Emerging affinity-based proteomic technologies for large scale plasma profiling in cardiovascular disease

Abstract

Plasma biomarkers that reflect molecular states of the cardiovascular system are central for clinical decision-making. Routinely used plasma biomarkers include troponins, natriuretic peptides and lipoprotein particles, yet interrogate only a modest subset of pathways relevant to cardiovascular disease. Systematic profiling of a larger portion of circulating plasma proteins (the plasma proteome) will provide opportunities for unbiased discovery of novel markers to improve diagnostic or predictive accuracy. In addition, proteomic profiling may inform pathophysiological understanding and point to novel therapeutic targets. Obstacles for comprehensive proteomic profiling include the immense size and structural heterogeneity of the proteome, as well as the broad range of abundance levels. Proteome-wide, untargeted profiling can be performed in tissues and cells with tandem mass spectrometry. However, applications to plasma are limited by the need for complex preanalytical sample preparation stages limiting sample throughput. Multiplexing of targeted methods based on capture and detection of specific proteins are therefore receiving increasing attention in plasma proteomics. Immunoaffinity assays are the workhorse for measuring individual proteins but have been limited for proteomic applications by long development times, cross-reactivity preventing multiplexing, specificity issues, and incomplete sensitivity to detect proteins in the lower range of the abundance spectrum (below picograms per milliliter). Emerging technologies to address these issues include nucleotide-labeled immunoassays and aptamer reagents which can be automated for efficient multiplexing of thousands of proteins at high sample throughput, coupling of affinity capture methods to mass spectrometry for improved specificity, and ultrasensitive detection systems to measure low-abundance proteins. In addition, proteomics can now be integrated with modern genomics tools to comprehensively relate proteomic profiles to genetic variants, which may both influence binding of affinity reagents and serve to validate the target specificity of affinity assays. The application of deep quantitative proteomic profiling to large cohorts has thus become increasingly feasible with emerging affinity methods. The aims of this article are to provide the broad readership of Circulation with a timely overview of emerging methods for affinity proteomics and recent progress in cardiovascular medicine based on such methods.

Introduction

Cardiovascular diseases are the leading causes of death and hospitalization globally, dominated by coronary artery disease, heart failure, atrial fibrillation and stroke.^{1, 2} Profiling of plasma proteins has been central for decision-making in cardiovascular medicine since the introduction of many immunoassays in the 1980s, most prominently for the diagnosis of myocardial infarction (creatine kinase, troponins) and heart failure (natriuretic peptides) and for cardiovascular risk stratification (lipoproteins). However, each of these markers suffers from limitations in diagnostic or predictive accuracy. Systematic assessment of a large portion of the entire range of proteins measurable in plasma (the plasma proteome) provides opportunities for unbiased discovery of novel markers to improve accuracy, generate pathophysiological insights and identify therapeutic targets. Most plasma proteins remain unexplored for relation to disease, but systematic proteome screens have been hampered by technical limitations, most importantly with regard to limitations in sensitivity, specificity, multiplexing and sample throughput (Table 1). Characteristics of the principal proteomic methods are summarized in Table 2. In this article we provide an overview of current and some of the most important emerging methods for proteomic profiling with particular emphasis on affinity-based methods, and the potential of such tools for cardiovascular medicine. We particularly emphasize new tools to improve the key properties of multiplexing, sample throughput, sensitivity, and specificity of affinity proteomics, but begin with an introduction to the proteome and traditional methods for plasma profiling. Finally, we discuss the integration of genetics with proteomic profiles.

Table 1.

Glossary of proteomic terms.

Affinity proteomics: methods for protein detection, based on protein isolation by an affinity reagent coupled to a reporter system for detection

Analyte: molecule undergoing measurement, e g a protein

Aptamer: nucleotide-based compounds with protein affinity, alternative to antibodies

Chromatography: method for protein separation from a mixture based on differential mobility through a medium

Coverage: total proportion of proteins identified from a complex mixture (for mass spectrometry also used to indicate the number of peptides identified from a given protein)

Dynamic range: the largest and smallest abundances of proteins in a mixture

Electrophoresis: method for protein separation from a mixture by mass and charge

iMRM: immunoaffinity coupled with multiple reaction monitoring, a targeted mass spectrometry based method

Limit of blank (LOB): measure of background noise with assay as the highest concentration at which false positive signal is detected in a sample without the analyte

Limit of detection (LOD): highest/lowest analyte concentration to be reliably distinguished from LOB with assay

Limit of quantification (LOQ): highest/lowest analyte concentration at which a detected analyte can be reliably quantified

Mass spectrometry (MS): method for protein detection by mass and charge

Microarray: technology for parallel testing of multiple analytes from mixture, based on a small glass or plastic slide to which multiple reagents are attached

Mixture: sample containing analyte of interest, also called matrix

Multiple reaction monitoring (MRM): MS method for quantification of targeted proteins by chemical labeling with stable isotopes.

Multiplexing: multiple analytes measured in parallel (for mass spectrometry also used to indicate number of samples analyzed in parallel)

Proximity extension assay: method for multiplexed protein detection based on antibodies linked to nucleotides for detection on a polymerase chain reaction system

Sandwich assay: immunoassay with dual antibodies for improved specificity

Sensitivity: measure of assay ability to capture analytes with low abundance

Specificity: measure of assay ability to measure particular analyte rather than others

Throughput: number of samples and/or analytes undergoing analysis in a certain timeframe

Ultrasensitive assay: assay with lower LOQ at concentrations below picogram per milliliter

Table 2.

Methods for highly parallel proteomic analysis.

	Mass spectrometry	Immunoaffinity arrays/beads	Aptamer array
Sample throughput	Low (untargeted) to moderate (targeted) due to multiple analytical steps	High	High
Analyte multiplexing	All proteins in sample: 10 to >5000	100 proteins	>1300 proteins
Sample volume	30 uL (targeted) – 100s of uL	1 uL (Proximity Extension Assay) – 100s of uL	65 ul
Dynamic range	High and medium abundance proteins. Targeted labelling required for low-abundance proteins, and depletion of high-abundance proteins.	Wide (mg/ml to pg/ml, down to fg/ml with ultrasensitive methods)	Wide (mg/ml to pg/ml)
Reproducibility	Modest intra-assay coefficient of variation (CV), modest to good with targeted methods	Good (low intra-assay CV)	Good (low intra-assay CV)
Specificity	High	Under investigation, higher with sandwich assays	Under investigation, high in preliminary analyses
Quantification	Relative, targeted labelling for absolute (i e mg/ml)	Absolute or relative	Relative
Characterization of coding DNA variants	Yes	No	No
Characterization of poststranslational modifications and isoforms	Yes	No	No
Protein digestion	Trypsin digestion	No	No
Protein separation	Multidimensional separation	No	No
Bioinformatics	Shotgun sequencing complicated, but multiple softwares available	Easy	Easy
Public access to reagents	Yes	For many immunoassays	Individual aptamers available

The plasma proteome

The human plasma proteome constitutes a complex mixture of proteins derived from all tissues, which makes plasma an attractive medium for clinical analysis as a dynamic representation of the molecular states of diverse systems.³ A wide range of proteins can thus be detected in plasma, including carrier proteins such as albumin, immune system effectors including immunoglobins and complement factors, hemostatic factors, tissue messengers such as natriuretic peptides and interleukins, and tissue leakage products such as troponin and creatine kinase. This diversity in plasma protein function is accompanied by a diversity in protein abundance, with reference intervals for known plasma proteins in healthy subjects spanning more than 11 orders of magnitude (Figure 1). The total protein content in a plasma sample can be estimated from nitrogen content using colorimetric tests, typically amounting to 60–80 mg/ml plasma. The most highly abundant protein in plasma is albumin (35–45 mg/ml) whereas low-abundance proteins include leakage products such as troponins and cytokines (a few pg/ml and below). Albumin constitutes 55% of the total proteome mass^{4, 5} and the 22 most highly abundant proteins in Figure 1, including immunoglobulins, together constitute approximately 99%, due to large molecular mass and low clearance. These proteins and others, up to 50–100 proteins identified by plasma separation with electrophoresis in the 20^th century, are referred to as the “classical” plasma proteins, are present in medium-high abundance and carry out their function in the circulation.^{4, 6–8} However, the total number of components of the plasma proteome remains unknown. Various systematic projects have been undertaken to catalog plasma proteins, and have individually discovered in the range of a few thousand proteins, with only limited overlap between projects depending on sample processing and analytical techniques.^9–15 It seems likely that a large proportion of the total human proteome may at some point in time be detectable in plasma, although often at low abundance, depending on a multitude of factors including age, sex, disease status, medications and sampling conditions. The total number of human proteins is expected to include from 20,000 (based on the number of human protein-coding genes) to millions of proteins¹⁶ resulting from alternative splicing, proteolytic processing and posttranslational modifications. Such modified proteins constitute important components of the proteome, as exemplified by processing of proBNP to BNP, D-dimers, glycated proteins (e g glycated hemoglobin) and phosphoproteins.

Figure 1. Dynamic range of the human plasma proteome and current detection methods.

Dynamic ranges for classical plasma proteins (high-medium abundance) and selected clinically relevant proteins. Intervals refer to published 95% reference ranges, or quartiles or ranges where reference intervals were unavailable. Lower ends of reference intervals have not been established for a few proteins (arrowhead). Affinity-based methods essentially capture proteins across the entire abundance spectrum but may be offset by specificity, whereas mass spectrometry (MS) has excellent specificity but is limited to proteins with high and medium abundance. Newer MS methods such as multiple reaction monitoring (MRM), immuno-MRM (iMRM) and isobaric tags (iTRAQ) may detect lower-abundance proteins and with additional separation steps reach as low as affinity methods. Ultrasensitive single molecule detection (SMD) assays are necessary to detect proteins at very low abundance, down to fg/ml. 2DE, 2-dimensional gel electrophoresis. LC-MS, liquid chromatography MS.

Assaying the Plasma Proteome

The complexity of the plasma proteome described above translates into analytical complexity for individual proteins, especially low-abundance proteins. For example, finding a troponin molecule present at 5 pg/ml among albumin molecules present at 45 mg/ml can be compared to finding one individual human while searching the entire human population. The quantitative predominance of the “classical” plasma proteins particularly complicates classical separation-based analytical methods such as electrophoresis and mass spectrometry.¹⁷ Electrophoresis and liquid chromatography were the earliest tools for protein detection, developed in the early 20^th century, and are based on separation of components of a mixture by characteristics such as hydrophobicity, mass and charge. The resolution of these methods is limited to the most abundant proteins in plasma (Figure 1), but they are commonly utilized in plasma proteomic experiments for preanalytical sample fractionation before mass spectrometry.

Mass spectrometry is the most powerful tool for systematic and unbiased detection of the full set of proteins present in tissues and cells, including the heterogeneous posttranslational protein modifications.^18–22 However, application to human plasma is limited by need for multiple sample preparation stages, typically including depletion of high-abundance proteins, liquid chromatography and trypsin digestion, resulting in multiple analytical steps.¹⁷ In addition, quantitative mass spectrometry with high precision requires protein labeling in the sample with stable isotopes. Instead, multiplexed affinity methods have recently received increased focus for plasma proteomics.²³

Affinity-based or “targeted” assays were first developed in the 1960–1980s,²⁴ based on antibodies to target specific proteins from a complex mixture,^25–27 and remain the golden standard for clinical protein analysis to date. The most important immunoassays in cardiovascular medicine – for troponins and natriuretic peptides – were developed in the late 1980s.^28–30 With such immunoassays, polyclonal or monoclonal antibodies are used to capture protein targets through establishment of diverse noncovalent bonds including electrostatic or hydrophobic interactions, hydrogen bonds, or van der Waals forces, specific to the individual antigen-antibody pair. Target proteins are then indirectly quantified based on either the signal intensity of luminescent, fluorescent, enzymatic (e g enzyme-linked immunosorbent assays, [ELISAs]), or radioactive (radioimmunoassays [RIAs]) antibody-labeled reporters or based on light scattering of antibody-protein complexes (turbidimetry and nephelometry). Protein abundance in plasma is typically estimated relative to a reference signal, from spiking unlabeled samples with known concentrations of isotopically labeled synthetic peptides or using serial dilution steps with a known amount of the protein of interest to develop calibration curves across the relevant dynamic range. The performance of the assay for each protein is further evaluated for limit of blank (see Table 1), limit of detection (LOD), and limit of quantification (typically defined as the limit below which the coefficient of variation exceeds 20%, which of necessity is higher than the LOD)³¹. For many known low-abundance plasma proteins, immunoassay detection and quantification in the lower end of the dynamic range is often limited by incomplete assay sensitivity, resulting in truncation of the dynamic range. To improve sensitivity and specificity of both single and multiplex assays, two antibodies are often combined for protein capture and detection (referred to as a sandwich assay).

The major obstacle confronting affinity proteomics is that immunoassays such as ELISAs cannot be readily multiplexed due to cross-reactivity of affinity reagents.³² The development times for individual antibodies are also generally long and costly (particularly for monoclonal antibodies). Further, specificity may be unclear, and sensitivity is often poor to detect proteins in the lower range of the abundance spectrum (below picograms per milliliter) which may be particularly informative.³³ There is currently intense focus on methods to address these central limitations. To generate a catalog of antibodies for human proteins, the Human Protein Atlas (HPA) project aims to assemble and validate antibodies for proteins encoded by all human genes in the genome, and characterize protein expression across multiple human tissues using immunohistochemistry. A recent HPA report describes results for antibodies targeting nearly 17 000 genes, which are being implemented in parallelized assays.¹⁵ To achieve multiplexing antibodies can be immobilized on planar microarrays³⁴ or lipid beads in solution³³, the latter of which allows parallel assessment of around 10–30 proteins, but antibody cross-reactivity prevents higher degrees of multiplexing with these tools.³⁵ Some of the recent key developments in affinity proteomics include systems for ultrasensitive detection of single molecules, nucleotide-labeled immunoassays and aptamer reagents for efficient multiplexing at high sample throughput, and coupling of affinity capture methods to mass spectrometry for improved specificity (Figure 2). Each of these methods are discussed in the following sections.

Figure 2. Schematic workflow for emerging affinity proteomic tools.

Abbreviations: DNA, deoxyribonucleic acid. m/z, mass to charge ratio. qPCR, quantitative polymerase chain reaction.

The selection of an affinity-based method or mass spectrometry for a particular experiment depends on several factors, as outlined in Table 2 and the following sections. In summary, advantages of affinity-based methods include a) high sensitivity for many low-abundance proteins, b) high sample throughput, c) ease of use, d) low instrument investment, and e) ability to target selected proteins of biological or clinical interest determined a priori. On the other hand, limitations include a) inability to discover proteins not targeted by the assay, b) inability to distinguish post-translational modifications and isoforms, c) potential influence of coding DNA variants on epitope structures and affinity of reagents (see below), d) batch effects that may be a source of significant variability in reagents and a source of irreproducibility, and e) lack of open science platforms for several of the available methods which complicates validation or customization of affinity reagents.

Increasing the sensitivity of immunoaffinity assays to single molecules

The most sensitive conventional immunoassays have a lower limit of quantification (LOQ) which enables detection of proteins present in down to a few picograms per ml. For example, several of the more than 30 commercially available immunoassays targeting epitopes on Troponin I or T are referred to as highly sensitive,³⁶ detecting plasma concentrations down to several pg/ml, below which level patients are unlikely to have an acute MI.^37–40 However, many proteins including locally acting cytokines and tissue damage leakage products, are present in plasma at concentrations below this level. Efforts are therefore ongoing to develop immunoassays with sensitivity to detect proteins present in femtograms per ml, referred to as ultrasensitive assays, and even down to the single molecule level.^{23, 41–43} One milliliter of a femtomolar solution would be expected to contain 600 000 single molecules.⁴¹ Ultrasensitive methods typically employ highly sensitive detectors, antibodies labeled with fluorescent, luminescent or enzymatic reporters for signal amplification as well as preanalytical enrichment steps. For example, a sandwich immunoassay based on preanalytical analyte enrichment with paramagnetic microparticles and counting of fluorescence events with laser-excitation in a capillary flow cell using a single-photon detector (Figure 2) is offered commercially by Singulex (Alameda, CA).⁴⁴ This assay has been shown to robustly detect a normal distribution of troponin in healthy individuals in the lower femtogram range with 10% coefficient of variation, and slightly increased levels in this range were associated with a graded increase in incidence of cardiovascular disease.^{45, 46} In addition, slight increments in response to exercise stress testing were associated with transient myocardial ischemia and higher incidence of adverse outcomes during follow-up.⁴⁷ Disadvantages of the method are that it requires relatively large sample volumes, up to 100 uL, and is limited to single proteins.²³ Another method referred to as digital ELISA, offered by Quanterix (Lexington, MA), has also been shown to detect troponin in this lower range⁴⁸ and is based on analyte enrichment with microbeads which are each loaded into a femtoliter microwell, avoiding antibody cross-reactions, and converted to fluorescence products using enzymatic methods.⁴⁹ Digital ELISA has been multiplexed to measure six low-abundance cytokines in parallel.⁵⁰ Other immune-based methods of interest are based on light emission at excitation (electrochemiluminescence) combined with microarray technology for multiplexing to up to 10 low-abundance proteins as offered commercially by Meso Scale Discovery (Rockville, MD),⁵¹ and nanofluidic systems⁵². These ultrasensitive methods open for precise quantification of proteins at the lower end of the abundance spectrum in plasma which may be the most informative.

Multiplexing with nucleotide-labeled antibodies

Another method with the aim of improving sensitivity includes the labeling of antibodies with a nucleic acid. Nucleic acids can then be amplified and quantified by Watson-Crick base pairing to fluorescently labeled primer fragments based on the polymerase chain reaction (PCR). This coupling of immunoaffinity to PCR is referred to as immuno-PCR.⁵³ One development of immuno-PCR referred to as a proximity extension assay (PEA) has proven particularly useful in multiplexing to reduce the problem of cross-reactivity. PEA is based on the incubation of the sample with dual antibodies targeting proximal epitopes on the analyte of interest. The antibodies are labelled with complementary DNA oligonucleotide sequences, which come in close proximity upon target binding and subsequently hybridize.⁵⁴ Oligonucleotides are then extended over the complementary probe to form a PCR amplicon through the addition of a DNA polymerase.⁵⁵ (Figure 2) The resulting amplicon is detected and quantified using quantitative real-time PCR, and thus only reports cross-matched DNA pairs and not cross-reactive antibody events. The method has been shown to be scalable without loss of specificity for multiplexed quantification of up to 92 proteins simultaneously in a commercialized version of the assay (Olink Biosciences, Uppsala, Sweden).³⁵ The assay quantifies proteins across 5 logs of abundance with a reported intra-assay variability below 10% and inter-assay variability below 15%.³⁵ In addition, the PCR-based amplification step results in uniquely small sample requirements with the PEA method of only 1 uL. The PEA assay thus offers several advantages including robust multiplexing, low sample consumption, and sensitivity, and is currently being developed for thousands of proteins.

Multiplexing with nucleic acids as affinity reagents (aptamers)

An alternative strategy to address the limitations of current immunoassays has been the development of other forms of affinity reagents than antibodies.⁵⁶ Candidate reagents to optimize affinity and specificity have been reviewed previously^57–59 and include engineered antibody fragments and a range of small synthetic proteins such as affibodies (based on three-helix scaffolds from the immunoglobin G binding domain of staphylococcal protein A)⁶⁰, DARPins (designed ankyrin repeat proteins)⁶¹, and peptoids (oligomers of N-substituted glycines)⁶². A particularly attractive type of affinity reagent is the oligonucleotide-based aptamer described in 1990,^{63, 64} which is simple and cost-efficient to design and has the major advantage that it may be amplified for improved sensitivity and easily detected using PCR and hybridization arrays. In aptamer development, large libraries of random oligonucleotide sequences of typically 10¹⁴–10¹⁵ molecules are mixed with target peptides or proteins in an iterative process to test for binding, while aptamers that bind to other targets are depleted from the pool.^{63, 64} The nucleotide aptamers fold into diverse shapes that interact physically with the protein surface. Aptamers have been shown to be stable and display high binding affinity which can be adjusted via sequence modifications.⁶⁵ Several groups have developed multiplexed aptamer platforms. A commercial platform has been developed by Somascan (SomaLogic Inc, Boulder, CO) based on strands of 40 nucleotides which have been chemically modified to expand the target repertoire, particularly with aromatic benzyl sidechains to enhance hydrophobic interactions, and for low reagent cross-reactivity, attributable, in part, to a charged phosphodiester backbone.⁶⁶ These modified aptamers, referred to as SOMAmers, are thus well suitable for multiplexing. Initial work has also indicated high specificity, but this issue requires further study.⁶⁵ Some level of unspecific binding has been reported⁶⁷ and an aptamer for the protein GDF11 was reported to also bind to GDF8.^68–71 SOMAmers have been developed for thousands of proteins which are detected and quantified on microarrays with high sample throughput. The current version of the assay quantifies over 1300 proteins through a multi-step capture, release and re-capture enrichment process.⁶⁵ (Table 2, Figure 2) The assay is easy to use (microarray reader), has been shown to be highly reproducible,⁷² can easily be automated to more than 20-fold faster than deep proteomic mass spectrometry,⁷² and to readily capture proteins across a wide dynamic range of at least 7 orders of magnitude: for example, capturing both albumin, interleukin-6, and natriuretic peptides.⁶⁵ The assay has been applied to multiple diseases including renal failure, muscular dystrophy and myocardial infarction, identifying both well-established markers such as cystatin, muscular creatine kinase and troponin, respectively, as well as new markers.^{65, 72, 73} As for antibodies, protein modifications may alter aptamer binding, and newer reagents may be developed that are targeted to specific post-translational modifications. Thus, the degree of multiplexing, sample throughput and ease of reagent development surpasses any other affinity method, although the specificity, as for antibodies, must be verified by “orthogonal” methods such as ELISA or immuno-MRM (see below).

Increasing specificity by coupling affinity methods to mass spectrometry

The target specificity of affinity reagents, whether antibodies or aptamers, remains the key determinant of assay performance. Unfortunately, there are no standardized methods to evaluate binding affinity of the reagent to its target protein versus other proteins, but application of MS methods after affinity pulldown is a powerful and increasingly utilized tool to understand reagent specificity.⁷⁴

Mass spectrometry (MS) has excellent specificity, provides protein sequence information, and is robust to detect posttranslational protein modifications. It can either be used as an unbiased tool to detect proteins in a sample (protein sequencing) or to detect and quantify specific targeted proteins.^19–22 In essence, MS can be considered both a separation and capture method.¹⁸ The principle of MS is to separate particles in a mixture (in the gas phase with ionized analytes) according to the ratio of mass and charge by acceleration through electric or magnetic fields (mass analyzer) for capture and detection of the number of particles (ion current intensity) with certain mass-to-charge ratios. Ion abundance within a selected mass spectrum is detected by a photomultiplier, microchannel plate or electron multiplier. For unbiased MS (protein sequencing), two mass analyzers are used in tandem to detect proteins present in a sample, typically one low-resolution mass analyzer coupled to a high-resolution analyzer, with a collision cell in between for peptide fragmentation.¹⁸ The identified ion current peaks are subsequently matched to proteins using bioinformatics approaches, requiring that the proteins under study are present in databases so >50% of peaks typically remain unmatched although this number can be improved with depletion, multidimensional separation, iMRM and large sample volumes. Unbiased MS, however, only provides rough abundance estimates based on ion current intensity. This limitation is due to the influence of a variety of factors such as protein digestion efficiency, recovery during sample preparation, ionization, matrix effects and various properties of the mass analyzer. Targeted methods instead provide precise quantitative information, by using isotopically labeled synthetic peptides with the mass analyzer in the MRM mode (MRM: multiple reaction monitoring) or through peptide labeling with isobaric chemical tags (iTRAQ: isobaric tags for relative and absolute quantification). The sensitivity of MRM-MS can be further improved by adding an antibody-based enrichment step before MS, referred to as immuno-MRM (iMRM). An iTRAQ method has been described to detect >5300 proteins in plasma, but with sample throughput limited to approximately one sample per day at considerable cost.⁷⁵ Similar to protein sequencing, targeted MS requires tandem mass analyzers, through which only user-selected specific ions are transmitted from the first analyzer and user-selected specific fragments arising from these ions are measured by the second analyser. Both unbiased and targeted, labeling-based MS methods are relatively labor-intensive and time-consuming limiting use of this technique to proteomic experiments with low to moderate sample throughput.¹⁸ Current platforms allow simultaneous analysis of up to 10 samples in parallel, and it is now technologically feasible to assess up to 40 samples in a single experiment.⁷⁶

When coupling MS to affinity-based assays, the affinity reagents are used to capture targets in plasma or cognate proteins in buffer which are then eluted from the reagent, digested to peptides, and sequenced to detect captured proteins using unbiased MS and quantified using MRM-MS (Figure 2).^{77, 78} Enrichment of the target protein after affinity pulldown can then be confirmed, and any other enriched proteins identified. In a recent study, we used this extension of iMRM to validate target capture for both cognate and endogenous proteins in all of eight tested SOMAmers,⁷² highlighting how MS-based approaches complement affinity-based methods to yield robust proteomic findings. We also observed strong correlation of aptamer-based and antibody-based reagents for selected proteins.

Coupling of affinity-based plasma proteomics to genomic tools

Coupling of proteomic information to genomic information may serve two purposes. As genetic factors influence both the structure and abundance of plasma proteins, they may therefore both influence the binding of affinity reagents and serve to validate the target specificity of affinity assays.

First, it has been well established that variants in the DNA sequence of protein-coding genes influencing the amino acid structure or processing of the expressed protein may alter epitopes and binding affinity for affinity reagents, resulting in apparent changes in protein abundance without influencing the true abundance or function of the protein. There are several published examples of this phenomenon,⁷⁹ for example the common polymorphism Val32Met (rs5063) located in the gene encoding atrial natriuretic peptide (ANP) and present in approximately 6% of subjects of European ancestry which disrupts the epitope for an immunoassay of N-terminal pro-ANP.⁸⁰ Although population-based genomic resequencing projects have established that the majority of amino acid altering variants are rare in human populations,^{81, 82} such common polymorphisms underscore the benefits of implementing population genetic information in the validation of affinity reagents.

In addition to identify affinity binding domains, genetic information can also illuminate variants that truly modify protein expression. Most genetic variation is common in the population (genetic polymorphisms) and not located in protein-coding but in intergenic regions of the genome, sometimes resulting in modulation of the expression of proximal genes when located in sequence that constitutes functional motifs such as genetic enhancers or silencers.⁸³ A genetic variant that modulates the abundance of mRNA expression of a particular gene is referred to as an expression Quantitative Trait Locus (eQTL) for that gene, and a variant that influences protein abundance is referred to as a protein Quantitative Trait Locus (pQTL). A heritable component to plasma abundance has been reported for many proteins.⁸⁴ The application of genomic tools such as genome-wide association studies (GWAS)⁸⁵ to comprehensively identify pQTLs for a protein measured with a specific affinity reagents may therefore provide information on the specificity of that reagent. In particular, the discovery in GWAS of pQTLs located near the gene encoding the measured protein (pQTL referred to as being in “cis”) suggests adequate target specificity of affinity reagents, whereas finding pQTLs in distant DNA regions or on other chromosomes (in “trans”, effects on expression mediated by indirect effects through other proteins) is less useful to validate target specificity. For example, plasma levels of lipoprotein(a) are known to be highly heritable, and genetic variation in cis has been estimated to explain up to >90% of variation.⁸⁶ Several recent studies have been undertaken to comprehensively discover pQTLs in cis and trans for many proteins, ^{84, 87–89} by performing GWAS of proteins measured by unbiased mass spectrometry,⁸⁷ parallellized immunoaffinity panels,⁸⁸ nucleotide-labeled antibody assays⁸⁴ and aptamer arrays.⁸⁹ Such studies have primarily identified cis-regulatory variants, at the genes encoding C-reactive protein, soluble interleukin-6 receptor, apolipoprotein E, γ-glutamyl transferase, haptoglobin, α-1-antitrypsin, and others. A substantial proportion of plasma protein variability was explained by sequence variants: in a study of 77 proteins measured with a PEA assay, protein abundance was heritable for 58 proteins and genetic variants robustly associated with 14 of them (of which 12 in cis), explaining up to 27% of variability for an intronic SNP in the alpha subunit of the interleukin-6 receptor.⁸⁴ For troponin and natriuretic peptide genes however, genetic variants in cis have been found to explain only a small proportion of variability, 1–2%.^{80, 90}

In addition, besides validating the specificity of affinity reagents, pQTLs in cis can also be used as instruments to infer causality of an observed protein-phenotype association. Central to this concept, termed Mendelian randomization, is that genetic sequence variants are not subject to confounding, but randomly assigned at gametogenesis following the laws of inheritance as described by Gregor Mendel and are not influenced by other traits under study. Therefore, association of a genetic variant, e g the variant conferring elevated lipoprotein(a), with a phenotype such as atherosclerotic vascular disease provides evidence of a causal link between lipoprotein(a) and vascular disease.^{91, 92} For Mendelian randomization experiments, plasma pQTL analyses are often more feasible than eQTL studies, which require access to the tissue of protein origin where the mRNA is expressed, although limited by uncertainty about tissue of origin. Such tissues may often be difficult to access, as exemplified by the liver, heart or bone marrow. On the other hand, in all pQTL studies, the potential for correlation of regulatory variants with nearby coding variants – which may influence affinity reagent binding – needs to be considered. pQTL studies also complement eQTL analyses to understand the determinants and mechanisms of gene expression, as not all pQTLs are necessarily eQTLs.⁹³ mRNA and protein expression in a tissue is often modestly correlated,^{94, 95} as protein abundance depends not only on gene transcription rate but also on mRNA stability, protein synthesis rate, and posttranslational regulation and degradation. Comparison of eQTLs and pQTLs may therefore inform our understanding of the differences between mRNA and protein regulation.

Recent progress in plasma proteomics of cardiovascular disease

The most prolific field in plasma proteomics has been cancer research, where a range of biomarkers have been identified such as prostate-specific antigen (PSA) for prostate cancer, CA-125 for ovarian cancer and carcinoembryonic antigen (CEA) for colon cancer. For cardiovascular disease, most proteomic studies to date have been based on mass spectrometric analyses in tissues and cells from experimental models. However, an increasing number of population cohorts are currently implementing commercially available aptamer microarrays (SomaLogic Inc, Boulder, CO), microbead-based multiplexed immunoassays (Luminex, Austin, TX) or the proximity extension assay (Olink, Uppsala, Sweden) to characterize plasma proteomes. A few initial studies have been published but a flurry of studies can be expected over the next few years, producing long lists of proteins for further testing.

For heart failure, a recent proteomic study based on the aptamer platform has received much interest.⁶⁸ The study used a so-called heterochronic parabiosis model, in which the circulation of a young mouse is surgically joined with that of an old mouse. Cardiac hypertrophy in the old mouse declined dramatically after four weeks of exposure to the circulation of a young mouse. By application of the aptamer platform to mouse plasma, the investigators were able to identify GDF11 as lower in old mice and increasing with heterochronic parabiosis, and also lower in older humans.⁹⁶ Furthermore, treatment with a synthetic form of GDF11 resulted in reversal of LV hypertrophy in old mice. This study highlighted the power of proteomics to systematically identify plasma proteins associated with heart disease. However, the study also highlighted the limitations in specificity inherent to affinity-based designs as subsequent studies have shown that the aptamer also binds myostatin (GDF8, sharing 90% sequence identity with GDF11) which may also contribute to the observed effects in mice.^{69, 70} Notably, a highly specific MS study observed lower levels of GDF8 but not GDF11 with aging.⁷¹

For myocardial infarction, many specific markers in well-established pathways have been studied for risk assessment and diagnosis but few large proteomic studies have been published.¹⁷ Initial work from our group has focused on diagnostic markers, using patients undergoing alcohol septal ablation for hypertrophic obstructive cardiomyopathy as a model for “planned” MI, with the benefit of uniform sampling at timepoints before and after onset of ischemia. In a study based on aptamer capture, we identified and validated 79 novel proteins that increased during 24 hours after planned MI including several non-troponin structural heart proteins but also other, several of which were validated with mass spectrometry to confirm on-target specificity. Of these 79 proteins, 25 increased within 10 minutes. Further work is ongoing to characterize the clinical relevance of these markers.⁷²

For stable coronary heart disease, a proteomic study was used to identify patients at risk of adverse outcomes based on an aptamer assay.⁹⁷ Identification of at-risk patients in this context is clinically relevant, as the benefits of an invasive strategy in the overall group have been shown to be limited. A total of 9 proteins were associated with adverse outcomes based on least absolute shrinkage and selection operator (LASSO) and stepwise regression analysis, including troponin I, matrix metalloproteinase-12 and angiopoietin-2. The 9 proteins were combined into a score which was reproducibly associated with a graded increase in risk.⁹⁷ The risk score performed better than the Framingham Risk Score, but still achieved only modest discrimination (area under the ROC curve of 0.70 compared to 0.64 for a clinical score) highlighting the complexities of clinical risk prediction.

The PEA assay is also increasingly applied to study determinants of cardiovascular disease in population cohorts. In a population-based study of elderly men, it was used to identify seven proteins associated with carotid plaque burden⁹⁸ and three proteins associated with ischemic stroke,⁹⁹ of which the latter were replicated in an independent cohort and resulted in a modest improvement in predictive accuracy.

Conclusions

The emergence of precise methods for automated, highly parallel affinity proteomics such as aptamer microarrays and proximity extension assays opens for unbiased discovery of novel biomarkers, biomarker profiles and therapeutic targets at a population-wide scale. By application of such tools to large patient cohorts, ongoing efforts aim to discover novel cardiovascular biomarkers and characterize the genetic and environmental determinants of protein profiles. Furthermore, the coupling of such tools to ultrasensitive detection systems, MS and genomics will improve the sensitivity and specificity of proteomic assays. In this review, we have outlined currently available methods in proteomics and recent progress in the field of cardiovascular medicine with such methods. These methods promise to extend the biomarker arsenal beyond troponins, natriuretic peptides and lipoprotein particles and improve clinical decision-making.