Longitudinal Protein Profiling using Targeted Proteomics and Recombinant Standards
Stable isotope standard generation and quality control
Heavy isotope-labeled recombinant protein standards termed SIS PrESTs were produced and purified using the workflow established within the Human Protein Atlas program previously described32 . Shortly, DNA fragments were cloned into an expression vector and transformed into an Escherichia coli strain auxotrophic for arginine and lysine. The cells were cultivated in minimal media with added heavy isotope-labeled (13C and 15N) versions of lysine and arginine. Harvested SIS PrESTs were purified using immobilized metal ion affinity chromatography and quantified by mass spectrometry using a non-labeled ultra-purified quantification tag standard previously quantified by amino acid analysis. All SIS PrEST standards were quantified in triplicates by mixing them in a 2:1 ratio with light quantification tag using the Bravo Automated Liquid Handling Platform (Agilent, Santa Clara, CA, USA). A mixture of Pierce mass spectrometry (MS) grade porcine trypsin (Thermo Scientific; Waltham, MA, USA) and Lysyl Endopeptidase (Lys C, Wako; Neuss, Germany) were added to the samples in 1:6 (w/w) (trypsin) and 1:12 (w/w) (Lys C) enzyme to substrate ratio (E:S). Proteins were digested at 37 °C and the reaction was quenched after 16 hrs by addition of trifluoroacetic acid (TFA) to a final concentration of 0.5% (v/v). The samples were vacuum-dried and analyzed on a Bruker impact II (Bruker Daltonics; Bremen, Germany) instrument in combination with UltiMate 3000 binary RS nano-liquid chromatography system (Thermo Scientific; Waltham, MA, USA) connected by a CaptiveSpray nanoBooster source (Bruker Daltonics). The samples were stored in a vacuum dried format and were resuspended by the autosampler prior to the analysis. Approximately 150 fmol of the sample was loaded onto an Acclaim PepMap 100 trap column (75 μm × 2 cm, C18, 3 μm, 100 Å, Thermo Scientific), washed for 5 min at 5 μl/min with 100% of solvent A (3% Acetonitrile (ACN), 97% water, 0.1% formic acid (FA)), and then separated by a PepMap RSLC C18 column (75 µm x 25 cm, 2 µm, 100 Å, Thermo Scientific). The peptides were eluted with a linear 9 min gradient of 4-40% solvent B (95% ACN, 5% water, 0.1% FA) at 0.500 µl/min flow rate followed by a 2 min increase to 95% of solvent B. The column was washed for 8 min with 95% of solvent B and then equilibrated for at least 10 min with 2% of solvent B. The analytical column was kept at 35 °C by the integrated liquid chromatography (LC) column oven. A data-dependent method was used where the cycle time was set to 2 s and the MS1 scan was done at 1 Hz covering a range between 400-2,200 m/z. Precursors were fragmented with the preferred charge states between 2-5 and spectra were acquired with dynamic rate (8-32 Hz) based on the number of counts (cts) per 1,000 summations (lower limit at 2,500 cts and higher at 25,000 cts). Active exclusion was set to exclude after acquiring one spectra and for 30 s, however, precursors were reconsidered if the intensity increased 3 times since the last MS/MS event. The MS data was evaluated in Skyline Proteomics Environment (version 3.7)33 and the SIS PrESTs were assigned an absolute concentration based on the median MS1 ratio calculated from three proteotypic peptides (ISEATDGLSDFLK++, DLQAQVVESAK++, DLQAQVVESAKK+++).
Spectral library and PRM method generation
A total number of 426 SIS PrESTs selected towards predicted secreted proteins were used for the initial spectral library and PRM method generation. All SIS PrEST standards were analyzed in equimolar amounts and digested by trypsin as described below. The samples were analyzed using a combination of UltiMate 3000 binary RS nano-LC system (Thermo Scientific) with an EASY‐Spray ion source connected to an on-line Q Exactive HF (Thermo Scientific) mass spectrometer. The peptides were resuspended in solvent A (3% ACN, 97% water, 0.1% FA) and approximately 100 fmol per SIS PrEST was injected. A Top5 method was used for spectral library generation where MS1 scans were acquired with 60,000 resolution at 200 m/z (AGC target 3e6, range 350 to 1,600 m/z, injection time 110 ms) was followed by 5 MS/MS scans at 30,000 resolution at 200 m/z (AGC target 1e5, range 200 to 2,000 m/z, injection time 150 ms) with isolation window 2 m/z, normalized collision energy (NCE) 27 and dynamic exclusion of 50 s. Raw files were searched in MaxQuant version 220.127.116.114, using the search engine Andromeda against SIS PrEST sequences (Table S1) with an E. coli background (BL21 UniProt ID: #UP000002032, 4,208 entries, accessed 2015-06-22). Arg10 and Lys8 were chosen as heavy labels (fixed) with a maximum of 3 labels per peptide. The enzyme specificity was set to trypsin and a maximum of 2 missed cleavages were allowed and the minimum peptide length was set to 7 amino acids. Oxidation on methionine and N-terminal acetylation were set as variable modifications and carbamidomethylation on cysteine was set as a fixed modification. The false discovery rate (FDR) was determined by a reverse database and a cutoff set to 0.01 at both the peptide and protein level. Identified peptides were further processed by only allowing proteotypic peptides originating from the PrEST sequence mapping to one single human gene (defined by SwissProt), also excluding peptides with more than 1 missed cleavage and with possible post-translational modifications. The peptides and genes for final PRM assay were chosen firstly by evaluating their endogenous signal in pooled plasma from the study participants and then limiting the number of peptides based on their retention time, so that the number of concurrent precursors would not surpass 22 to allow for at least 10 points across the chromatographic peak (approx. 20 s) with MS method cycle time of 1.4 s.
Additionally, the signal linearity and peptides’ coefficient of variations in the measured range were analyzed. Standard curves were established between the SIS PrEST standard pool used for quantification and their corresponding endogenous peptides. Pooled plasma was serially diluted (reverse standard curve) in relation to an SIS PrEST standard pool in three-fold dilution steps representing 27:1, 9:1, 3:1, 1:1, 1:3 and 1:9 where the 1:1 represents the spike in level used in this study. The plasma pool consisted of 15 randomized female and 12 randomized male samples (the female fraction made up 57% of the total volume). The samples were prepared following the same workflow as the samples of the main study and analyzed using the same LC-MS/MS method (described in the Quantification by PRM section below).
A semi-automated sample preparation
All samples and SIS PrESTs were processed by a semi-automatic liquid handler system Bravo (Agilent) equipped with a 96LT tip head. Plasma samples were diluted 10 times in 100 mM Tris buffer (pH 8). 5 µl of the diluted sample corresponding to 0.5 µl of raw plasma was mixed with a SIS PrEST mixture representing an estimated 1:1 (Light to Heavy, L:H) peptide ratio with the endogenous plasma levels. Proteins were first reduced and denatured at 90°C for 10 minutes in a buffer containing 1% sodium deoxycholate (SDC) (w/v) and 10 mM dithiothreitol (DTT), Thereafter alkylated with 50 mM chloroacetamide (CAA) and incubated in dark for 20 min. A mixture of Pierce MS grade porcine trypsin (Thermo Scientific) and Lys-C (Wako) was added manually in respective E:S ratios (1:50 for trypsin and 1:100 for Lys-C) (w/w). After 16 hrs of incubation at 37 °C the digestion was quenched by adding TFA to a final concentration of 0.5% (v/v). SDC was precipitated for 30 min at RT and then centrifuged for 10 min at 3,273 rcf on Allegra X 12R centrifuge (Beckman-Coulter; Brea, CA, USA). The samples were desalted using in-house prepared StageTips packed with Empore C18 Bonded Silica matrix (3M, Saint Paul, MN, USA) 35 . Briefly, three layers of octadecyl membrane were placed in 200 μl pipette tips. The membrane was activated by addition of 100% ACN and then equilibrated with 0.1% TFA. 15 µg of peptides from the acidified sample (half of the sample volume) was added to the StageTip membrane and then washed twice with 0.1% TFA. The peptides were eluted in two-step elution with 30 µl of elution buffer (80% ACN, 0.1% FA). In between every buffer addition during the desalting the samples were centrifuged for 3 min at 931 rcf. Desalted peptides were vacuum‐dried and stored at -20 °C before being subjected to LC‐MS/MS analysis.
Samples were measured on the same MS setup as for the method generation, an online system of Ultimate 3000 nano-LC with an EASY‐Spray ion source connected to Q Exactive HF mass spectrometer. All plasma samples were stored vacuum-dried and were resuspended by the autosampler. A total of 1 µg peptide weight was loaded onto an Acclaim PepMap 100 trap column (75 μm × 2 cm, C18, 3 μm, 100 Å, Thermo Scientific), washed 5 min at 5 μl/min with 100% of solvent A (3% ACN, 97% water, 0.1% FA), and separated by a PepMap RSLC C18 column (75 µm x 25 cm, 2 µm, 100 Å, Thermo Scientific). The peptides were eluted with a linear 33 min gradient of 1-30% solvent B (95% ACN, 5% water, 0.1% FA) at a flow rate of 0.400 µl/min followed by a 2 min increase to 43% and then 1 min increase to 99% of solvent B. The column was washed for 8 min with 99% of solvent B at a flow rate of 0.750 µl/min and then equilibrated for 10 minutes min with 3% of solvent B at a flow rate of 0.400 µl/min. The analytical column was kept at 35 °C by the in-source temperature controller.
Quantification by PRM
For sample analysis, previously developed PRM methods were used. Each MS cycle comprised of one full MS scan performed at 60,000 resolution (AGC target 3e6, mass range 350-1,600 m/z and injection time 110 ms) followed by 20 MS/MS scans at 30,000 resolution (AGC target 2e5, NCE 27, isolation window 1.5 m/z and injection time 55 ms) defined by a scheduled (2 min windows) PRM isolation list that contained 174 paired light and heavy peptide precursors (npeptides = 87) from 55 SIS PrESTs covering 52 human proteins.
MS data evaluation and protein quantification
The raw MS-files from all study samples were processed in Skyline (version 3.7) 33 and all raw files and extracted chromatograms were uploaded to the Swedish National Data Service https://snd.gu.se/en (Longitudinal Plasma Protein Profiling using Targeted Proteomics and Recombinant Protein Standards Identifier: SND 1292), a data repository certified by Core Trust Seal (see the Data and materials availability). For each peptide the ratio between endogenous and SIS PrEST peptide was calculated from the summed area intensity over the retention time. The peak integration was done automatically at the MS2 level, however peak boundaries were adjusted to cover the whole peak area in cases where the peak apex had not been fully integrated by Skyline. The ten most intensive fragments (in m/z range of 200 to 1,600) from the spectral library per precursor were used for quantification, whereas y1 and b1 fragments were excluded. Data were exported from Skyline and analyzed in R (version 3.4.1). Firstly, all results with ratio dot product (rdotp) value less than 0.75 were excluded. This eliminated chromatographic peaks with low signal-to-noise and peaks affected by contaminants or interfering signals from other peptides. Also, peptides with more than 25% CV calculated from 15 replicates quantified in five plates were excluded. Finally, peptides with possible PTM’s were removed and two peptides from two proteins were excluded due to low signal intensity and the considerably higher CV’s observed compared to other peptides from the same proteins. In total, 17 out of 87 measured peptides were removed from the final data analysis resulting in 70 peptides that were used for quantification of 47 proteins with 50 SIS PrESTs (Table S2). As the samples were run in 6 plates, the ratios from 5 plates were normalized using median centric peptide-specific normalization factors based on the replicates from each plate (Table S3). The pools from one plate were not prepared together with the samples, thus the results were normalized using quantile normalization against the other 5 plates as reference. The absolute peptide concentrations were calculated using the absolute and known amounts of spiked-in SIS PrEST standards. Medians of normalized peptide values were stated as the absolute protein concentrations. The results were further mean-centered based on each visit.