Development and Clinical Evaluation of a Multiplexed Health Surveillance Panel Using Ultra High-Throughput PRM-MS in an Inflammatory Bowel Disease Cohort

Abstract

Despite advances in clinical proteomics, translating protein biomarker discoveries into clinical use remains challenging due to the technical complexity of the validation process. Targeted MS-based proteomic approaches such as parallel reaction monitoring (PRM) offer sensitive and specific assays for biomarker translation. In this study, we developed a multiplex PRM assay using the Stellar mass spectrometry platform to quantify 57 plasma proteins, including 24 FDA-approved biomarkers. Loading curves (11 points) were performed at 4 sample throughputs (100, 144, 180, and 300 samples per day) using independently optimized and scheduled PRM methods. Following optimization, an inflammatory bowel disease (IBD) cohort of plasma samples (493 IBD, 509 matched controls) was analyzed at a throughput of 180 samples per day. To monitor system performance, the study also included over 1000 additional injections for system suitability tests, low-, middle-, and high-quality controls, washes, and blanks. Using this approach, we observed high quantifiability (linearity, sensitivity, and reproducibility) in the PRM assay and consistent data acquisition across a large cohort. We also validated the candidate IBD markers, C-reactive protein and orosomucoid protein, identified in a recent discovery experiment.

Keywords: Clinical biomarker translation; Inflammatory bowel disease; Targeted peptides; Validation proteomics.

Figure 1

Schema of scheduled PRM methods development, evaluation, and establishment of preferred method for large scale and high‐throughput analysis.

Figure 2

Technical evaluation of PRM methods with repeated injections. The performance of the four SPD methods was evaluated using 30 fmol of the 83 SIL peptides spiked into 300 ng of digested control plasma (n = 10 technical). a) Indicates the time required for 10 injections with each method. b) The median cross‐peak points. c) The median peak intensities. d) The median coefficient of variation (CV) of the signal intensities. e) Correlation between the retention times of 83 SIL across any two throughput conditions.

Figure 3

Evaluation of the sensitivity and linearity of each peptide in the PRM assay for the different throughput methods. 0.002–120 fmol 83 SIL/300 ng pooled plasma peptides was evaluated for each SPD method (11‐point dilution curve; n = 5 technical). The LLOD a) and LLOQ b) was calculated using the Skyline bilinear regression fit. The results are binned into five ranges as indicated. c) Median linear response (R²) for all 83 SIL peptides over the dilution curve. d) Examples of linear regression curves for A2GL peptide ALGHLDLSGNR are presented for 100, 144, 180, and 300 SPD. The peaks displayed are from Skyline showing integrated peaks of the same peptides. LLOD/Q data for all peptides and throughputs can be found in Table S5.

Figure 4

Evaluation of PRM assay technical reproducibility. a) Schematic outlining the analysis of five unique pooled plasma samples collected over 5 different days using the 180 SPD method (n = 5 biological; n = 5 technical). Each plasma pool was created by combining eight individual plasma samples; a total of 40 samples used to generate five distinct plasma pools. Summary of the median intra‐day b) and inter‐day c) CV for all 83 SIL peptides. d) Total variability was assessed using the mean intra‐day CV (across 5 days) and the mean inter‐day CV (across all five replicates) with the following formula: CV total​ = (CV intra²​ + CV inter²​)^1/2. The total median variability is reported for each pool. Reliability data for each peptide is presented in Table S6.

Figure 5

Design of 57‐Plex PRM assay for patient cohort analysis. a) Schema for processing and assignment of randomized blocks of subject plasma samples in 12 96‐well plates for analysis. b) Schematic of the plate layout for the patient and QC samples. Each plate included LC column washes, solvent banks, and PRTC blanks for every eight IBD cohort samples. System suitability tests (SST) were monitored for every eight samples. Quality controls were implemented at three levels: high, mid, and low (100, 20, or 5 fmol SIL/300 ng plasma), assessed at the beginning, middle, and end of each plate. The complete acquisition process for 1002 samples (12 plates) took over 12 days and >2000 injections. c) PRM peak integration and data analysis for the SST, QC, and patient samples were performed using Skyline.

Figure 6

Assessment of PRM assay for patient cohort acquisition. a) Plot of the mean RT (above) and %CV (below) for the SST (black), QC (greys), control (blue), and IBD (red) samples for the 83 SIL peptides (error bars = SD; n = 311 SST; n = 36 QCs; n = 509 Control; n = 493 IBD). The peptides are ordered by column elution. Side panel shows distribution of CVs for each sample type. b) Plot of the mean SIL intensity (above) and CV (below) for the QC, control, and IBD samples. Side panels show distribution of SIL intensities and CVs for each sample type.

Figure 7

Evaluation of technical and biological variability in PRM assay quantitation for QC and patient cohort acquisition. Plot of the mean ratio of light to heavy peptide intensities a) and CVs b) for each of the SIL peptides (error bars = SD; n = 36 QCs; n = 509 Control; n = 493 IBD). The peptides are arranged by order of their column elution. Side panels show distribution of ratios and CVs by sample type. Technical (Tech) variability (QCs CVs) is lower than combined (Comb) technical and biological variability in the patient samples for all but seven peptides. Arrows in b) indicate seven peptides with CV > 40% in the QC samples. c) Plot of the mean light peptide intensity for the mid QC samples (error bars = SD; n = 36). Peptides are ordered by descending intensity. The seven peptides indicated in panel b) are labeled along with their column elution position.

Figure 8

Detection and validation of candidate IBD biomarkers. PRM measurement of CRP precursor ESDTSYVSLK over the dilution series (11‐points) using MS² a) and MS³ c) detection to measure LOD (purple) and LOQ (teal) for each approach (n = 5 technical). Panels to the right show summary of the light to heavy peptide ratios for MS² b) and MS³ d) in the QC samples and patient cohorts. CV values for the QC samples are listed below points. The MS³ approach increased sensitivity and reduced technical variability for CRP measurements. CRP was significantly increased in IBD patients compared to controls (n = 509 control; n = 493 IBD; P = 0.0001) e) PRM detection of A1AG1 precursor SDVVYTDWK in MS² mode over the dilution series (11‐point) to measure LOD and LOQ (n = 5 technical). f) Summary of A1AG1 QC and patient cohort analysis. A1AG1 was significantly increased in IBD patients (n = 509 control; n = 493 IBD; P = 0.00625).

Figure 9

Comparison of PRM with clinical immunoassay measurements for CRP. Plot of the correlation of CRP PRM‐MS based analysis (area ratio of light/heavy) and immuno‐based assay (ng mL⁻¹) in IBD patient plasma (n = 91). A Spearman correlation test revealed a strong correlation between the antibody‐based ELISA assay and the mass spectrometry based PRM assay (r = 0.907, p (two tailed) < 0.0001, 95% confidence interval 0.8603–0.9384).