NEW YORK (GenomeWeb News) – With the recent publication of a paper detailing the company's mass spec process, researchers from Integrated Diagnostics have provided an inside look at what is the first multiplexed multiple-reaction monitoring mass spec proteomics assay to make it into the clinic.
Detailed in a study published in January in Clinical Proteomics, the approach features a relatively novel twist for an MRM mass spec-based clinical proteomics assay, using not only traditional stable isotope-labeled peptide standards to control for assay variation, but also a series of endogenous protein standards.
Combining the endogenous protein standards with SIL peptides allows Indi to control for both the workflow's analytical variability and pre-analytical variation caused by, for instance, differences at sample collection sites, Paul Kearney, senior author on the paper and Indi's president and CSO, told GenomeWeb.
Proteomics researchers and test developers have long viewed mass spec, and triple quadrupole-based MRM-MS in particular, as a platform potentially well suited to clinical use. Compared to immunoassays, mass spec offers advantages including improved accuracy and, with the advent of large protein multiplexes, lower costs.
Mass spec-based proteomics has faced a number of technical obstacles on its way to the clinic, including issues of reproducibility, sensitivity, and throughput.
Nonetheless, in recent years, a number of researchers/vendors/etc have made strides toward solving thes issues. For instance, in November 2012, researchers from SISCAPA Assay Technologies and Agilent published on a MRM-MS-based peptide quantitation workflow with a sample cycle time as short as seven seconds.
The field's advances culminated in Indi's launch in 2013 of its mass spec-based Xpresys Lung test. Intended to aid doctors in identifying lung nodules detected via CT scans as likely benign, Xpresys uses MRM-MS to quantify the levels of 11 proteins in patient blood samples. As the first multiplexed proteomic product to hit the market using MRM-MS, the test was something of a milestone.
Mass spec-based clinical proteomics is still in its early days, however. And, as MRM-MS proteomic tests like Xpresys and others under development at companies such as Applied Proteomics and Sera Prognostics move towards commercialization, the field has drawn increased scrutiny from the clinical chemistry world.
As SISCAPA Assay Technologies CEO Leigh Anderson told GenomeWeb at the 2014 Mass Spectrometry Applications to the Clinical Laboratory annual meeting, "all of this original proteomics research has moved enough in the direction of [clinical chemistry] that [clinical chemists] are now asking questions that they care about, and not just saying 'OK guys, come back to me when this is really working.'"
A main area of concern with mass spec-based proteomic tests is how best to ensure that their measurements are accurate and reproducible, particularly when ported across different sites and applied to samples collected at diverse locations.
One of the primary issues considered with regard to this question is the appropriate use of standards – specifically, what sort of standards to use and when in the workflow to add them.
Typically, mass spec-based targeted quantitation assays use as standards either SIL peptides or full-length SIL proteins. While SIL peptides are more commonly used due in large part to the relative expense and difficulty of obtaining full-length SIL protein standards, a number of recent studies have suggested that SIL proteins allow for more reproducible and more accurate assays.
For instance, in a study published last year in the Journal of Proteome Research, a team from Leiden University Medical Center found that use of SIL proteins was key to reducing inaccuracy introduced by trypsin digestion.
This is a problem inherent in targeted proteomic workflows wherein peptide levels are used as proxies for whole protein levels. Because the measured peptides are created by digesting the protein of interest prior to mass spec analysis, differences in digestion efficiency that lead to differences in the amounts of peptides released can affect an assay's accuracy.
Because SIL peptide standards don't undergo digestion, they can't account for these effects. By using full-length SIL protein standards, however, researchers can identify and correct for biases in trypsin digestion, Irene van den Broek, a LUMC researcher and first author on the JPR paper, told GenomeWeb in an interview following its publication.
A study published last month by researchers from the National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium similarly found that use of full-length labeled proteins significantly improved the accuracy and reproducibility of MRM-MS assays.
Neither SIL proteins nor SIL peptides, however, control for the pre-analytical variables that are perhaps the most significant contributors of variation in a commercial test like Xpresys, Kearneysaid.
"When running a commercial test, samples are coming from all over the country, possibly the world," he said. "And so the major source of variability is pre-analytic variability."
"Protein standards that people use can help control for variations that occur during sample preparation, and labeled peptide [standards] can control for variation after [digestion], but those don't control for variability from individual to individual, from site to site," he added. "And anyone who has really had experience in managing samples that come from many different sites knows that that is where the real variation in samples occurs."
To address this issue, Indi turned to endogenous protein standards. Of the 11 proteins the Xpresys test measures, only five actually have diagnostic value. The remaining six are used as normalizers, which, as the Clinical Proteomics authors wrote, are "processed and analyzed together with the target proteins of interest" and so should effectively "serve as monitors for systematic variation in both pre-analytical and analytical procedures."
Combining use of these six endogenous normalizers with the addition of SIL peptide standards post-digestion to control for variation in desalting and mass spec analysis, the Indi researchers found they were able to achieve a median CV of all proteins measured of 9.3 percent, compared to 13.3 percent using SIL peptides alone.
Van den Broek, who was not involved in the Indi research, told GenomeWeb this week that the method is "a very elegant approach" and is the first effort she was aware of to combine SIL peptide standards and endogenous normalizer proteins in this way.
The endogenous proteins, she said, served much the same function as the full-length proteins she and her colleagues use in their targeted proteomics work but with the bonus of controlling for pre-analytical variation, as well.
However, she noted, with full-length protein standards, the researchers can be sure that their standards in fact reflect the protein-specific variation of corresponding endogenous targets. The Indi approach, on the other hand, relies on careful selection of normalizers whose behavior can be used to control for that of the assay's actual diagnostic analytes.
"That is a key question," she said. "How did they select these six proteins?"
Essentially,Kearney said, it was done empirically, by measuring large numbers of proteins and looking for those that did the best job of reducing variation in control samples.
"If there are five proteins that you are measuring for diagnostic value, at the same time you measure a whole bunch of other proteins, and if there is a collection of, say, four that do the best job of minimizing variation then you use those four," he said. "It's a very straightforward, practical way of finding the set of endogenous normalizers that work the best."
In development of the Xpresys Lung test, Indi researchers identified the six normalizer proteins from a pool of 371 candidates.
Also key,Kearney noted, is selecting normalizer proteins that are affected by sample prep steps like depletion and digestion in the same way as the test's diagnostic proteins.
However, he added, the company in significant part tackled the problem of analytic variability simply by avoiding it.
"With the diagnostic proteins we have in our assay, we looked to see how much they varied, and those that had a lot of analytical variation whether due to depletion or digestion or anything, we just didn't include in our assay, rather than trying to make peptide digestion better than it is,"Kearney said.
"[Digestion] does affect different proteins differently, so we picked the proteins that were less affected," he added. "We looked at 377 proteins and how they varied when we modified the depletion column, how they varied when we varied our digestion parameters. So you find ones that are very sensitive to [these steps], and you just eliminate them from consideration."