The Dynamic Platelet Transcriptome in Obesity and Weight Loss

Abstract

Objective:

Adiposity is associated with oxidative stress, inflammation, and glucose intolerance. Previous data suggest that platelet gene expression is associated with key cardiometabolic phenotypes, including body mass index (BMI) but stable in healthy individuals over time. However, modulation of gene expression in platelets in response to metabolic shifts (e.g., weight reduction) is unknown and may be important to defining mechanism.

Approach and Results:

Platelet RNA sequencing and aggregation were performed from 21 individuals with massive weight loss (>45 kg) following bariatric surgery. Based on RNA sequencing data, we measured the expression of 67 genes from isolated platelet RNA using high-throughput quantitative reverse transcription quantitative PCR (RT-qPCR) in 1,864 Framingham Heart study participants.

Many transcripts not previously studied in platelets were differentially expressed with bariatric surgical weight loss, appeared specific to platelets (e.g., not differentially expressed in leukocytes), and were enriched for a nonalcoholic fatty liver disease (NAFLD) pathway. Platelet aggregation studies did not detect alteration in platelet function after significant weight loss. Linear regression models demonstrated several platelet genes modestly associated with cross-sectional cardiometabolic phenotypes, including body mass index. There were no associations between studied transcripts and incident diabetes or cardiovascular endpoints.

Conclusion:

In summary, while there is no change in platelet aggregation function after significant weight loss, the human platelet experiences a dramatic transcriptional shift that implicates pathways potentially relevant to improved cardiometabolic risk post-weight loss (e.g., NAFLD). Further studies are needed to determine the mechanistic importance of these observations.

Graphical Abstract

INTRODUCTION

Obesity is a central determinant of metabolic and inflammatory diseases, including type 2 diabetes, cancer, and cardiometabolic disease¹. Weight loss can result in a reduction of inflammatory stimuli and a decrease in obesity-associated comorbidity^{2, 3}. While most studies of obesity and its complications have focused on phenotypes (e.g., non-alcoholic fatty liver disease, NAFLD) or metabolomic/proteomic dysregulation⁴, previous work from our group suggested that platelet transcripts are cross-sectionally associated with body mass index (BMI)⁵. Indeed, increased platelet function has been variably noted in patients with obesity and has been suggested to contribute to the risk of atherothrombosis⁶. In addition to their role in maintaining hemostasis, platelets regulate complex vascular processes, RNA transfer, and mitochondrial secretion, in aggregate impacting hemostasis, infection, thrombosis, and immunity^{7, 8}. Specifically, platelets may be linked to hepatic inflammation and metabolism^9–12. A recent study of five individuals investigated transcriptional change in platelets as a result of bariatric surgery and demonstrated that obesity is associated with an altered platelet transcriptome, but without definitive mechanistic pathways.¹³ Here, we investigated a series of obese individuals undergoing clinically indicated weight loss (bariatric) surgery to determine the platelet transcriptome before and after significant weight loss to understand if platelet transcriptional patterns are reversible, related to specific pathways, and associated with alterations in platelet function.

METHODS

RT-qPCR data have been made publicly available at dbGaP [Study Accession: phs000325.v1.p5] and can be accessed here (www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000325.v1.p5). Fully analyzed sequencing data is provided in supplementary tables. The summary of the study design is shown in Figure 1. The subsequent description of methods is stratified by the cohort studies (“bariatric surgery” or “Framingham Heart Study”).

Figure 1.

Overview of study. To determine if massive weight loss alters thrombotic functions, platelet aggregation was measured using complementary agonists (thrombin receptor activating peptide and collagen) in samples from 21 individuals with massive weight loss before and after bariatric surgery. To determine the effect of significant weight loss on platelet mRNA expression, transcripts were measured in an unbiased fashion using RNA sequencing. Measurement of significantly altered genes (N=38) was completed in leukocytes derived from the same patients to confirm gene expression changes seen after weight loss are specific to platelets. The selected genes (N=67) that significantly changed after bariatric surgery (including those involved in NAFLD pathway) were measured using platelet RNA from 1,864 participants of a large prospective cohort study (Framingham Offspring Cohort, Exam 8).

Bariatric surgery sample

A total of 121 patients were prospectively recruited for a study of bariatric surgery at Tufts Medical Center as previously described⁴. Subjects were screened at baseline and at approximately 3 and 12 months post-operatively. Subjects lost weight ranging from 0.75-48.5% (median 28%) after one year. This analysis focuses on a subset of 21 patients from the overall bariatric surgery study with the greatest weight loss to amplify differences between pre- and post-operative transcriptional profiles for discovery (>45 kilograms between pre-operative to 12 months post-operative). The prospective study was approved by the Tufts University and the University of Massachusetts Medical School institutional review boards. All participants gave written informed consent.

Sample Collection

Non-fasting peripheral venous blood was drawn into Mononuclear Cell Preparation Tube (CPT; Becton Dickenson) tubes and centrifuged at 1,800 g for 30 minutes at room temperature with no brake. These tubes are pre-prepared Ficoll tubes including a gel barrier which makes peripheral blood mononuclear cells (PBMCs) separation easier compare to classic Ficoll method. After centrifugation Plasma was removed from the surface layer. The layer above the gel contains PBMCs and platelets. This layer is transferred into an Eppendorf tube and centrifuged at 376 g for 15 minutes at room temperature. When centrifugation completed PBMCs were pelleted in the bottom of the tube and platelets were in the supernatant. Then platelets transferred to another tube and pelleted by centrifugation. Pelleted platelets and PBMCs were lysed with 350 μl RLT solution (Qiagen Germantown, MD) and stored at −80°C until RNA isolation.

Platelet Aggregation

Whole blood was drawn into BD Vacutainer Sodium Citrate ACD (Solution A) tube (Fisher Scientific Inc.) and centrifuged at 150 g for 17 minutes. The supernatant containing platelet rich plasma (PRP) was collected. Aggregation of platelets in platelet rich plasma was induced using thrombin-receptor activating peptide (TRAP) and collagen at the concentration of 2.5 μM and 5 μg/ml added to PRP. After 10 minutes aggregation was measured using an aggregometer (Bio/Data Corporation, Horsham, PA).

RNA Isolation and Sequencing

Total RNA from platelets and PBMCs was isolated using RNeasy Mini Kits (Qiagen, Germantown, MD). The concentration of isolated RNA, the integrity and purity of total RNA were assessed using an Agilent 2100 Bioanalyzer system (Agilent, Santa Clara, California). The RNA samples were stored at −80°C until library preparation. Libraries were created using the AmpliSeq Transcriptome Human Gene Expression Kit (Thermofisher, Carlsbad, CA) according to the manufacturer’s protocol, and as previously published¹⁴. Briefly, 10 ng of total platelet RNA (RNA Integrity Numbers greater than 7) was reverse transcribed, and cDNA was amplified for 12 cycles by adding PCR Master Mix, and the AmpliSeq human transcriptome gene expression primer pool (targeting 18,574 protein-coding mRNAs and 2,228 non-coding ncRNAs) (based on UCSC hg19). As described in the protocol, cDNA in samples with lower concentration (~ 0.1-1 ng, 1-10 ng) were amplified for 16 and 14 cycles respectively. The limited number of PCR cycles reduces biased amplification of highly abundant molecules and also reduces dropout of low abundance transcripts, as well as minimizes the confounding effect of undesirable PCR duplicates produced by higher cycle numbers. This facilitates a precise and sensitive linear range of measurement of gene expression over 6 orders of magnitude.

Amplicons were digested with the proprietary FuPa enzyme, then barcoded adapters were ligated onto the target amplicons. The library amplicons were bound to magnetic beads, and residual reaction components were washed off. Libraries were eluted and individually quantitated by an Agilent Bioanalyzer (Agilent, Santa Clara, California). Individual libraries were diluted to a 100 pM concentration, then combined in batches for further processing. Emulsion PCR, templating and PI chip loading was performed with an Ion Chef Instrument (Thermo-Fisher, Carlsbad, CA). Sequencing of platelet RNA was performed on an Ion Proton sequencer (Thermo-Fisher, Carlsbad, CA), with Ion PI Hi-Q sequencing chemistry.

RNA-seq analysis

RNA-seq FASTQ files from different samples were uploaded on the NIH Genboree platform (https://www.genboree.org/site/) to generate read counts^{15, 16}. Patients were categorized in two different groups (“baseline” [pre-operative] and “12 months” [post-operative]). We performed differential expression analysis (paired and unpaired) using DEBrowser¹⁷ with DESeq2¹⁸ (https://debrowser.umassmed.edu/). To determine the functional ontologies associated with the genes that are differentially expressed post-surgery, we performed functional annotation of the mapped genes using DEBrowser; we used only the categories with Benjamini-Hochberg adjusted P < 0.05 for analyses. In paired analysis, transcripts of each study participant were compared at 2 different time points using normalized read counts in DESeq2.

We performed over-enrichment analyses on the differentially expressed genes using gene-sets defined in WikiPathways and the Kyoto Encyclopedia of Genes and Genomes (KEGG). The clusterProfiler, enrichplot, pathview, and org.Hs.eg.db R packages were used to identify the gene-set and pathway enrichments¹⁹. Significantly up-regulated genes (fold change > 1 from pre- to post-surgery) were considered separately to down-regulated genes and enrichments were calculated relative to the background (or universe) of the complete set of 12,061 genes detected in these samples. DEBrowser was also used as a secondary pathway analysis¹⁷ for unpaired analysis.

Framingham Heart Study

A total of 1864 Framingham Heart Study Offspring cohort participants who attended Exam 8 (April 2005-January 2008) in a fasting state and had platelet transcripts measured (with a housekeeper glyceraldehyde-3-phosphate dehydrogenase detectable at <30 Cq PCR cycles) were eligible for this study⁵. Platelets were isolated on site at the Framingham Heart Study and subsequently sent to the University Massachusetts Medical School for transcriptional analysis as described⁵. Clinical covariates and computed tomography-based exposures were measured as previously described^20–22. The study protocol was reviewed and approved by the Boston University and University of Massachusetts Medical School Institutional Review Board. All participants provided written informed consent.

High-throughput RT-qPCR of bariatric surgery and Framingham Heart Study samples

RT-qPCR reactions were performed as described previously⁵. Briefly, isolated RNA samples from platelets (Framingham Heart Study) converted to cDNA by using High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA) and PBMCs (bariatric surgical study) RNA samples were converted to cDNA with SuperScript cDNA Synthesis kit (Thermofisher, Carlsbad, CA). Prior to qPCR, the cDNA samples were pre-amplified using TaqMan Pre-Amp Master Mix (Applied Biosystems, Foster City, CA) and 0.2x TaqMan assays. The pre-amplified cDNA samples were stored at −20°C. Pre-amplified cDNA samples were mixed with TaqMan Universal Master Mix and Sample Loading Reagent and loaded into sample inlets of Dynamic Array 96.96 chips (Fluidigm, San Francisco, CA). TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA) were diluted 1:1 with Assay Loading Reagent (10x concentration) and then pipetted into the assay inlets of the Dynamic Array. The Dynamic Array was placed into the IFC controller to distribute the assays and samples into the reaction wells of the chip through microfluidic delivery. The qPCR reactions were performed in the BioMark Real-Time PCR system using the following protocol: 2 minutes at 50°C, 30 minutes at 70°C, 10 minutes at 25°C, 10 minutes at 96.5°C, and 15 seconds at 96°C and 1 minute at 60°C for 40 cycles.

For fold-change analysis (PBMCs) in the bariatric surgery sample, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a reference gene in the normalization of the RT-qPCR results⁵. RT-qPCR data analysis of PBMCs were performed using the software by Gene Globe Data Analysis Center (https://geneglobe.qiagen.com/us/analyze/). The software uses the ΔΔCT method of relative quantification, and the statistical method is a basic Student’s t-test^{23, 24}. For the FHS, we used GAPDH as a reference gene (model construction details below).

Association of candidate transcripts with cardiometabolic disease risk factors in the Framingham Heart Study

Of the 269 genes found to be statistically significantly differentially expressed pre- to post-surgery, 64 transcripts had been previously studied in the FHS and were analyzed for association with incident cardiovascular disease (CVD), diabetes and cardiometabolic disease. An addition 3 genes that were not different were included due to their previous association with cardiometabolic disease.

We first analyzed the association of each transcript with triglycerides, high-density lipoprotein (HDL) concentration, insulin level, BMI, and waist circumference. Linear models for each trait were built separately for each gene, excluding participants with below-detection limit expression for that gene (to limit leverage in continuous regressions) and adjusted for age, sex, and GAPDH expression. Each dependent variable was log-transformed for analyses. Age was standardized. Regressions for lipid phenotypes (triglyceride and HDL) were further adjusted for the use of lipid-lowering therapies. Regressions for insulin were performed on a subset of the overall sample who did not have diabetes at Exam 8. Analyses for imaging phenotypes (liver fat, subcutaneous fat, etc.) were carried out in a similar fashion (linear models, adjusted for age, sex, GAPDH). A Benjamini-Hochberg (BH) false discovery rate (FDR) adjustment was performed across all regressed RNAs for each cardiometabolic trait separately, with a 5% FDR threshold used to assess statistical significance. R version 4.0.0 was used for these analyses (R project, https://www.r-project.org).

Association of candidate genes with incident diabetes and cardiovascular disease in the Framingham Heart Study

Logistic models were fit on incident diabetes to evaluate association with each gene expression, where diabetes was defined as a plasma glucose at least 200 mg/dL or a fasting plasma glucose of at least 126 mg/dL or having diabetes treatment. Fasting status is defined as fasting for at least 8 hours. Proportional hazards models were used on incident CVD against each gene expression. CVD was defined as the following events: myocardial infarction, atherothrombotic infarction of brain, cerebral embolism, intracerebral hemorrhage, subarachnoid hemorrhage and congestive heart failure. Individuals were included for diabetes regressions if they had a follow-up at Exam 9 with adjudication of this endpoint. The number of FHS participants at risk of CVD and diabetes were 1769 and 1255, respectively. A total number of 214 and 63 participants developed incident CVD and diabetes, respectively. Both the logistic models and the proportional hazards models were adjusted for age, sex, BMI and GAPDH. The analysis on incident diabetes and CVD were run in SAS 9.4 (SAS Institute, Cary, NC), with a chi-square test to assess statistical significance (P<0.05).

RESULTS

Clinical characteristics of bariatric surgical cohort

The characteristics of 21 participants in the bariatric surgery study included in this analysis are shown in Table 1. Participants were primarily young female (median 34 years, 57% female), with a mean BMI 51.4 kg/m² pre-operatively (baseline) and 32.4 kg/m² at approximately 12 months post-operatively. Hypertension was present in 38% of participants at baseline, and diabetes was present in 14%. After bariatric surgery, we observed improvement in lipid profile, blood pressure, and dysglycemia²⁵.

Table 1.

Demographic data Values are reported as mean or number (percentage). P values refer to χ2 comparisons (categorical) or Wilcoxon comparisons (continuous).

Characteristics	N (%)
Sex	n=12 (57% female), n=9 (43% male)
Age; Median (range)	34 (23-67)
Race
White	19 (90%)
Black	2 (10%)
Type of surgery
Sleeve gastrectomy	14 (67%)
Gastric bypass	7 (33%)
Clinical Variable (%)	Before surgery (Baseline)		After surgery (12 Months)		P value
Hypertension (%)	8 (38%)		3 (14%)		0.079
Type 2 diabetes (%)	3 (14%)		1 (5%)		0.293
Hypercholesterolemia (%)	5 (24%)		1 (5%)		0.077
Obstructive sleep apnea (%)	8 (35%)		0 (0%)		0.001
Smoker (%)	0 (0%)		0 (0%)
Alcohol history (%)	0 (0%)		0 (0%)
Clinical Variable	Mean	Standard Deviation	Mean	Standard Deviation
Weight (lbs)	348.9	53.7	220.4	45.3	< 0.001
Body mass index (kg/m2)	51.4	7.1	32.4	5.6	< 0.001
Systolic BP (mm Hg)	128.7	11.6	119.7	17.9	0.026
Diastolic BP (mm Hg)	79.9	8.3	71.8	7.6	0.002
Heart Rate (beats per minute)	81.7	18.0	69.7	9.5	0.016
Hemoglobin A1c	6.1	1.3	5.5	0.9	0.002
Total Cholesterol (mg/100 ml)	184.1	34.0	172.9	34.4	0.049
LDL (mg/100 ml)	103.5	21.5	103.7	26.3	0.666
HDL (mg/100 ml)	44.8	12.6	53.5	13.7	0.001
Triglycerides (mg/100 ml)	127.0	50.4	77.8	37.7	0.006
Platelet Count	264.4	69.4	239.3	60.6	0.026
Glucose (mg/dL)	109.1	37.7	94.2	24.4	0.048

Effect of weight loss on platelet aggregation

To determine if significant weight loss alters thrombotic function, platelet aggregation was measured using complementary agonists. With the use of either thrombin or collagen to induce activation (Figure 2), no significant changes were seen when comparing platelet reactivity pre- versus post-operatively, suggesting that weight loss may not have a direct effect on pathways involved in platelet aggregation/thrombotic function.

Figure 2.

Platelet aggregation was measured in 21 individuals with massive weight loss (over 45 kg) pre- (baseline) and post-bariatric surgery (12 months). Aggregation of platelets in platelet rich plasma (PRP) was induced using thrombin-receptor activating peptide (TRAP) and collagen at the concentration of 2.5 μM and 5 μg/ml. Wilcoxon signed rank test was used to calculate p values. No significant changes were observed in platelet aggregation as a result of massive weight loss. BL: Baseline, 12MO: 12 months after surgery.

Effect of weight loss on platelet gene expression

A total of 269 protein-coding genes were significantly differentially expressed pre- vs. post-operatively in unpaired analysis (Supplementary Tables IA, IB). A total of 110 genes were up-regulated, and 159 genes were down-regulated (Figure 3a). Differentially expressed genes (BH adjusted P < 0.05) were selected for pathway analysis (Figure 4). Enrichments against WikiPathways revealed the non-alcoholic fatty liver disease (NAFLD) pathway as significantly altered (BH adjusted P < 0.05) one year after bariatric surgery in the obese subjects (Supplementary Table IIA, Figure 4). Pathway analysis using DEBrowser also confirmed that NAFLD is the most significant pathway altered with weight loss. A total of 11 genes involved in the NAFLD pathway were altered: EIF2AK3 (Eukaryotic Translation Initiation Factor 2 Alpha Kinase 3) and PIK3R1(Phosphoinositide-3-Kinase Regulatory Subunit 1) were up-regulated and SOCS3 (Suppressor of cytokine signaling 3), RXRA (Retinoid X Receptor Alpha), BAX (Bcl-2-associated X protein), COX (Cyclooxygenase) 6B1, COX8A, NDUFA10 (NADH:Ubiquinone Oxidoreductase Subunit A10), INSR (Insulin Receptor), BID (BH3 Interacting Domain Death Agonist), and MAP3K11(Mitogen-activated protein kinase kinase kinase 11) were down-regulated (Figure 5).

Figure 3.

Differentially expressed genes comparing platelet mRNA in 21 individuals with massive weight loss before (baseline) and after bariatric surgery (12 months). Differential expression analysis of RNA-seq data performed with DESeq2. a. Unpaired analysis: A total of 269 transcripts were significantly altered one year after bariatric surgery (Benjamini-Hochberg adjusted pValue < 0.05; y-axis) of which 159 were down-regulated and 110 were up-regulated.

b. Paired analysis: A total of 416 transcripts were significantly altered one year after bariatric surgery (Benjamini-Hochberg adjusted pValue < 0.05; y-axis) of which 275 were down-regulated and 141 were up-regulated (FC: fold change, NS: Not significant).

Figure 4.

Metabolic pathways significantly altered after surgery and massive weight loss as determined using WikiPathways analysis. Pathways related to nonalcoholic fatty liver disease were found to be most significantly changed when comparing platelet gene expression pre- to post-surgery. Bar size (x-axis) corresponds to the number of differentially expressed genes overlapping each pathway.

Figure 5.

Non-alcoholic fatty liver disease (NAFLD) represents a spectrum of conditions ranging from simple steatosis to more severe steatohepatitis with hepatic inflammation and fibrosis, known as nonalcoholic steatohepatitis (NASH). NASH may further lead to cirrhosis and hepatocellular carcinoma (HCC). Altered genes in platelets are highlighted in blue (down regulated genes) and red (up regulated genes). The orange circle highlights genes that are also expressed in PBMC and their expression follow the same direction in both platelets and PBMCs. SOCS3 (Suppressor of cytokine signaling 3), PI3K (Phosphoinositide-3-Kinase Regulatory Subunit 1), INSR (Insulin Receptor), TNF-α (Tumor necrosis factor alpha), LXR-α (Liver X receptor alpha ), RXRA (Retinoid X Receptor Alpha), FFA (Free Fatty Acids), FAS, (Fas receptor) BAX (Bcl-2-associated X protein), CHOP (C/EBP homologous protein), ATF4 (Activating Transcription Factor 4) , ROS (Reactive oxygen species), AMPK (activated protein kinase), GSK-3 (Glycogen synthase kinase 3), COX8A (Cyclooxygenase 8A), ACDC (adipocyte C1q and collagen domain-containing), SREBP (Sterol regulatory element-binding protein), ER (Endoplasmic reticulum), PPAR-α (Peroxisome proliferator-activated receptor alpha). The upregulated component of PI3K is PIK3R1. The downregulated components of CxI and CxIV are NDUFA10 and COX8A/COX6B1, respectively.

In additional paired analysis (to examine intraparticipant changes in gene expression), we found a total number of 416 genes differentially expressed (Supplementary Table IC, Figure 3b) of which 141 were significantly up-regulated and 275 were down-regulated. Pathway analysis, based on genes significantly altered at individual level pre and post-surgery, also demonstrated NAFLD pathways (Supplementary Table IIB).

Effect of weight loss on PBMC gene expression

To understand if weight loss-associated changes in gene expression are specific to platelets or part of a global change in transcripts, we performed RT-qPCR on PBMCs accrued from the same patients during the same blood draw. We selected the 33 genes significantly differentially expressed in platelets and 5 genes without significant change. Before performing RT-qPCR on the selected genes, we utilized public databases to ensure that genes selected were known to be expressed in PBMCs. In measurements of 38 selected genes, 26 (68%) of the differential gene expression changes observed were found to be platelet-specific (Table 2) and 7 (18%) were noted to be concordant between platelets and PBMCs (Table 3). There were no significant differences in platelet and PBMC expression for TLR4 (Toll-like receptor 4), BCL2 (B-cell lymphoma 2) and CFD (Complement Factor D). The expression of VAMP3 (vesicle-associated membrane protein 3) and STAT4 (Signal Transducer and Activator of Transcription 4) did not significantly change in platelets but was over expressed in PBMCs (Supplementary Table III).

Table 2.

Changes in gene expression in platelets (based on RNA-seq data; natural scale) and PBMCs (based on RT-qPCR data) measured in 21 patients pre- (baseline) and post-bariatric surgery (12 months). The table shows discordant gene expression patterns in platelets and PBMCs after surgery. Differential expression analysis of RNA-seq data performed with DESeq2. RT-qPCR data analysis used the ΔΔCT method of relative quantification, and the statistical method is a basic Student’s t-test.

Gene	Fold Change (PBMCs by RT-qPCR)	P value (PBMCs by RT-qPCR)	Fold Change (Platelets by sequencing)	P value (Platelets by sequencing)
PAFAH1B2	1.27	<0.001	0.35	<0.001
AP2A1	1.1	0.055	0.66	<0.001
TMCC3	1.12	0.051	0.59	<0.001
H2AFY	1.12	0.014	0.73	<0.001
INSR	1.02	0.623	0.65	0.002
MAP3K11	1.14	0.004	0.70	<0.001
BAX	1.17	<0.001	0.73	0.002
COX6B1	1.12	0.002	0.81	0.005
RXRA	1.02	0.244	0.70	0.010
BID	1.08	0.043	0.70	0.022
NDUFA10	1.18	<0.001	0.78	0.029
COX8A	1.19	<0.001	0.76	0.036
ITGB2	1.06	0.147	0.74	0.036
FADS1	1.03	0.399	0.64	0.007
TLR9	1.34	<0.001	0.80	0.025
MRAS	0.82	0.181	0.53	0.003
CDKN1C	0.87	0.787	0.46	0.047
FCN1	0.95	0.908	0.55	0.006
HMOX1	0.94	0.690	0.66	0.032
SERPINA1	0.91	0.3021	0.67	0.026
NCOR2	1.26	<0.001	0.81	0.025
AGPAT2	0.94	0.347	0.70	0.027
PCK2	1.15	<0.001	0.81	0.046
CD63	1.08	0.058	0.80	0.017
PLEC	1.13	0.053	0.64	0.007
CD68	0.98	0.846	0.72	0.018

Table 3.

Changes in gene expression in platelets (based on RNA-seq data) and PBMCs (based on RT-qPCR data) measured in 21 patients pre- (baseline) and post-bariatric surgery (12 months). The table shows concordant gene expression patterns in platelets and PBMCs after surgery (i.e. both upregulated or both downregulated). Differential expression analysis of RNA-seq data performed with DESeq2. RT-qPCR data analysis used the ΔΔCT method

Gene	Fold Change (PBMCs by RT-qPCR)	P values (PBMCs by RT-qPCR)	Fold Change (Platelets by sequencing)	P value (Platelets by sequencing)
XPC	1.27	0.002	1.47	<0.001
EIF2AK3	1.24	0.005	1.26	0.039
SOCS3	0.66	0.012	0.45	0.017
PIK3R1	1.39	<0.001	1.29	0.002
PCCB	1.18	<0.001	1.24	0.040
ALMS1	1.28	<0.001	1.53	0.026
MTERFD3	1.34	<0.001	1.40	0.009

Association of candidate genes with cardiometabolic traits

To determine the association of platelet genes affected by weight loss with prevalent cardiometabolic traits and incident diabetes and CVD, we measured platelet gene expression in the FHS Study Offspring Cohort (Supplementary Table V). SOCS3, ICAM4 (Intercellular Adhesion Molecule 4), AP2A1 (Adaptor Related Protein Complex 2 Subunit Alpha 1), and TOP2A (DNA topoisomerase 2-alpha) were associated with BMI, and SOCS3 was associated with waist circumference and insulin after FDR (Supplementary Table VI; all FDR < 0.05). We did not observe a significant association (at an FDR < 0.05) between platelet transcripts and visceral fat, subcutaneous fat, liver attenuation and liver/phantom ratio (Supplementary Table VI), incident CVD or incident diabetes (Supplementary Table VII; includes cerebral hemorrhage and subarachnoid bleeding).

DISCUSSION

In this study, we extend previous work from our group exploring role of the platelet in obesity by studying functional and transcriptional shifts in platelets with weight loss. Our primary results indicated, in an unbiased manner, the identification of 269 significantly altered transcripts after massive weight loss in 21 subjects. The paired analysis comparing transcript of each individual after massive weight loss resulted in identification of 416 altered transcripts. Although we did not observe changes in platelet aggregation function after significant weight loss, a dramatic transcriptional shift that implicates pathways relevant to improved cardiometabolic risk post-weight loss (e.g., NAFLD). In a large, community-based sample, we found modest association of platelet expression of several genes (including SOCS3) and relevant cross-sectional cardiometabolic traits (e.g., BMI). Collectively, these findings suggest a potential role for platelet transcriptional shifts with cardiometabolic disease.

Previous studies have shown that significantly increased weight may alter pathways contributing to thrombotic events potentially enhancing increased risk of heart diseases^{26, 27}. Another study identified 170 differentially expressed genes before and post-surgery in 5 subjects and demonstrated obesity is associated with an altered platelet transcriptome and increased platelet activation¹³. Despite platelets canonical role in hemostasis and thrombosis, in our study of patients with massive weight loss, we did not observe changes in platelet aggregation (Figure 2), in agreement with prior rodent data suggesting that platelet aggregation may not be important in obesity-induced liver disease pathogenesis²⁸. Our previous study investigating protein expression in plasma of bariatric surgery patients showed that P-selectin expression in plasma, demonstrating platelet shedding of this surface receptor, was significantly decreased after the bariatric surgery,⁴ suggesting that platelet inflammatory function is decreased with weight loss. However, platelet transcriptome analysis showed no significant change in the expression of SELP (P-selectin coding gene) in platelets as a result of interventional weight loss over one year. Interestingly, as stated above, platelet gene expression changes did not include thrombotic transcripts, but notably included genes related to nonalcoholic fatty liver disease (NAFLD), a key factor in obesity-induced cardiometabolic stress. However, paired analysis has shown the significant alterations in transcripts involved in tuberculosis and in NAFLD, we focused on NAFLD as it is well known that obesity is linked to an increased risk of NAFLD²⁹.

Nonalcoholic steatohepatitis (NASH) is the consequence of insulin resistance, fatty acid accumulation, oxidative stress, and lipotoxicity³⁰. High fat induced mouse models have shown a reverse in hepatic transcription and activity due to subsequent weight loss³¹. In our study, we observed changes in expression of platelet genes implicated in NAFLD, including down-regulation of SOCS3, BID, RXRA, NDUFA10, BAX, COX6B1, INSR, MAP3K11, COX8A, and up-regulation of EIF2AK3 and PIK3R1. As seen in Figure 5, these specific NAFLD genes are altered in insulin resistance, leading to a defect in insulin suppression of free fatty acid (FAA) disposal, a key first step in the progression of NAFLD to non-alcoholic steatohepatitis (NASH), attendant oxidative and ER stress, lipid peroxidation, and irreversible hepatic injury. Most of the genes shown in Table 2 are also differentially expressed in PBMCs but primarily in the opposite direction while, in Table 3, concordant gene expression alterations are also seen. These complex observations suggest that changes in platelet gene expression after weight loss are not simply an epiphenomenon of global reduction in inflammation and oxidative stress with weight loss as most appear to be platelet specific. The mechanistic reasons for the divergent changes in gene expression will require future investigation.

Inflammatory transcripts in platelets previously associated with obesity¹⁴ were not found to be significantly altered in platelets after significant weight loss due to bariatric surgery. However, several genes well-known to be connected to obesity are linked to alterations in gene expression in our cohorts. For instance, IL-6 may contribute to insulin resistance in obesity via activation of SOCS3, which inhibits the insulin signaling pathway³². In our study, platelet SOCS3 expression was decreased after bariatric surgery, and a higher SOCS3 PCR cycle value (lower SOCS3 expression) was associated with lower insulin, BMI, and waist circumference in FHS, consistent with a deleterious role for SOCS3 in obesity. Coupled with our previous findings of platelet IL-6 transcript association with BMI⁵, the current observations suggest the possibility that that IL-6 signaling via SOCS3 pathway in platelets may mediate the inflammatory response in obesity resulting in the progression of NAFLD. Mechanistic studies are needed to investigate this potential pathway.

Platelet transcriptome have shown to be stable in healthy individuals up to four years³³. It is speculated that due to anucleate nature and short lifespan, moderate RNA changes in platelets promotes a stable and defined healthy gene expression signature. A limitation of our study is the lack of a weight-stable control group. To examine this closer, we specifically compared the expression of platelet transcripts between our study and that of Rondina et al.³³. Using their data from a cohort with similar age and race, we noted that none of the 47 genes detected as different in our overlapping group from the bariatric and FHS cohorts were significantly different over time in this weight stable cohort.

Our findings raise the question as to whether platelets directly impact the development or resolution of NAFLD independent of thrombotic function. Based on our data, as well as previous observations⁷, changes seen in platelet gene expression profiles after massive weight loss could influence hepatic inflammation via direct transfer of transcriptomic information to cells. A recent study on potential interventional targets for NASH and subsequent liver cancer demonstrated the interaction of platelets with Kupffer cells, the importance of platelet cargo, adhesion, and activation; but not platelet aggregation in relation to the development of NASH and subsequent hepatocarcinogenesis²⁸. Additionally, pathophysiological changes due to obesity may impact the cellular microenvironment of megakaryopoiesis resulting in the release of specific platelet transcriptomic profiles. These changes could possibly influence hepatic function and health.

Our study has several strengths and limitations. Our bariatric surgical cohort is of modest size (21 individuals), minimal ethnic/racial diversity, and age (median age 34 years old). Nevertheless, this was primarily a discovery sample for further downstream functional target identification (e.g. NAFLD). Future, targeted studies across race and sex would be important. In addition, sampling serially early after bariatric surgery to see whether these platelet changes are weight-dependent or weight-independent (and more metabolic in nature) is of ongoing interest. The cohort was a referral cohort with significant weight loss; whether similar changes are seen in the range of weight reduction (and gain) seen in medical weight loss is unclear. In addition, whether there is a “memory effect” of platelet gene expression (epigenetic control over obesity-related platelet abnormalities) remains important. Since the surgery group did not have any known underlying liver diseases, formal testing of liver fibrosis and steatosis (e.g., fibroscan or MRI) had not been performed, which is a non-invasive standard for liver disease. In order to limit leverage bias in regression, we restricted our analysis for traits in FHS to participants with detectable expression level for each gene of interest, which may bias against discovery of expression-phenotype association for those genes in which low expression may be important. In addition, we chose GAPDH as a housekeeper gene; the choice of alternate housekeeping genes (with a different range in expression) may limit pre-analytical or technical bias. In the obesity cohort, only RNAseq data was available for study due to limitations in platelet RNA quantity. In FHS, we studied cross-sectional phenotypes in a primarily Caucasian cohort, and associations with several cardiometabolic traits and outcomes did not withstand type 1 error correction. Future studies with serial changes in gene expression over time (especially in multi-ethnic cohorts) may be further relevant to refocusing efforts on those transcripts that may be most relevant.

In summary, here, we demonstrate that the platelet transcriptome undergoes a dramatic shift after bariatric surgical weight loss, implicating pathways relevant to liver disease and obesity. Several platelet transcripts uncovered in discovery efforts appeared to be modestly associated with cardiometabolic traits in a large cohort study. Further studies are needed to understand whether these platelet changes are functional in the pathogenesis of NAFLD during obesity and reversible.

HIGHLIGHTS.

• Our primary results indicated, in an unbiased manner, the identification of significantly altered transcripts in platelets after massive weight loss.

• We demonstrate that changes observed in platelet gene expression after surgical weight loss, are platelet specific and implicate pathways relevant to improved cardiometabolic risk post-weight loss (e.g., NAFLD).

• Several platelet transcripts uncovered in discovery efforts appeared to be associated with cardiometabolic disease in a large cohort study.

ABBREVIATION

BMI

Body mass index

NAFLD

Nonalcoholic fatty liver disease

RT-qPCR

reverse transcription quantitative PCR

CPD

Citrate phosphate dextrose

PBMC

Peripheral blood mononuclear cells

PRP

Platelet rich plasma

TRAP

Thrombin-receptor activating peptide

mRNA

messenger RNA

cDNA

complementary DNA

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

FC

Fold change

FHS

Framingham heart study

CVD

Cardiovascular disease

HDL

High-density lipoproteins

FDR

False discovery rate

BH

Benjamini-Houchberg