The Chicken Chorioallantoic Membrane Tumor Assay as a Relevant In Vivo Model to Study the Impact of Hypoxia on Tumor Progression and Metastasis

Round 1

Reviewer 1 Report

Harper and colleagues described a well known technique allowing grafting of cells or tissues on chorioallantoic membrane of chick embryo. This robust technique is used since decades (MacLachlan 1982 ?). Authors should clearly indicate the novelty they bring with this manuscript. In addition, I have some concerns regarding the study itself.

1) Authors used matrigel to embedded cells before grafting. Some authors (i.e. Hagedorn et al 2005 - https://doi.org/10.1073/pnas.0408622102) did not use matrigel, while some other do (Peulen et al 2013 - https://doi.org/10.1371/journal.pone.0075102). Authors should give a comment about these alternatives and explain why they choose to use matrigel. They should indicate in material and methods section if the matrigel contains growth factors.

2) In the previously mentioned articles (Peulen et al 2013 - https://doi.org/10.1371/journal.pone.0075102), authors were unsuccessful to grow Panc-1 on CAM. Authors should discuss this point. At their opinion, what is the main point explaining this discrepancy?

3) Authors sacrificed embryos at ED16 and harvest tumors. Hatching is supposed to occur at ED20-21. Is it possible to maintain tumors until ED19-20, and therefore increase the tumoral volume ?

4) Authors should prove that blood vessels inside tumors are functional by showing the presence of red blood cells.

5) How were the tumor volume at ED16? Was there any correlation between tumor volume and hypoxia ?

6) It could be of interest to show abundance of HIF1a in the tumors grown on CAM.

7) In figure 1A, authors showed red staining supposed to be blood vessels labeled with LCA lectin. To be convincing, high magnification of blood vessel section should be showed, including vessel lumens.

8) In figure 2, LCA stained the periphery of the harvested tumors indicating an unspecific staining. A more specific marker should be used (i.e. CD31).

9) Line 279 - authors used PDK1 as marker of hypoxia. Full name should be stated at first appearance.

Author Response

Reviewer #1:

Harper and colleagues described a well-known technique allowing grafting of cells or tissues on chorioallantoic membrane of chick embryo. This robust technique is used since decades (MacLachlan 1982 ?). Authors should clearly indicate the novelty they bring with this manuscript. In addition, I have some concerns regarding the study itself.

REPLY:  The reviewer is right, the use of the CAM model in oncology has been recognized for many years.  However, it remains to be determined whether this model recreates the hypoxic microenvironment of solid tumors. We have put more emphasis to the novelty of the manuscript in the last paragraph of the introduction section (p2,3 lines 94-100) “Here, we provide an original demonstration that the CAM model efficiently supports the development of hypoxic zones in a variety of human tumor cell line-derived xenografts as well as human cancer patient-derived xenografts. These findings suggest that the CAM xenograft model is a relevant platform to test therapeutic interventions targeting hypoxia that can be applied to various cancer types.”

1) Authors used matrigel to embedded cells before grafting. Some authors (i.e. Hagedorn et al 2005 - https://doi.org/10.1073/pnas.0408622102) did not use matrigel, while some other do (Peulen et al 2013 - https://doi.org/10.1371/journal.pone.0075102). Authors should give a comment about these alternatives and explain why they choose to use matrigel. They should indicate in material and methods section if the matrigel contains growth factors.

REPLY: As it is the standard procedure for xenograft assays, we are using standard Matrigel matrix for implantation.  Our main raison for the use of Matrigel is to prevent cell-dispersion on the CAM during the implantation procedure. We also tried the ring method to encircle the implanted cells (with or without Matrigel) but this procedure can induce some alterations of underlying blood vessels so it was discarded.  These specifications were added to the updated material and methods section of the manuscript (p3, line 136.

2) In the previously mentioned articles (Peulen et al 2013 - https://doi.org/10.1371/journal.pone.0075102), authors were unsuccessful to grow Panc-1 on CAM. Authors should discuss this point. At their opinion, what is the main point explaining this discrepancy?

REPLY: In the Peulen study, the lack of growth of PANC-1 cell lines in the CAM assay could be due to the very larger number of PANC-1 cells (3.5 million cells) used for inoculation.  In our experience, 0.3 to 0.5 million cells are normally optimal for the CAM xenograft assay and that is true for a variety of cell lines. Increasing this number often results in a loss of cell viability likely during the procedure where cells are forced to stay at very high density during a variable period of time before inoculation. For clarification, we added in the Materials and Methods section (p3, lines 137-139) the reason justifying the number of cells used for implantation in our assays. “ The number of cells used for implantation was chosen based on preliminary concentration-dependent experiments where we found that high cell concentrations during the implantation procedure can result in a loss of cell viability.”

3) Authors sacrificed embryos at ED16 and harvest tumors. Hatching is supposed to occur at ED20-21. Is it possible to maintain tumors until ED19-20, and therefore increase the tumoral volume ?

REPLY: This possibility would in fact be interesting. In our studies we are using the ex ovo method for culturing CAM because even though the long-term viability of embryos is slightly lower than for the in ovo method, it gives us greater access to the CAM vasculature for cell implantation and most importantly intravenous injection of reagents such as pimonidazole.  Thereby, to increase the robustness of the assay, we need to perform the assay over a shorter experimental time window.  So, even though we might be able to increase the tumor volume by allowing the ex ovo CAMs to develop for a longer time-period, the robustness of the assay would likely suffer.

4) Authors should prove that blood vessels inside tumors are functional by showing the presence of red blood cells.

REPLY: We agree with the reviewer that functional blood vessels are essential for our model and we now show in a new Figure 2F H&E staining of CAM xenografts where we can observe nucleated chick erythrocytes within blood vessels inside these tumors (p7,8, lines 271-274).

5) How were the tumor volume at ED16? Was there any correlation between tumor volume and hypoxia ?

REPLY: Thanks for the suggestion. We evaluated the correlation between tumor size and the most differentially expressed hypoxic marker, CA9. A statistically significant negative association between CA9 expression and tumor volume was observed for vehicle- and sorafenib-treated xenografts while VEGF-treated xenografts display no relationship. This information is now added as a new supplementary Figure 1 and in result section (p8, lines 289-292).

6) It could be of interest to show abundance of HIF1a in the tumors grown on CAM.

REPLY: Yes, as the reviewer mentioned, results from HIF-1 staining/western blotting would be interesting, and we initially thought to include this measurement in our experimental design. Because of the particularly short half-life of the HIF-1 proteins, we finally decided to measure the expression of HIF-1-dependent genes instead, as this approach is more robust than HIF-1 detection per se.

7) In figure 1A, authors showed red staining supposed to be blood vessels labeled with LCA lectin. To be convincing, high magnification of blood vessel section should be showed, including vessel lumens.

REPLY: In order to provide more convincing evidence of the blood vessel within CAM xenografts we now show, in a new Figure 2E and an associated video, a 3D reconstruction of the CAM vasculature that reveals the 3D vascular network (p8, lines 289-292; see also previous response to comment 4).

8) In figure 2, LCA stained the periphery of the harvested tumors indicating an unspecific staining. A more specific marker should be used (i.e. CD31).

REPLY: Thanks for the critical insight of the reviewer.  In figure 2B, thick 30 mm tissue sections were used to quantify angiogenesis. In that case, it is the CAM vasculature surrounding the tumor that appears denser and more prominent compared to a tumor slice cut into standard 7mm sections (see for comparison Figure 1A).  This is especially marked in the case where VEGF is used as a treatment. 

Unfortunately, the CD31 marker, which is the gold standard marker of blood vessel, does not detect blood vessels of chick origin and thereby cannot be used in the CAM xenograft assay. So we are using the LCA reagent which is the gold standard to detect CAM vasculature ( Jilani et al., J. Histochem Cytochem, 2003, PMID: 12704207 ; Deryugina et al., Methods Enzymol, 2009, PMID: 19007659;  Subauste et al., Clin Exp Metastasis, 2009, PMID: 19842048 ; Mangir et al., in vivo, 2018 PMID: 29695547; Mangir et al., 2019 ACS Biomaterials Science and Engineering, PMID: 33405582). LCA reagent is also a good option when using 30 mm-thick tissue slices to quantify angiogenesis.  We now refer to the above-mentioned studies in the material and methods section (p4, lines 177,178) of the revised manuscript.

9) Line 279 - authors used PDK1 as marker of hypoxia. Full name should be stated at first appearance.

REPLY: Full name is now found in the revised version of the manuscript (p8, line 287), as suggested by the reviewer.

Reviewer 2 Report

The authors explore in their work the use of the chicken chorioallantoic membrane assay (CAM) as a model to study hypoxia as well as its applicability as a drug testing platform.  In the study, different cancer cell lines and patient-derived xenografts (from renal carcinomas) were transplanted and tumor growth and levels of hypoxia were evaluated. The authors demonstrate the presence of hypoxic microenvironments that can also be modulated by VEGF and clinically relevant drugs such as sorafenib, as indicated hypoxia stainings expression changes of hypoxia-related genes. In particular, the differences observed for different patient-derived xenografts are interesting. In general, the authors provide a well-written report, and the study shows the applicability of the model for exploring new therapeutic drugs. There are some points, however, that should be improved:

Introduction:

The authors have provided a nice overview of how the CAM model has been used over the past 20 years for cancer research. Regarding the fact that this technique has been out for a while, I found the sentence: “the CAM model is emerging as alternative in vivo model” a bit misleading.

Results:

Fig. 1 To which in vitro culture condition was the expression of CAIX in xenografts compared to? 2D or 3D culture in Matrigel?

Fig 1b: the figure caption is somewhat misleading “n=3 cell lines, n=6-13 xenografts”. The xenografts in 1B are also from cell lines, correct? 2D culture versus xenografts?

Please indicate statistical test performed for datasets presented in 1B. Although the dimensions of the right fluorescence pictures can be derived from the rectangles on the left, it would be helpful to also provide a scale bar on the right.

In line 222-224 the authors state that the mRNA expression correlates with the pimonidazole stainings. Was the level of staining quantified? It appears from the methods that some quantification of the hypoxyprobe-positive region was performed. The quantitative data should be shown, otherwise it should be pointed out that it is only a qualitative observation.

Fig 2B, same as above, indicate scale bar also in zoom-ins. The figure legend should state which statistical test was used. In the method section only t-tests are mentioned. However, this would in the case of 2A, C and D not be appropriate since there are more than two conditions compared.

It is not clear from the methods, how anastomoses and capillaries were quantified. The authors just refer to a previous paper. A brief statement in the methods would be helpful.

Fig. 3 and 4 see above comment for statistical test.

The observation that after Sora treatment increased metastasis was found is interesting and should be discussed in the discussion section.

Fig.4G Was the level of hypoxia quantified? Did the stainings for patient 2 and 3 not show any changes as expected from the mRNA levels of hypoxia-related genes?

How many fragments were studied for each patient? The number should be given in the legend. Since it is about the establishment of a model, I would expect some discussion on the variations that are observed from fragment to fragment.

The finding that the three patient-derived xenografts show different responses to drug treatments is interesting. Is it possible to correlate this result to the patient outcome (if the patients were treated with sora)?

Author Response

Reviewer #2

Comments and Suggestions for Authors

The authors explore in their work the use of the chicken chorioallantoic membrane assay (CAM) as a model to study hypoxia as well as its applicability as a drug testing platform.  In the study, different cancer cell lines and patient-derived xenografts (from renal carcinomas) were transplanted and tumor growth and levels of hypoxia were evaluated. The authors demonstrate the presence of hypoxic microenvironments that can also be modulated by VEGF and clinically relevant drugs such as sorafenib, as indicated hypoxia stainings expression changes of hypoxia-related genes. In particular, the differences observed for different patient-derived xenografts are interesting. In general, the authors provide a well-written report, and the study shows the applicability of the model for exploring new therapeutic drugs. There are some points, however, that should be improved:

Introduction:

1) The authors have provided a nice overview of how the CAM model has been used over the past 20 years for cancer research. Regarding the fact that this technique has been out for a while, I found the sentence: “the CAM model is emerging as alternative in vivo model” a bit misleading.

REPLY:   Thanks for this insightful comment.  To make sure that we don’t mislead the readers, we changed this sentence for “  The chicken ChorioAllantoic Membrane (CAM) model provides an attractive alternative in vivo model for cancer research” (p 2, lines79,80).

Results:

2) Fig. 1 To which in vitro culture condition was the expression of CAIX in xenografts compared to? 2D or 3D culture in Matrigel?

REPLY:  We used the standard 2D culture method without Matrigel. This is now specified within the revised legend of Figure 1.

3) Fig 1b: the figure caption is somewhat misleading “n=3 cell lines, n=6-13 xenografts”. The xenografts in 1B are also from cell lines, correct? 2D culture versus xenografts?

REPLY: Thanks for noticing this. The revised version of the legend of Figure 1B has been corrected for “cell lines in 2D culture n=3, cell line-derived xenografts n=6-13 “.

4) Please indicate statistical test performed for datasets presented in 1B. Although the dimensions of the right fluorescence pictures can be derived from the rectangles on the left, it would be helpful to also provide a scale bar on the right.

REPLY: The statistical tests for the datasets in Figure 1B are now indicated in the figure legend (p7, line 255).  Scale bars have also been added to the zoomed images in Figure1A, and their length is indicated in the figure legend (p7, line 251).  

5) In line 222-224 the authors state that the mRNA expression correlates with the pimonidazole stainings. Was the level of staining quantified? It appears from the methods that some quantification of the hypoxyprobe-positive region was performed. The quantitative data should be shown, otherwise it should be pointed out that it is only a qualitative observation.

REPLY: Pimonidazole staining has indeed not been quantified in that particular experiment. We thereby modified the results section accordingly to reviewer’s comment: “Furthermore, the relative pimonidazole staining intensity in cell line-derived xenografts appeared to be consistent with the level of CAIX mRNA expression” (p.5, lines 225,226)

6) Fig 2B, same as above, indicate scale bar also in zoom-ins. The figure legend should state which statistical test was used. In the method section only t-tests are mentioned. However, this would in the case of 2A, C and D not be appropriate since there are more than two conditions compared.

REPLY: Scale bars have been added to the zoomed images of Figure 2B, and their length is indicated in the figure legend.  Thanks for bringing this to our attention, the methods section has now been updated to include one-way ANOVA  (p. 5, line 203) and the statistical tests are now indicated in the figure legend of figure 2.

7) It is not clear from the methods, how anastomoses and capillaries were quantified. The authors just refer to a previous paper. A brief statement in the methods would be helpful.

REPLY: We added additional information in the revised version of the Material and Methods section (p 4, lines 195-199) on the number and size of the counting frames used and the method used to count the number of capillaries.

8) Fig. 3 and 4 see above comment for statistical test.

REPLY: The statistical tests for figure 3 and 4 are now indicated in the corresponding figure legends.

9) The observation that after Sora treatment increased metastasis was found is interesting and should be discussed in the discussion section.

REPLY: We don’t know the exact mechanism by which sorafenib treatment increase metastasis but this is in keeping with the fact that hypoxia can reduce cell-cell attachment allowing cancer cells to invade, thereby increasing the risk of both local invasion and metastatic spread. This statement was added to the revised manuscript (p8, lines 302-305)

10) Fig.4G Was the level of hypoxia quantified? Did the staining for patient 2 and 3 not show any changes as expected from the mRNA levels of hypoxia-related genes?

REPLY: The reviewer is correct, we did not observe any changes in pimonidazole staining for patient # 2 and #3.  To support this, representative immunofluorescence images from these patients are now included in the updated Fig 4G and associated result section. In Fig.4D-F, we analyzed a set of hypoxia-responsive genes but did not quantify pimonidazole staining in Fig 4G.

11) How many fragments were studied for each patient? The number should be given in the legend. Since it is about the establishment of a model, I would expect some discussion on the variations that are observed from fragment to fragment.

REPLY: The number of fragments tested per group is now indicated in the legend of Figure 4. In fact, a higher tumor size and gene expression variability were observed when using tumor fragment-derived xenografts (Fig 4A-C) compared to xenografts derived from cell lines (Fig 2A), an observation likely due to intra-tumoral heterogeneity of ccRCC tumors. This is now mentioned in the result section associated with Fig.4 (p3, lines 318,319).

12)The finding that the three patient-derived xenografts show different responses to drug treatments is interesting. Is it possible to correlate this result to the patient outcome (if the patients were treated with sora)?

REPLY:  Thanks for the comment.  We would have liked to be able to do such correlation. Unfortunately for our study, but very good for the patients, the 3 patients treated with Sorafenib in CAM assays were found to be low grade ccRCC and didn’t need adjuvant therapy. Such precision is now included in the result section “ We could not retrospectively analyze the clinical response to sorafenib as the 3 patients tested in the CAM assay did not receive sorafenib treatment” (p8, lines 314-316).

Reviewer 3 Report

In this manuscript the authors aim to demonstrate that CAM-based tumor model is a relevant in vivo platform to 20 further understand the pathological responses to hypoxia and test therapeutic interventions aimed 21 at targeting hypoxic cancers.

The rationale behind the paper is very interesting with very important therapeutic clinical implications. The paper is well writeen, the results and conclusions clearly supported by the data. Still, there are minor points to be clarified:

1. Did the authors studied other metastatic sites? Why did you choose specifically the CAM liver?
2. Are the results for the metastasis only performed for the HT-1080 cells, or this effect is also observed in other cancer models? If it was only for HT-1080, please specify that information in the main text.
3. Did the authors tried to treated the patient-derived Sorafenib resistant xenografts to the Sorafenib

Author Response

Reviewer #3

Comments and Suggestions for Authors

In this manuscript the authors aim to demonstrate that CAM-based tumor model is a relevant in vivo platform to further understand the pathological responses to hypoxia and test therapeutic interventions aimed at targeting hypoxic cancers.

The rationale behind the paper is very interesting with very important therapeutic clinical implications. The paper is well written, the results and conclusions clearly supported by the data. Still, there are minor points to be clarified:

• Did the authors study other metastatic sites? Why did you choose specifically the CAM liver?

REPLY: In fact, we also tried the lung and intestine but more reproducible results were found in the liver.  In addition, at chick embryo development day 16, the liver is bigger and easier to collect compared to some other chick organs.

• Are the results for the metastasis only performed for the HT-1080 cells, or this effect is also observed in other cancer models? If it was only for HT-1080, please specify that information in the main text.

REPLY: Metastasis were indeed measured in HT-1080-derived xenografts. As suggested by the reviewer, this information is now specified in the results section (p8, line 298). 

• Did the authors tried to treated the patient-derived Sorafenib resistant xenografts to the Sorafenib

REPLY: Thanks for this interesting comment. Unfortunately for the study, but very good for the patients, the 3 patients treated with Sorafenib in CAM assays were found to be low grade ccRCC and didn’t need adjuvant therapy. Such precision is now included in the result section “ We could not retrospectively analyze the clinical response to sorafenib as the 3 patients tested in the CAM assay did not receive sorafenib treatment” (p8, lines314-316).

Round 2

Reviewer 1 Report

Authors have addressed all my concerns.

Concerning HIF1a western-blot, addition of proteasome inhibitor MG132 while protein extraction would protect HIF1A from degradation, increasing robustness of HIF1A western-blot.

Point 8 - I guessed it was 30 µm section as reported in the main manuscript and not 30 mm section (!).