ITLN1 modulates invasive potential and metabolic reprogramming of ovarian cancer cells in omental microenvironment

Advanced ovarian cancer usually spreads to the omentum. However, the omental cell-derived molecular determinants modulating its progression have not been thoroughly characterized. Here, we show that circulating ITLN1 has prognostic significance in patients with advanced ovarian cancer. Further studies demonstrate that ITLN1 suppresses lactotransferrin’s effect on ovarian cancer cell invasion potential and proliferation by decreasing MMP1 expression and inducing a metabolic shift in metastatic ovarian cancer cells. Additionally, ovarian cancer-bearing mice treated with ITLN1 demonstrate marked decrease in tumor growth rates. These data suggest that downregulation of mesothelial cell-derived ITLN1 in the omental tumor microenvironment facilitates ovarian cancer progression.

Summary: In this manuscript, the authors showed that mesothelial cells in the omental microenvironment modulate the invasive potential and proliferation of ovarian cancer (OC) cells via ITLN1. The expression level of ITLN1 was down-regulated in mesothelial cells enriched in the OC omental adipose tissue compared to normal controls. The authors showed that pro-inflammatory cytokines released from the OC-associated omental microenvironment decrease the expression level of ITLN1 in mesothelial cells. They also demonstrated that ITLN1 suppresses the motility and invasive potential of OC cells via the ITLN1-LTF-MMP1 axis. In addition, ITLN1 suppresses the proliferation of OC cells by increasing glucose uptake in adipocytes through upregulation of GLUT4, which was supported by both in vitro and in vivo observations. The present manuscript makes a valuable contribution by providing a mechanistic insight into the cross-talk between mesothelial cells in the OC-associated omental microenvironment and OC cells. Furthermore, they demonstrated the feasibility of using ITLN1 as a therapeutic agent for OC treatment in vivo. However, their proposed mechanism regarding the ITLN1-LTF axis requires further in vivo validation.
Major points: 1. Supplementary Figure 1a: The authors showed that mesothelial cells are the major source of ITLN1 by comparing its expression levels among diverse cell types enriched in the omental adipose tissue. However, they used cell lines whose genetic background and culture conditions are not identical, which possibly affects the expression levels of genes. A more direct comparison using cells isolated from the omental adipose tissue (e.g. single-cell RNA-seq) should be applied. Figure 1b-c: The in vitro data suggest that pro-inflammatory cytokines such as TNF-alpha and TGF-beta downregulate the expression of ITLN1 in mesothelial cells of OC patients. The cross-talk between immune cells and mesothelial cells via cytokines is one of the key findings in this manuscript, but this was not strongly supported by in vivo data. I'm wondering whether these pro-inflammatory cytokines are more highly expressed in OC patients than in normal controls (using serum or the omental adipose tissue). What are the major cell types secreting these cytokines in the omental microenvironment? Are the corresponding cytokine receptors are upregulated in mesothelial cells of OC patients? The repertoires of cytokines affecting the expression level of ITLN1 in mesothelial cells should be systematically examined by analyzing the transcriptome data.

Supplementary
3. Figure 2e: I'm wondering whether the AUC score was calculated using cross-validation. I could not find the details of the logistic regression analysis. Without cross-validation, the prediction model with more features are more prone to over-fitting.
4. LTF and MMP1: Which cell types in the omental microenvironment are the major source of LTF? The authors mentioned that neutrophils might be a candidate, but this should be validated. Are the levels of LTF up-regulated in the omental adipose tissue or peritoneal fluid of OC patients compared to normal controls? Is there any correlation between circulating LTF and ITLN1 in OC patients? I'm also wondering whether we can use LTF as a diagnostic marker in combination with CA125 and ITLN1. 5. Figure 6: ITLN1 seems to increase glucose uptake in adipocytes by upregulating GLUT4 in a LTFdependent manner. I'm wondering whether ITLN1 can upregulate GLUT4 in adipocytes cultured in SFM without LTF.

Reviewer #2 (Remarks to the Author):
This is a very comprehensive study that includes observational data from human tissues, blood and cell lines, functional and biochemical data and in vivo studies to show the cellular interplay in the tumor microenvironment of mesothelial cells and adipocytes and their impact on ovarian cancer cells. Specifically, the authors show that circulating ITLN1 is low in women with high grade ovarian cancer, that high ITLN1 levels decrease ovarian cancer cell proliferation, motility and invasion and that ITLN1 is produced only in mesothelial cells. They go on to show that ITLN1 increases glucose uptake in mature adipocytes and this leads to glucose starvation of ovarian cancer cells and the resulting decreased invasive potential.
The study is novel and will be of great interest to the ovarian cancer and general cancer research community. My concerns are related to the lack of detail in the Methods throughout the manuscript that made it difficult to interpret the data. Some examples of concerns are detailed below.
In Results it states that mesothelial cells were isolated from the peritoneal fluid of healthy women. There is no reference to this in the methods and it is unclear how this is possible. Please clarify and correct.
The manuscript includes the use of tissues from women. However, no ethics information is supplied. Please add in the appropriate ethics protocol identifiers.
Figure 1e -tissues were from non-ovarian cancerous and ovarian cancerous tissues, but not normal tissue as stated. Please correct here and throughout.
Confirmation of preadipocytes and mature adipocytes needs to be shown.
There is no information on how the co-culture experiments (Fig 1f,g) were performed. Please clarify. Conditions and timings for the MTT assays need to be included (SI Fig 4) Co-culture conditions (Fig 4 and SI fig 6) need to be included.
They report from the literature but do not demonstrate that ITLN1 increases insulin-dependent glucose update exclusively in adipocytes. They show that ITLN1 in the presence of mature adipocytes decreased the growth of ovarian cancer cells. However, the methods state that the only cells that were grown with insulin were the adipocytes, thus these were essentially the only cells tested under the appropriate conditions. Their conclusion that only mature adipocytes play a role in mediating ITLN1's growth suppressive effects on cancer cells requires that all cells be grown in the same conditions. Similarly, GLUT4 is regulated by insulin, so different results would be expected in the cells grown with and without insulin.
Reviewer #3 (Remarks to the Author): The manuscript by Au-Yeng and colleagues provides evidence that the expression of intelectin-1 (ITLN1), an intestinal lactoferrin (LTF) receptor, is down-regulated in ovarian cancer (OC) associated mesothelial cells. Decreased ITLN1 expression is found in serum samples of OC patients and in mice bearing OC. Serum ITLN1 levels have prognostic significance in OC patients. Mechanistically, they show that ITLN1 represses OC migration/invasion and the expression of collagenase (MMP1).
Apparently, the ITLN1-LTF interaction prevents LTF to interact with its receptor LRP1, which is responsible for the ERK1/2 mediated activation and upregulation of MMP1 expression, thus resulting in the attenuation of OC invasive potential. Co-culture experiments of OC cells with adipocytes in the presence of ITLN1 indicated a growth-suppressive effect of ITLN1 on OC cells that was abrogated by the addition of glucose. These findings suggested that the ITLN1-induced glucose uptake in adipocytes restricted glucose utilization and limited growth of OC cells. Consistently, they show that ITLN1 upregulates GLUT4 expression in the adipocytes and show that GLUT4 is required to abrogate OC cells growth by ITLN1. Interestingly, LTF downregulates GLUT4 expression and glucose uptake leading to enhanced OC growth. Analysis of the glycolytic flux by determination of the lactate released in the medium and by the production of labeled pyruvate and lactate by GC-MS, indicated that ITLN1 diminished the glycolytic flux of OC cells co-cultured with adipocytes, suggesting that ITLN1-treated adipocytes inhibit the glycolytic of OC cells. Finally, they show that ITLN1 administration to mice bearing OC suppresses MMP1 expression and arrests tumor growth in vivo. By using a cutting edge MALDI-IMS approach to analyze metabolites in tissue sections from the omental tumors derived in mice, they report a rapid 1h reduction of glucose-6-phosphate and lactate content in tumor areas whereas the adjacent adipocytes showed the opposite trend in both metabolites in response to ITLN1 administration. Overall, the manuscript supports both in vitro and in vivo that ITLN1 suppresses tumor growth by limiting the glucose available to OC cells by an unfavorable competition with the nearby adipocytes. I find this paper solid, addressing mechanistic aspects and very convincing. The MALDI-imaging mass spectrometry approach nicely documents a rapid opposite effect of ITLN1 administration in glucose utilization and glycolytic flux in adipocytes when compared to the nearby tumor cells.

Reviewer #1
Comment: Supplementary Figure 1a: The authors showed that mesothelial cells are the major source of ITLN1 by comparing its expression levels among diverse cell types enriched in the omental adipose tissue. However, they used cell lines whose genetic background and culture conditions are not identical, which possibly affects the expression levels of genes. A more direct comparison using cells isolated from the omental adipose tissue (e.g. single-cell RNA-seq) should be applied.

Response:
In order to demonstrate that mesothelial cells are the major source of ITLN1 and the spatial distribution of ITLN1 expressing cells, we performed immunolocalization of ITLN1 and calretinin (a known mesothelial cell marker) on formalin-fixed paraffin-embedded (FFPE) omental tissue sections from healthy women and patients with HGSC instead of using singlecell RNA-seq. We found that ITLN1 is highly expressed in normal adipose tissues but not in cancer-associated adipose tissues. The expression of ITLN1 is also highly co-localized with calretinin positive mesothelial cells covering the omental adipose tissue but not in other cell types. Representative microscopic images are presented as Fig. 1e and are described in the Results section on page 4 of the revised manuscript. Figure 1b-c: The in vivo data suggest that pro-inflammatory cytokines such as TNF-alpha and TGF-beta downregulate the expression of ITLN1 in mesothelial cells of OC patients. The cross-talk between immune cells and mesothelial cells via cytokines is one of the key findings in this manuscript, but this was not strongly supported by in vivo data. I'm wondering whether these pro-inflammatory cytokines are more highly expressed in OC patients than in normal controls (using serum or the omental adipose tissue).

Response:
The expression levels of TNF-α and TGF-β in serum samples from healthy women and HGSC patients are measured using commercially available ELISA kit (BE69211 and BE69206, respectively; IBL America) according to the manufacturer's protocol. The mean levels of both TNF-α and TGF-β are higher in serum from HGSC patients compared to that from healthy women (TNF-α mean level: 116 vs 104 pg/mL, TGF-β mean level: 252 vs 154 pg/mL). However, the difference does not reach significance. The results are presented in Supplementary Fig. 1d and 1e, and are described in the Results section on page 5 of the revised manuscript.

Comment:
What are the major cell types secreting these cytokines in the omental microenvironment?
Response: Cytokines like TNF-α and TGF-β are secreted by many cell types. The major cell type that produce TNF-α is macrophages. Besides, natural killer cells, neutrophils, mast cells, endothelial cells, adipose tissues and fibroblasts that are present in the omental microenvironment also produce TNF-α. (Parameswaran and Patial, 2010) For TGF-β, it is also mainly expressed in the immune system including macrophages. (Wrzesinski et al., 2007, Yang et al., 2013 Comment: Are the corresponding cytokine receptors are upregulated in mesothelial cells of OC patients? The repertoires of cytokines affecting the expression level of ITLN1 in mesothelial cells should be systematically examined by analyzing the transcriptome data.

Response:
The RNA-seq data on ovarian cancer-associated mesothelial cells and normal mesothelial cells showed that TNF-alpha (TNFRSF1A and TNFRSFR1B) and TGF-beta (TGFBR2 and TGFBR3) receptors are upregulated in mesothelial cells from OC patients. Meanwhile, TNF-alpha and TGF-beta were also shown to be upregulated in cancer-associated mesothelial cells compared to normal mesothelial cells. The data are summarized as Supplementary Table 2 in the revised manuscript. Although the circulating levels of TNF-alpha and TGF-beta are not significantly higher in HGSC patients than in healthy women, the localized levels of these cytokines in mesothelial cells are increased, so do the levels of their corresponding receptors. This suggests that the expression level of ITLN1 in mesothelial cells in HGSC patients is regulated by these cytokines in an autocrine and paracrine manner in the tumor microenvironment. The results are described in the Results section on page 5 and Discussion section on pages 14 and 15 of the revised manuscript. Figure 2e: I'm wondering whether the AUC score was calculated using crossvalidation. I could not find the details of the logistic regression analysis. Without cross-validation, the prediction model with more features are more prone to over-fitting.

Response:
The cross-validation is a good technique to optimize hyperparameters and reduce the chance of overfitting for a model with high dimensional feature space. However, in our case, we only tested the diagnostic value of one or two features, i.e., the ROC curves were generated based on the gene expression value of one feature (ITLN1) or two features (ITLN1 and CA125). Also, we did not include any hyperparameters in our model because it was not necessary to constrain the weights of the one or two given features. Cross-validation will be used for calculation if we want to select multiple markers from feature space containing hundreds or thousands of genes in the feature.
Comment: LTF and MMP1: Which cell types in the omental microenvironment are the major source of LTF? The authors mentioned that neutrophils might be a candidate, but this should be validated.
Response: LTF has been reported to be mainly found in human neutrophils. We have this observation validated in omental tumor tissues from HGSC patients using Opal multiplex immunohistochemistry. Cancer-associated omental tissues were stained with CD11b, CD66b and LTF together with DAPI (nuclear stain). CD11b and CD66b are used as neutrophil markers. We found that most of the LTF present in the tissues are co-localized with CD11b and CD66b. This confirms that neutrophil is the major source of LTF. Representative images are presented in Supplementary Fig. 2a-2f and are described in the Results section on page 7 of the revised manuscript.
Comment: Are the levels of LTF up-regulated in the omental adipose tissue or peritoneal fluid of OC patients compared to normal controls?

Response:
The expression levels of LTF were examined in cancer-associated adipose tissues from HGSC patients and normal adipose tissues from healthy women using immunohistochemical analysis. The results showed a significant upregulation of LTF in ovarian cancer-associated adipose tissue compared to normal adipose tissues (n=7 for each group, p=0.001). The results are presented in Supplementary Fig. 2g and are described in the Results section on page 7 of the revised manuscript. In addition, the level of LTF in sera from healthy women, patients with benign diseases and HGSC patients, and that in ascites from HGSC patients are measured using a commercially available ELISA kit (ORG 527; IBL America) according to the manufacturer's protocol. We found that there is a trend of increasing levels of LTF in sera from healthy women to HGSC patients although the change does not reach significance. However, the level of LTF in ascites, which is rich in neutrophils, from HGSC patients is significantly higher than that in sera from any of the group examined. The results are presented in Supplementary Fig. 2h and are described in the Results section on page 8 of the revised manuscript.

Comment: Is there any correlation between circulating LTF and ITLN1 in OC patients?
Response: Human LTF concentrations in sera were measured using a commercially available ELISA kit (ORG 527; IBL America) according to the manufacturer's protocol. Human sera from healthy women, patients with benign gynecologic disease and patients with HGSC were obtained from the ovarian cancer repository of the Department of Gynecologic Oncology and Reproductive Medicine under protocols approved by MD Anderson Cancer's institutional review board. The ELISA results showed that there were no significant correlation between circulating LTF and ITLN1 levels in OC patients (r=0.021; p=0.738). The results are presented in Supplementary Fig. 3a and are described in the Results section on page 8 of the revised manuscript.
Comment: I'm also wondering whether we can use LTF as a diagnostic marker in combination with CA125 and ITLN1.
Response: Receiver operating characteristic curves were constructed for LTF, LTF in combination with CA125, and with CA125 and ITLN1 to test for discriminatory ability between healthy women and women with OC. We found that LTF alone had a significantly smaller AUC than CA125 alone (p=2.2e-16), and that for LTF with CA125 does not have a significant larger AUC than CA125 alone (p=0.448) (Supplementary Fig. 3c). However, ROC curve for LTF in combination with ITLN1 and CA125 showed a significant larger AUC than CA125 alone (p=0.005) or CA125 with ITLN1 (p=0.037) (Fig. 2e). These data suggest that LTF in combination with ITLN1 complements CA125 in identification of OC patients. The ROC curves for LTF and LTF with CA125 are presented in Supplementary Fig. 3c while that for LTF in combination with ITLN1 and CA125 are presented in Fig. 2e, and are described in the Results section on page 8 of the revised manuscript.

Response:
The cell treatments and conditions for Fig. 5 are as follows, Fig. 5a: SKOV3 and A224 cells were incubated in SFM with or without 100 μg/mL LTF for 24 h. Fig. 5b: SKOV3 and A224 cells were treated with 100 μg/mL LTF in SFM with or without 500 ng/mL ITLN1 for 24h. Comment: Conditions and timings for the MTT assays need to be included (SI Fig 4) Response: Conditions and timings for the MTT assays in Supplementary Fig. 4 (a, b, d-f, i) (now Supplementary Fig. 6a, 6c-6f and 7e in the revised manuscript) are included in the figure legends of Supplementary Fig. 6 and 7 in the revised manuscript.
Comment: Co-culture conditions (Fig 4 and SI Fig 6) need to be included.

Response:
The method of co-culture experiments in Figure 6 and SI Fig 4 are as follows, Mature adipocytes, preadipocytes or mesothelial cells were grown on the 0.4 µm pore size transwell insert (Thermo Fisher Scientific) and treated with ITLN1 and/or LTF for 24 h. Transwell inserts with cells were then put together with ovarian cancer cells that had been grown on the bottom well of the transwell. After 72 h, MTT assay and lactate secretion assay were performed to determine cell viability and lactate secretion, respectively. The co-culture experimental protocol is also described in the Methods section under the "Coculture conditions" subsection in the revised manuscript. Comment: They report from the literature but do not demonstrate that ITLN1 increases insulindependent glucose uptake exclusively in adipocytes. They show that ITLN1 in the presence of mature adipocytes decreased the growth of ovarian cancer cells. However, the methods state that the only cells that were with insulin were the adipocytes, thus there were essentially the only cells tested under the appropriate conditions. Their conclusion that only mature adipocytes play a role in mediating ITLN1's growth suppressive effects on cancer cells requires that all cells be grown in the same conditions. Similarly, GLUT4 is regulated by insulin, so different results would be expected in the cell grown with and without insulin.

Response:
We also include the effect of ITLN1 on glucose uptake in both mature adipocytes and preadipocytes in the absence of insulin in the revised Fig. 6b. The results show that ITLN1 has no significant effect on glucose uptake in both mature adipocytes and preadipocytes in the absence of insulin while significant increase in glucose uptake was only observed in mature adipocytes in the presence of insulin but not in preadipocytes. These results demonstrate that ITLN1 increases insulin-dependent glucose uptake exclusively in mature adipocytes. Based on this finding, all the other experiments using ITLN1-treated mature adipocytes to demonstrate the effect of glucose uptake on ovarian cancer cell growth, GLUT4 mRNA and protein expressions, lactate secretion and GC-MS are performed in the presence of insulin unless otherwise specified. The related figure legend was revised to address reviewer's concern in the revised manuscript. The effect of ITLN1 on GLUT4 mRNA expression in mature adipocytes was also examined in the absence of insulin. We found that there is no significant increase in GLUT4 mRNA expression after ITLN1 treatment in the absence of insulin. The results are presented in Supplementary Fig. 7a and are described in the Results section on page 11 of the revised manuscript.
This reviewer has focused primarily on the MALDI MSI data part of the manuscript, as this is my area of expertise. There are a number of questions concerning the MALDI MSI data which should be addressed (but I am confident that they are addressable) before the manuscript may be considered for publication.
MALDI Imaging data i) Figure 7H and Supplementary Figures 8f and 8g. The figures claim to report the normalized abundances of glucose-6-phosphate, lactate and ATP. It is not explained how the data was normalized, as such it is not possible to understand what metric is being displayed in the MSI figures. iv) Figure 7. The m/z given for the glucose-6-phospate [M-H]¬-anion is not correct. The mass they have given is that of the [M-H] neutral molecule, but the species measured by the mass spectrometer is the negatively charged molecule. And so the species they measure does not have a mass of 259.0219 but rather an m/z of 259.0224 because the additional electron has a mass of .00054858. v) Figure 7G. It is not clear why the authors have assigned the data to glucose-6-phospate when its isomer, fructose-6-phosphate, has identical mass. vi) Figure 7G. The images are interpolated and the color scales saturated. This is not accepted practice in the MSI field. The merged images (bottom row) do not add to the manuscript as the underlying histological images are barely visible. Please specify how the MSI and histological images co-registered, and provide mass accuracy for all assignments. Note: this reviewer is concerned how the authors can state in the methods section that "Metabolite identifications were made based on accurate mass, which typically has a discrepancy of less than 1 ppm" when the calculated masses referred to the figure captions are all incorrect (for instance the mass of an electron contributes approximately 2ppm mass error for glucose-6-phosphate). vii) Figure 7G. The authors have based their analysis on a single circular ROI for the tumor and the adipocyte regions for each animal. However it is not clear how these ROI's were selected or if the results would change if different ROIs were selected. For instance the adipocyte ROIs indicated in Figure 7G for the 0h and 1h time points have clearly very different morphological characteristics). Please provide clear criteria used for ROI selection, and average multiple ROIs from each animal for both tumor and adipocyte regions.
viii) Please refer to the MetaSpace program for help with confident assignment of metabolites to MSI data.

Response:
The monoisotopic molecular weight of ATP (C10H16N5O13P3) is 506.9957 and that of the [M-H] ion adduct (C10H15N5O13P3) is 505.9879 (or 505.9885 with an extra electron). For the same reason mentioned in the previous response, due to limitation of the TOF mass spectrometer and even with mass calibrated to within 5ppm error prior to acquiring the IMS data, the difference in mass with or without an electron would not be distinguishable, which the measured mass error for these runs are typically around 10ppm. There was a mistake in the corresponding Methods section in the previous version of the manuscript. The discrepancy should be rated to less than 10 ppm instead of 1 ppm. The corresponding Methods and Figure Legends section is revised according to reviewer's comment.
Comment: Figure 7. The m/z given for the glucose-6-phospate [M-H]-anion is not correct. The mass they have given is that of the [M-H] neutral molecule, but the species measured by the mass spectrometer is the negatively charged molecule. And so the species they measure does not have a mass of 259.0219 but rather an m/z of 259.0224 because the additional electron has a mass of .00054858.

Response:
The monoisotopic molecular weight of glucose-6-phosphate (C6H13O9P) is 260.0297 and that of the [M-H] ion adduct (C6H12O9P) is 259.0219 (or 259.0224 with an extra electron). For the same reason mentioned in the previous response, due to limitation of the TOF mass spectrometer and even with mass calibrated to within 5ppm error prior to acquiring the IMS data, the difference in mass with or without an electron would not be distinguishable, which the measured mass error for these runs are typically around 10ppm. There was a mistake in the corresponding Methods section in the previous version of the manuscript. The discrepancy should be rated to less than 10 ppm instead of 1 ppm. The corresponding Methods and Figure Legends section is revised according to reviewer's comment. Figure 7G. It is not clear why the authors have assigned the data to glucose-6phospate when its isomer, fructose-6-phosphate, has identical mass.

Response:
The labels in Figure 7g and h are corrected to "Normalized hexose-6-phosphate abundance" in the revised manuscript. The corresponding Results and Figure Legends sections are also revised. In this study, we aim at using the intermediate product of glycolysis as a measure of glucose consumption. Although we cannot distinguish glucose-6-phosphate from fructose-6phosphate by their masses, the decreases in hexose-6-phosphate (glucose-6-phosphate and fructose-6-phosphate) in tumor cells and their increases in adipocytes adjacent to tumor cells after ITLN1 injection suggest that ITLN1 treatment decreases glucose uptake of metastatic OC cells but increases glucose consumption of neighboring adipocytes in the omental microenvironment.
Comment: Figure 7G. The images are interpolated and the color scales saturated. This is not accepted practice in the MSI field. The merged images (bottom row) do not add to the manuscript as the underlying histological images are barely visible. Please specify how the MSI and histological images co-registered, and provide mass accuracy for all assignments. Note: this reviewer is concerned how the authors can state in the methods section that "Metabolite identifications were made based on accurate mass, which typically has a discrepancy of less than 1 ppm" when the calculated masses referred to the figure captions are all incorrect (for instance the mass of an electron contributes approximately 2ppm mass error for glucose-6-phosphate).

Response:
The heat map images were produced by High Definition Imaging software (HDI; Waters) when the tissue sections were scanned using the mass spectrometer. The interpolated images are for better visualization and the saturated color scale gives a better insight of the distribution of compounds and assists in coregistering with the histological images. However, quantification of the abundance of the compound of interest is done using Progenesis QI by importing the ROI raw data and is independent of the visual of the images. The merged images are removed from Fig. 7g in the revised manuscript according to the reviewer's comment. The same tissue section was washed and stained with hematoxylin and eosin (H&E) after data acquisition with the mass spectrometer. The image of H&E staining and heat map image were overlaid using HDI by the shape of the tissue section. This is described in the Method section under the "MALDI-IMS" subsection in the revised manuscript. All masses of compounds of interest are revised accordingly as well.
Comment: Figure 7G. The authors have based their analysis on a single circular ROI for the tumor and the adipocyte regions for each animal. However it is not clear how these ROI's were selected or if the results would change if different ROIs were selected. For instance the adipocyte ROIs indicated in Figure 7G for the 0h and 1h time points have clearly very different morphological characteristics). Please provide clear criteria used for ROI selection, and average multiple ROIs from each animal for both tumor and adipocyte regions.
Response: Fig. 7g shows representative images from MALDI-IMS of one ROI for tumor and adipocyte regions. Three ROIs for each region (tumor and adipocyte) were randomly selected based on histology of cell types for the analysis. Average of the three ROIs was taken before calculating for statistical significance. This is described in the Method section under the "MALDI-IMS" subsection in the revised manuscript. Adipocytes are identified as rounded cells with over 90% cell volume taken up by a single fat droplet. The ROIs were selected by pixel for analysis and the circles on the representative images in Fig. 7g are shown for better visualization purposes. They are revised in the current version of the manuscript for better representation of the ROIs.
Comment: Please refer to the MetaSpace program for help with confident assignment of metabolites to MSI data.

Response:
We appreciate the reviewer's suggestion for using the MetaSpace program. We will consider using MetaSpace in a future opportunity as it does offer another method of validation but we believe the current method of using Progenesis QI for the engine is sufficient for the time being.