Review Reports - Computational Microscopy Reveals Compound-Specific Flickering Phenotypes of Red Blood Cells Under Flavonoid Exposure

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Editor, dear Authors, I’ve reviewed the paper entitled “Computational microscopy reveals compound-specific flickering phenotypes of red blood cells under flavonoid exposure”, written by Carlos del Pozo-Rojas et al., and below are my questions about it.

The authors investigate, through a novel computational microscopy approach, the flickering of erythrocytes membrane and the effect of flavonoid (quercetin, apigenin and rutin) exposure. The authors evidence the existence of different flickering spectra depending on the molecular characteristics of the treatment.

The subject of study is interesting, and the approach contains innovative elements. Although still at the stage of proof of concept, the methodology seems to have the potential to deliver more insights on the relationship between cell morphology, membrane dynamical behavior and metabolism in a variety of condition. I liked the idea of the manuscript and the experimental approach, but at the present state I believe some points have to be clarified and some statements have to be better specified before the paper can be considered for publication. In the following I report in detail my doubts (and some minor points to be fixed).

Overall: the introduction section is clear and contains sufficient information and references to properly follow the remaining part of the manuscript. The written English is proper and understandable. The conclusions have been properly presented. Some aspects of the figures need to be improved (see below)

1) On the base of the reported literature, the effect of apigenin on RBCs is not clear. Does this molecule accelerate degradative cell pathways, or does it play a protective role against oxidation-induced cell degradation?

2) At line 112-113 the reference Moleiro et al is reported as 2023, while at line 119 it is reported as a paper of 2025. Moreover, why not reporting the entire author list? Sometimes the author list is fully reported, other times there are only few authors followed by et al. Please, check the references and standardize.

3) A figure reporting the chemical formula for the three tested flavonoids would help the readability of the text (especially when discussing the molecules characteristics and the expected effects).

4) In figure 1 the label (C) is missing. Please check.

5) I need explanation about the data of figure 2 on the hemolysis rate. The figure shows that the treatment induces moderate hemolysis rate in the time window of the experiment, however it is not clear to me why the authors do not observed hemolysis with Rutin after 90’. I also need clarification about Apigenin and Quercetic, which seem to have only a transient peak at, respectively, short or intermediate incubation. It is definitely unclear the sense of a decrease in hemolysis over time, as lysed cells are lost.

6) Still on the same point: How, operatively, did the authors measure the hemolysis rate? How many times have this measurement been repeated? Why the error bars in the graph have not been reported?

7) Figure 3. Please, adjust the parameters list reported in the caption according to the 6 panels composing the image: they are not listed in the same order and sometimes have different names.

8) At lines 404-406 the authors states: “After 1 h of incubation, all analysed parameters exhibited systematic shifts consistent with metabolic reactivation following the osmotic and mechanical stress…”.

The question is: What kind of morphometric variation the authors associate to the reprise of metabolic activity? And, more interestingly, why?

9) Lines 411-412. The authors cite a progressive recovery of membrane dynamics. I just see and evolution between C0 and C1, but I don’t know if we can properly speak of “recovery” as the original condition is not defined (nor measured). Please comment.

10) The data have been collected essentially (if not exclusively) on discocytic cells. Although this choice is understandable, a non-negligible fraction of the cells have non biconcave shape (e.g. crenated cells, echinocytes or spherocytes), and this fraction is important in physiological and pathological conditions. Have some of these cells been included in the analysis (or, at least, tested)? What kind of differences the authors observed, or expect, in the dynamic phenotype and in the flickering spectra from these cells? Please, comment in the text

11) the ethic clearance statement need, probably, a more detailed statement. Please, check.

Author Response

Reviewer #1

The subject of study is interesting, and the approach contains innovative elements. Although still at the stage of proof of concept, the methodology seems to have the potential to deliver more insights on the relationship between cell morphology, membrane dynamical behavior and metabolism in a variety of condition. I liked the idea of the manuscript and the experimental approach, but at the present state I believe some points have to be clarified and some statements have to be better specified before the paper can be considered for publication. In the following I report in detail my doubts (and some minor points to be fixed).

Overall: the introduction section is clear and contains sufficient information and references to properly follow the remaining part of the manuscript. The written English is proper and understandable. The conclusions have been properly presented. Some aspects of the figures need to be improved (see below)

Response: We thank Reviewer 1 for the careful evaluation of our manuscript and for the constructive feedback. We appreciate the reviewer’s positive assessment of the novelty and potential of our computational microscopy framework, while also acknowledging the need for clarifications and improved precision before publication. In the revised version, we have addressed the reviewer’s concerns comprehensively, with particular emphasis on (i) strengthening and clarifying the hemolysis assay reporting (including presentation and interpretation of the results), (ii) correcting citation inconsistencies and standardizing reference formatting, (iii) revising figure labels and captions to ensure full alignment with the displayed content, and (iv) completing and verifying the required ethics/informed consent statements. We have also carefully considered and incorporated the remaining suggestions throughout the manuscript to improve clarity, rigor, and overall presentation.

On the base of the reported literature, the effect of apigenin on RBCs is not clear. Does this molecule accelerate degradative cell pathways, or does it play a protective role against oxidation-induced cell degradation?

Response: The literature indicates that apigenin is not uniformly “protective” or “deleterious”; rather, it shows regime-dependent behavior. Under acute oxidative challenges (H₂O₂ or tBHP), apigenin reduces hemolysis and lipid peroxidation and preserves membrane protein thiols and structural integrity, supporting a cytoprotective antioxidant role (An et al., 2015; Savko et al., 2023). Conversely, under prolonged exposure at sustained micromolar concentrations, apigenin can induce canonical markers of eryptosis (Ca²⁺i rise, shrinkage, phosphatidylserine exposure) without immediate hemolysis, consistent with activation of a degradative/clearance pathway (Zbidah et al., 2012). Therefore, the apparently unclear effect in the literature reflects a strong dependence on dose, incubation time, and oxidative context, and both outcomes are compatible within a unified framework.

We will clarify this explicitly in the manuscript and emphasize the importance of the experimental window, noting that dietary plasma levels are typically in the nanomolar range (Meyer et al., 2006).

References:

An F, Cao X, Qu H, Wang S. Attenuation of oxidative stress of erythrocytes by the plant-derived flavonoids vitexin and apigenin. Die Pharmazie. 2015;70(11):724–732. PMID: 26790189.

Savko AI, Ilyich TV, Veiko AG, Kovalenia TA, Lapshina EA, Zavodnik IB. … inhibit oxidative processes in erythrocytes. Biomed Khim. 2023;69(5):281–289. doi:10.18097/PBMC20236905281. PMID: 37937430.

Zbidah M, Lupescu A, Jilani K, et al. Apigenin-induced suicidal erythrocyte death. J Agric Food Chem. 2012;60(1):533–538. doi:10.1021/jf204107f. PMID: 22132906.

Meyer H, Bolarinwa A, Wolfram G, Linseisen J. Bioavailability of apigenin from apiin-rich parsley in humans. Ann Nutr Metab. 2006;50(3):167–172. doi:10.1159/000090736. PMID: 16407641.

At line 112-113 the reference Moleiro et al is reported as 2023, while at line 119 it is reported as a paper of 2025. Moreover, why not reporting the entire author list?

Response: We thank the reviewer for pointing out this inconsistency. We have now added the missing reference Moleiro et al., 2023 to the final reference list, which was inadvertently omitted in the initial submitted version. In addition, we have revised and standardized all citations and reference entries throughout the manuscript to ensure full consistency with the journal’s editorial guidelines, including the prescribed reference formatting and author-list presentation.

New Cite: L. H. Moleiro, M. T. Martín-Romero, D. Herráez-Aguilar, J. A. Santiago, N. Caselli, C. Dargel, R. Geisler, T. Hellweg, and F. Monroy, Dual mechanical impact of β-escin on model lipid membranes, Front. Soft Matter 3, 1240878 (2023).

Sometimes the author list is fully reported, other times there are only few authors followed by et al.

Response: Thank you for noting this. We have now standardized all references to follow a consistent author-list format throughout the manuscript, in full compliance with the journal’s editorial guidelines.

A figure reporting the chemical formula for the three tested flavonoids would help the readability of the text (especially when discussing the molecules characteristics and the expected effects).

Response: We thank the reviewer for this helpful suggestion and fully agree that it improves readability. Accordingly, we have added a new figure at the beginning of the manuscript showing the chemical structures of the three flavonoids investigated (apigenin, quercetin, and rutin), to facilitate comparison when discussing their molecular features and expected effects.

In figure 1 the label (C) is missing. Please check.

Response: We thank the reviewer for this observation. The caption of Figure 2 (formerly Figure 1) has been corrected to match the actual figure content, which comprises two panels only. Panel A describes the workflow used for the morphometric analyses, and Panel B highlights the time-series extraction procedures employed in the spatial analysis of membrane flickering.

I need explanation about the data of figure 2 on the hemolysis rate. The figure shows that the treatment induces moderate hemolysis rate in the time window of the experiment, however it is not clear to me why the authors do not observed hemolysis with Rutin after 90’. I also need clarification about Apigenin and Quercetic, which seem to have only a transient peak at, respectively, short or intermediate incubation. It is definitely unclear the sense of a decrease in hemolysis over time, as lysed cells are lost.

Response: We thank the reviewer for raising this key point. We agree that, in the previous version, the hemolysis data presentation could suggest transient “peaks” and apparent decreases over time (e.g., rutin at 90 min), which is counterintuitive if interpreted literally, since lysed cells are not recovered. This ambiguity stemmed from an insufficient description of the assay workflow and from not explicitly communicating the method’s uncertainty and detection limits under compound-dependent optical cross-talk.

To address this, we have rewritten the relevant sections in Methods (Section 2.5) and Results (Section 3.1), and we have added Appendix B detailing the uncertainty propagation and limit-of-detection (LoD) estimation for the spectrophotometric hemolysis determination.

In addition, Figure 3 (formerly Figure 2) has been reconfigured to improve interpretability: controls and effectors are now displayed separately, each time point includes error bars (SD) derived from triplicate measurements, and the LoD for each effector is explicitly reported. The LoD is computed from the inherent uncertainty of the spectrophotometric method (including blank correction and residual compound-dependent optical contribution; see Appendix B). Under these conditions, small non-monotonic fluctuations near zero (e.g., 0.0–0.8%) fall within the method’s resolution and should not be interpreted as true time-dependent hemolysis.

Finally, Figure 3 also includes a panel reporting the percentage of normocytes detected in each sample, to document that the cell population remains predominantly normocytic during the experiments.

Overall, the revised analysis shows that, within the 0–180 min window, hemolysis remains below the LoD for all flavonoid-treated conditions, supporting that the experiments were conducted in a sub-hemolytic regime and that the effectors do not induce detectable hemolytic damage under the selected assay conditions.

Still on the same point: How, operatively, did the authors measure the hemolysis rate? How many times have this measurement been repeated? Why the error bars in the graph have not been reported?

Response: We thank the reviewer for requesting these operational details. We have now clarified the hemolysis workflow in Section 2.5 (Methods): after each incubation time point, samples were centrifuged (5000×g, 10 min, 4 °C), the supernatant was collected, and hemoglobin was quantified by scanning spectrophotometry using the Harboe approach (primary absorbance at 415 nm with corrections at 380 nm and 450 nm), applying condition-matched blank subtraction (vehicle blank and flavonoid-vehicle blank) prior to hemoglobin/hemolysis calculation (see Appendix B).

Regarding replication, each condition and time point (controls and flavonoid-treated samples) was performed in triplicate (independent replicates). In the revised version, we now report error bars (SD) derived from these triplicates in the hemolysis plot (Figure 3, formerly Figure 2). We also include the limit of detection (LoD) computed from replicate blank measurements and uncertainty propagation (Appendix B), to facilitate interpretation of low hemolysis values.

Figure 3. Please, adjust the parameters list reported in the caption according to the 6 panels composing the image: they are not listed in the same order and sometimes have different names.

Response: We thank the reviewer for this remark. We have revised the caption accordingly by reordering the listed parameters to match their appearance in Figure 4 (formerly Figure 3) across the six panels, and we have also unified the terminology so that parameter names are fully consistent with both the figure labels and the descriptions in the main text.

At lines 404-406 the authors states: “After 1 h of incubation, all analysed parameters exhibited systematic shifts consistent with metabolic reactivation following the osmotic and mechanical stress…”. The question is: What kind of morphometric variation the authors associate to the reprise of metabolic activity? And, more interestingly, why?

Response: We thank the reviewer for this thoughtful question and agree that the term “metabolic reactivation” requires a clearer operational link to the specific morphometric shifts we observe between C0h (post-isolation) and C1h (after 1 h incubation). In our dataset, the C0h→C1h transition is characterized by (i) a reduction in equivalent radius (consistent with partial restoration of volume/osmotic homeostasis after isolation-induced stress), (ii) a decrease in circularity while elongation remains essentially unchanged (reflecting a subtle relaxation/redistribution of contour geometry rather than a gross anisotropic deformation), and (iii) a mechanically coherent change in fluctuation-derived descriptors (increased fluctuation amplitude/shape factor together with a decrease in effective tension) indicative of a more compliant membrane state (often referred to as metabolic softening).

Mechanistically, this interpretation is supported by prior work showing that RBC membrane mechanics and flickering are modulated by metabolism and ATP-dependent processes: ATP availability and metabolic activity can alter the effective mechanical response of the RBC envelope and the fluctuation spectrum (Betz, Lenz, Joanny, & Sykes, 2009; Turlier et al., 2016), and metabolic remodeling has been directly linked to changes in RBC membrane properties (Park et al., 2010). In addition, cytoskeleton-related active forces have been shown to produce membrane softening in RBCs (Rodríguez-García et al., 2015), and modulation of the metabolic environment (e.g., glucose availability) measurably impacts flickering dynamics (Tapia et al., 2021). Taken together, these studies provide a rationale for why a post-isolation incubation period can yield systematic shifts toward lower effective tension and higher fluctuation amplitudes, consistent with recovery from osmotic/mechanical stress and partial restoration of physiologically relevant cell-to-cell variability. We have added this clarification to the Results text to make explicit which morphometric variations are being associated with this interpretation and why.

Lines 411-412. The authors cite a progressive recovery of membrane dynamics. I just see and evolution between C0 and C1, but I don’t know if we can properly speak of “recovery” as the original condition is not defined (nor measured). Please comment.

Response: We agree with the reviewer that the term “recovery” may be misleading, as a true pre-isolation or in vivo baseline is not defined or measured in this study. We have therefore replaced “recovery” with more conservative wording. The revised text now states that the observed changes between C0 and C1 reflect an incubation-dependent relaxation from post-isolation perturbation, associated with increased membrane fluctuations and mechanical compliance after sample preparation. This wording avoids implying a return to an original baseline and more accurately reflects the experimental design.

The data have been collected essentially (if not exclusively) on discocytic cells. Although this choice is understandable, a non-negligible fraction of the cells have non biconcave shape (e.g. crenated cells, echinocytes or spherocytes), and this fraction is important in physiological and pathological conditions. Have some of these cells been included in the analysis (or, at least, tested)? What kind of differences the authors observed, or expect, in the dynamic phenotype and in the flickering spectra from these cells? Please, comment in the text

Response: We thank the reviewer for this important point. We agree that non-discocytic morphotypes (e.g., echinocytes/crenated cells, spherocytes, stomatocytes) can represent a relevant fraction of RBC populations under physiological variability and, especially, in pathological conditions. In the present study, however, our goal was to establish a controlled, phenotype-level proof-of-concept and to compare compound-specific signatures under sub-hemolytic conditions using a morphometrically homogeneous population. Accordingly, downstream morphometric and flickering analyses were restricted to discocytic (normocytic) cells.

A key practical reason is that non-discocytic shapes frequently place portions of the membrane out of the focal plane, which degrades the definition of the peripheral halo and introduces systematic errors in contour detection and height reconstruction. This strongly affects the reliability of derived morphometric parameters and, for flickering, it can bias mode decomposition and the inferred mechanical descriptors. To make this selection explicit, we have now included a condition-wise quantification of the normocyte fraction in Figure 3B, together with a brief reference in the Results section; the remaining population is predominantly echinocytic, with a residual fraction of stomatocytes.

Regarding dynamic phenotypes, we now comment in the text that we would expect non-discocytic cells to exhibit systematically different flickering spectra because both geometry and dominant mechanical constraints differ from discocytes. In particular, stomatocytic shapes are typically associated with altered tension/curvature constraints that can suppress long-wavelength modes and reshape the static spectrum, whereas echinocytes often involve membrane–cytoskeleton remodeling and spicule-associated heterogeneities that can strongly modulate fluctuation amplitudes, mode content, and relaxation kinetics. Because these factors would confound the compound-to-compound comparisons targeted here, these morphotypes were not included in the primary analysis; their dedicated characterization constitutes a natural next step for future extensions of the framework, especially in multi-donor and pathology-oriented studies

The ethic clearance statement need, probably, a more detailed statement. Please, check.

Response: We thank the reviewer for this comment. We have revised and expanded the ethics statement in the manuscript to clarify that blood was obtained from a healthy adult volunteer under approval of the Research Ethics Committee of Universidad Francisco de Vitoria (Ref. 60/2024), with written informed consent obtained prior to donation. The Ethics Committee documentation and the informed consent form are included with the revised submission files.

Reviewer 2 Report

Comments and Suggestions for Authors

The article by Carlos del Pozo-Rojas et al. describes an advanced biophysical study using computational microscopy to analyze how flavonoids influence the mechanics of red blood cell membranes. Specifically, this study presents a computational microscopy framework that integrates morphometry and high-throughput spectroscopy to analyze red blood cell flickering as an indicator of membrane mechanics. The authors' results demonstrate that, although discocyte morphology remains unchanged, exposure to flavonoids (quercetin, apigenin, rutin) alters the dynamic parameters in a manner specific to each chemical structure. Fourier mode decomposition reveals distinct mechanical phenotypes between aglycone and glycosylated compounds, validating this label-free approach for screening drug-membrane interactions.

Comments:

Figure 2: Why do hemolysis values fluctuate so significantly between samples at different incubation times? For example, samples incubated with quercetin show an increase in hemolysis of 0.80% after 30 minutes, 0% after 90 and 180 minutes; while rutin induces 0.50% hemolysis after 30 minutes, 0% after 90 minutes, and 0.80% after 180 minutes. The authors are asked to clarify the expression of the results to facilitate understanding, or to correct any errors. Even considering this increase in hemolysis as not very significant (it is less than 3-4%), the graph shown by the authors seems to highlight a notable difference between the various samples and between the same ones after 30, 90 and 180 minutes.

Figure 3: I would suggest also performing statistical analysis between the treated samples and the control (C0h). It could help clarify how apigenin, quercetin, and rutin influence erythrocyte properties compared to C0h.

Figure 4: Why are the control values at time 0 (C0h) not reported in figure 4?

Discussion: I advise the authors not to divide discussions into subparagraphs. Furthermore, the section, while clear and straightforward, seems to describe the results obtained and partially discussed in the previous section. I would recommend further exploration of the biochemical properties and potential molecular interactions of flavonoids with the phospholipid bilayer.

Author Response

REVIEWER #2

Response: We sincerely thank Reviewer 2 for this thorough and insightful assessment of our work. We appreciate the reviewer’s accurate summary of the computational microscopy framework and the emphasis on the compound-specific flickering phenotypes induced by quercetin, apigenin, and rutin under sub-hemolytic conditions. We have carefully addressed all comments and suggestions raised by the reviewer and revised the manuscript accordingly to further strengthen the presentation, statistical reporting, and interpretation of the mechanical phenotypes observed.

Comments:

Figure 2: Why do hemolysis values fluctuate so significantly between samples at different incubation times? For example, samples incubated with quercetin show an increase in hemolysis of 0.80% after 30 minutes, 0% after 90 and 180 minutes; while rutin induces 0.50% hemolysis after 30 minutes, 0% after 90 minutes, and 0.80% after 180 minutes. The authors are asked to clarify the expression of the results to facilitate understanding, or to correct any errors. Even considering this increase in hemolysis as not very significant (it is less than 3-4%), the graph shown by the authors seems to highlight a notable difference between the various samples and between the same ones after 30, 90 and 180 minutes.

Response: We thank the reviewer for this important observation, and we agree that, in the previous version, both the presentation of Figure 2 and the accompanying description of the hemolysis assay could be confusing because the methodological context and measurement limitations were not explained with sufficient detail.

To address this, we have substantially revised Section 2.5 (Methods) to provide a clearer and more complete description of the hemolysis protocol, including sample handling, blank correction, and how absorbance contributions from the flavonoids are accounted for. In addition, we have added an Appendix (B) at the end of the manuscript describing how the measurement uncertainty associated with the estimated hemolysis percentage is calculated and propagated, so that the reader can interpret small variations in a quantitative and transparent way.

We have also reconfigured the hemolysis figure (now Figure 3, formerly Figure 2). The results are now displayed separately for each effector to improve readability, and each condition/time point includes the corresponding standard deviation (SD). Importantly, we now explicitly report the limit of detection (LoD) in hemolysis percentage for each effector, which depends on the residual cross-talk/optical contribution of the flavonoid and the associated blank-correction uncertainty. We have also included a plot indicating the % of normocytes within the selected conditions.

Finally, in the Results&Discussion we clarify that, although small non-monotonic fluctuations may appear when values are near the assay resolution (e.g., 0.0–0.8%), all estimated hemolysis values within the studied time window remain below the LoD once uncertainty and compound-specific cross-talk are considered. Therefore, the apparent fluctuations should not be interpreted as robust time-dependent hemolytic effects, but rather as variations within the measurement noise floor under sub-hemolytic conditions.

Figure 3: I would suggest also performing statistical analysis between the treated samples and the control (C0h). It could help clarify how apigenin, quercetin, and rutin influence erythrocyte properties compared to C0h.

Response: We thank the reviewer for this valuable suggestion. We have now included a new table reporting the statistical outcomes of the comparative analysis across all experimental groups, explicitly including comparisons against both C0h and C1h controls. In addition, we have added a brief reference to this expanded statistical comparison in the Results section to guide the reader to the new material.

Figure 4: Why are the control values at time 0 (C0h) not reported in figure 4?

Response: We thank the reviewer for this question. For the flickering analysis, we did not include C0h in Figure 4 because this measurement is experimentally and computationally demanding (high-speed imaging acquisition, long recordings, and intensive downstream processing). Therefore, we prioritized the control condition that is directly comparable to the effector-treated samples, i.e., the 1 h vehicle control (C1h), which matches the incubation time used for apigenin, quercetin, and rutin.

In contrast, the morphometric analysis is comparatively “low-cost” in both experimental throughput and computational processing, and we therefore included C0h there to explicitly capture the time-dependent recovery trajectory after isolation (C0h → C1h) and to contextualize the baseline shifts induced by the isolation procedure.

Discussion: I advise the authors not to divide discussions into subparagraphs. Furthermore, the section, while clear and straightforward, seems to describe the results obtained and partially discussed in the previous section.

Response: We thank the reviewer for this suggestion. We agree that the Discussion should go beyond a restatement of the Results and focus on interpretation and broader implications. Our use of labeled subsections was intended to streamline the discussion by explicitly linking each interpretive block to the corresponding set of results, thereby improving readability for an interdisciplinary audience.

In the revised manuscript, we have retained a limited number of clearly labeled sections but have reworked the Discussion to reduce descriptive repetition and to place greater emphasis on mechanistic interpretation, comparison with prior literature, and the implications of the observed compound-specific phenotypes. We believe this preserves structural clarity while providing a more insightful and integrative discussion.

I would recommend further exploration of the biochemical properties and potential molecular interactions of flavonoids with the phospholipid bilayer.

Response: We thank the reviewer for this suggestion. In the revised Discussion, we have expanded the biochemical interpretation of our results by explicitly linking the observed mechanical phenotypes to structure-dependent membrane interactions, including flavonoid polarity, hydroxylation, glycosylation, and expected depth of interfacial insertion within the phospholipid bilayer. We further discuss how these interactions may modulate lipid packing, bilayer order, and bilayer–cytoskeleton coupling, and we outline specific experimental directions (e.g., cholesterol modulation and membrane order probes) to directly test these mechanisms in future work.

Reviewer 3 Report

Comments and Suggestions for Authors

General Assessment
The manuscript titled “Computational microscopy reveals compound-specific flickering phenotypes of red blood cells under flavonoid exposure” provides a compelling and relevant contribution to membrane biophysics. Its combination of high-throughput morphometry with single-cell flickering analysis is technically sound, biologically significant, and well aligned with the study of membrane–active small molecules. The work effectively shows compound-specific phenotypes under sub-hemolytic conditions and offers a methodological framework with clear potential for screening and mechanistic studies.
The writing is clear, the figures are informative, and the scholarly foundation is solid. The work is suitable for publication with minor revisions.
Major Remarks
1. Clarify scope and future generalization

Regarding the single-donor limitation, the authors correctly acknowledge the single-donor basis and avoid broad population claims (Section 2.2.9 & Discussion). To improve clarity, the Introduction or Conclusions should explicitly frame the current work as:
• establishing phenotype- level feasibility
• validating the method
• providing a framework for future expansion
This will clarify expectations for subsequent multi-donor studies, dose–response analyses, and cholesterol-modulation experiments discussed in Section 4.4.6.

2. Contextualize mechanistic interpretation with membrane partitioning
The Discussion provides strong conceptual references around:
• glycosylation versus aglycone chemistry
• cholesterol dependence
• escin as a benchmark comparison
To enhance impact, consider briefly noting:
• whether phenotypes align more with partition depth, lipid ordering, cytoskeleton interactions, or metabolic/active processes
• how these could be tested experimentally (e. g., Laurdan GP, cholesterol depletion/loading, AFM indentation, RBC ATP modulation)
This would require only interpretive context, not new experiments.

3. Clarify the link between “effective tension” and “rigidity.”
In flickering metrics, two mechanical readouts are used:
• ensemble tension (from static spectra)
• local rigidity, friction, and relaxation rate (from time- resolved data)
While the manuscript notes they measure different mechanical aspects (Section 3. 3.4 & 4.2. 2), readers unfamiliar with RBC flickering may confuse them. Including a brief schematic or explanation of:
“more tension” ≠ “more rigidity”; a priori would improve understanding and prevent literal interpretation of mechanical terms.

4. I suggest that authors include a discussion of two additional publications in the Introduction.

A. RBC flickering / mechanics and quantitative phase imaging

Park YK, Best CA, Badizadegan K, Dasari RR, Feld MS, Kuriabova T, Henle ML, Levine AJ, Popescu G.
Measurement of red blood cell mechanics during morphological changes.
Proc Natl Acad Sci U S A. 2010;107(15):6731–6736. doi:10.1073/pnas.0909533107

This study employs non-contact optical interferometry to measure RBC membrane fluctuations with nanometer precision and to determine mechanical parameters during the transitions from discoid to echinocyte to spherocyte. You may cite this work when discussing RBC flickering, deformability, and the advantages of full-field, label-free optical methods for assessing membrane mechanics.

B. Flavonoid–membrane interactions (with erythrocyte models)

Selvaraj S, Krishnaswamy S, Devashya V, Sethuraman S, Krishnan UM.
Influence of membrane lipid composition on flavonoid–membrane interactions: Implications on their biological activity.
Prog Lipid Res. 2015;58:1–13. doi:10.1016/j.plipres.2014.11.002

A thorough review of how flavonoids engage with lipid membranes, such as erythrocyte model membranes, and how the lipid composition influences their biophysical and biological impacts. This aligns with your discussion of flavonoids as membrane-active agents, offering a solid conceptual foundation for your analysis of compound-specific flickering phenotypes.

Minor corrections (typos/style)
Two minor typographical issues:
1. Table 1 typo
“Log P (hidofobicity)” should be “Log P (hydrophobicity)”
(page 4);
2. Section title formatting
“High- thoughput morphmechanical phenotyping” should be “High- throughput morphomechanical phenotyping” (page 6);
Recommendation
Accept with minor revisions: Accept with Minor Revisions

Author Response

REVIEWER #3

General Assessment

The manuscript titled “Computational microscopy reveals compound-specific flickering phenotypes of red blood cells under flavonoid exposure” provides a compelling and relevant contribution to membrane biophysics. Its combination of high-throughput morphometry with single-cell flickering analysis is technically sound, biologically significant, and well aligned with the study of membrane–active small molecules. The work effectively shows compound-specific phenotypes under sub-hemolytic conditions and offers a methodological framework with clear potential for screening and mechanistic studies.

The writing is clear, the figures are informative, and the scholarly foundation is solid. The work is suitable for publication with minor revisions.

Response: We sincerely thank Reviewer 3 for this positive and constructive assessment. We appreciate the recognition of the technical soundness and biological relevance of our combined high-throughput morphometry and single-cell flickering framework, as well as its potential for compound screening and mechanistic studies under sub-hemolytic conditions. We have carefully addressed all minor revision requests and revised the manuscript accordingly to further improve clarity, consistency, and presentation

Major Remarks

Clarify scope and future generalization

This will clarify expectations for subsequent multi-donor studies, dose–response analyses, and cholesterol-modulation experiments discussed in Section 4.4.6.

Response: We thank the reviewer for this helpful recommendation. To further improve clarity and manage expectations, we have explicitly framed the present study as establishing phenotype-level feasibility, validating the end-to-end computational microscopy pipeline under controlled, sub-hemolytic conditions, and providing a methodological framework for future expansion. These clarifications have been incorporated in the Abstract, Introduction, and Conclusions, and we have also revised the wording throughout the manuscript to soften any statements that could be interpreted as broad population generalizations, ensuring that claims remain appropriately scoped to this single-donor proof-of-concept.

Contextualize mechanistic interpretation with membrane partitioning

The Discussion provides strong conceptual references around: a) glycosylation versus aglycone chemistry b) cholesterol dependence c) escin as a benchmark comparison

To enhance impact, consider briefly noting: a) whether phenotypes align more with partition depth, lipid ordering, cytoskeleton interactions, or metabolic/active processes b) how these could be tested experimentally (e. g., Laurdan GP, cholesterol epletion/loading, AFM indentation, RBC ATP modulation).

This would require only interpretive context, not new experiments.

Response: We thank the reviewer for this helpful suggestion. In the revised Discussion, we have expanded the mechanistic interpretation of the observed phenotypes by explicitly linking them to structure-dependent membrane partitioning and interfacial interactions of flavonoids. We now discuss how aglycones versus glycosylated compounds are expected to differ in depth of insertion and interfacial localization, with consequences for lipid packing, bilayer order, and bilayer-cytoskeleton coupling, providing a plausible basis for the distinct rigidity-friction phenotypes observed.

In addition, we now explicitly outline testable experimental directions, such as cholesterol depletion/loading, membrane order probes (e.g., Laurdan GP), and metabolic modulation. Further experimentation could distinguish between these mechanistic contributions in future studies, without implying that such measurements are required to support the present conclusions.

Clarify the link between “effective tension” and “rigidity.”

In flickering metrics, two mechanical readouts are used:a) ensemble tension (from static spectra) b) local rigidity, friction, and relaxation rate (from time- resolved data)

While the manuscript notes they measure different mechanical aspects (Section 3. 3.4 & 4.2. 2), readers unfamiliar with RBC flickering may confuse them. Including a brief schematic or explanation of: “more tension” ≠ “more rigidity”; a priori would improve understanding and prevent literal interpretation of mechanical terms.

Response: We thank the reviewer for this helpful suggestion. To prevent potential confusion between the two mechanical readouts, we have added a brief, reader-oriented clarification in Section 3.4 (Integrative phenotypic analysis) and reinforced it in Section 4.5 (Interpretation of extracted mechanical descriptors). In these additions we explicitly state that the ensemble/static “effective tension” and the time-resolved descriptors (effective rigidity, friction, and relaxation rate) probe different physical regimes and are not equivalent a priori: i.e., “more tension” does not necessarily imply “more rigidity”. We also briefly summarize how tension predominantly constrains long-wavelength (low-mode) undulations, whereas rigidity and friction shape shorter-scale responses and relaxation kinetics, to guide interpretation for readers less familiar with RBC flickering mechanics.

I suggest that authors include a discussion of two additional publications in the Introduction.

RBC flickering / mechanics and quantitative phase imaging. Park YK, Best CA, Badizadegan K, Dasari RR, Feld MS, Kuriabova T, Henle ML, Levine AJ, Popescu G.
Measurement of red blood cell mechanics during morphological changes.
Proc Natl Acad Sci U S A. 2010;107(15):6731–6736. doi:10.1073/pnas.0909533107

Response: We thank the reviewer for this valuable suggestion. We have now incorporated the authoritative Park et al. (2010) in the Introduction in two places: first, within the overview of established approaches for RBC flickering/mechanics quantification, where we explicitly cite it as a representative example of full-field, label-free quantitative phase imaging used to measure membrane fluctuations and infer effective mechanical parameters; and second, in the discussion of morphological changes and recovery after incubation, where we use this work to contextualize how transitions among discocyte / echinocyte / spherocyte states are associated with systematic shifts in fluctuation behavior and mechanical descriptors. These additions strengthen the methodological background and support our interpretation of the incubation-related morphometric trends.

Flavonoid–membrane interactions (with erythrocyte models). Selvaraj S, Krishnaswamy S, Devashya V, Sethuraman S, Krishnan UM.
Influence of membrane lipid composition on flavonoid–membrane interactions: Implications on their biological activity. Prog Lipid Res. 2015;58:1–13. doi:10.1016/j.plipres.2014.11.002

Response: We thank the reviewer for this recommendation. The review by Selvaraj et al. (Prog. Lipid Res., 2015) was already included in the original manuscript (Ref. 15). In the revised version, we have highlighted this reference more explicitly in the Discussion when addressing flavonoid partitioning, lipid composition dependence, and membrane-mediated effects, to strengthen the conceptual grounding of our interpretation.

Minor corrections (typos/style)

Two minor typographical issues:

Table 1 typo: “Log P (hidofobicity)” should be “Log P (hydrophobicity)” (page 4)

Thank you for noting this typographical error. We have corrected the label in Table 1 to “Log P (hydrophobicity)

Section title formatting “High- thoughput morphmechanical phenotyping” should be “High- throughput morphomechanical phenotyping” (page 6);

Thank you for pointing this out. We have corrected the section title to “High-throughput morphomechanical phenotyping”.

Recommendation
Accept with minor revisions

Reviewer 4 Report

Comments and Suggestions for Authors

This manuscript presents a computational microscopy framework that combines high throughput, brightfield morphometry with flickering spectroscopy to characterize compound specific mechanical phenotypes of RBCs under flavonoid exposure. By integrating static contour statistics and high-speed time-resolved fluctuation analysis, the authors propose that membrane flickering provides a sensitive mechanical readout capable of resolving differences between aglycone flavonoids (quercetin, apigenin) and a glycosylated flavonoid (rutin), even when classical morphological descriptors and hemolysis assays remain unchanged. The work is conceptually interesting and methodologically sophisticated. However, several aspects require clarification before the work can be considered for publication.

Major concerns

The entire study is based on erythrocytes from a single donor. While the authors appropriately state this as a limitation, the manuscript still relies heavily on statistical comparisons (p-values) that may be misleading under these conditions.This should be emphasized more strongly in Results and Discussion.

The ergodic approach (spatial ensemble ≈ temporal fluctuations) is central to the static analysis, but its justification is largely based on previous literature rather than demonstrated here. For readers unfamiliar with this methodology, this appears as a strong assumption. Should include a short validation plot or comparison showing that static spectra and dynamic spectra yield comparable effective tension for at least the control group in this dataset.

The manuscript occasionally uses terminology such as “mechanically cooled” or interprets effective parameters in ways that may be misread as absolute material properties. Could move part of the interpretative caution from Discussion into Results where these parameters are first introduced.

The central interpretation is that aglycones alter membrane mechanics via insertion and bilayer–cytoskeleton coupling, while rutin acts as an interfacial antioxidant. This is plausible but remains inferred, not measured. There are no independent measurements of:membrane order (e.g., Laurdan GP),cholesterol modulation or membrane partitioning.

Hundreds of cells are analyzed for static contours, but only ~5–10 cells per condition for high-speed flickering. Given that the main conclusions rely strongly on dynamic descriptors, this asymmetry weakens confidence.

Author Response

REVIEWER #4

Major concerns

The entire study is based on erythrocytes from a single donor. While the authors appropriately state this as a limitation, the manuscript still relies heavily on statistical comparisons (p-values) that may be misleading under these conditions. This should be emphasized more strongly in Results and Discussion.

Response: We thank Reviewer #4 for highlighting this important limitation. We agree that, because the study is single-donor, statistical testing should be interpreted strictly as within-donor, condition-level discrimination rather than population-level inference. In the revised manuscript we now emphasize this more explicitly in both the Results (where p-values are first reported; Figure 4) and the Discussion (Section 4.6), clarifying that p-values quantify separability under matched acquisition/analysis conditions and do not support claims about inter-individual variability. We also added explicit language that multi-donor replication is required for population inference and clinical generalization.

The ergodic approach (spatial ensemble ≈ temporal fluctuations) is central to the static analysis, but its justification is largely based on previous literature rather than demonstrated here. For readers unfamiliar with this methodology, this appears as a strong assumption. Should include a short validation plot or comparison showing that static spectra and dynamic spectra yield comparable effective tension for at least the control group in this dataset.

Response: We thank the reviewer for raising this important methodological point. We agree that the ergodic (ensemble-based) approximation treating the population of discocytic cells at a fixed time point as an ensemble proxy for fluctuation statistics. This should be certainly interpreted carefully, particularly in an active membrane system. In the present proof-of-concept study, we use the ergodic approach strictly as a comparative estimator to quantify condition-dependent shifts under matched acquisition and analysis, rather than as a route to absolute material constants or to single-cell temporal inference.

Hence, we did not include a dataset-specific internal cross-validation between ensemble-derived and time-resolved tension estimators because the two analysis branches were designed to probe different physical regimes (low-mode/long-wavelength static spectra versus time-resolved local descriptors), and the manuscript does not aim to infer the same parameter by two independent routes for the same set of cells. Instead, to avoid overinterpretation, we have explicitly strengthened the limitation statement in the Discussion, clarifying (i) the scope of the ergodic assumption, (ii) that it is expected to be most reliable under the restricted conditions used here (normocytic, in-focus contours, sub-hemolytic regime, homogeneous acquisition pipeline), and (iii) that a dedicated internal cross-validation is a natural extension for future work and future pipeline releases.

Importantly, our main conclusions do not rely on absolute values from the ergodic estimator, but on consistent within-donor, within-protocol separability across conditions. We believe that making the scope and limitation explicit addresses the reviewer’s concern while keeping the manuscript aligned with its stated phenotype-level validation objective.

The manuscript occasionally uses terminology such as “mechanically cooled” or interprets effective parameters in ways that may be misread as absolute material properties. Could move part of the interpretative caution from Discussion into Results where these parameters are first introduced.

Response: We agree that terminology such as “mechanically cooled” and the use of effective parameters can be misread as absolute material constants. We have therefore (i) tightened the language to consistently refer to effective descriptors of the projected dynamics, (ii) clarified earlier in the Results (at first introduction of inferred parameters) that these are comparative markers under the adopted model assumptions, and (iii) retained the more detailed caveats in the Discussion. We also explicitly state that “mechanically cooled” is a purely phenomenological descriptor of reduced fluctuation timescales, not a thermodynamic temperature change.

The central interpretation is that aglycones alter membrane mechanics via insertion and bilayer–cytoskeleton coupling, while rutin acts as an interfacial antioxidant. This is plausible but remains inferred, not measured. There are no independent measurements of:membrane order (e.g., Laurdan GP),cholesterol modulation or membrane partitioning.

Response: We thank the reviewer and agree. Our mechanistic interpretation is inferred from structure–phenotype consistency and prior literature, and we do not intend to claim direct molecular measurement of insertion depth, membrane order, or cholesterol dependence in the present dataset. To avoid over-interpretation, we have revised the Discussion to frame these points explicitly as plausible mechanistic hypotheses consistent with the observed phenotypes, and we have added a concise list of concrete validation experiments (e.g., Laurdan GP, cholesterol depletion/loading, partitioning assays) as future directions. We also toned-down wording that could be read as assigning rutin a specific antioxidant localization mechanism, keeping our claims strictly phenotype-level.

Hundreds of cells are analyzed for static contours, but only ~5–10 cells per condition for high-speed flickering. Given that the main conclusions rely strongly on dynamic descriptors, this asymmetry weakens confidence.

Response: We agree that the asymmetry between static (high-throughput) and dynamic (high-speed) sample sizes should be clearly justified. In the revised manuscript, we explicitly state that high-speed flickering is acquisition- and processing-intensive, and was designed here as a deep phenotyping subset to demonstrate feasibility and extract time-resolved descriptors under tightly controlled imaging conditions. We now emphasize this limitation in Results/Discussion and clarify that the dynamic results should be interpreted as proof-of-concept, motivating future extensions with larger dynamic n and multi-donor replication. Where possible, we also report the per-condition cell numbers prominently and include variability estimates to make the confidence level transparent.

In line with the other reviewers, Reviewer #4 raised important points about scope, assumptions, and sample-size balance. We have clarified that the study is single-donor and supports within-donor phenotype discrimination, not population inference; added earlier cautions in the Results regarding the interpretation of inferred quantities as effective descriptors; explicitly clarified the scope and limitations of the ensemble/ergodic approximation (used here as a comparative estimator, without introducing a dataset-specific internal cross-validation in this proof-of-concept paper); softened mechanistic language to hypothesis-level; and made the dynamic sample-size limitation explicit while reporting per-condition n clearly.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Editor. Dear Authors

I’ve read the revised version of the manuscript and I’m please to say that all my previous questions have been considered and answered. In my opinion this improved the text readability and clarified some possible doubt on the scientific content.

In the light of this, my feeling is that the paper can be published although a few sentences (see below) should be checked for English or other kind of typos:

Please check the last (highlighted) sentence in the caption of figure 3 which states “Sample size: n cells, m videos”. It has probably remained in the text by mistake.
Please check also the sentence at lines 310-312.

Author Response

Question 1: Please check the last (highlighted) sentence in the caption of Figure 3 which states “Sample size: n cells, m videos”. It has probably remained in the text by mistake.
Answer 1: Thank you for pointing this out. The sample size and replication details were described in the Methods, but we inadvertently omitted them from the figure caption. We have now updated the caption to report the sample sizes (n cells, m videos) as requested.

Question 2: Please check also the sentence at lines 310–312.
Answer 2: We have revised the text for clarity as follows:
“In the ensemble-based (“ergodic”) approach, long-wavelength fluctuation statistics are estimated by treating the population of discocytic cells, acquired under matched conditions at a given time point, as an ensemble proxy for membrane states.”

Reviewer 2 Report

Comments and Suggestions for Authors Manuscript can now be accepted for publication in its current form

Author Response

Question: Manuscript can now be accepted for publication in its current form
Answer: We thank the editor for this assessment and are pleased to confirm our agreement with acceptance of the manuscript in its current form.

Reviewer 4 Report

Comments and Suggestions for Authors

Author Response

We thank the reviewer for this possitive assessment and are pleased to confirm our agreement with acceptance of the manuscript in its current form.