Review Reports - A Putative Hsa-miR-582-5p–CD81 Relationship Identified by Integrative Transcriptomic Analysis in Osteosarcoma

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study suggests a novel regulatory axis involving CD81 and hsa-miR-582-5p in osteosarcoma using public transcriptome data. From DEG identification to PPI network analysis and miRNA target prediction, the methodology is typically sound. Nevertheless, the current version of the manuscript seems more like an early report than a completed study. It is necessary to solve the issue of relying on "putative" relationships without experimental proof, either by conducting more experiments or by considerably reducing the assertions made in the debate. Additionally, quick attention must be given to the statistics' quality and the citation styles' observance.

1-Abstract :

The results "support CD81 as a candidate prognostic biomarker," according to the conclusion, although neither the hazard ratios (HR) nor the p-values from the survival study are included in the abstract.

2-Introduction :

Although the introduction gives a general review of miRNAs and osteosarcoma, it is not very detailed about the particular need this study covers. Why pay particular attention to cell adhesion?

3-Methodology :

A-It is unclear what criteria were used to choose the particular datasets (GSE12865, GSE14359, and GSE246405). Did other osteosarcoma datasets on GEO have any exclusion criteria?

B-According to Figure 1D, the control groups for GSE12865 and GSE14359 consist of n=1 sample each.It is not statistically valid to perform a standard t-test or ANOVA with a sample size of 1 in the control group.

C-You used a "medium confidence" score (0.400) for the STRING analysis. For a purely bioinformatic study, a 0.400 cutoff is too low.

4-Results :
A-Figure 3 : Panels 1A and 1B (labeled A and B in the figure) are currently unacceptable for publication. The resolution of the PPI network in Figure 3A is too low; the node labels (gene names) are unreadable.

B-Figure 1D: The table formatting appears to be a raw Excel screenshot and is not conventional. Instead of using this as an inset inside a figure, please format it as a formal table in the main text.

c-Figure 5 A: When comparing Osteosarcoma (n = 5) and Normal (n = 1), you report a p-value of 0.00000495. As previously indicated, a highly significant p-value against a n=1 control indicates that the wrong statistical test was applied.

D-Comparing plasma miRNA levels to tissue mRNA levels is a weak proxy for direct regulation. Circulating miRNAs often do not reflect the tumor microenvironment's intracellular content.

5- Discussion :

A- The discussion strongly suggests that hsa-miR-582-5p and CD81 have a functional interaction, but this connection is based only on inverse expression correlation and database predictions (miRDB/TargetLink).

6-References :

A-Formatting: The reference style needs to be thoroughly checked against the journal's rules because it is inconsistent.

B-Inconsistency: The DOI is included in some references but not in others.

C-Journal names are inconsistently abbreviated (e.g., "Nucleic Acids Res" vs. full names).

D-Punctuation: Verify the author lists' punctuation and spacing.

Author Response

1-Abstract:

Response:

Thank you for this helpful suggestion. We agree that the Abstract should explicitly report the key survival statistic supporting the prognostic relevance of CD81. We have revised the Abstract to include the Bonferroni-adjusted p-value from the R2 KaplanScan analysis (Bonferroni-adjusted p = 0.043) and maintained conservative wording (e.g., “candidate” and “putative”) to reflect the hypothesis-generating nature of this in silico study as following: “Results: We identified 107 overlapping DEGs and prioritized eight hub genes. CD81 was significantly associated with overall survival (Bonferroni-adjusted p = 0.043) and showed reduced expression in OS tissues and cell line models. (page 1, lines 30-32)”.

2-Introduction:

Although the introduction gives a general review of miRNAs and osteosarcoma, it is not very detailed about the particular need this study covers. Why pay particular attention to cell adhesion?

Response:

We thank the reviewer for this important comment. We agree that the rationale for focusing on cell adhesion–related processes was not sufficiently explicit in the original Introduction. We have revised the Introduction to explicitly link these biological and clinical considerations, thereby clarifying the specific gap and motivation addressed by this study as following: " In this study, we applied a multi-dataset integrative bioinformatic framework to systematically investigate molecular networks relevant to osteosarcoma progression. Our emphasis on cell adhesion– and microenvironment-related processes was data-driven: across independent osteosarcoma transcriptomic datasets, consistently dysregulated genes converged on extracellular matrix remodeling, focal adhesion–associated signaling, and other invasion-related programs, which are central to metastatic dissemination. Accordingly, we integrated publicly available mRNA expression datasets to identify reproducibly altered genes, constructed protein–protein interaction networks, and prioritized candidate hub genes using complementary network-topology algorithms, followed by prognostic association assessment in an independent survival cohort. To explore a putative post-transcriptional regulatory layer, we further combined miRNA target prediction with circulating miRNA profiling to nominate testable miRNA–mRNA relationships. Collectively, this study aims to prioritize candidate prognostic biomarkers and generate hypothesis-driven directions for subsequent validation in clinical specimens and functional assays. (pages 2-3, lines 93-107)".

3-Methodology:

A-It is unclear what criteria were used to choose the particular datasets (GSE12865, GSE14359, and GSE246405). Did other osteosarcoma datasets on GEO have any exclusion criteria?

Response:

We thank the reviewer for this important comment. We have revised the Methods section to explicitly describe the dataset selection strategy as following: " Publicly available osteosarcoma-related gene expression datasets were systematically screened in the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). Datasets were considered eligible if they met the fol-lowing criteria: (i) inclusion of osteosarcoma samples with clearly defined non-malignant bone-related controls; (ii) availability of processed expression data suitable for differential expression analysis; (iii) sufficient annotation to distinguish biological samples from technical replicates; and (iv) relevance to primary osteosarcoma biology rather than treat-ment response or unrelated experimental perturbations. Based on these criteria, three mRNA expression datasets—GSE12865, GSE14359, and GSE246405—were selected for integrative transcriptomic analysis. Other osteosarcoma-related GEO datasets were ex-cluded if they lacked appropriate non-malignant controls, focused exclusively on drug-treated conditions, had insufficient sample annotation, or were not suitable for cross-dataset comparison. GSE12865 (last updated July 26, 2018) was generated using the Affymetrix Human Gene 1.0 ST Array (GPL6244) and includes six osteosarcoma tumor samples and one normal human osteoblast (HOB) sample, each profiled in technical du-plicate. GSE14359 (GPL96; last updated August 10, 2018) contains five conventional pri-mary osteosarcoma tissue samples, four osteosarcoma lung metastasis samples, and non-neoplastic primary osteoblast cells (HOBc), all profiled in technical duplicate. For differential expression analysis, only conventional primary osteosarcoma samples and HOBc controls were included, while metastatic samples were excluded to reduce biologi-cal heterogeneity. GSE246405 (last updated April 17, 2024), generated using the NextSeq 2000 platform (GPL30173), comprises four osteosarcoma RNA-seq models (NRHOS, OSKG, OSRH20115, and I063021) and one osteoblast control model (OB). Three osteosar-coma models and the osteoblast control were available with two replicates, whereas one osteosarcoma model (I063021) had a single replicate. For circulating miRNA analysis, GSE65071 (GPL19631; last updated April 25, 2016) was selected because it includes plasma samples from osteosarcoma patients and healthy controls profiled using a stand-ardized qPCR-based platform. In the present study, 15 healthy control plasma samples and 10 localized osteosarcoma plasma samples were analyzed, while metastatic plasma samples were not included (Figure 7). Together, these datasets were selected to enable cross-dataset validation of consistently dysregulated genes and miRNAs while balancing data quality, biological relevance, and platform compatibility. (page 13, lines 400-431)".

B-According to Figure 1D, the control groups for GSE12865 and GSE14359 consist of n=1 sample each. It is not statistically valid to perform a standard t-test or ANOVA with a sample size of 1 in the control group.

Response:

We thank the reviewer for raising this important statistical concern. We agree that conventional group-comparison tests (e.g., a standard t-test or ANOVA) are not appropriate when the biological control group size is n = 1, because such tests rely on estimating within-group variance from multiple independent samples. We would like to clarify that our differential expression analyses for GSE12865 and GSE14359 were conducted using GEO2R (limma) rather than a standard t-test/ANOVA. limma uses moderated t-statistics with empirical Bayes variance shrinkage, which borrows information across genes to stabilize variance estimation and is commonly used for exploratory transcriptomic analyses in small-sample settings. Importantly, however, this does not eliminate the fundamental limitation that a single biological control sample restricts the strength of inference from any single dataset. Therefore, we treat the DEG results from each dataset as exploratory, and we avoid over-interpreting p-values derived from one dataset alone. To mitigate this limitation, our workflow emphasized cross-dataset reproducibility: we retained only DEGs that were consistently identified across three independent datasets (GSE12865, GSE14359, and GSE246405) generated from different sample sources and platforms, and subsequent PPI/hub analyses were based on this overlapping DEG set. We have revised the Methods and Discussion to clarify the statistical framework used and explicitly state that these findings are hypothesis-generating and require validation in appropriately powered cohorts and/or experimental systems as following: “Differential expression analyses were performed using GEO2R, which implements the limma package and applies moderated t-statistics with empirical Bayes variance estimation. This framework is specifically designed for high-dimensional transcriptomic data and improves variance estimation by borrowing information across genes, thereby allowing exploratory inference even under conditions of limited sample size or unequal group distributions. Importantly, the use of limma does not eliminate the constraints imposed by small biological control cohorts; rather, it enables statistically principled screening of candidate genes at the exploratory stage. Multiple-testing correction was applied using the Benjamini–Hochberg method as implemented by the platform. Survival analyses were conducted using the R2 Genomics Analysis and Visualization Platform with log-rank testing and the KaplanScan procedure, with multiple-testing correction applied as reported by the platform. Unless otherwise specified, statistical significance was defined as p < 0.05. Where summary statistics were reported, values are presented as mean ± standard deviation (SD). (page 16, lines 521-534)” and “Some of the GEO datasets included in this study contained a limited number of biological control samples, which constrains statistical power at the level of individual dataset comparisons. Although differential expression analyses were conducted using GEO2R with limma-based moderated t-statistics and empirical Bayes variance estimation—an approach that improves stability and is appropriate for exploratory transcriptomic analyses—results derived from any single dataset should not be interpreted as definitive. Accordingly, the analytical strategy of this study emphasized cross-dataset reproducibility rather than reliance on individual comparisons. Only genes consistently identified across multiple independent datasets, platforms, and biological contexts were retained for downstream network, survival, and miRNA analyses. This integrative design reduces dataset-specific bias and supports the interpretation of the findings as hypothesis-generating rather than confirmatory. (page 12, lines 375-387)”.

C-You used a "medium confidence" score (0.400) for the STRING analysis. For a purely bioinformatic study, a 0.400 cutoff is too low.

Response:

We thank the reviewer for this valuable suggestion regarding STRING interaction confidence thresholds. In the revised analysis, we reconstructed the protein–protein interaction network using a high-confidence interaction score (STRING combined score ≥ 0.700) to increase stringency and reduce potential false-positive interactions. Hub gene identification was re-performed using MCC, Closeness, Degree, and EPC algorithms in CytoHubba. Under these stricter conditions, eight hub genes were consistently identified, with CD81 remaining a central hub gene across all algorithms. Compared with the original analysis, only THY1 was excluded, while the remaining hub genes were preserved. These results indicate that the identification of CD81 as a hub gene is robust and not dependent on lower-confidence protein–protein interactions. The revised network analysis and updated results have been incorporated into the manuscript as following: “A protein–protein interaction (PPI) network was constructed using the STRING database (https://string-db.org/), with the minimum required interaction score set to high confidence (0.700). (page 6, lines 176-178)” and “To identify the most influential genes within the network, CytoHubba in Cytoscape was applied using four topological algorithms—Maximal Clique Centrality (MCC), Closeness, Degree, and Edge Percolated Component (EPC)—each ranking gene by network centrality and connectivity. The top 10 hub genes from each algorithm were compared using a Venn diagram, identifying eight shared hub genes: CD81, STAT1, NCAM1, CD36, VCAM1, TIMP1, THBS1, and FN1 (Figure 3C–G). (page 6, lines 181-187)” and “Figure 3. Construction and Hub-gene Analysis of the Protein–Protein Interaction (PPI) Network. (A) A PPI network was constructed from the 107 DEGs using the STRING database with the minimum required interaction score set to high confidence (0.700). Disconnected nodes were omitted from the visualization for clarity. (B–F) The network was visualized and analyzed in Cytoscape v3.10.2. Network modules were identified using the MCODE plugin, and key hub genes were determined with the CytoHubba plugin based on Maximal Clique Centrality (MCC), Closeness, Degree, and Edge Percolated Component (EPC) algorithms. (page 7, lines 195-202)” and “Figure 4. Prognostic Evaluation of Hub Genes in Osteosarcoma. The prognostic significance of eight hub genes (NCAM1, THBS1, STAT1, VCAM1, CD36, CD81, TIMP1, and FN1) was analyzed using the R2 Genomics Analysis and Visualization Platform. (page 8, lines 205-207)” and “Interactions with a combined confidence score greater than 0.7 were considered significant and included in the analysis [1,2]. For visualization purposes, disconnected nodes were hidden in STRING to improve net-work readability. The resulting interaction data were imported into Cytoscape v3.10.2 for network visualization and topological analysis [3,4]. Within Cytoscape, the CytoHubba plugin was applied to evaluate node connectivity and identify the most influential hub genes using four established topological algorithms: Maximal Clique Centrality (MCC), Closeness, Degree, and Edge Percolated Component (EPC). The MCODE plugin (version 1.5.1) was used to detect densely connected submodules under the parameters: Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 2, and Max Depth = 100 [3,4]. Genes consistently ranked highly across the four CytoHubba algorithms were considered key nodes within the network. Based on this integrative approach, eight hub genes—neural cell adhesion molecule 1 (NCAM1), thrombospondin-1 (THBS1), signal transducer and activator of transcription 1 (STAT1), vascular cell adhesion molecule 1 (VCAM1), CD36, CD81, tissue inhibitor proteinases 1 (TIMP1), and fibronectin 1 (FN1)—were identified for subsequent validation and prognostic assessment. (page 15, lines 478-493)”.

4-Results:

A-Figure 3: Panels 1A and 1B (labeled A and B in the figure) are currently unacceptable for publication. The resolution of the PPI network in Figure 3A is too low; the node labels (gene names) are unreadable.

Response:

We thank the reviewer for this important comment. We have replaced the original PPI network figure with a higher-resolution version generated directly from Cytoscape/STRING export settings to ensure that node labels are clearly readable and publication-quality. The updated figure now presents legible gene symbols and improved layout clarity.

B-Figure 1D: The table formatting appears to be a raw Excel screenshot and is not conventional. Instead of using this as an inset inside a figure, please format it as a formal table in the main text.

Response:

We agree with the reviewer. The sample size summary has been reformatted as a formal table and moved from the figure inset to the main text to improve presentation quality and compliance with journal style.

C-Figure 5 A: When comparing Osteosarcoma (n = 5) and Normal (n = 1), you report a p-value of 0.00000495. As previously indicated, a highly significant p-value against a n=1 control indicates that the wrong statistical test was applied.

Response:

We thank the reviewer for pointing out this important concern. We agree that statistical inference based on a comparison against a single biological control sample (n = 1) should be interpreted with extreme caution, and that an apparently highly significant p-value may be misleading in this setting. In the revised manuscript, we have removed the p-value from Figure 5A and present the comparison as a descriptive visualization only, explicitly stating that GSE14359 contains only one biological osteoblast control sample and therefore does not support robust between-group inferential testing as following: “In the GSE14359 dataset, CD81 expression was markedly lower in conventional osteosarcoma tissues than in the osteoblast control sample (Figure 5A). (pages 8-9, lines 213-215)”. Importantly, our conclusions do not rely on this single comparison; instead, we emphasize cross-dataset consistency (overlapping DEGs across three independent datasets) and the independent survival association of CD81.

D-Comparing plasma miRNA levels to tissue mRNA levels is a weak proxy for direct regulation. Circulating miRNAs often do not reflect the tumor microenvironment's intracellular content.

Response:

Thank you for this important comment. We agree that comparing circulating miRNA levels with tissue mRNA expression provides only an indirect line of evidence and cannot be interpreted as proof of direct intracellular regulation within the tumor microenvironment. In the revised manuscript, we have therefore reframed the plasma miRNA analysis as biomarker-oriented and hypothesis-generating, and we explicitly state that circulating miRNAs may originate from multiple tissues/cell types and may not reflect the tumor-intracellular miRNA pool as following: “Notably, miR-582-5p expression was significantly elevated in the plasma of OS patients relative to healthy controls (GSE65071), which is consistent with a circulating miRNA signal and, together with in silico target prediction, nominates a testable hypothesis that miR-582-5p may be related to CD81 regulation in osteosarcoma; however, circulating miRNA levels cannot be used as a surrogate for tumor-intracellular miRNA activity. (page 11, lines 319-324)” and “Importantly, this comparison is indirect: circulating miRNAs may originate from multiple tissues and cell types and may not reflect the intracellular miRNA pool within tumor cells or the tumor microenvironment. (page 12, lines 340-342)” and “Therefore, the proposed miR-582-5p–CD81 relationship should be regarded as hypothesis-generating, and will require validation in matched tumor tissues (and ideally paired plasma/exosomal fractions) as well as functional assays (e.g., luciferase reporter and gain-/loss-of-function experiments). (page 12, lines 342-345)” and “Accordingly, we do not claim that plasma miR-582-5p directly regulates CD81 in osteo-sarcoma tissues based on the present data alone. (page 12, lines 356-357)”. We further clarified that the proposed miR-582-5p–CD81 link requires validation using matched tumor tissues and functional assays. In addition, we revised the Figure 6 legend to avoid causal wording as following: “Figure 6. Identification of hsa-miR-582-5p as a circulating miRNA candidate linked to CD81 target prediction in osteosarcoma. (A) Candidate miRNAs predicted to target CD81 were retrieved from miRDB and miRTargetLink 2.0, and overlapping candidates were identified by Venn analysis. (B) Differentially expressed circulating miRNAs were analyzed in GSE65071 by comparing localized osteosarcoma plasma with healthy control plasma. (C) Intersection of predicted CD81-related miRNAs with dysregulated plasma miRNAs. (D) Plasma hsa-miR-582-5p levels in GSE65071. Statistical significance was set at adj. p < 0.01. (page 10, lines 258-264)”.

5- Discussion:

Response:

We appreciate the reviewer’s concern and agree that the previous wording could be interpreted as implying a functional interaction. In the revised manuscript, we substantially tempered the statements regarding miR-582-5p and CD81 and consistently describe the relationship as putative/hypothesis-generating, supported by in silico target prediction and indirect expression patterns across independent public datasets, rather than experimental evidence. We also added explicit language stating that direct binding and causality must be established by experimental validation as following: “Notably, miR-582-5p expression was significantly elevated in the plasma of OS patients relative to healthy controls (GSE65071), which is consistent with a circulating miRNA signal and, together with in silico target prediction, nominates a testable hypothesis that miR-582-5p may be related to CD81 regulation in osteosarcoma; however, circulating miRNA levels cannot be used as a surrogate for tumor-intracellular miRNA activity. (page 11, lines 319-324)” and “Therefore, the proposed miR-582-5p–CD81 relationship should be regarded as hypothesis-generating, and will require validation in matched tumor tissues (and ideally paired plasma/exosomal fractions) as well as functional assays (e.g., luciferase reporter and gain-/loss-of-function experiments). (page 12, lines 342-345)” and “Accordingly, we do not claim that plasma miR-582-5p directly regulates CD81 in osteosarcoma tissues based on the present data alone. Instead, we present circulating miR-582-5p as a non-invasive biomarker candidate and propose the miR-582-5p–CD81 pair as a testable regulatory hypothesis to be examined in tissue- and cell-based systems. (page 12, lines 356-360)”.

6-References:

A-Formatting: The reference style needs to be thoroughly checked against the journal's rules because it is inconsistent.

Response:

We thank the reviewer for this comment. The entire reference list has been thoroughly reviewed and revised to ensure full compliance with the journal’s reference formatting guidelines.

B-Inconsistency: The DOI is included in some references but not in others.

Response:

We agree with the reviewer. To ensure consistency, DOIs have now been uniformly presented in all references throughout the manuscript, in accordance with the journal’s reference style.

C-Journal names are inconsistently abbreviated (e.g., "Nucleic Acids Res" vs. full names).

Response:

We have standardized journal titles across the reference list. All journal names are now presented consistently using the journal-approved abbreviated format.

D-Punctuation: Verify the author lists' punctuation and spacing.

Response:

We have carefully reviewed and corrected author name formatting, punctuation, and spacing throughout the reference list to ensure consistency and adherence to journal standards.

Szklarczyk, D.; Franceschini, A.; Kuhn, M.; Simonovic, M.; Roth, A.; Minguez, P.; Doerks, T.; Stark, M.; Muller, J.; Bork, P.; et al. The STRING database in 2011: functional interaction networks of proteins, globally integrated and scored. Nucleic Acids Res. 2011, 39, D561-568, doi:10.1093/nar/gkq973.
Szklarczyk, D.; Kirsch, R.; Koutrouli, M.; Nastou, K.; Mehryary, F.; Hachilif, R.; Gable, A.L.; Fang, T.; Doncheva, N.T.; Pyysalo, S.; et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res. 2023, 51, D638-646, doi:10.1093/nar/gkac1000.
Shannon, P.; Markiel, A.; Ozier, O.; Baliga, N.S.; Wang, J.T.; Ramage, D.; Amin, N.; Schwikowski, B.; Ideker, T. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13, 2498-2504, doi:10.1101/gr.1239303.
Chin, C.H.; Chen, S.H.; Wu, H.H.; Ho, C.W.; Ko, M.T.; Lin, C.Y. cytoHubba: identifying hub objects and sub-networks from complex interactome. BMC Syst Biol 2014, 8 Suppl 4, S11, doi:10.1186/1752-0509-8-S4-S11.

Reviewer 2 Report

Comments and Suggestions for Authors

This is an interesting and important study, relevant to the field. The article is well written, clear, and easy to read. The experimental design, data analysis, and data availability statements are adequate. The reference list is appropriate and adequate. This article may be accepted for publication with minor revisions.

The authors write: "The apparent paradox—elevated plasma miR-582-5p levels in OS patients despite CD81 downregulation—likely reflects fundamental differences in miRNA dynamics between systemic circulation and the tumor microenvironment." It is unclear what the paradox is—high plasma miR-582-5p levels can lead to low tissue CD81 levels if miR-582-5p suppresses CD81 expression by binding to its mRNA. The authors are asked to clarify their thinking.
The authors write: "Despite advances in chemotherapy and limb-sparing surgery, overall survival has plateaued since the 1980s, especially in metastatic or relapsed cases." It would be helpful to add information on this issue—provide survival figures, describe hypotheses for the stagnation of progress in this area.

Author Response

The authors focus on the identification of prognostic biomarkers for osteosarcoma. This in silico study identified CD81 as a potential prognostic biomarker in osteosarcoma and hsa-miR-582-5p as a candidate microRNA putatively targeting CD81. The findings provide new information, but they require further experimental confirmation. This is an interesting and important study, relevant to the field. The article is well written, clear, and easy to read. The experimental design, data analysis, and data availability statements are adequate. The reference list is appropriate and adequate. This article may be accepted for publication with minor revisions.

The authors write: "The apparent paradox—elevated plasma miR-582-5p levels in OS patients despite CD81 downregulation—likely reflects fundamental differences in miRNA dynamics between systemic circulation and the tumor microenvironment." It is unclear what the paradox is—high plasma miR-582-5p levels can lead to low tissue CD81 levels if miR-582-5p suppresses CD81 expression by binding to its mRNA. The authors are asked to clarify their thinking.

Response:

We appreciate this comment and agree that the term “paradox” was unclear. Our intention was not to imply a biological contradiction, but rather an interpretational limitation: the miRNA signal was derived from plasma, whereas CD81 downregulation was observed in tissue/cell transcriptomes. Because circulating miRNAs may originate from multiple cell types and may not reflect the tumor-intracellular miRNA pool, we revised the wording to explicitly state this caveat and to avoid over-interpreting plasma miR-582-5p as direct evidence of tissue-level CD81 regulation as following: "An important interpretational caveat is that the miRNA and mRNA signals were derived from different biospecimens (circulating plasma miRNAs vs. tissue transcriptomes). Therefore, elevated plasma miR-582-5p cannot be assumed to reflect the tumor-intracellular miR-582-5p pool or to indicate direct regulation of CD81 in tumor tissues without matched tissue validation. (page 12, lines 345-349)".

The authors write: "Despite advances in chemotherapy and limb-sparing surgery, overall survival has plateaued since the 1980s, especially in metastatic or relapsed cases." It would be helpful to add information on this issue—provide survival figures, describe hypotheses for the stagnation of progress in this area.

Response:

Thank you for this helpful suggestion. We have revised the Introduction to provide quantitative survival figures and to better contextualize the long-recognized survival plateau in osteosarcoma as following: “Historically, osteosarcoma outcomes improved substantially after the introduction of multi-agent chemotherapy; however, survival gains have largely plateaued since the late 1980s–1990s. Contemporary series report ~60–70% 5-year survival for localized disease treated with standard multimodality therapy, whereas outcomes for patients presenting with metastasis remain poor (approximately 10–40% survival), with pulmonary (micro)metastatic disease being a major driver of treatment failure. These limitations highlight the need for improved risk stratification and molecular biomarkers to better understand aggressive biology and guide future translational studies [1]. (page 2, lines 45-53)”.

Belayneh, R.; Fourman, M.S.; Bhogal, S.; Weiss, K.R. Update on Osteosarcoma. Curr Oncol Rep 2021, 23, 71, doi:10.1007/s11912-021-01053-7.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have no further comments and believe the manuscript is now ready for publication.