Circulating Tumor Cell Models Mimicking Metastasizing Cells In Vitro: Discrimination of Colorectal Cancer Cells and White Blood Cells Using Digital Holographic Cytometry
, Yuecheng Zhang 2,3,4, Jenny L. Persson 2,3,5, Jan Balvan 1, Zahra El-Schich 2,3 and Anette Gjörloff Wingren 2,3,*Round 1
Reviewer 1 Report
Feith and colleagues present a succinct and interesting report of the impressive accuracy by which DHC can classify tumor cells from blood cells in mixed populations. I sincerely thank the authors for the excellent study design, transparent and complete methods, clear data presentation, and exciting results. We really need more manuscripts like this - easily the most enjoyable and high quality review experience I have had with MDPI in years. It is fun to have the opportunity to write a positive and enthusiastic review once in awhile.
One request - for any classifier, the limit of detection is an important parameter. Can the authors comment on the expected percentage of circulating tumor cells in cancer patient serum? If it is less than 5% (which I suspect it might be), can the authors repeat the experiment in Fig 2, with a further cancer cell dilution series down to what we might expect to see in patient serum? It may require capturing more ROIs per condition, but to show DHC can detect those cancer cells at the rare frequency at which they occur in patients would be exceptionally powerful.
I think the above experiment is important to show the LOD of the system, but regardless of the result, I highly recommend this article for publication and would encourage the authors to use these data to justify future studies where serum from cancer patients and healthy controls are compared by DHC.
Minor comments:
Line 195: Reads figure 6A - I suspect this is a typo (I don't see a figure 6).
Figure 2 legend and Table 1 both list the unit of area as um3 - either the descriptor should be volume or the unit should be squared.
Line 61, listed references [17-19] are missing a very relevant comparable article: https://www.mdpi.com/2076-3417/10/13/4439
Author Response
Minor comments:
Line 195: Reads figure 6A - I suspect this is a typo (I don't see a figure 6).
Reply: This has now been changed to Figure 2A.
Figure 2 legend and Table 1 both list the unit of area as um3 - either the descriptor should be volume or the unit should be squared
Reply: This has now been changed to um2 in Figure 2, Figure 2 legend and in Table 1.
Line 61, listed references [17-19] are missing a very relevant comparable article: https://www.mdpi.com/2076-3417/10/13/4439
Reply: The suggested reference by Barker et al has been added.
Reviewer 2 Report
The manuscript of M. Feith et al. is concerning the perspective use of digital holographic cytometry in liquid biopsy. It is an interesting approach indicating the morphological differences between colorectal cancer cells (CRC) and peripheral blood mononuclear cells (PBMCs). This was also validated in the case of the in vitro study, where it was possible to differentiate the CRC cells and PBMCs based on the multi-parametric morphological phenotyping of these cells by DHC.
The issues addressed in the manuscript overlap with the scope of the Photonics journal. The language used in the manuscript should be understandable to a wide range of readers. The logical structure and organization of the manuscript have also raised no major concerns.
Generally, appropriate and relevant scholarly literature is utilized to place the study in a framework that comprises building novel understanding upon previously investigated directions.
The fundamental downside of the paper is that the morphometric analysis of the cells was only carried out via DHC. The authors chose not to use and validate the results obtained by alternative microscopic techniques or ultrastructural studies (electron microscopy). In my opinion, it is this study limitation.
A certain shortcoming is an absence in the introduction of references to the latest developments in digital holography, namely digital holographic tomography techniques, which have been playing an increasingly important role in recent years through their ability to determine the spatial distribution of the refractive index. The authors should refer to these techniques in relation to CRC studies and applications in liquid biopsy. quantitative phase imaging is playing an increasingly important role in quantitative cell phenotyping and liquid biopsy, so in my opinion, the authors should mention something about it.
I did not notice that the authors provided any demographics of these patients (age, sex, race, etc.) and their numbers (from how many samples were the cells isolated?).
Reagent part numbers missing (antibodies)
Some formatting errors ( see line 125).
Figures' quality is poor. Please improve the resolution figures ( fig.2/3) and axes description.
Author Response
The authors should refer to these techniques in relation to CRC studies and applications in liquid biopsy. Quantitative phase imaging is playing an increasingly important role in quantitative cell phenotyping and liquid biopsy, so in my opinion, the authors should mention something about it.
Reply: We thank the reviewer for these important comments. We have added a part (see below) in the introduction about digital holographic tomography techniques and applications in liquid biopsy and included the following references:
Merola et al, Methods, 136: 108; 2018
Balasubramani et al, Appl Optics, 60:B65; 2021
Balasubramani et al, J Imaging, 7:252; 2021
Merola et al, Light: Science & Appl, 6: e16241; 2017
Nissim et al, Cytometry Part A, 99A:511; 2021
Habaza et al, Adv Science, 4:1600205; 2017
Dudaie et al, J Biophotonics, 13:e202000151; 2020
Added text in the introduction:
Indeed, optical tomography provide a more complete mapping of the biological sample in combination with QPI (Merola 2018). The 3D refractive index will allow studies of subcellular and micrometer structures of the cells (Balasubramani Appl Optics 2021; Balasubramani J Imaging 2021). By combining all time-lapse quantitative phase maps, each orientation of the cell can be obtained (Merola 2017). For high-throughput 3D cell measurements, microfluidic devices have been integrated to tomographic phase microscopy including work on red blood cells and CTCs (Merola 2018). Nissim et al recently showed in their system that blood cells mixed with colorectal cancer cells could be automatically discriminated on the level of individual cell types, by using label-free holographic flow cytometry (Nissim 2021). Moreover, cell sorting was included in a tomographic interferometry approach by rotating cells in a flow with dielectrophoretic forces (Habaza 2017; Dudaie 2020).
I did not notice that the authors provided any demographics of these patients (age, sex, race, etc.) and their numbers (from how many samples were the cells isolated?).
Reply: Since this is a proof-of-concept study, only the result of one donor of PBMCs is shown in the paper, but the experiments have been performed and analysed with similar results on a total of three healthy donors.
Reagent part numbers missing (antibodies)
Reply: We have added the following information regarding antibodies in the Material and methods section at page
Anti-EpCAM-PE Miltenyi Cat No 130-113-264
anti-CD44 R&D Cat No MAB6127
anti-rat-FITC R&D Cat No F0104B
Some formatting errors ( see line 125).
Reply: 1 x 106 cells, has been corrected to 1 x 106 cells
Figures' quality is poor. Please improve the resolution figures ( fig.2/3) and axes description.
Reply: The figure No 2 and 3 have now been revised for better resolution and axes description.
Reviewer 3 Report
In this manuscript, the authors used digital holographic cytometry to discriminate cellular morphological properties and the marker EpCAM and CD44 were analyzed to differentiate colorectal cancer cell line and blood mononuclear cells. As a quantitative phase imaging technology, DHC was used to determine the differences of cell morphology, such as cell area, cell thickness, cell volume, cell irregularity between CRC cells and the PBMCs. Combining digital holographic cytometry and biomarker method to identify circulating tumor cells is a valuable research work. Multimodal imaging method can integrate the advantages of various imaging technologies. However, I still have several questions about this manuscript.
1. About the DHC imaging process in part 2.4, some information is needed, such as pixels of camera, magnification of microscope objective, time required to acquire 30 images, etc.
2. Besides, the author should show the picture of the DHC system, so that readers can better understand the obtaining process of experimental images.
3. In Figure 1A, the distribution difference morphological characteristics of COLO 205 cell line and PBMCs can be seen. However, the author should explain how to identify these cells in practical application, and how accurate the identification is?
4. In Figure 3, the author should indicate what the black line and red line represent respectively in the figure.
5. In part 3, although COLO 205 cell line seemed to have the most similar morphological pattern to PBMCs than the rest of the cell lines, the volume characters shown in Figure 3C are quite different between COLO 205 cell line and PBMCs. Why not use volume to distinguish them?
6. The morphological properties of COLO 205 cell line and PBMCs acquired by DHC was able to discriminate the COLO 205 and PBMC cells. Why consequent biomarker analysis is needed? Is it necessary to combine DHC and biomarker analysis in this task?
7. The morphological difference between CRC cell lines and PBMCs seems to be big. So, it is easy to distinguish them with DHC. Whether all cancer cells are morphological different from PBMCs?
Author Response
About the DHC imaging process in part 2.4, some information is needed, such as pixels of camera, magnification of microscope objective, time required to acquire 30 images, etc.
Reply: We have now added the following information in part 2.4: Image capture was performed with a low-intensity 635-nm diode laser, which prevents phototoxicity. The optical magnification used was 10x and the resolution was 0.5 μm.
Besides, the author should show the picture of the DHC system, so that readers can better understand the obtaining process of experimental images.
Reply: We have now prepared a graphical abstract to highlight and clarify the DHC system in a better way.
In Figure 1A, the distribution difference morphological characteristics of COLO 205 cell line and PBMCs can be seen. However, the author should explain how to identify these cells in practical application, and how accurate the identification is?
Reply: The morphological characteristics of COLO205 and PBMCs were shown in Figure 2. We assume that the referee mean Fig 2A. We have added text referring to figure 2B as well.
For identification with DHM, we have added the following sentence in part 2.4: “The cells in the images are segmented to extract information on each imaged cell separately, and thereby cellular parameters such as area, optical thickness and optical volume were obtained.”
As a proof of concept study, we have not been able to specify how accurate the identification would be. Ideally the idea is to take patient’s blood and analyse it with DHM and discriminate cancer cells according to their morphological parameters. It would be also beneficial to include machine learning and advanced image analysis for better detection of cancer cells which would be eventually our next step.
In Figure 3, the author should indicate what the black line and red line represent respectively in the figure.
Reply: we have now clarified this in the text, and not only in the figure text.
In part 3, although COLO 205 cell line seemed to have the most similar morphological pattern to PBMCs than the rest of the cell lines, the volume characters shown in Figure 3C are quite different between COLO 205 cell line and PBMCs. Why not use volume to distinguish them?
Reply: The volume characters of COLO205 and PBMCs were shown in Figure 1C. We assume that the referee mean Fig 1C.
First, the referee´s comment is correct. For this specific cell line and PBMC, the difference is enough. However, for other types of CRC cells, or even for other types of cancer, it may be less difference in volume between the cancer cell and the PBMCs. Moreover, the volume calculations are a result of both area and thickness, so we are careful to not only use the volume as the main result.
The morphological properties of COLO 205 cell line and PBMCs acquired by DHC was able to discriminate the COLO 205 and PBMC cells. Why consequent biomarker analysis is needed? Is it necessary to combine DHC and biomarker analysis in this task?
Reply: The referee´s comment is relevant. In this case, the biomarker analysis would give an additional proof that we are indeed analysing cancer cells and differ them from PBMCs, since there are cases where it is much harder to differ between the blood cells and cancer cells.
We would like to add that under the Cell Search detection method (the first FDA-approved method for CTC detection), CTCs must possess the following properties: a round to oval shape by light scatter, an evident nucleus by 4′,6-diamidino-2-phenylindole (DAPI) staining, epithelial cell adhesion molecule positivity (EpCAM+), and cytokeratin (CK)-8+, -18+, -19+, and CD45− by immunofluorescence (see for example reference in the paper: Allen et al, Curr Colorectal Cancer Rep. 2010, 6(4): 212–220).
The morphological difference between CRC cell lines and PBMCs seems to be big. So, it is easy to distinguish them with DHC. Whether all cancer cells are morphological different from PBMCs?
Reply: We think this question (and answer) goes back to both question 5 and 6. For this specific cell line, COLO 205, and PBMC, the difference in volume is enough. However, for other types of CRC cells, or even for other types of cancer, it may be less difference in volume between the cancer cell and the PBMCs. Moreover, the volume calculations are a result of both area and thickness, so we are careful to not only use the volume as the main result.
Round 2
Reviewer 2 Report
I would like to thank the authors for making the changes and taking my comments into account.
Additional comments:
1) Since the sample size is limited (one donor of PBMCs/three healthy donors) please indicate in the text of the manuscript that this is a proof-of-concept study and further examination are necessary to validate the results. You can also consider the change of the article type on Case Report.
2) The authors could only have considered the issue of the added references regarding digital holographic tomography, as the added items yes refer to the proposed technique, but do not directly address the analysis and characterisation of colorectal cancer cells, their structures, or circulating tumour cells addressed in this article. They do not even refer to selected colorectal disorders. I therefore suggest that these references be better tailored so that they are relevant to the subject matter of the manuscript.
3) Figure 2 is outside the page area.
4) Table 1 is added in the fragment of caption of Figure 2.
Author Response
We thank the reviewer and are pleased to respond to the additional Reviewer Comments for the manuscript “Circulating tumor cell models mimicking metastasizing cells in vitro: discrimination of colorectal cancer cells and white blood cells using digital holographic cytometry” submitted by Feith et al.
Please find our response below after each comment from the reviewer:
I would like to thank the authors for making the changes and taking my comments into account.
Additional comments:
1. Since the sample size is limited (one donor of PBMCs/three healthy donors) please indicate in the text of the manuscript that this is a proof-of-concept study and further examination are necessary to validate the results. You can also consider the change of the article type on Case Report.
REPLY: We have now added in the text, at page 8, in the Discussion part: “Since this was performed as a proof-of-concept study with small sample number, further examinations are necessary to validate the results.”
2. The authors could only have considered the issue of the added references regarding digital holographic tomography, as the added items yes refer to the proposed technique, but do not directly address the analysis and characterisation of colorectal cancer cells, their structures, or circulating tumour cells addressed in this article. They do not even refer to selected colorectal disorders. I therefore suggest that these references be better tailored so that they are relevant to the subject matter of the manuscript.
REPLY:The added text from revision round 1 has now been adjusted. The text in red has now been added in revision round 2.
Indeed, optical tomography provide a more complete mapping of the biological sample in combination with QPI (Merola 2018). The 3D refractive index will allow studies of subcellular and micrometer structures of the cells (Balasubramani Appl Optics 2021; Balasubramani J Imaging 2021). By combining all time-lapse quantitative phase maps, each orientation of the cell can be obtained (Merola 2017).
For high-throughput 3D cell measurements, microfluidic devices have been integrated to tomographic phase microscopy including work on red blood cells and CTCs (Merola 2018). Nissim et al recently showed in their system that blood cells mixed with colorectal cancer cells could be automatically discriminated on the level of individual cell types, by using label-free holographic flow cytometry (Nissim 2021). By their custom-built optical system operated under flow, they received single-cell holograms in real time and applied image processing and machine learning. The database created found features that could differ between cancer cells and various blood cells.
Moreover, cell sorting was included in a tomographic interferometry approach by rotating cells in a flow with dielectrophoretic (DEP) forces (Habaza 2017; Dudaie 2020). Interestingly, here the authors used three types of CRCs, HT29, SW-480 and SW-620 together with blood cells (Dudaide 2020). The CRC cell type mixed with blood cells, i.e. the liquid biopsy, were first enriched by filtration and then the CRCs were cells during flow using machine learning, and then isolated by using activating DEP. Most elegantly, the authors also built in a fluorescence imaging system for an external validation where only the cancer cells emitted fluorescence light. Moreover, a recent study with an approach aiming for clinical use, also showed classification of cancer cells based on the cell spatial and temporal fluctuations (Baruch 2021). The classification method can both be used to detect CTCs from blood and analyse cancer cells from tissue or solid tumors.
3. Figure 2 is outside the page area.
REPLY: We have now adjusted Figure 2.
4. Table 1 is added in the fragment of caption of Figure 2.
REPLY: We have now adjusted Table 1.
