Analysis of 1508 Plasma Samples by Capillary-Flow Data-Independent Acquisition Profiles Proteomics of Weight Loss and Maintenance

Abstract

Comprehensive, high throughput analysis of the plasma proteome has the potential to enable holistic analysis of the health state of an individual. Based on our own experience and the evaluation of recent large-scale plasma mass spectrometry (MS) based proteomic studies, we identified two outstanding challenges: slow and delicate nano-flow liquid chromatography (LC) and irreproducibility of identification of data-dependent acquisition (DDA). We determined an optimal solution reducing these limitations with robust capillary-flow data-independent acquisition (DIA) MS. This platform can measure 31 plasma proteomes per day. Using this setup, we acquired a large-scale plasma study of the diet, obesity and genes dietary (DiOGenes) comprising 1508 samples. Proving the robustness, the complete acquisition was achieved on a single analytical column. Totally, 565 proteins (459 identified with two or more peptide sequences) were profiled with 74% data set completeness. On average 408 proteins (5246 peptides) were identified per acquisition (319 proteins in 90% of all acquisitions). The workflow reproducibility was assessed using 34 quality control pools acquired at regular intervals, resulting in 92% data set completeness with CVs for protein measurements of 10.9%.The profiles of 20 apolipoproteins could be profiled revealing distinct changes. The weight loss and weight maintenance resulted in sustained effects on low-grade inflammation, as well as steroid hormone and lipid metabolism, indicating beneficial effects. Comparison to other large-scale plasma weight loss studies demonstrated high robustness and quality of biomarker candidates identified. Tracking of nonenzymatic glycation indicated a delayed, slight reduction of glycation in the weight maintenance phase. Using stable-isotope-references, we could directly and absolutely quantify 60 proteins in the DIA.In conclusion, we present herein the first large-scale plasma DIA study and one of the largest clinical research proteomic studies to date. Application of this fast and robust workflow has great potential to advance biomarker discovery in plasma.

Keywords: Absolute quantification; Clinical proteomics; Label-free quantification; Plasma or serum analysis; SWATH-MS; data-independent acquisition; high throughput; single shot; stable isotope standards.

Graphical abstract

Fig. 1.

Capillary flow data-independent acquisition plasma proteomics. A, A plasma pool sample was acquired in ten re-injections on the established capillary flow DIA setup. The total identifications and the identifications with CV below 10% or 20% were calculated. B, The CV of each protein was calculated. C, The reproducibility of protein identification was analyzed by counting the repeated identification of the individual proteins across the 10 runs. D, Quantification reproducibility of six plasma proteins spanning nearly six orders of magnitude (ALB (serum albumin), HP (haptoglobin), APOC3 (apolipoprotein C-III), PROS1 (vitamin K-dependent protein S), F11 (coagulation factor XI), and PI16 (Peptidase inhibitor 16)). E, The identification and quantification of FDA approved biomarkers was assessed. 47 of 52 identified biomarkers were quantified with CV below 20% (of a total of 109 described by Anderson et al. (38)).

Fig. 2.

Exploratory analysis of the DiOGenes dataset. A, Design of the European-wide weight loss and maintenance study called DiOGenes (http://www.diogenes-eu.org). B, The 1508 samples and 68 pools were randomized into 17 96-well plates for sample preparation and LC-MS acquisition. C, The missing values for the 34 replicates of the CID1-pool were calculated. D, The CVs for proteins were calculated for the 34 replicates of the CID1-pool. E, The missing values for proteins of the whole sample set of 1474 quality-controlled DiOGenes samples was calculated. F, The CVs for proteins were calculated for the full dataset. G, The CVs for proteins were calculated for each CID. H, For the analysis of intra individual CVs of proteins, only individuals who were sampled at all 4 CIDs were considered and compared with inter individual CVs.

Fig. 3.

Stable isotope standard reference peptides for absolute quantification in DIA. A, The missing values in the dataset for the SIS peptides spiked into the 1508 samples of the DiOGenes sample set. B, Number of missing values for each SIS reference peptides was calculated. 90 SIS were identified in 90% of the acquisitions. C, The CVs for the SIS reference peptides were calculated.

Fig. 4.

Biological processes during weight loss and maintenance. A, PLS-DA analysis of the DiOGenes dataset. B, The proteins significantly differentially expressed in the dataset were clustered in a heat map. The CID1 time point was set to zero. The annotated protein legend was colored by biological function: (lipid metabolism/transport (green); acute phase response/inflammation (red); blood coagulation (blue); steroid hormone transport/regulation (violet) angiogenesis (brown) and immune system components (yellow). C, IPA pathway analysis of the weight loss and weight maintenance to baseline time point, significantly enriched pathways are shown. Orange indicates an activated pathway, white no activity change and blue an inactivated pathway. A dot means no significant enrichment of the pathway.

Fig. 5.

Comparison of biological findings to related large-scale proteomic weight loss maintenance studies. A, The significantly differentially abundant proteins in this study resulting from the CID3 to CID1 comparison (151 candidate proteins) were compared with the candidate list published by Moreno et al. (13) (32 candidate proteins) using the same DiOGenes samples of CID1 and CID3, but analyzing them with the ASAP² pipeline including depletion, isobaric labeling, and DDA acquisition with an Orbitrap Elite mass spectrometer. B, Comparison of the significantly differentially abundant proteins between CID2 and CID1 resulting from this study (from 271 candidate proteins) to the results from the independent weight loss study by Geyer et al. (22) with the same layout of 8 weeks of weight loss (from 68 candidate proteins), acquired with a Q Exactive HF mass spectrometer.

Fig. 6.

Non-enzymatic glycation analysis of plasma proteins. A library was generated containing high quality glycation sites and was then applied to the targeted analysis of the DIA DiOGenes dataset. A, All identified glycation sites changes were calculated comparing to CID1 and corrected by the protein abundance change. Subsequently, the data was visualized in a clustered heat map (HC, average linkage). B, The glycation sites significantly differentially abundant in the dataset were processed as described above.