Quantitative Clinical Proteomics by Liquid Chromatography–Tandem Mass Spectrometry: Assessing the Platform

Clinical laboratories use electrospray ionization and mass spectrometry (MS)1 every day to measure dozens of different small molecule analytes. The capacity to precisely quantify these biomarkers has improved patient care. To build on the successful implementation of novel MS methodologies, many clinical mass spectrometrists hope to expand the list of measurable analytes to include peptides and proteins, several examples of which have been reported and reviewed (1). One can also envision the translation of novel biomarkers from the research-grade MS assays, with which they are discovered, to clinically acceptable MS assays for diagnosis, prognosis, and evaluation of therapeutic efficacy. An argument can be made that investigators should use clinically acceptable MS assays from the outset (2), which would facilitate the transition from research/discovery to clinical practice. Given the successes over the last few years in precisely measuring proteins by MS in clinically relevant matrices, this outcome may not be too much to ask. Traditionally, the word proteomics has encompassed efforts aimed at simultaneously measuring many proteins in a given complex sample. Shotgun proteomic experiments, which attempt to identify all of the proteins in a complex sample, and targeted proteomic experiments, which attempt to quantify a subset of the proteins known to exist in a complex sample, can both be carried out using liquid chromatography–tandem MS (LC-MS/MS). Shotgun proteomics is already being used clinically to identify the proteinaceous component of amyloid deposits (3). For the quantification of multiple protein concentrations in a complex sample, however, targeted methods are more precise because data-dependent shotgun proteomic methods variably sample fewer ions across chromatographic peaks. For this reason, targeted approaches are much more likely to be employed in the quantification of proteins in clinical specimens. Targeted methods typically use proteolysis with trypsin to digest proteins into peptides, which are more amenable to analysis by MS. These peptides can be quantified by use of selected reaction monitoring, which specifically selects the mass-to-charge ratio (m/z) of an enzyme-generated peptide and then detects a specific fragment ion of the peptide generated within the mass spectrometer. The specific pair of the mass-to-charge ratio of the enzyme-generated peptide and the mass-to-charge ratio of the specific fragment ion is termed a transition. Reduced imprecision of quantification in targeted methods can be achieved by using isotope dilution, in which chromatographic peak areas of endogenous peptide-specific transitions are normalized to peak areas from stable-isotope–labeled internal standard peptide-specific transitions. Whether shotgun or targeted, proteomic experiments have contributed substantially to basic science research. However, no proteomic experiments thus far have given the clinical community what it so desperately wants: a new and better biomarker that improves patient care. Even so, clinical chemists remain hopeful that one day proteomics will have a major impact on the clinical laboratory. Quantitative clinical proteomics, or the accurate and precise multiplexed measurement of proteins in clinical samples, is alive and well on several different bead-based immunoassay platforms. The move to an LC-MS/MS–based multiplexed protein assay platform in the clinical laboratory requires that a number of technical milestones be met in regard to the imprecision and accuracy necessary for clinical utility. The recent report by Addona et al. in Nature Biotechnology (4) provides preliminary multilaboratory benchmark data for the measurement of spiked exogenous proteins in a natural human plasma protein matrix, an important prelude to the measurement of endogenous proteins in clinical samples. Eight independent sites participated in the 3-part study of Addona et al. (4). In study I, 11 synthetic peptides were spiked into a tryptic plasma digest and sent to each site for analysis. This experiment tested the reproducibility of detecting known amounts of peptides in a complex digest. Further assessing imprecision, study II used digested proteins spiked into a tryptic plasma digest. The samples for the first 2 experiments were formulated centrally at the National Institute of Standards and Technology and distributed for MS analysis. In study III, sites received plasma spiked with exogenous proteins, which were digested at each site according to a standard protocol and analyzed by use of selected reaction monitoring LC-MS/MS. In all 3 experiments, linear dilutions of analytes were used to determine sensitivity, limits of detection, and recovery. The results showed that LC-MS/MS of peptides might be generalizable across laboratories. In studies I and II, in which the plasma and protein digestions were done centrally, median intralaboratory CVs across the linear dilutions were all substantially <20% for the 10 detectable peptides measured at each of the 8 sites. Importantly, the interlaboratory CVs were calculated unconventionally, with CV defined as the median intralaboratory CV (across 8 sites) of 4 (studies I and II) or 12 (study III) replicate injections of each peptide near the limit of quantification. However, the authors posted their raw data on a publicly accessible website, making it possible to recalculate interlaboratory imprecision for each peptide at 2 different concentrations (Table 1 ). The results were compelling. For the 10 detectable peptides in study I, the interlaboratory CV was still <20% near the limit of detection for their assay (2.9 nmol/L). Interlaboratory CVs from a multisite, 3-part study of quantitative proteomics using LC-MS/MS.1 Interlaboratory CVs (expressed as percentages) were recalculated from the raw data supplied by the authors for 7 of the 8 sites (www.proteomecommons.org). Reliable data for study II were available for only 5 sites and were therefore not included in the analysis. The overall CV was calculated as the SD of the measured peptide concentrations divided by the mean concentration determined across all sites for each of the peptides at 2 different concentrations of analyte. For study I, the measured concentrations of peptide at each site were calculated as the mean concentrations of peptide determined from 4 replicate injections. For study III, the measured concentrations of peptide at each site were calculated as the mean concentrations of peptide determined from 3 replicate digestions, each of which was injected 4 times. The interlaboratory imprecision of measuring a specimen by use of a single digestion and single injection, which might be expected to more closely predict performance in a clinical laboratory, was calculated from the between-measurement variability across all injections of the 3 digests in study III across all sites (N = 84). Peptide YLASASTMDHAR was not reliably detected in study III and was excluded from the analysis. Interlaboratory CVs from a multisite, 3-part study of quantitative proteomics using LC-MS/MS.1 Interlaboratory CVs (expressed as percentages) were recalculated from the raw data supplied by the authors for 7 of the 8 sites (www.proteomecommons.org). Reliable data for study II were available for only 5 sites and were therefore not included in the analysis. The overall CV was calculated as the SD of the measured peptide concentrations divided by the mean concentration determined across all sites for each of the peptides at 2 different concentrations of analyte. For study I, the measured concentrations of peptide at each site were calculated as the mean concentrations of peptide determined from 4 replicate injections. For study III, the measured concentrations of peptide at each site were calculated as the mean concentrations of peptide determined from 3 replicate digestions, each of which was injected 4 times. The interlaboratory imprecision of measuring a specimen by use of a single digestion and single injection, which might be expected to more closely predict performance in a clinical laboratory, was calculated from the between-measurement variability across all injections of the 3 digests in study III across all sites (N = 84). Peptide YLASASTMDHAR was not reliably detected in study III and was excluded from the analysis. With digestions performed independently at each site, study III identified the important contribution of the variability of the digestion step to the overall variability of the assay. The reported median intralaboratory CVs across the linear dilution increased compared to study I for the 9 detectable peptides, ranging from 7% to 55%. More importantly, as shown in Table 1, the interlaboratory imprecision at the limit of detection increased for 5 of 9 peptides to above 20% for samples digested in triplicate and measured by LC-MS/MS in quadruplicate. This unexpected variability was due to between-site variability of peptide recovery during protein digestion. When interlaboratory imprecision was estimated for a single digestion and single injection, similar to what might happen in clinical practice, imprecision ranged from 22% to 60%. Indeed, even with a detailed standard operating procedure in place, assay behavior relied heavily on the location performing the digestion, highlighting the considerable work investigators still must do before LC-MS/MS protein assays will be considered reference methods. Study III also provided data to suggest the QC parameters that will likely be necessary to clinically implement LC-MS/MS assays for proteins. Many of these parameters were developed from experience with small molecules and are an integral part of current clinical practice. The use of internal standard peptides that are chemically identical to the endogenous peptides but have different mass can help avoid 2 important sources of error. First, it is possible to normalize signal changes due to ion suppression, which is caused by nonanalyte ions coeluting during the chromatographic run and competing for charge at the surface of droplets (5). Given the interindividual heterogeneity of plasma, this process will be important. Second, it is possible to normalize for variable mass spectrometer performance over time. The promising results obtained by Addona, et al. (4) in their study I were attributable to the inclusion of internal standard peptide. The analytical recovery of most peptides in study III was well below 100% (median 47.3%; range 0%–93%), greatly compromising the ability of the assay to provide absolute quantification of proteins in plasma. A calibration curve made in an appropriate matrix and included with each batch of samples would permit the signal of each peptide in each sample to be related to an actual concentration in a multiplexed fashion. Because incomplete digestion due to the high concentration of matrix proteins likely played a large role in the reduced yield of peptide in study III, it may also be important to use an exogenously spiked protein in each sample to gauge and perhaps normalize digestion efficiency. Spiking a properly folded, stable isotope–labeled version of the protein could have the added benefit of serving as a source of internal standard peptides and potentially as an internal calibrator for concentration determination. These approaches may have substantially reduced the observed interlaboratory variability in study III. Three transitions were monitored for all 10 analyte peptides and all 10 internal standard peptides. This process turned out to be extremely advantageous because the authors were able to identify and exclude transitions of several peptides that were not linear across dilutions at 1 or more sites. This lack of linearity was caused by interference in either the analyte or the internal standard channel and came from solvents, reagents, or the samples themselves. Interestingly, several interferences were site specific, making it important to monitor 1 or more qualifying ions, and to use ion ratios as QC parameters to detect nonspecificity in individual samples. Quantitative clinical proteomics has the potential to make substantial contributions to laboratory medicine. So much is expected from the powerful methodology embodied in LC-MS/MS: hopefully, it will reduce the cost and limitations of immunoassays (1) and will one day deliver novel protein biomarkers or panels of biomarkers to improve clinical care. The meticulous development of robust assay procedures and QC parameters based on the results of collaborative multisite clinical studies similar to that discussed herein (4) will provide the necessary data for clinically relevant implementation of this novel methodology. Clearly, future studies aimed at improving the interlaboratory reproducibility of the proteolytic digestion of plasma and practical methods for monitoring that digestion are imperative. Nonstandard abbreviations: MS, mass spectrometry; LC-MS/MS, liquid chromatography–tandem MS.