Obesity, adipokines, and C-peptide are associated with distinct plasma phospholipid profiles in adult males, an untargeted lipidomic approach

Obesity is associated with dysregulated lipid metabolism and adipokine secretion. Our group has previously reported obesity and adipokines are associated with % total fatty acid (FA) differences in plasma phospholipids. The objective of our current study was to identify in which complex lipid species (i.e., phosphatidylcholine, sphingolipids, etc) these FA differences occur. Plasma lipidomic profiling (n = 126, >95% Caucasian, 48–65 years) was performed using chromatographic separation and high resolution tandem mass spectrometry. The responses used in the statistical analyses were body mass index (BMI), waist circumference (WC), serum adipokines, cytokines, and a glycemic marker. High-dimensional statistical analyses were performed, all models were adjusted for age and smoking, and p-values were adjusted for false discovery. In Bayesian models, the lipidomic profiles (over 1,700 lipids) accounted for >60% of the inter-individual variation of BMI, WC, and leptin in our population. Across statistical analyses, we report 51 individual plasma lipids were significantly associated with obesity. Obesity was inversely associated lysophospholipids and ether linked phosphatidylcholines. In addition, we identify several unreported lipids associated with obesity that are not present in lipid databases. Taken together, these results provide new insights into the underlying biology associated with obesity and reveal new potential pathways for therapeutic targeting.


UPLC Parameters
Three MS functions were used to obtain MS/MS spectra and correct for mass drift: Function1 was used to obtain a parent ion spectra of lipids; Function 2 was used to obtain a fragmentation spectra of parent lipid ions; and Function 3 was used to measure leucine enkephalin as lock mass for mass correction.

Time of Injection Effect
After initial data processing of mass defect filtered dataset, we investigated batch and time of injection effects in the data. The variable time of injection was created by multiplying the samples file number by 15 mins, thus, yielding the samples time of injection relative to the to the first injection of the analysis. We observed a time of injection effect after plotting principal components (Pc) 1 and 2. There were distinct clusters of samples based on their time of injection (mins) during the mass spectrometric analysis. In addition, the Pc 1 and 2 score were significantly associated with time of injection in regression analyses.
Principal components derived from patient (n=126) lipidomic profiles. Number listed next to samples indicate the samples time of injection in mins relative to the first injection of analysis. Samples injected earlier in the analysis clustered tightly together, while samples injected later in the analysis drifted.
Next, to further examine the time of injection effect, we plotted the range of intensities for each plasma lipid in the 126 samples.
Figure displays time of injection and the range of plasma lipid intensities for each sample (n=126) analyzed. As time of injection increased, so did the range of metabolite intensities, indicating ion suppression throughout analysis. We determined the time of injection effect was due to ion suppression by trifluoroacetate contamination in the mass analyzer. Since a majority of plasma phospholipids are PCs, we normalized our entire data matrix by the IS PC(8:0/8:0). IS normalization of the data matrix removed the time of injection effect on plasma lipids.
Principal components derived from patient (n=126) IS normalized lipidomic profiles. Number listed next to samples indicate the samples time of injection in minutes. Normalization of the data matrix with IS PC(8:0/8:0) removed the time of injection effect. Next, to further examine if normalizing the data matrix removed the time of injection effect, we plotted the range of intensities for each plasma lipid in the 126 samples Above figure displays time of injection and the range of plasma lipid intensities for each sample (n=126) analyzed. After IS normalization of the data matrix, the time of injection effect was removed. Supplementary and LPE(18:2) were significantly correlated with the % of LA (-6, C18:2 9Z,12Z ), whereas LPC(18:2) and LPE(18:2) were not correlated with the % of trans-isomer linoelaidic acid (-6, C18:2 9E,12E ). Numeric values represent the Spearman correlation coefficients and are bolded if p<0.05. The plasma analyzed from both data sets were collected at the same time point from the same patients. a The significant PL associated with the responses that were structurally characterized in our study. The experimental methodology employed was the UPLC-ESI-MS E analysis of crude lipid extracts outlined in this manuscript b The geometric and positional FA isomers that were targeted in our previous study (Pickens et al. PLEFA. 2015.). The experimental methodology employed was the FAME analysis of isolated plasma PL by GC-FI # For more information on using BGLR refer to https://github.com/gdlc/BGLR-R ### Creating a directory to store file outputs from BGLR ### dir.create("BMI"); setwd("BMI") getwd() # Verify working directory ### Setting response, lipidomic data, iterations, and burn in ### X=pareto.IS # 1,745 pareto scaled lipidomic data y<-H$BMI # Response: body mass index values for each respective patient nIter=200000 # Long Markov Chain of 200,000 iterations burnIn=50000 # Number of iterations to discard for burn in ## Computing the metabolomic similarity matrix ## L<-sum(apply(X=X,FUN=var,MARGIN=2)) G<-tcrossprod(X)/L # The G matrix is an nxn matrix of distances to measure similarities # between participants with respect to their lipid profiles ## RKHS model parameters ## # ETA # ETA.FixMet=list(Met=list(K=G,model="RKHS"), # The G matrix represents the lipidome, # and the RKHS kernel is specified Fix=list(~H$age+factor(H$smoking), model="FIXED")) # The fixed effects of the model are age and smoking ## RKHS regression ## fmGBLUP<-BGLR(y=y,ETA=ETA.FixMet, nIter=nIter, burnIn=burnIn, saveAt="GBLUP_") # RKHS model is: BMI = fixed effects + lipidomic data ### Variance of Best Linear Unbiased Predictor (BLUP) and variance of error ### # inference was done based on one of every 5 samples of the last 150,000 # therefore, since the burn in was 50,000 we need to remove the first 10,000 samples list.files() VarU=scan("GBLUP_ETA_Met_varU.dat") #load in variance of the lipidome BLUP model (varU