clarification.

Response 5: We sincerely thank the reviewer for pointing out this error in our terminology. You are entirely correct that RECIST 1.1 is used for radiological assessment and is inappropriate for pathological grading. The therapeutic responses of the surgical specimens in our study were evaluated pathologically according to the histological therapeutic response criteria defined by the JBCS. We have removed all inappropriate references to “RECIST” throughout the manuscript and figures and replaced them with the “JBCS histological therapeutic response criteria” (e.g., Section 3.2, lines 253–254). We have also added Supplementary Table S1 defining JBCS Grades 0–3, and cited Shien T et al., Breast Cancer Res. Treat. 2024, 208, 145–154 (reference 19).

Comments 6: The culture conditions include charcoal-treated FBS supplemented with 1 nM E2 + 10 nM progesterone. Please provide the rationale for these specific hormone concentrations.

Response 6: We adopted supplementation of the culture medium with E2 and progesterone based on a previous study that examined the effects of progestogens on human breast tumor cells under defined hormonal conditions [23]. Because PgRMC1 is a progesterone-associated protein whose function may depend on progesterone stimulation, we cultured the cells under these hormone-supplemented conditions to provide a relevant hormonal environment for the chemosensitivity assays. We have cited reference [23] (Chen FP et al., Climacteric 2011, 14, 345–351) in the Methods (Section 2.5, line 174).

Comments 7: Table 1 has inconsistent lymph node data. Positive (n=66) + Negative (n=34) = 100, not 112. The remaining 12 patients are unaccounted for.

Response 7: We thank the reviewer for pointing this out. We re-examined the clinical records and confirmed the nodal status and the other clinicopathological factors for all patients. The corrected figures are 73 node-positive and 39 node-negative cases (total, 112). Table 1 has been updated accordingly (lines 126–following).

Comments 8: Reference 18 (cited for IHC methodology) is a neuroscience paper on Alzheimer’s disease/colivelin and has no apparent relevance to breast cancer IHC methodology. Reference 20 (cited for qPCR) is an Alzheimer’s/JAK-STAT hippocampal neuron paper.

Response 8: We thank the reviewer for pointing this out. We originally cited these papers because they contained the detailed IHC and real-time qPCR protocols routinely used in our laboratory (established by the corresponding author, T.C.). However, we agree that citing neuroscience papers in a breast cancer study may cause confusion. We have therefore replaced them with methodological references more relevant to breast cancer research: Ueno T et al., Tumour Biol. 2018, 40, 1010428318811025 (reference 20, for IHC, cited at line 128); and Yang CM et al., Horm. Mol. Biol. Clin. Investig. 2012, 10, 241–248 (reference 22, for qPCR, cited at line 147).

Comments 9: The text in Section 3.5 references the luminal B survival plot as “Figure 7C” but it is Figure 8C in the manuscript. Also, the paragraph beginning “We further examined the effects of forced PgRMC1 expression on the gene expression profiles of MCF7 cells…” appears twice with near-identical content.

Response 9: Thank you for catching these errors. We have corrected the figure reference (now Figure 8C, line 358) and deleted the duplicated paragraph in the Results section.

Comments 10: The primary conclusion — that elevated PgRMC1 expression is associated with poor NAC response — is supported by the data, but the strength of that support is moderate given the small cohort. The claim that “PgRMC1 contributes to chemoresistance” from in vitro data may be overstated, since forced overexpression under supraphysiological hormone conditions does not necessarily recapitulate the tumor microenvironment. The wording should be softened to “may contribute.”

Response 10: We agree that in vitro conditions do not perfectly recapitulate the complex in vivo tumor microenvironment. Following your recommendation, we have softened our wording throughout the manuscript — including the Abstract (line 40), Results (lines 315, 318), Discussion (lines 377–380, 413, 418–420, 421, 493–494), and Conclusions (lines 522–528) — changing definitive claims to “may contribute to” or “is associated with.”

Comments 11: The proposed clinical utility of PgRMC1 as a predictive biomarker in pre-treatment CNBs is a reasonable hypothesis, but the study did not test this directly (IHC was performed on post-NAC specimens, not pre-treatment biopsies). This should perhaps be presented as a future research direction rather than a current finding.

Response 11: We completely agree. We have revised the Conclusions (lines 522–528) and the Discussion (Limitation section, lines 457–488) to frame the utility of PgRMC1 evaluation in pre-treatment CNBs as a hypothesis and an important future research direction, rather than as an established finding of the current study.

Comments 12: Figure 5 — there is a discrepancy between the figure legend, figure header, and the text that must be resolved.

Response 12: We have carefully reviewed and corrected the Figure 5 legend (lines 300–310), header, and the main text (lines 293–298) to ensure complete consistency.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript has improved substantially, and most of my previous concerns have been addressed. A few minor points should still be addressed before the manuscript is accepted:

Although the authors now refer to PgRMC1 as a “putative biomarker,” the statement that it could serve as a biomarker for assessing NAC sensitivity remains somewhat stronger than the clinical data support. PgRMC1 was measured in residual tumor tissue after chemotherapy rather than in pretreatment biopsies. Therefore, the study demonstrates an association between PgRMC1 expression in residual disease and pathological response but does not establish its ability to predict chemotherapy sensitivity before treatment.

I suggest revising the Abstract and Conclusions to state that PgRMC1 is a candidate marker associated with pathological response whose predictive value requires validation in pretreatment specimens.

In Figure 7 please specify the number and type of replicates, the statistical test used, or the exact p-values.

Author Response

Comments 1: Although the authors now refer to PgRMC1 as a “putative biomarker,” the statement that it could serve as a biomarker for assessing NAC sensitivity remains somewhat stronger than the clinical data support. PgRMC1 was measured in residual tumor tissue after chemotherapy rather than in pretreatment biopsies. Therefore, the study demonstrates an association between PgRMC1 expression in residual disease and pathological response but does not establish its ability to predict chemotherapy sensitivity before treatment.  I suggest revising the Abstract and Conclusions to state that PgRMC1 is a candidate marker associated with pathological response whose predictive value requires validation in pretreatment specimens.

Response 1: We completely agree with your insightful assessment. You are entirely correct that since our evaluation was performed on post-NAC residual tissues, our findings demonstrate an association rather than a predictive ability. Following your excellent suggestion, we have revised both the Abstract and the Conclusions sections to accurately reflect this. We have removed the phrase "putative biomarker" and explicitly stated that PgRMC1 is a "candidate marker associated with pathological response, whose predictive value requires validation in pretreatment specimens."

Comments 2: In Figure 7 please specify the number and type of replicates, the statistical test used, or the exact p-values.

Response 2: We apologize for the omission of these important statistical details in the Figure 7 legend. We have revised the legend for Figure 7 to clearly specify that the data are presented as the mean ± SD from three independent biological replicates (n = 3), and that the relative expression was normalized to the internal control GAPDH. We have also clarified that Student's t-test was used to determine the statistical significance of these continuous variables, and we have provided the exact p-values for the significant changes in gene expression. Accordingly, we also revised our methods on statistical analyses.