Proteomic profiling of human omental and subcutaneous adipose tissue in individuals with a broad range of BMI

Figures

• 

  Fig. 1. Clinical study design, experimental scheme, data acquisition and processing workflow. (a) Matched OM and SQ adipose tissues were collected during elective surgery from participants with and without obesity. (b) Histological analysis was performed to estimate the adipocyte size in each adipose tissue depot. (c) Samples were randomized and analyzed in a single batch of 58. Yeast enolase was added as a process control to each adipose protein lysate before digestion. After digestion, Pierce PRTC synthetic peptide mixture was added as a system suitability control. (d) An on-column chromatogram library was built from six gas-phase fractionated LC-MS/MS injections using narrow-window DIA (4 m/z overlapping windows) of a pooled mixture of all individual samples. Single injections of each individual sample were acquired for quantitative analysis using wide-window (12 m/z overlapping windows) DIA. Peptide detection and scoring was done using EncyclopeDIA and extracted ion chromatograms produced and integrated with Skyline. (e) Peptide and protein peak areas were normalized using directLFQ prior to data analysis and visualization using python. (f) The proteomics data is publicly available…

• 

  Fig. 2. Adipose tissue histology suggests that OM and SQ adipocyte size correlates with BMI and that SQ adipocytes are bigger in size than OM. (a) Representative histology images of adipocytes. (b) Weighted linear regression with log adipocyte size plotted against BMI (SQ = 31 and OM = 28). The data weight/dot size was calculated based on the number of squares available for adipocyte counting out of all the squares drawn. (c) Histogram for the size difference between SQ and OM adipocytes with the kernel density estimation. On average, the SQ adipocytes are 1,065 µm2 greater than the OM adipocytes (Wilcoxon signed-rank test: p-value < 0.05). The estimated mean adipocyte size in each individual sample is included in the associated metadata for the adipose tissue biopsies analyzed.

• 

  Fig. 3. Overview of peptide variance in process control (yeast enolase), internal LC-MS/MS quality control (PRTC), and external system suitability control (PRTC). (a) Box and Whisker plot of coefficient of variation of enolase (added as whole protein) and PRTC peptides spiked into all adipose samples. Each point represents an individual peptide. (b) Coefficient of variation of PRTC peptides measured in sample-independent injections interspersed between the sample injections.

• 

  Fig. 4. Peptides detected in individual samples. Box and Whisker plots (left) for each individual sample and kernel density plots (right) show the distribution of (a) raw peptide peak area, (b) log2 transformed peptide peak area, (c) median normalized log2 transformed peptide peak area. Sample groups are highlighted with red (OM) and blue (SQ) in the log2 transformation and median normalization plots. Sample presentation order was organized for clarity and does not reflect sample run order, which was randomized. Figure is generated with level 2 and level 3a data.

• 

  Fig. 5. Box and Whisker plots of peptide and protein detections in different sample groups. (a) The number of peptides identified by EncyclopeDIA in each LC-MS/MS injection. (b) The number of proteins identified by EncyclopeDIA identified per LC-MS/MS injection. Samples are grouped by type: OM, SQ, with obesity or without obesity. Each sample is a member of two groups: either OM or SQ and either with obesity or without obesity. Data is presented at 1% FDR. All values are shown in Supplementary Table S1 (see Supplementary_Table_S1.xlsx).

• 

  Fig. 6. Overview of protein level adipose proteomics data. (a) Principal component analysis (PCA) on the entire dataset. The analysis was performed on directLFQ normalized protein peak areas without feature selection (level 3b data). The shaded area represents 95% confidence ellipses for OM and SQ adipose tissues. (b) Dynamic range plot of ranked, directLFQ normalized protein peak area in OM adipose tissue (level 3b data). (c) Dynamic range plot of ranked, directLFQ normalized protein peak area in SQ adipose tissue (level 3b data). Adipose tissue-specific proteins are labeled in both OM and SQ adipose tissues.

• 

  Fig. 7. Summary data of proteomic differences observed between adipose tissue depots. (a) Table of numbers of proteins that are significantly enriched (adjusted p-value < 0.05 and fold change > 2 or <2) in each depot or show no statistically meaningful difference in normalized protein peak area at 5% FDR. (b) Volcano plot of paired log2 fold changes in protein abundance between OM and SQ adipose tissues for the 28 study participants that have matched adipose tissue collected. A Wilcoxon signed-rank test was used to generate p-values. Statistically significantly different proteins are shown using a cut-off of log2 fold-change > 1 or < −1 and Benjamini-Hochberg adjusted p-value < 0.05. (c) Hierarchical clustering of the differentially expressed proteins in OM and SQ adipose tissue that were statistically significantly different as highlighted in b. A single sample was removed from (c) due to low protein abundance (see Methods). Weight stable (grey) means weight change was < 1 kg while weight change (pink) means ≥ 1 kg weight change was reported within 6 weeks prior to surgery. For visualization, Z scores were capped at −4 and +4 to limit the influence of extreme outliers on the color…

• 

  Fig. 8. Differential expression of proteins in OM and SQ adipose tissues. (a) Box and Whisker plot for the log2 fold-change of the top 10 proteins that are enriched in OM adipose tissue and in SQ adipose tissue between 28 study participants. The proteins were filtered based on the adjusted p-value (lowest 10) and log2 fold-change (<−1 or >1) for significantly different proteins in each depot. Adjusted p values are shown in the bar chart above each protein shown. (b) Box and Whisker plot of all-trans retinoic acid- (atRA-) synthesizing protein peak area in all adipose tissue samples using data independent acquisition-mass spectrometry collected in the current study. (c) Previously published atRA- synthesizing enzymes’ expression in a subset of adipose tissue samples using targeted MRM method46. The subset of samples analyzed in the previous study were from the same clinical study as the current work. The original observations from the 2022 study showing enrichment of the three atRA-synthesizing enzymes in OM adipose tissue were reproduced in the current study.

• 

  Fig. 9. Summary data of proteomic signatures correlated with adipocyte size in OM and SQ adipose tissues. (a) Table of numbers of proteins that statistically correlated with log(adipocyte size) (adjusted p-value < 0.05) in each depot. (b–e) Scatter plots of protein peak area against log(adipocyte size) for PWP1, NBEA, RAD52, and EST1 that showed correlations with log(adipocyte size) in OM but not in SQ adipose tissues. The line and the shaded region represent fitted linear regression and 95% confidence intervals, respectively, for each panel.

• 

  Fig. 10. Summary data of proteomic signatures correlated with the level of adiposity observed in OM and SQ adipose tissues. (a) Table of numbers of proteins that statistically correlated with BMI (adjusted p-value < 0.05) in each depot. (b–e) Scatter plots of protein peak area against BMI for ITSN1, AGFG1, LDHD, and TXTP that showed correlations with BMI in both OM and SQ adipose tissues. The line and the shaded region represent fitted linear regression and 95% confidence intervals, respectively, for each panel.