Macrophage Phenotypes and Gene Expression Patterns Are Unique in Naturally Occurring Metabolically Healthy Obesity

Round 1

Reviewer 1 Report

The authors expose a meta-analysis for showing the epicardial fat reduction with pharmacological treatment, foccused their attention on glucose and lipid-lowering drugs. A flow chart based on PRISMA was defined and a few studies were included in the analysis. They considered variability inter publications, etc...Although the meta-analysis is quite interesting, the authors should include the clinical characteristics of patients according different treatments. Moreover, they should discuss more the druges effects on epicardial fat metabolism, inflammation on different groups of patients. The novelty is scarce since there is a more complete meta-analysis already published by Launbo N, Zobel EH, von Scholten BJ, Faerch K, Jørgensen PG, Christensen RH. Targeting epicardial adipose tissue with exercise, diet, bariatric surgery or pharmaceutical interventions: A systematic review and meta-analysis. Obes Rev. 2021 Jan;22(1):e13136.

Author Response

Reviewer 1

Major points:

1. Are the gene expression data deposited in a publicly available database?

Data will be made available upon submission of the revised manuscript and prior to publication.

2. The authors acknowledge that some of the transcriptomic changes may be related to the underlying pathologic changes or cellular composition, but this should be explored more completely in the discussion. Are any of the transcripts enriched in specific cellular types based on available literature?

Inflammatory transcripts that are enriched in the MUO SQ AT are related to pro-inflammatory macrophages and general adipocyte and whole tissue dysfunction (PMID: 23824685). Others are not related to specific cell types.

3. Is there a difference between Abdominal VIS volume between MHO and MUO?

There was no difference in the abdominal VIS volume between the MHO and the MUO groups. Like and unlike letters in Figure 1 denote the statistical outcomes.

4. While it is acceptable not to show the gene names in the heatmaps due to space constraints, gene details included in the heatmaps should be given in Supplementary Tables.

The heatmaps are included to provide visual representation of the transcript expression differences between health groups. The genes in the tables provided in the Supplement are included in the heatmaps and their statistical scrutiny is included in the tables. The complete gene lists will be published publicly upon submission of the revised manuscript.

5. This sentence in Methods is convoluted and needs to be rewritten with more clarity—" The animals’ healthy environment led to 36% of screened African green monkeys demonstrating zero MetS risk factors [15], which is threefold higher than the prevalence of American adults without MetS risk”

Thank you for this note. We have revised this sentence.

1. 1.       Table 1 legend: “increased” repeated twice.

We have revised this.

2. The upper case and lower case used in the Figure legends is inconsistent (for example, see , Fig 3 legend)

We have revised these errors.

3. Please check the following references as the formatting is incorrect: 1, 19, 51

We have revised these.

4. Indicate what the cross bars show in Fig 1e.

The cross bars indicate the mean values. This information has been added to the legend.

Reviewer 2 Report

This is a very interesting and rigorous study that uses a non-human primate model to determine the various obesity-related phenotypes [Obesity (MUO, MHO) and Lean (MUL, MHL)]. The authors used a cohort of 44 AGMs to categorize into these 4 groups and assessed adipocyte size, macrophage subtype distribution, transcriptome and vascularity, and fibrosis. They found that the adipose tissue from metabolically healthy obese (MHO) had higher levels of the anti-inflammatory M2 macrophages and upregulation of fatty acid oxidation-related terms and transcripts. Notably, these phenotypes were developed in absence of an unhealthy diet. However, some additional clarifications are needed.

Major points:

1.       Are the gene expression data deposited in a publicly available database?

2.       The authors acknowledge that some of the transcriptomic changes may be related to the underlying pathologic changes or cellular composition, but this should be explored more completely in the discussion. Are any of the transcripts enriched in specific cellular types based on available literature?

3.       Is there a difference between Abdominal VIS volume between MHO and MUO?

4.       While it is acceptable not to show the gene names in the heatmaps due to space constraints, gene details included in the heatmaps should be given in Supplementary Tables.

5.       This sentence in Methods is convoluted and needs to be rewritten with more clarity—" The animals’ healthy environment led to 36% of screened African green monkeys demonstrating zero MetS risk factors [15], which is threefold higher than the prevalence of American adults without MetS risk”

Minor points:

1.       Table 1 legend: “increased” repeated twice.

2.       The upper case and lower case used in the Figure legends is inconsistent (for example, see , Fig 3 legend)

3.       Please check the following references as the formatting is incorrect: 1, 19, 51

4.       Define AU where appropriate.

5.       Indicate what the cross bars show in Fig 1e.

6.       Fig 2b Use more contrasting colors for M1 and intermediate cells

Author Response

Reviewer 2

Although the meta-analysis is quite interesting, the authors should include the clinical characteristics of patients according different treatments. Moreover, they should discuss more the drugs effects on epicardial fat metabolism, inflammation on different groups of patients.

In this work, the animals were placed into metabolic health groups according to nonhuman primate-specific metabolic syndrome criteria. No drugs were administered during this study. We did not see differences in total body fat between groups, which includes epicardial adipose, though we were not able to sample this adipose depot and plan to in future work. We were able to evaluate immune cell difference, specifically macrophage differences, in the subcutaneous adipose of our four health groups. We also evaluated circulating and local cytokines as markers of inflammation. We identify the M2/M1 macrophage ratio and upregulated lipid handling transcripts as markers of healthy obesity.

Reviewer 3 Report

The comments are in attachment.

Comments for author File: Comments.pdf

Author Response

Reviewer 3

Major comments

1. In the abstract and in several places in the text there is a claim that M2 macrophages are increased (or more accumulated) in MHO animals. However, this is not accurate, according to the results M2 macrophages are not increased as such but rather the M2/M1 ratio.

We thank the reviewers for this point of clarification. We have revised the text so that it more accurately states that the differences observed between the metabolically healthy obese (MHO) and the metabolically unhealthy obese (MUO) subcutaneous (SQ) adipose tissue (AT) macrophages was between the M2/M1 ratio (lines 127-130). Importantly, this ratio correlates with insulin resistance (PMID: 33479582, PMID: 30382157), and this detail has been included in the text (line 276-277).

2. Visceral fat volume appears to be higher in animals with MHO compared to MUO. In addition, the VAT/SAT ratio is higher in lean animals compared to obese animals. This is a big difference from the phenotypes observed in humans. Given the close association of the amount of VAT to the development of metabolic syndrome in humans (e.g., DOI:10.1007/s00125-012-2639-5, DOI: 10.4082/kjfm.18.0122), this limitation of the model should be discussed.

Thank you for highlighting this point. The lean animals demonstrated comparable and significantly less visceral (VIS) and SQ AT volumes compared to the obese (MHO and MUO) animals. Our ratio is lower in obesity as adipose expands more in the SQ spaces than intra-abdominal depots, which is in part due to the healthy low-fat, low-sugar, high-fiber, and high-protein chow diet consumed. Additionally, in the African green monkey, gluteal and appendicular stores of fat are much lower than they are in human primates. The expansion of SQ adipose with obesity, coupled with healthy and unhealthy consequences, is the premise for evaluating its’ properties described in this manuscript. Previous work has shown that human females without cardiovascular disease have an abdominal VIS/SQ ratio under 1 (PMID: 17576866). We are able to monitor the progression of metabolic syndrome abnormalities in our animals, and to date, the progression of metabolic syndrome abnormalities in this animal model mirrors what is observed people. These details have been added to the Discussion section (lines 289-296).

3. Did the authors find any correlation between phenotype (VAT, FM, IS, lipid levels) and M2/M1 ratio in AT or with the transcriptome? This would be interesting.

The SQ AT M2/M1 ratio did not correlate with total adipose accumulation, adipose tissue distribution, circulating insulin concentrations, or any metabolic syndrome risk factors. However, the SQ AT M2/M1 ratio did positively correlate with expression levels of ATP binding cassette subfamily C member 2 (ABCC2) (r=0.315, p=0.0448), and negatively correlated with expression levels of cyclin dependent kinase inhibitor 1A (CDKN1A) (r=-0.419, p=0.00646), immunoglobulin heavy constant delta (IGHD) (r=-0.398, p=0.00988), and heat shock transcription factor 4 (HSF4) (r=-0.436, p=0.00436). These associations indicate that increases in the M2/M1 ratio correspond with decreased antigen binding, cellular senescence, and stress response, and improved movement of molecules across membranes. These correlations support the reported differentially expressed transcripts and gene ontology terms found between health groups, and have been added to the text (lines 205-213). A complete list of statistically significant associations between the SQ AT M2/M1 ratio and identified transcript expression levels have been added to the supplement (now Supplemental Table 4).

4. Missing is a discussion of transcriptomic data in the Discussion section (e.g. in the context of published human or mice data found in obesity/MetS or in the context of the association of inflammation/macrophages and AT metabolism).

A previous assessment of obesity-discordant monozygotic twin pairs concluded that metabolically healthy obesity is associated with upregulated mitochondrial oxidative phosphorylation and fatty acid oxidation transcript expression, decreased levels of markers of chronic inflammation, and reduced liver adipose accumulation (PMID: 24100782). Additionally, the MHO phenotype has been shown to demonstrate expression of solute carrier family members in circulation or saliva (PMID: 32788640; PMID: 30120429). Our work supports these findings and adds vital information regarding local immune cell differences between metabolic health groups. Work is currently being done to target mitochondrial fatty acid oxidation to improve health with obesity (PMID: 33397994). This information has been added to the text (lines 265-277). 

5. The authors write that “our tissue cytokines tracked with our observed macrophage

phenotypic shifts.” Does this mean that the authors found some correlations between

cytokines and macrophages?

A number of macrophage measures associated with cytokines. The SQ AT M2/M1 ratio negatively correlated with local SQ AT concentrations of TGF-β (r=-0.351, p=0.045), and circulating concentrations of PAI-1 negatively correlated with the percentages of M2 macrophages (r=-0.29, p=0.06). Additionally, M2 macrophages were negatively associated with circulating concentrations of MCP-1. In contrast, M1 percentages in SQ AT positively correlated with circulatory PAI-1 (r=0.27, p=0.09). These data are presented in the text (lines 129-131; 208-223)

Minor comments

6. The number of animals in each group should be listed in the heading of the tables

The number of animals in each group has been added to the table headings.

7. The description of the results should not be in the legend of tables and results but in the main text.

Descriptions of the results have been moved from the table legends to the main text.

8. In the legend of the figure 1 it should be defined what significance the individual letters represent and what the comparison refers to.

This information has been added to the first line of the figure legend.

9. The graphic layout of the figures should be modified - reduce the size of the graphs, arrange

a), b) next to each other, etc.

The main figure layouts have been modified.

10. In figure 2b) (pie graphs) it would be good to show the percentages of individual macrophage subtypes numerically

The percentages of each macrophage subtype by group are numerically presented in Supplementary Table 2.

11. The tables should have smaller fonts

All table font sizes have been reduced.

12. In my opinion, it would be better to show all genes that differ between MHO and MUO.

The complete list of differentially expressed transcripts has been added to the supplement (Supplemental Table 3).

13. In Supplemental Fig 7, it would be useful to show the figures of staining for all 4 groups.

Representative staining images from each health group have been added to Supplemental Figure 7.

14. I would rather assume a discussion of macrophage polarization in relation to AT metabolism, or in relation to metabolic syndrome or obesity in human studies.

We have added additional statements to the Discussion that focus on treatments that may drive M2 AT macrophage polarization or increase AT fatty acid oxidation (lines 298-312).

15. Regarding the results and discussion of fibrosis, it might be useful to focus on the likely remodeling - collagen and ECM composition in regarding composition in relation to obesity and adipocyte size.

We have added additional statements regarding extracellular matrix (ECM) transcript expression between health groups. We observed upregulation of COL4A3 and downregulation of MMP7 in the MUO SQ AT compared to the MHL. These changes in ECM transcripts were not coupled with increases in adipocyte size, as adipocyte sizes were comparable between obese groups. No other differences in the expression levels of collagens, matrix metalloproteinases, laminins, or fibronectins were observed between groups. Additionally, we did not observe differences in the expression levels of HIF1α. This information has been added to the text (lines 254-257).

Round 2

Reviewer 1 Report

The authors have described the role of macrophages phenotype M2 in a obese healthy phenotype in a non human primate model. They classified the animals according obesity and healthy and unhealthy metabolism. The authors have performed analysis of adipose tissue (visceral and subcutaneous) by immunohystochemistry, RNA-seq. Fat and lean mass was also determined by CT image. The field has special interest in the current society. However, the phenotypes characterization of macrophages on adipose tissue is not a novel issue. There are several points that should be addressed: 

a) How the authors explain the high intermediate macrophages phenotype in the healthy lean group. This is one of the most inflammatory phenotype. In fact, there is a reduction of M2. 

Perhaps, there is a missinterpretation. How many slides of each tissue was analyzed?

b) Why the Visceral/SQ ratios is lower in obese than in lean? The authors should discuss more this result.

c) Other question is regarding IL-10 SQ adipose tissue. Why the lower levels are found in healthy obese.

d) Have the authors performed RNA-seq in samples from all included animales? It has to be more clear into text.

e) Dot plots should be more accurate.

f) A brief flowchart regarding methods could help to understand the analysis performed on adipose tissue, blood, immunohistochemistry, RNA-seq, ELISA, etc...

Author Response

a) How do the authors explain the high intermediate macrophages phenotype in the healthy lean group? This is one of the most inflammatory phenotype. How many slides of each tissue was analyzed?

Thank you for this question. Macrophage immunohistochemical analyses were performed on a single slide, though the entirety of each adipose tissue section was quantitated. Adipose tissue sections were large and averaged 0.5 cm² (exactly 48993121 μm²). The metabolically healthy lean (MHL) and the metabolically unhealthy lean (MUL) animals demonstrated similar percentages of M2 macrophages. The lean animals demonstrated higher percentages of undefined macrophages in their subcutaneous adipose tissue (SQ AT) than the obese animals. This indicates that, in the absence of obesity and poor diet consumption, fewer inflammatory signals result in less overall macrophage polarization in lean adipose tissue.

The intermediate macrophage populations comprise a small percentage of the total macrophages quantified (<5%) and was not different between animals classified by any criteria. The intermediate macrophage subtype is poorly defined and could indicate a transition either from the M2 polarization state to the M1 polarization state or the inverse. Accordingly, it was not possible to determine the directionality of the macrophage polarization transition. The metabolically healthy lean subcutaneous adipose tissue samples did not demonstrate increases in pro-inflammatory cytokine protein expression or increases in inflammation-related transcript expression so we do not think that the low percent, but numerically higher intermediate macrophages, are contributing to local inflammation. We have added these details to the Discussion section (Lines 304-309).

b) Why are the Visceral/SQ ratios is lower in obese than in lean?

Thank you for this inquiry. The obese animals demonstrated primarily subcutaneous adipose tissue expansion with obesity onset. This type of expansion corresponds with the healthy environment and diet that the animals consumed, as Western dietary components are known to shift adipose tissue to ectopic depots (Kavanagh et al. 2007, PMID: 17636085; Kavanagh et al. 2013, PMID: 23783298). These details have been added to the text (lines 296-300).

c) Why are lower levels of SQ IL-10 found in healthy obese?

IL-10 is an anti-inflammatory cytokine that plays a role in limiting the host immune response (Lyer and Cheng 2012; PMID: 22428854). We interpret these lower levels as being consistent with metabolically healthy obese (MHO) SQ AT having lower inflammation-related transcripts (while the (metabolically unhealthy obese) MUO SQ AT displayed higher expression of these transcripts). We also feel this is consistent with the unique finding was that MHO SQ AT increases M2 macrophage polarization, whereas MUO SQ AT does not. This has been added to the Discussion section (Lines 358-362).

Obesity results in a necessary adipose tissue hypertrophy and hyperplasia and an immune response to maintain tissue homeostasis. Our SQ AT macrophage data illustrate that both the MHO and the MUO groups have increased immune responses, though these responses are dichotomous. Our IL-10 measures are from whole tissue protein extracts, such that obese animals may have more of the adipocyte fraction (non-IL-10 secreting) than the lean animals, where the protein fraction deriving from the stromal vascular fraction and immune cells may be higher, leading to numerically higher concentration in the lean state. In contrast, MCP-1 can be derived from adipocytes as well as other cell types and, thus, we see a much clearer shift from lean to obese, and healthy to unhealthy.

d) Have the authors performed RNA-seq in samples from all included animals?

RNA-seq was performed on subcutaneous adipose tissue samples from 43 of the 44 animals. We have added this detail to the methods (Line 481).

e) Dot plots should be more accurate.

All continuous endpoints met assumptions of normality with or without transformation. No data points were identified as outliers, or were three or more standard deviations from the mean. As no data points were outlying, we believe that the data are presented with transparently portrayed variation. If the reviewers and the editorial team would prefer that we alter the graphs, we will be more than happy to do so.

f) A brief flowchart regarding methods could help to understand the analysis performed on adipose tissue, blood, immunohistochemistry, RNA-seq, ELISA, etc.

A figure depicting the methods performed on each biological sample type has been added to the Supplement (Supplementary Figure 8; lines 436-437 and 454-455).

Author Response File: Author Response.docx

Reviewer 3 Report

Comment to several previous Minor comments

7. The description of the results should not be in the legend of tables and results but in the main text.

Descriptions of the results have been moved from the table legends to the main text.

REW2: Some of the results have been added to the main text, but the description of the differences between the groups, i.e. the "results", are still present in the legend of the table1. This is unnecessary. In addition, the p-values that are in the table do not need to be restated in the text. The same applies to the figure legends (I apologize for the mistake in the text in the last version... I meant “in the legend of tables and figures”).

8. In the legend of the figure 1 it should be defined what significance the individual letters represent and what the comparison refers to.

This information has been added to the first line of the figure legend.

REW2: The information about statistics in figure legends is still superficial.

In Fig1: The statement ..” and unlike letters denote post-hoc testing for group differences that reached statistical significance (p<0.05).” is not clear.

Authors should describe which analysis was used (e.g. Two-way ANOVA with Bonferroni post-hoc analysis), what statistical value the letters express (e.g. a p<0.05, b: p<0.01, c: p<0.001 ?) and/or what comparison they describe (like the post-hoc comparison of the groups to MHL? Or what does it mean?) This statement should be at the end of the figure legend.

In Fig 4 the correlation coefficient and p-values should be written within the graphs. The type of corel. coefficient should be listed (Pearson/ Spearman) in the legend.

10. In figure 2b) (pie graphs) it would be good to show the percentages of individual macrophage subtypes numerically

The percentages of each macrophage subtype by group are numerically presented in Supplementary Table 2.

REW2: I understand, but then the figure 2 is only illustrative and not so informative. The inclusion of numbers would greatly increase the telling value of the figure. The display of populations in Supplementary Table 2 is OK, but then you dont´need the figure. From my point of view the Figure is better.. I also understand that you probably want to show statistics, however you state in the text that the populations do not differ between groups, so you can truncate or omit the table.

Author Response

Some of the results have been added to the main text, but the description of the differences between the groups, i.e. the "results", are still present in the legend of the table1. In addition, the p-values that are in the table do not need to be restated in the text. The same applies to the figure legends.

The Table 1 legend no longer includes results, and p-values presented in Table 1 have been removed from the text. Results have been removed from the figure legends.

In Figure 1, authors should describe which analysis was used (e.g. Two-way ANOVA with Bonferroni post-hoc analysis), what statistical value the letters express (e.g. a p<0.05, b: p<0.01, c: p<0.001 ?) and/or what comparison they describe (like the post-hoc comparison of the groups to MHL? Or what does it mean?) This statement should be at the end of the figure legend.

We have clarified the statistics used for the endpoints presented in Figure 1 (lines 708-712).

In Fig 4 the correlation coefficient and p-values should be written within the graphs. The type of correl. coefficient should be listed (Pearson/ Spearman) in the legend.

We have made these changes.

The inclusion of numbers would greatly increase the telling value of figure 2b.

We have altered Figure 2b so that it is easier to compare macrophage subtype percentages between groups. Figure 2b now depicted stacked bars that demonstrate the percentage that each macrophage subtype makes up. We have left the numerical descriptions of each macrophage subtype in the Supplement.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

The authors have addressed all the reviewer's suggestions