Translational Molecular and Fluid Biomarkers for Age-Related Macular Degeneration: Practical Insights from Animal Models and Humans

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The review manuscript provides a comprehensive summary of molecular changes reported in animal models and patients with age-related macular degeneration (AMD). In particular, it highlights novel biomarkers in tear fluid and the gut microbiome, as well as genetic variations that may influence disease progression. Overall, the manuscript is highly informative and well-organized. Below are a few minor suggestions for the authors to consider:

• Definition of biomarkers: Systemic biofluid- or gut-derived biomarkers are important for monitoring AMD progression and assessing therapeutic outcomes. However, several of the described molecular changes—such as those in Sections 2.1 and 3.1—are not directly measurable in patients or animal models. While these are clearly relevant to disease pathogenesis, they cannot serve as practical biomarkers. The authors may consider revising the section titles, as well as the overall manuscript title, to better reflect this distinction.
• miRNA specificity (Lines 115–116): Some miRNAs may be up- or downregulated in hypoxia-induced retinal damage. Could the authors comment on whether these changes are cell-type specific, and if so, provide examples?
• IRAK-M model: The studies involving IRAK-M utilized a rather unique light toxicity model. A brief discussion of the rationale and limitations of this model would be helpful for readers.
• Complement system differences (Lines 227–229): Since rodents and humans differ substantially in their complement systems, this may partly explain why complement components are often not measured in rodent models. The authors may wish to comment on this distinction.

Author Response

Author's Reply to the Review Report (Reviewer 1)

The review manuscript provides a comprehensive summary of molecular changes reported in animal models and patients with age-related macular degeneration (AMD). In particular, it highlights novel biomarkers in tear fluid and the gut microbiome, as well as genetic variations that may influence disease progression. Overall, the manuscript is highly informative and well-organized. Below are a few minor suggestions for the authors to consider:

 

1. Definition of biomarkers: Systemic biofluid- or gut-derived biomarkers are important for monitoring AMD progression and assessing therapeutic outcomes. However, several of the described molecular changes—such as those in Sections 2.1 and 3.1—are not directly measurable in patients or animal models. While these are clearly relevant to disease pathogenesis, they cannot serve as practical biomarkers. The authors may consider revising the section titles, as well as the overall manuscript title, to better reflect this distinction.

We thank the Reviewer for the insightful and constructive comment regarding the definition of biomarkers in our manuscript. We appreciate the observation about the importance of differentiating systemic biofluid- or gut-derived biomarkers, which are directly measurable and clinically applicable, from molecular changes described in Sections 2.1 and 3.1 that, while relevant to disease pathogenesis, are not practical biomarkers measurable in patients or animal models.

In response, we have revised the manuscript title and section headings to better reflect this important distinction. The new manuscript title is: "Translational Molecular and Fluid Biomarkers for Age-Related Macular Degeneration: Practical Insights from Animal Models and Humans."

Additionally, to improve clarity, we have modified the section titles as follows:

• sections that discuss molecular and tissue-level mechanistic markers not directly measurable in vivo have been clearly labeled to reflect their mechanistic and exploratory nature,
• sections that focus on systemic, biofluid, and stool biomarkers that can be directly measured in patients or models have been titled to emphasize their practical biomarker potential.

These changes help to distinguish the underlying molecular mechanisms from the practical biomarkers that are suitable for monitoring disease and assessing therapy.

We hope these revisions satisfactorily address your concern and improve the clarity and translational relevance of our review. We thank you again for your valuable guidance.

 

2. miRNA specificity (Lines 115–116): Some miRNAs may be up- or downregulated in hypoxia-induced retinal damage. Could the authors comment on whether these changes are cell-type specific, and if so, provide examples?

We appreciate the valuable comment of the Reviewer regarding the cell-type specificity of miRNA regulation in hypoxia-induced retinal damage. We have clarified this aspect in the revised manuscript by adding the following text in page 5, paragraph 3.1.1:

“Notably, miR-21 is predominantly upregulated in retinal pigment epithelium (RPE) cells, where modulates necroptosis; miR-183 is chiefly expressed in retinal neurons, whereas miR-210 is induced in photoreceptors and retinal vascular endothelial cells under hypoxic conditions.”

This addition aims to highlight the distinct cellular expression patterns and functional implications of these miRNAs within the hypoxic retina. We trust this revision addresses your comment adequately.

 

3. IRAK-M model: The studies involving IRAK-M utilized a rather unique light toxicity model. A brief discussion of the rationale and limitations of this model would be helpful for readers.

Thank you for your valuable suggestion regarding the IRAK-M studies and the light toxicity model employed. We agree that a brief discussion of the rationale and limitations of this model will enhance reader understanding. Accordingly, we have added the following text at page 8 lines 336 in the revised manuscript:

“The studies involving IRAK-M utilized a rather unique light toxicity model. The rationale for this model is that light exposure induces acute oxidative and immune stress in the retinal pigment epithelium, simulating environmental risk factors for AMD and allowing the assessment of IRAK-M’s role in regulating retinal homeostasis and vulnerability. However, the model represents an acute injury rather than the chronic and gradual progression typical of human AMD. Moreover, the intensity of light exposure does not reflect normal physiological aging conditions. These limitations should be considered when interpreting the results.”

We hope that the changes we have implemented effectively address your comment and improve the clarity of the manuscript.

4. Complement system differences (Lines 227–229): Since rodents and humans differ substantially in their complement systems, this may partly explain why complement components are often not measured in rodent models. The authors may wish to comment on this distinction.

We appreciate the insightful comment regarding the differences between the complement systems in rodents and humans, which may influence biomarker measurement in mouse models. To address this, we have added the following text to the relevant paragraph in lines 363–371 (page 9):

"Rodents exhibit genetic, structural, and functional differences in key complement proteins and their regulators, which affect the localization, expression, and activation pathways of complement components. These interspecies disparities reduce the translational comparability of rodent data to human AMD and likely contribute to the limited focus on complement biomarkers in murine studies. Such differences must be carefully considered when interpreting complement-related data and when developing complement-targeted therapies based on mouse models."

We believe this addition highlights the critical translational limitations and enhances the clarity of the discussion section.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript by Intonti et al. is a well-written, comprehensive review cited with recent work in this field. Authors have done a nice work compiling all the molecular biomarkers reported in different tissues in preclinical mouse models and human patients. However, I think summarizing the different biomarkers in the form of a Table would be extremely helpful for the readers/fellow researchers.

Authors are requested to address the following issues:

1. I would emphasize on summarizing the reported findings in preclinical mouse models and human patients in Table format; either single or split into two.
2. The whole review could be divided into ‘Preclinical mouse models’ and ‘Human studies’. Currently there is no clear demarkation which affects the flow.
3. Since human AMD section has been written as dry and wet-AMD, mouse models section could be divided in similar way instead of Genetic and laser-neovascularization models.
4. For consistency and clarity, I would suggest changing the section titles as: (2) Biomarkers in chemically-induced AMD mouse models, (3) Biomarkers in dry-AMD mouse models, (4), Bio……wet AMD mouse models (instead of Laser-induced neovasc. model). The numbering would change if the review were divided as per second comment.

5. Include a brief paragraph of major advantages and limitations for each mouse model (chemically-induced and dry AMD mouse models), just like in laser-neovascularization model.
6. In Biomarkers in wet-AMD mouse model (i.e. laser-neovascularization model; section 4): include Vldr^-/- mouse model (line 178-180) too and add the high variability associated with laser-CNV mouse model (line 264-266). If separating the dry-AMD and wet-AMD mouse models (as per comment #4), remove Vldr^-/- mouse model information from line 178-180. 
7. Use the correct mouse gene/strain nomenclature, genus name and species epithet nomenclature throughout the manuscript.
8. Section 5.2: Add serum markers (cytokines, C-reactive proteins, miRNAs etc.) in text and heading.
9. Figure 4: Add ‘genetic’ component of dry-AMD in humans
10. Replace section 6.6 with 6.5 to maintain consistency throughout.
11. Fig. 5: Remove stool from ‘Gut and stool’, since stool-biomarkers are not yet reported.
12. Figures 1-3: Replace with pigmented mice, since albino mice are not used as AMD models.
13. The figures look repetitive; they could be more concise and clearer. I would suggest merging of the mouse model figures (Fig. 1-3), showing biomarkers for different AMD models (from various tissue) alongside each other. The same should be done for human studies; merge Fig. 4-5. That would be two figures in total.
14. All figures: correct ‘stoole’ to “stool” and ‘vitreous umor’ to “vitreous humor”.

Author Response

Author's Reply to the Review Report (Reviewer 2)

The manuscript by Intonti et al. is a well-written, comprehensive review cited with recent work in this field. Authors have done a nice work compiling all the molecular biomarkers reported in different tissues in preclinical mouse models and human patients. However, I think summarizing the different biomarkers in the form of a Table would be extremely helpful for the readers/fellow researchers.

 

Authors are requested to address the following issues:

1. I would emphasize on summarizing the reported findings in preclinical mouse models and human patients in Table format; either single or split into two.

We thank the Reviewer for the helpful suggestion. We agree that tables can provide a clear and concise overview of the discussed data. However, since the manuscript already contains detailed figures presenting this information, creating a table would largely reiterate the same data.

To avoid redundancy and preserve the flow of the main text, we propose to include the below summary table in the supplementary material only if you or the editor consider it necessary. This approach allows us to maintain manuscript clarity while providing an additional resource upon request.

2. The whole review could be divided into ‘Preclinical mouse models’ and ‘Human studies’. Currently there is no clear demarkation which affects the flow.

We acknowledge the significance of clear structural organization in facilitating comprehension and enhancing the coherence of the manuscript. In accordance with the recommendation of the Reviewer, a comprehensive reorganization of the manuscript has been undertaken, distinctly delineating the sections pertaining to the discussion of ‘Preclinical Mouse Models’ and those addressing ‘Human Studies.’ This restructured framework substantially improves the narrative flow and enables a more precise juxtaposition of findings from these two domains.

3. Since human AMD section has been written as dry and wet-AMD, mouse models section could be divided in similar way instead of Genetic and laser-neovascularization models.

We appreciate the Reviewer's suggestion. In the revised manuscript, the section regarding mouse models has been reorganized to parallel the structure used for human AMD. Specifically, the models are now categorized into dry and wet AMD models. This organization allows for enhanced consistency between the preclinical and clinical parts of the review and improves the overall clarity and logical flow of the manuscript.

4. For consistency and clarity, I would suggest changing the section titles as: (2) Biomarkers in chemically-induced AMD mouse models, (3) Biomarkers in dry-AMD mouse models, (4), Bio……wet AMD mouse models (instead of Laser-induced neovasc. model). The numbering would change if the review were divided as per second comment.

We thank the Reviewer for the helpful recommendation regarding section titles. As suggested, we have adjusted the titles to improve clarity and uniformity. The updated section titles are now: “Biomarkers in chemically induced dry AMD mouse models,” “Biomarkers in genetic dry AMD mouse models,” and “Biomarkers in laser-induced neovascular wet AMD mouse models.” The numbering of these sections has been revised to reflect these modifications and to maintain consistency throughout the manuscript.

5. Include a brief paragraph of major advantages and limitations for each mouse model (chemically-induced and dry AMD mouse models), just like in laser-neovascularization model.

We acknowledge the importance of clearly presenting the strengths and limitations of the various mouse models used in AMD research. In response to your suggestion, we have added brief but comprehensive paragraphs summarizing the major advantages and limitations of the chemically-induced and genetic dry AMD mouse models, comparable to the description provided for the laser-neovascularization model section.

At Lines 148 - 159, the chemically-induced AMD mouse model section now includes the following: “These models have several advantages, including the ability to rapidly induce disease and study the effects of oxidative stress and inflammation on retinal structure and function. However, chemically induced models primarily cause acute damage and may not accurately replicate the chronic, multifactorial pathophysiology of human dry AMD, particularly with regard to the disease's progressive and age-related nature. Furthermore, pathological features such as drusen formation, geographic atrophy progression and systemic metabolic alterations are not fully mirrored by these models. Despite these limitations, chemically induced models remain valuable tools for screening candidate therapeutics, investigating acute molecular pathways and establishing biomarkers in preclinical research.”

Similarly, at Lines 287- 301, the genetic dry AMD mouse model section now contains:

“Despite their usefulness, genetically engineered mouse models have both strengths and limitations. The main advantage of these models is the ability to manipulate specific genes in order to elucidate causal mechanisms and to recreate distinct pathological features of dry AMD in a controlled environment. These models are also employed in the preclinical testing of potential drugs and to study cell-cell interactions within the retinal microenvironment. However, significant differences in ocular anatomy and lifespan between mice and humans limit the direct application of findings, particularly with regard to the macula, which is absent in rodents. Furthermore, most models only reproduce selected aspects of AMD, rather than the complete disease spectrum that evolves over decades in humans. Nevertheless, these models remain fundamental resources for unravelling disease mechanisms, identifying therapeutic targets, and advancing translational research in the AMD field.

6. In Biomarkers in wet-AMD mouse model (i.e. laser-neovascularization model; section 4): include Vldr^-/-mouse model (line 178-180) too and add the high variability associated with laser-CNV mouse model (line 264-266). If separating the dry-AMD and wet-AMD mouse models (as per comment #4), remove Vldr^-/- mouse model information from line 178-180. 

Following your suggestion, we have removed the Vldr^-/- mouse model information from lines 178-180, as these details pertain to genetic models inducing dry AMD and are therefore out of context in the wet-AMD model section.

7. Use the correct mouse gene/strain nomenclature, genus name and species epithet nomenclature throughout the manuscript.

The manuscript has undergone a thorough review to ensure the accurate use of mouse gene and strain nomenclature, as well as the correct formatting of genus and species epithets throughout the text. All nomenclature now complies with the current international guidelines and standards for mouse genetics, as outlined by the Mouse Genome Informatics (MGI) database and relevant nomenclature committees.

8. Section 5.2: Add serum markers (cytokines, C-reactive proteins, miRNAs etc.) in text and heading.

We appreciate the insightful recommendation to include serum markers such as cytokines, C-reactive proteins, and miRNAs in Section 4.1.2. In response, we have expanded this section to incorporate these important biofluid biomarkers in the text. Additionally, the section heading has been updated to accurately reflect the expanded scope.

The added content discusses the significance of serum cytokines (e.g., IL-6, IL-8, TNF-α), C-reactive protein as a marker of systemic inflammation and circulating miRNAs such as miR-146a and miR-34a as potential non-invasive biomarkers for AMD progression. We also highlight emerging evidence from tear and aqueous humor biomarkers, along with novel insights into stool metabolomic profiles implicating the gut-retina axis in AMD pathology.

We believe these additions provide a more comprehensive and clinically relevant overview of biofluid biomarkers in dry AMD.

9. Figure 4: Add ‘genetic’ component of dry-AMD in humans.

We thank the reviewer for this helpful suggestion. The genetic component of dry AMD in humans has now been added to the figure as requested.

10. Replace section 6.6 with 6.5 to maintain consistency throughout.

We appreciate your suggestion to reorder Sections 6.5 and 6.6 to improve consistency. Accordingly, the manuscript has been updated to present the paragraph on the gut microbiome before the genetic biomarker section, thereby aligning the structure with the rest of the review.

11. 5: Remove stool from ‘Gut and stool’, since stool-biomarkers are not yet reported.

Following your recommendation, we have removed "stool" from all paragraphs and section headings referencing the gut microbiome and related biomarkers, as stool biomarkers have not yet been reported for AMD. This change enhances the accuracy and clarity of the manuscript.

12. Figures 1-3: Replace with pigmented mice, since albino mice are not used as AMD models.

We have replaced the images of albino mice with those of pigmented (black) mice in Figures 1-3, as pigmented strains are more appropriate models for AMD research.

13. The figures look repetitive; they could be more concise and clearer. I would suggest merging of the mouse model figures (Fig. 1-3), showing biomarkers for different AMD models (from various tissue) alongside each other. The same should be done for human studies; merge Fig. 4-5. That would be two figures in total.

We have followed your suggestion and merged Figures 1-3 into a single comprehensive figure displaying biomarkers across different mouse AMD models and tissues. Similarly, Figures 4-5 have been combined to present human study biomarkers in a unified format. This reduction to two figures improves clarity and conciseness.

14. All figures: correct ‘stoole’ to “stool” and ‘vitreous umor’ to “vitreous humor”.

As per your recommendation, “stool” has been completely removed from all text and figure references throughout the manuscript, as stool biomarkers have not been reported for AMD.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript assembles a broad, clinically grounded survey of molecular biomarkers for AMD across biofluids, tissues, and animal models. However, to fulfill its translational promise, several corrections are needed to be performed to enhance papers clarity.   

 

Major points

1. The manuscript is framed as a broad “translational overview,” but no methods are provided for literature identification, study selection, or evidence grading, which prevents readers from assessing selection bias and evidential weight. It is recommended that a brief Methods section be added specifying databases, dates, search strings, inclusion/exclusion criteria, and any risk-of-bias appraisal appropriate for a structured narrative review (PRISMA-lite), given the stated scope as a review.
2. Findings are presented qualitatively (e.g., complement fragments, CRP/IL-6/TNF-α, oxidative stress adducts, miRNAs) without effect sizes, discrimination metrics, or calibration, which limits translational interpretability. It is recommended that, for each biomarker, effect estimates (fold-change or standardized difference), diagnostic/prognostic performance (AUC, sensitivity, specificity, LR+/LR− with CIs), and model-added value over genetics/imaging be reported in a summary table.
3. The gut-microbiome narrative combines mouse and human associations without adequately discussing fragility to diet/medication confounding and causal-inference limits. The duplicated metabolomics text also appears, signaling insufficient synthesis. It is recommended that the microbiome section be tempered by adding a subsection on confounding/MR assumptions, and that duplicated sentences be removed while emphasizing which taxa/metabolites replicate across independent cohorts.
4. The blood/plasma section intermixes proteins and miRNAs without indicating incremental value over genetic risk or imaging and without clarifying which markers survive multivariable models. It is recommended that nested-model performance be reported (genetics – genetics+imaging - +molecular marker) to demonstrate additive predictive utility.

 

Minor points

1. Tear-film proteomics are highlighted based on a small discovery study (n=22) using 2D-GE/MALDI-TOF, which is hypothesis-generating and subject to reproducibility constraints; the manuscript over-generalizes the translational readiness. It is recommended that the tear section be explicitly labeled as discovery-stage and that a verification – validation pipeline be proposed (targeted PRM/MRM or ELISA with pre-registered endpoints, standardized collection/diurnal control, and powered multicenter cohorts).
2. The oxidative-stress markers (CEP, MDA, 8-OHdG) are emphasized via older literature without reconciling mixed replication and context (storage artifacts, batch effects, comorbidities). It is recommended that an evidence synthesis be added indicating which oxidative markers remain significant after multivariable adjustment across modern, well-controlled cohorts, and that inconsistent findings be explicitly summarized.
3. Redundancy/duplication appears in the gut-metabolomics description, which undermines synthesis quality and may confuse readers about evidence hierarchy. It is recommended that duplicated passages be removed and that a concise evidence hierarchy be inserted prioritizing replicated taxa/metabolites and prospective designs.
4. Several promising discovery signals (e.g., candidate tear proteins; selected miRNAs) are presented without a proposed threshold or clinical cut-point strategy, limiting decision utility. It is recommended that preliminary cut-points be proposed for validation (Youden-optimized or clinically anchored), with plans for recalibration and clinical-impact modeling.

Author Response

Author's Reply to the Review Report (Reviewer 3)

The manuscript assembles a broad, clinically grounded survey of molecular biomarkers for AMD across biofluids, tissues, and animal models. However, to fulfill its translational promise, several corrections are needed to be performed to enhance papers clarity. 

  

Major points

1. The manuscript is framed as a broad “translational overview,” but no methods are provided for literature identification, study selection, or evidence grading, which prevents readers from assessing selection bias and evidential weight. It is recommended that a brief Methods section be added specifying databases, dates, search strings, inclusion/exclusion criteria, and any risk-of-bias appraisal appropriate for a structured narrative review (PRISMA-lite), given the stated scope as a review.

Thank you for your valuable comment regarding the absence of a detailed Methods section describing literature identification, study selection, and evidence appraisal. We agree that such transparency is important for assessing potential biases and weighting of evidence in a translational overview.

In response, we have included a dedicated Methods section titled "METHODS" in the revised manuscript. This section specifies the databases searched (PubMed, Scopus, Web of Science), the date range covered (from inception to July 2025), and the key search terms employed. We also describe the inclusion and exclusion criteria for selecting studies focused on molecular biomarkers in AMD, encompassing both human clinical and preclinical animal model research. We further elaborated on the qualitative approach to evidence grading, outlining factors considered such as study design, cohort size, validation status of biomarkers, and replication across independent cohorts. Our narrative synthesis approach is also detailed, emphasizing the integration of mechanistic and translational insights across multiple biological sample types. While this is a structured narrative review rather than a systematic review or meta-analysis, we have taken care to follow principles consistent with a "PRISMA-lite" approach, ensuring methodical literature identification and transparent reporting of our selection and appraisal strategy.

We believe the added Methods section enhances the rigor, reproducibility, and interpretability of our review and satisfactorily addresses the reviewer’s concern.

 

2. Findings are presented qualitatively (e.g., complement fragments, CRP/IL-6/TNF-α, oxidative stress adducts, miRNAs) without effect sizes, discrimination metrics, or calibration, which limits translational interpretability. It is recommended that, for each biomarker, effect estimates (fold-change or standardized difference), diagnostic/prognostic performance (AUC, sensitivity, specificity, LR+/LR− with CIs), and model-added value over genetics/imaging be reported in a summary table.

We appreciate the reviewer’s insightful suggestion to present effect sizes and diagnostic/prognostic metrics to enhance the translational interpretability of our review.

Given the heterogeneous nature of the literature on AMD biomarkers, data reporting varies widely. While some clinical studies and preclinical models provide quantitative measures such as fold-changes, receiver operating characteristic (ROC) curves (AUC), sensitivity, specificity, and likelihood ratios, many others present qualitative or semi-quantitative findings without standardized metrics.

To address this, we have compiled a summary table highlighting key biomarkers for which such quantitative data are available, including effect size estimates and diagnostic performance metrics where reported. This table also provides information on the biological source and any known added value beyond existing genetic or imaging markers.

For biomarkers lacking robust quantitative performance data, we provide a concise narrative summary in the main text, acknowledging the need for further validation.

This approach aims to balance comprehensiveness with clarity, offering readers a transparent and practical overview of biomarker utility while identifying gaps for future research.

We hope this revision satisfactorily addresses the reviewer’s recommendation.

 

 

3. The gut-microbiome narrative combines mouse and human associations without adequately discussing fragility to diet/medication confounding and causal-inference limits. The duplicated metabolomics text also appears, signaling insufficient synthesis. It is recommended that the microbiome section be tempered by adding a subsection on confounding/MR assumptions, and that duplicated sentences be removed while emphasizing which taxa/metabolites replicate across independent cohorts.

We thank the reviewer for this insightful comment. In response, we have condensed the relevant paragraph and introduced a new subsection (4.2.5.1) that specifically addresses potential confounding factors and Mendelian randomization assumptions. This new section also highlights the common microbial taxa identified across cohorts to enhance clarity and coherence.

 

4. The blood/plasma section intermixes proteins and miRNAs without indicating incremental value over genetic risk or imaging and without clarifying which markers survive multivariable models. It is recommended that nested-model performance be reported (genetics – genetics+imaging - +molecular marker) to demonstrate additive predictive utility.

We thank the Reviewer for this constructive comment. We agree that assessing the incremental predictive value of systemic biomarkers beyond genetic or imaging markers, and clarifying which remain significant in multivariable models, is crucial for evaluating translational potential. As our work is a narrative review and not a primary data analysis, we were unable to perform nested-model comparisons directly. To address this, we have revised the Discussion section and highlight that few available datasets formally test additive predictive performance. We now note that, while some circulating markers such as complement C3, factor B, and IL-6 show associations independent of genetic risk, quantitative nested-model data (e.g., ΔAUC or reclassification metrics) are not yet systematically available in either the NaIO₃ model or human AMD cohorts. This limitation has been explicitly acknowledged in the revised manuscript.

 

Minor points

1. Tear-film proteomics are highlighted based on a small discovery study (n=22) using 2D-GE/MALDI-TOF, which is hypothesis-generating and subject to reproducibility constraints; the manuscript over-generalizes the translational readiness. It is recommended that the tear section be explicitly labeled as discovery-stage and that a verification – validation pipeline be proposed (targeted PRM/MRM or ELISA with pre-registered endpoints, standardized collection/diurnal control, and powered multicenter cohorts).

Thank you for this valuable comment and for highlighting the need to avoid overgeneralization of discovery-stage findings from small tear-film proteomics studies. We fully agree that these results are hypothesis-generating and require further robust validation prior to translational application.

In response, we have undertaken the following actions:

• In all relevant paragraphs (Sections 3.3.3, 4.1.3, 4.2.3), we have explicitly labeled findings from tear-fluid proteomic studies as “discovery-stage.” We now clarify that these studies are exploratory, with conclusions restricted by the limited sample size (n=22) and use of 2D-GE/MALDI-TOF techniques.
• We have removed any statements that imply immediate translational readiness and instead emphasized that current evidence is hypothesis-generating and subject to reproducibility constraints.
• We now outline a proposed verification–validation pipeline in each section that discusses tear biomarkers. Specifically, we recommend targeted quantification using PRM/MRM or ELISA, registration of prespecified endpoints, harmonized and diurnally controlled sample collection, and inclusion of powered, multicenter cohorts as essential next steps.
• The language throughout these sections has been revised to explicitly state the need for further validation and to provide clear guidance for future studies, as per your recommendation.

These changes are found on pages 12, 19, 23 in paragraphs 3.3.3 (Mouse Model, Tear Fluid Biomarkers), 4.1.3 (Human, Dry AMD, Tear Fluid Biomarkers), and 4.2.3 (Human, Wet AMD, Tear Fluid Biomarkers), with all insertions and modifications clearly indicated in the tracked changes version. 

 

2. The oxidative-stress markers (CEP, MDA, 8-OHdG) are emphasized via older literature without reconciling mixed replication and context (storage artifacts, batch effects, comorbidities). It is recommended that an evidence synthesis be added indicating which oxidative markers remain significant after multivariable adjustment across modern, well-controlled cohorts, and that inconsistent findings be explicitly summarized.

We thank the reviewer for highlighting the need to reconcile existing literature on oxidative stress markers with mixed replication results and methodological contexts. We recognize that this is a critical issue for advancing biomarker research in AMD.

To thoroughly address this point, we have expanded the manuscript with two focused statements in the relevant sections:

• In Section 4.1.1. (Non directly Measurable Tissue Mechanistic: markers from pathology, page 17, lines 689- 698), we discuss the role of tissue-level oxidative markers such as 8-OHdG and CEP adducts, emphasizing their biological relevance but also the significant technical and sampling challenges that complicate data interpretation and reproducibility at the tissue level.
• In Section 4.1.2 (Directly Measurable Serum and Plasma Biomarkers, page 18 lines 716- 722), we synthesize current evidence from recent well-controlled studies, noting that while systemic markers like MDA, CEP, and 8-OHdG were frequently elevated in older studies, their association with AMD after rigorous adjustment for storage artifacts, batch effects, and comorbidities is inconsistent. We stress the importance of large multicenter cohorts, harmonized sample processing, and multivariable analysis for future validation.

These additions provide the intended evidence synthesis, explicitly acknowledge inconsistent findings, and align the manuscript with contemporary standards of biomarker validation. We believe these revisions substantially improve the rigor and clarity of our discussion on oxidative stress markers in AMD.

We appreciate this opportunity to clarify these points and trust that these modifications satisfactorily address the reviewer's concerns.

 

3. Redundancy/duplication appears in the gut-metabolomics description, which undermines synthesis quality and may confuse readers about evidence hierarchy. It is recommended that duplicated passages be removed and that a concise evidence hierarchy be inserted prioritizing replicated taxa/metabolites and prospective designs.

We have revised the description of the gut metabolomics accordingly. We have modified the text to improve readability and highlight recurring taxa in the new section 4.2.5.1. Additionally, we have reported the recurring taxa in the figures to facilitate visualisation of the differences and overlaps between the various models.

 

4. Several promising discovery signals (e.g., candidate tear proteins; selected miRNAs) are presented without a proposed threshold or clinical cut-point strategy, limiting decision utility. It is recommended that preliminary cut-points be proposed for validation (Youden-optimized or clinically anchored), with plans for recalibration and clinical-impact modeling.

We thank the reviewer for this insightful suggestion. As our manuscript is a review summarizing existing literature, we do not present new patient-level data and cannot therefore define novel cut-points ourselves. However, we agree that establishing thresholds is critical for clinical translation. In response to this comment, we have added a paragraph to the Discussion section to highlight this point. We emphasize that, although numerous candidate biomarkers for AMD, including tear proteins and microRNAs, have been identified, few studies have established standardized clinical thresholds. We also discuss the importance of defining preliminary cut-points using approaches such as Youden index optimisation or clinically anchored thresholds, and we outline our plans for recalibration and clinical impact modelling in future studies. These additions clarify the current state of the field and emphasise the need for further validation to improve translational utility.