Hypoxia and DNA-Repair Radiosensitivity Signatures Are Associated with Radiotherapy-Modified Survival in TCGA Breast Cancer, with External Prognostic Validation of the Hypoxia Score in METABRIC

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study addresses radiotherapy as a main treatment for breast cancer to decrease recurrence and improve long-term outcomes of patients in many stages of this disease. Research highlights different responses between patients.

 

Retrospective analysis was conducted. It is emphasised that both tumour hypoxia and intrinsic radiosensitivity play a vital role in patients’ responses to radiotherapy. Researchers built a 563-gene hypoxia meta-signature. Cox screening was conducted. CP, GPC3, STC1 - hypoxia score was determined. In addition, key regulators of intrinsic radiosensitivity were assessed. Hypoxia scores were calculated.  

 

Clear inclusion and exclusion criteria were provided, and statistical analyses were thoroughly described and conducted.

 

Future studies are proposed as a follow-up to this research to confirm independent RT-annotated datasets while maintaining similar timing, dose, and systemic therapy, to provide a rationale for biomarker trials.

 

More recent references are needed, please; the latest is dated 2023 (the others are ‘older’).

Author Response

Comment: More recent references are needed, please; the latest is dated 2023 (the others are ‘older’).

Response: We thank the reviewer for this helpful suggestion. We have now updated the manuscript to include more recent literature relevant to tumour radiosensitivity biomarkers, breast-cancer radioresistance, and biomarker-guided radiotherapy interpretation. Specifically, we added and cited Bleaney et al. (2024), a systematic review of clinical biomarkers of tumour radiosensitivity and radiotherapy benefit, and Mageau et al. (2025), a systematic review of molecular mechanisms of radiation resistance and radiosensitisation strategies in breast cancer. These references were incorporated into the Introduction to strengthen the contemporary rationale for the work and into the Discussion to place our findings in the context of current translational radiobiology and radiosensitisation strategies.

Changes made in manuscript: Introduction revised to include recent radiotherapy biomarker and radiosensitisation references.

• Discussion revised to place findings in the context of recent breast cancer radiosensitivity literature.
• Reference list updated to include new references [50] and [51].

Reviewer 2 Report

Comments and Suggestions for Authors

Interesting article; the data collected could have a future impact on a future strategy for exacerbation or descaling of radiotherapy in breast cancer.
The work would be more complete, however, if you could trace data on doses and timing of radiotherapy. A useful table.
Furthermore, talking about breast cancer in general is misleading; it's a huge entity. I would also like a table showing the histology of the cases in question. I assume your study only included carcinomas, which in turn are divided into ductal, lobular, and not otherwise specified. Sarcomas, on the other hand, have a different radiobiology and are much more radioresistant (it's worth mentioning in the introduction that you excluded this type of tumor and citing at least one reference DOI: 10.23937/2378-3419/1410114).
If possible, please include more details that would enrich your interesting work.
Thank you.

Author Response

Comment: The work would be more complete, however, if you could trace data on doses and timing of radiotherapy. A useful table.

Response: We agree and thank the reviewer. We revisited the TCGA clinical XML-derived tables and extracted available radiotherapy-related metadata from the nested radiation fields. We added a new descriptive table summarising the availability of radiotherapy annotation within the RT-defined subset. The revised manuscript now reports that the RT-defined subset contained 170 patients, of whom 143 had evidence of RT and 27 had no evidence of RT. Within this subset, radiation type was available for 150 patients, radiation dosage for 141, days to RT start for 146, and days to RT end for 147. We also harmonised recorded dose units where possible and summarised the harmonised dose distribution descriptively. After unit harmonisation, the median recorded dose was 60 Gy (IQR 50.4–64 Gy). Because the TCGA fields still showed heterogeneous and occasionally extreme values, these metadata were used descriptively only rather than for formal dose-response modelling.

Changes made in manuscript: Materials and Methods, Section 2.2.4 substantially expanded.

• Added Table 2: Availability and descriptive summary of radiotherapy metadata in the RT-defined subset (n = 170).
• Discussion updated to clarify that RT metadata were suitable for descriptive characterisation but not sufficiently standardised for formal dose-response modelling.

 

Comment: Furthermore, talking about breast cancer in general is misleading; it's a huge entity. I would also like a table showing the histology of the cases in question. I assume your study only included carcinomas, which in turn are divided into ductal, lobular, and not otherwise specified. Sarcomas, on the other hand, have a different radiobiology and are much more radioresistant (it's worth mentioning in the introduction that you excluded this type of tumor and citing at least one reference DOI: 10.23937/2378-3419/1410114).

Response: We agree with the reviewer. We revised the manuscript to make clear that this analysis concerns primary breast carcinoma cases represented within TCGA-BRCA rather than breast malignancies in the broadest possible sense. We also extracted and summarised histological subtype information from the TCGA clinical annotation. In the analysed overall OS cohort (n = 266), the cohort was dominated by carcinoma subtypes, including infiltrating ductal carcinoma (187/266, 70.3%) and infiltrating lobular carcinoma (50/266, 18.8%), with the remaining 29 cases (10.9%) classified as other carcinoma types. These data were added as a clinicopathologic summary table. We also revised the Introduction to explicitly state that the study was restricted to primary breast carcinoma cases within TCGA-BRCA and that breast sarcomas were not the target disease context of this analysis. The reviewer-suggested angiosarcoma reference has been added to support the point that secondary breast angiosarcoma is a biologically and clinically distinct entity from the carcinoma cases analysed here.

Changes made in manuscript: Introduction revised to clarify carcinoma-specific scope and non-generalisability to all breast malignancies.

• Added breast angiosarcoma reference as new citation [52].
• Added Table 1: Clinicopathologic characteristics of the TCGA-BRCA overall OS cohort (n = 266), including histology and stage summary.
• Results Section 3.1 revised accordingly.
• Discussion revised to reiterate that the analysed cohort consisted of breast carcinoma subtypes rather than the full spectrum of breast malignancies.

 

Comment: If possible, please include more details that would enrich your interesting work.

Response: We thank the reviewer. In response, we enriched the manuscript by adding a clinicopathologic summary table for the overall OS cohort, a radiotherapy metadata table for the RT-defined subset, clearer explanation of the structure and limitations of TCGA radiotherapy annotations, improved contextualisation of disease scope and radiobiological interpretation, and more recent literature relevant to radiotherapy biomarkers and radiosensitisation.

Changes made in manuscript: Added Table 1 for cohort histology and stage summary.

• Added Table 2 for radiotherapy metadata availability and descriptive statistics.
• Expanded Materials and Methods and Discussion for additional clinical and annotation detail.

Reviewer 3 Report

Comments and Suggestions for Authors

Review on "Hypoxia and DNA-Repair Radiosensitivity Signatures Are 2 Associated with Radiotherapy-Modiied Survival in TCGA Breast Cancer, with External Prognostic Validation of the Hypoxia Score in METABRIC" by Osei et al.

This study reports statistically significant associations between post–radiation therapy survival and gene expression. For the hypoxia analysis, genes were selected using Cox regression with determination of λm, resulting in three genes (CP, GPC3, and STC1). Based on these analyses, the authors conclude that high hypoxia is strongly associated with poorer survival (C-index = 0.70), suggesting that radiation therapy (RT) outcomes are adversely affected by hypoxia. These findings are consistent with current understanding.
DDR genes were analyzed similarly, and ATR, BLM, and RPA2 were selected using a LASSO-penalized model. Based on their expression profiles, the authors report that the combined interaction of hypoxia and radiosensitivity yields a strong predictive model. The results were further validated using the METABRIC database.
Kaplan–Meier analyses showed that the hypoxia signature was significantly associated with survival. The non-RT/high hypoxia group showed the poorest survival, followed by the non-RT/low hypoxia and RT/high hypoxia groups, whereas the RT/low hypoxia group showed the best overall survival, consistent with the authors’ prediction.
Overall, the results are consistent with the current understanding of the relationship among radiation, the oxygen effect, and DNA double-strand break repair. The identification of three genes each for hypoxia and DDR is interesting and may warrant further investigation toward potential clinical application. The following questions and comments are raised by this reviewer.

Comments:

The authors should examine the TCGA records and clarify whether the tissues used for RNA-seq were procured prior to radiation therapy. In clinical practice, biopsies are typically obtained at the time of diagnosis, before the initiation of RT. However, it is not clear whether all RNA-seq analyses in this study were performed using biopsies collected before RT. This point has important implications for the interpretation and prognostic value of the findings. The authors should therefore clarify the timing and source of the specimens used for RNA-seq.

The hypoxia and DDR gene expression patterns may simply reflect differences in cancer stage. Since TCGA contains clinical information on tumor stage, the authors should examine whether the expression of these gene sets is associated with BRCA stage. This analysis would help clarify whether the observed patterns are independent of disease stage.

Line 204-210: May want to include MRN (MRE11, NBS1, RAD50) genes in the set for the gene analysis.

Line 329 and 376: Are these blank parentheses for references or functions()?

Line 503: Please elaborate the mechanism more carefully. What does "fixing radiation induced DNA damage" mean exactly? The oxygen effect in radiation biology is usually understood as electrons released due to the photo-electron effect, generating reactive oxygen species that then result in clustered DNA damage along with DNA strand breaks directly generated by the high energy radiation.

 

Author Response

Comment: The authors should examine the TCGA records and clarify whether the tissues used for RNA-seq were procured prior to radiation therapy. In clinical practice, biopsies are typically obtained at the time of diagnosis, before the initiation of RT. However, it is not clear whether all RNA-seq analyses in this study were performed using biopsies collected before RT. This point has important implications for the interpretation and prognostic value of the findings. The authors should therefore clarify the timing and source of the specimens used for RNA-seq.

Response: We agree that this is an important interpretive point. We revisited the TCGA data structure and clarified this issue in the manuscript. The RNA-seq data used here were restricted to TCGA primary tumour specimens (sample type code “01”). In TCGA-BRCA, such biospecimens are generally obtained at diagnosis or surgery rather than after radiotherapy. We therefore interpreted the transcriptomic profiles as reflecting primary tumour biology prior to RT exposure. At the same time, we now state explicitly that TCGA does not provide a fully harmonised specimen-level field linking RNA-seq procurement timing directly to RT timing for every case, so we cannot formally verify pre-RT sampling for every individual patient. We have therefore framed this as a likely but not individually confirmed pre-treatment sampling context.

Changes made in manuscript: Materials and Methods, Section 2.2.1 revised to clarify specimen source and likely pre-RT interpretation.

• Discussion updated to reflect the retrospective and annotation-limited nature of TCGA sampling information.

 

Comment: The hypoxia and DDR gene expression patterns may simply reflect differences in cancer stage. Since TCGA contains clinical information on tumor stage, the authors should examine whether the expression of these gene sets is associated with BRCA stage. This analysis would help clarify whether the observed patterns are independent of disease stage.

Response: We thank the reviewer for this important suggestion. We performed an additional exploratory analysis to assess whether the derived hypoxia and radiosensitivity scores were associated with pathologic stage in TCGA-BRCA. Using Kruskal–Wallis testing across available pathologic stage categories, we found that the hypoxia score was associated with stage (P = 0.0288), whereas the radiosensitivity score was not significantly associated with stage (P = 0.1056). We therefore revised the manuscript to acknowledge that the hypoxia score may partly overlap with stage-related disease severity, whereas the radiosensitivity score appears less clearly stage-dependent in this cohort. Because of the modest sample size and limited event count, these analyses are presented as exploratory rather than definitive.

Changes made in manuscript: Added a new Results subsection on association of hypoxia and radiosensitivity scores with pathologic stage.

• Discussion revised to interpret the hypoxia signal more cautiously with respect to stage dependence.

 

Comment: Line 204–210: May want to include MRN (MRE11, NBS1, RAD50) genes in the set for the gene analysis.

Response: We thank the reviewer for this constructive mechanistic suggestion. We revised the radiosensitivity candidate set to include the MRN-complex genes MRE11A, NBN, and RAD50. After updating the candidate list and refitting the RT-trained penalised Cox model, the revised RS signature retained four genes with non-zero coefficients: ATR, RPA2, BLM, and MRE11A. The manuscript has been updated throughout to reflect this revised four-gene RS signature.

Changes made in manuscript: Materials and Methods, Section 2.5.1 updated to include MRN-complex genes.

• Results, Section 3.4 updated to report the revised four-gene RS signature.
• Abstract, Methods, Results, Discussion, Conclusion, and relevant figure text updated accordingly.

 

Comment: Line 329 and 376: Are these blank parentheses for references or functions()?

Response: Thank you for catching this. These were editing artefacts in the draft. They have been removed and the corresponding sentences have been corrected in the revised manuscript.

Changes made in manuscript: Blank parentheses removed and affected sentences edited for clarity.

 

Comment: Line 503: Please elaborate the mechanism more carefully. What does “fixing radiation induced DNA damage” mean exactly? The oxygen effect in radiation biology is usually understood as electrons released due to the photo-electron effect, generating reactive oxygen species that then result in clustered DNA damage along with DNA strand breaks directly generated by the high energy radiation.

Response: We thank the reviewer for this important clarification. We revised the mechanistic wording to avoid oversimplified phrasing. Rather than stating only that oxygen “fixes” radiation-induced DNA damage, we now explain that oxygen enhances the chemical fixation of radiation-induced DNA radical lesions, helping convert transient radical damage into more persistent, biologically consequential damage. We also state that under hypoxic conditions this process is reduced, thereby contributing to radioresistance alongside broader hypoxia-associated transcriptional programmes. This revised wording appears in both the Introduction and Discussion.

Changes made in manuscript: Introduction revised to clarify oxygen-mediated chemical fixation of radiation-induced radical damage.

• Discussion revised to explain the oxygen effect more carefully and mechanistically.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors, the changes you've made make the article more complete and strengthen your message.
Congratulations.