Assaying and Classifying T Cell Function by Cell Morphology

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

2. Questions for General Evaluation

Reviewer’s Evaluation

Response and Revisions

Does the introduction provide sufficient background and include all relevant references?

Yes

Are all the cited references relevant to the research?

Yes

Is the research design appropriate?

Yes

Are the methods adequately described?

Yes

Are the results clearly presented?

Can be improved

Thank you for the suggestions on clarity, which are addressed below.

Are the conclusions supported by the results?

Yes

3. Point-by-point response to Comments and Suggestions for Authors

Comment 1: “1 - Abstract: Please, cite the machine learning algorithms used.”

Response: We have updated the Abstract to include the machine learning algorithms that were used. Details on these algorithms, including citations of the underlying report, are too extensive for the Abstract and thus cited in the Materials and Methods section.

Comment 2: “2 - Figure 3: Please, remove Figure 1B which has the same information regarding the explained variance of PC1 and PC2.” and “Please, choose one of the figures 3C and 3D and remove it. They have almost the same information.

Response: We thank the reviewer for urging this simplification in presentation and have removed the previous Figures 3B and 3C. A bar graph to show loadings on PC2 has been added to Figure 3D.

Comment 3: “4 - Tables 1, 2.1 and 2.22: Please, add Matthews's correlation coefficient and discuss the results including this classification metric.”

Response: Matthew’s correlation coefficient has been added to the tables.

Comment 4: “5 - Tables 2.1 and 2.2 Please, merge these two tables in one table, or replace "2.1" and "2.2" with "2" and "3".”

Response: Tables 2.1 and 2.2 have been relabeled as Tables 2 and 3. We thank the reviewer for this point of clarity in presentation.

1. Summary

Thank you very much for this thoughtful manuscript review. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

2. Questions for General Evaluation

Reviewer’s Evaluation

Response and Revisions

Does the introduction provide sufficient background and include all relevant references?

Yes

Are all the cited references relevant to the research?

Yes

Is the research design appropriate?

Must be improved

We have added experimental details and analysis to this manuscript. In addition, please see the discussion below for motivation on the experiments.

Are the methods adequately described?

Can be improved

We have provided additional details to the Methods section.

Are the results clearly presented?

Must be improved

Please see discussion below.

Are the conclusions supported by the results?

Must be improved

Please see discussion below.

3. Point-by-point response to Comments and Suggestions for Authors

Comment 1: “On the other hand, it is not clear whether the T cell function assessed in vitro reflects the function of cellular immunotherapy in vivo. We believe that more detailed descriptions are needed.”

Response 1: We thank the reviewer for encouraging additional context and background. The Introduction and Discussion sections have been augmented to provide this information.

Comment 2: “What is the goal for the future with the conclusion of this paper? T cells from healthy or CLL donors can be machine-readable by the authors' method. Do T cells from CLL donors clearly function poorly in vivo? If this point is not clarified, what is the point of analyzing them?”

Response 2: We have clarified the Introduction section to better specify the differences between cells from Healthy bs. CLL donors. The Discussion section has been likewise expanded to describe uses of this technology.

1. Summary

Thank you very much for this thoughtful manuscript review. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

2. Questions for General Evaluation

Quality of English Language: Please see point-by-point responses.

Reviewer’s Evaluation

Response and Revisions

Does the introduction provide sufficient background and include all relevant references?

Can be improved

The introduction has been expanded.

Are all the cited references relevant to the research?

Yes

Is the research design appropriate?

Yes

Are the methods adequately described?

Can be improved

The Methods section has been expanded

Are the results clearly presented?

Can be improved

The Results section has been streamlined and augmented.

Are the conclusions supported by the results?

Yes

3. Point-by-point response to Comments and Suggestions for Authors

Major Revision:

Comment 1: “Based on experimental methods, authors used CD4/CD8 mixed cells. These cells, so called as Naïve T cells were replated on CD3/CD28 coated surface and collected the data at 40 min on soft vs hard matrix. However, in-vivo infection persists for couple of days to become Teff cells (effective against tumor killing). How author mimic their condition relevant to in-vivo or traditional T cells expansion and study approach. In short, why author studied Naïve T cells but not Teff.”

Response: T cells were isolated from healthy donors and CLL patients using negative selection methods that provide a mix of naïve, memory, and effector cells representing the composition present in the patients as is used in immunotherapy. We added clarification of this issue in the methods and results sections.

Comment 2: “I would prefer author use these cell attachment assay when the cells are in resting state. Do cells morphologically (Cell Diameter) differ in two different states, resting vs activated on CD3/CD28. OR These results are based on source of T cells (Resting or Naïve activated T cell).”

Response: A core concept underlying this project is that functional measures of cell response reflect the T cell state. As such, we are examining high-level responses of cells following activation, capturing behaviors not observable in unactivated/resting cells. Moreover, this study seeks to use cell response to specify materials for activation. These experiments require observation of T cell interaction with the materials.

Comment 3: “Your study is relevant comparing soft vs hard matrix to study healthy vs disease state. But I assume that normal tissue culture plate has highest stiffness among all. Have you looked these cells over normal culture plate coated with ligand and replate cells? Sometimes matrix coating is not uniform over the plate. How do you control that uniformity?”

Response: While we agree that tissue culture plates provide very high stiffness, differences in chemistry between those plates and polydimethylsiloxane (PDMS) elastomer influence adsorption of the activating antibodies. Most prominently, we observe that activation and growth of T cells on tissue culture plates is highly variable between replicates, likely reflecting the issue of uniformity raised by the reviewer. Consequently, we do not routinely include tissue culture plates with PDMS.

We did inspect the antibody coating on PDMS to make sure it was uniform. Specifically, anti-CD3 and anti-CD28 antibodies were fluorescently labeled for visualization microscopy and observed to be uniform across the field of view (Figure A). Moreover, quantification of fluorescence revealed a consistent amount of antibody coating across different stiffness (Figure B).

Comment 4: “There is no description of Live-cell-imaging in method section. How the acquisition set up? Have you used perfect focusing system (PFS) to control focus cell surface towards the plate? Imaging seems performed under 20x magnification! Have you looked and studied spreading under higher magnification as cell morphology appears better resolution at high level?”

Response: Thank you for urging inclusion of these conditions. Additional information on imaging has been added to the Methods section describing live- and fixed-cell imaging. Unfortunately, the relatively thick layer of PDMS interfered with our automatic focusing system. Focus was updated manually over the course of live-cell imaging.

Regarding imaging of cell morphology, we agree that a higher magnification image provides finer detail, as illustrated in the images below collected at 100X. However, we used 40X imaging to collect cells for analysis, as each field provided detail sufficient for morphological analysis while including over six times the number of cells per field of view.

Comment 5: “I was just curious if author could stain membrane and study under TIRF microscopy to look for spreading area. Happy to hear your thoughts.”

Response: We tried membrane staining for CD2. However, fixed staining for actin provided more definition of the cell-substrate interface, free of membrane images far away from the PDMS surface. We also tried TIRF and IRM imaging, but the presence of the glass-PDMS interface prevented high-quality illumination of the cell interface in these modalities.

Comment 6: “Author explains mechano-sensing effect on heathy T cells over diseased T cells. Author did look over any mechano-based candidate Piezo or other alternative to validate their model.”

Response: This work focuses on assaying T cell health by morphology with less emphasis on an underlying mechanism. With an emphasis on cell spreading associated with cytoskeletal effects, the inhibitors we used targeted actin dynamics.

Comment 7: “I am not fully understanding the transition from experimental to machine learning approach here, although author explained with sufficient information. There are two methods authors could go through classification vs discrete approach of analysis. Why author choose classification approach in their study but not other.”

Response: This project seeks to provide an assay of T cell functionality and state, rather than identify differences in any specific morphological feature. As such, our goal is to classify a donor based on images. Allowing the algorithm to use all available morphological features optimizes this analysis. In future work, we aim to use T-cell morphology to predict proliferation capacity, which would be a regression approach.

Comment 8: “Based on my observation, Accuracy supposed to be >85% for best performance model but here I do see author got about 75%. How author co-relate their accuracy (with AUC) with standard ML models.”

Response: The accuracy we showed in the manuscript is based on a very stringent data splitting approach, the “leave one out” method, to avoid data leakage. Specifically, in each fold of validation, we randomly select 1 healthy and 2 CLL as testing dataset, and the other 2 healthy and 4 CLL as training dataset. This prevents the algorithm from seeing images from the testing donors, but can result in lower performance measures.

Using a less stringent splitting method where 70% of Healthy/CLL samples are used for training and the other 30% of Healthy/CLL samples for testing, we get much higher accuracy, as shown below.

With the goal of classifying cells from new individuals which did not appear in the training set, we believe that the “leave one out” method is the ideal way to analyze the model. We anticipate that large-scale collection of data in subsequent studies will improve the performance of this approach, reaching the often-used 85% criteria.

Comment 9: “Elaborate on Feature-based classification section in the methods.”

Response: We have added information in the Methods section.

Comment 10: “Why author prefer Ap2/3 inhibitor over cytochalasin D for cell spreading assay!”

Response: Branched actin polymerization, the target of the Arp2/3 inhibitor, has a major role in early cell spreading and lamellipodia formation. The overall cell morphology suggests a large role of this extended structure in determining cell morphology.

Comment 11: “Author prepared valuable contribution to their work. However, I would recommend elaborating the discussion more for broader audience and relevance to clinical.”

Response: We thank the reviewer for encouraging this additional perspective. Please see the additions to the Introduction and Discussion sections, which have been expanded with clinical applications.

 Minor Comments:

Comment 1: “The text is generally clear and well-structured, with a logical flow from the introduction to the methods, results, and conclusion. Consider providing a brief roadmap at the beginning of the abstract to guide readers through the key objectives and findings or Workflow figure to improve the readability.”

Response: A Workflow figure has been added to the supplementary material.

Comment 2: “Consider including more visual/figures, to illustrate key concepts and findings. This can enhance the reader's comprehension, especially when dealing with complex methodologies.”

Response: We thank the reviewer for urging this. Please see the Workflow figure added to supplementary material.

Comment 3: “Consider including more visual/figures, to illustrate key concepts and findings. This can enhance the reader's comprehension, especially when dealing with complex methodologies.”

Response: Make sure to use consistent terminology and formatting throughout the manuscript.

4. Response to Comments on the Quality of English Language

Comment 1: “Minor correction is required with detailed discussion.”

Response: We appreciate the attention on this, and have corrected inconsistencies throughout the document.