Development of protein biomarkers in cerebrospinal fluid for secondary progressive multiple sclerosis using selected reaction monitoring mass spectrometry (SRM-MS)

Multiple sclerosis (MS) is an inflammatory demyelinating disease of the CNS. During MS, myelin sheaths surrounding axons and the axons themselves are damaged as a result of chronic inflammation. Such inflammatory damage creates characteristic focal plaques in the white matter of the brain and spinal cord [1, 2]. Patients with MS initially present with a host of non-specific neurological symptoms and there is a high degree of variability in the early symptoms that patients experience. For almost all patients, however, early MS associated symptoms will occur as a series of relapses and remissions whereby a patient will remain almost entirely asymptomatic for an extended period of time between temporary symptom “relapses.” Over a variable period of time, MS will typically evolve from relapsing-remitting multiple sclerosis (RRMS) into a slowly progressive form of neurological dysfunction termed secondary progressive multiple sclerosis (SPMS) [3].

For a complex disease like MS, there is a good reason to believe that a change in the protein expression profile occurs long before clinical symptoms are established. Therefore, protein biomarkers differentially expressed in MS patients as compared to healthy individuals or patients suffering from other neurological disorders have great clinical potential - not only as early diagnostics, but as potential prognostic markers to be used for monitoring disease course and evaluating treatment efficacy [4–6]. Identification of such markers would also further our current understanding of MS disease pathology, pinpointing novel protein targets or signalling pathways for therapeutic intervention.

Mass spectrometry-based proteomic technologies have become preferred laboratory strategies for the discovery of diagnostic, prognostic, and therapeutic biomarkers [7, 8]. Since the early 1990s, a large number of potential protein biomarkers have been discovered every year in labs across the country - in both the academic and the industry sectors. However, only a handful of these identified targets are selected for further clinical investigation, fewer still have been formally validated, and of the thousands of markers that were identified between 2000 and 2005, only five new protein markers were approved by the US Food and Drug Administration (FDA) for measurement in sera or plasma in that time span [9]. Indeed, since 1998, the introduction of new protein biomarkers approved by the US FDA has fallen to an average of one per year; and this trend continues to be true today despite intensified interest and investment from both academia and industry. While to some extent, these types of statistics reflect the inherent time-line needed to conduct discovery-based research, it has long been suggested that development of faster and more efficient biomarker verification strategies could greatly reduce the time and cost of biomarker-based diagnostic development and that the greatest bottleneck to the current biomarker pipeline occurs at the verification/validation stage [10], that is the stage at which purported biomarkers, obtained through any number of discovery strategies, are tested quantitatively on samples derived from a statistically significant number of subjects and controls.

There have been two commonly adopted strategies for obtaining these kinds of quantitative assays for protein biomarkers: the enzyme-linked immunosorbent assay (ELISA) utilizing antibodies to the proteins and mass spectrometry-based methods that target the predictable fragmentation of surrogate peptides. Immunoassays enjoy high specificity and high throughput; however, they depend critically on the availability of highly specific antibodies, the development of which is usually long and costly. In addition, there are technical limitations for multiplexing of immunoassays. In contrast, quantitative approaches using selected reaction monitoring (SRM) mass spectrometry, have recently drawn intense interest to protein biomarker verification/validation for advantages of enhanced sensitivity and specificity, ease of high throughput, and relatively low cost [8, 9, 11–15]. Using a tandem mass spectrometer in which the mass analysers act as two mass filters, one for the peptide molecular ion mass and the other for a known sequence fragment, SRM-MS provides a selectivity and specificity that in many cases enables one to simplify sample preparation by avoiding the initial immunoprecipitation or fractionation common in many other mass spectrometry-based strategies.

CSF has been a valuable diagnostic resource for neurodegenerative diseases because its composition directly reflects the metabolic process of the brain. Five proteins including interleukin (IL)-17 have been recently identified by Trojanowaski and colleagues as CSF biomarkers capable of differentiating between two forms of frontotemporal lobar degeneration [16]. Similarly, amyloid beta (1–42), total tau and phosphorylated tau have been established as biomarkers for diagnosis of Alzheimer’s Disease (AD). Mattsson and colleagues found that combining biomarkers into a panel leads to a better predictive value than using individual biomarker for the diagnosis of AD [17]. Interestingly, while the three AD biomarkers identified by Mattsson, et al. [17] had diagnostic value when patient CSF was interrogated, the proteins were either not detected in the peripheral blood or their measured concentration in plasma did not provide useful information for diagnosis. Given the results from the Mattsson study and how varied MS clinical presentation and disease progression can be across patients, it is highly unlikely that a single protein biomarker will be sufficient for a conclusive diagnosis of SPMS. It is equally unlikely that a single protein would define the pathophysiology of SPMS. It is far more probable that a biomarker dependent diagnosis will require a panel of disease indicators whereby different indicators will have greater or lesser diagnostic utility depending upon individual patients or stages of disease evolution. As such, our overall objective is to find a panel of CSF protein biomarkers that show differential expression profiles amongst SPMS patients, patients suffering from non-inflammatory neurological disorders, and disease-free controls. For the studies presented here, we also included subjects from the placebo-controlled clinical trial for lamotrigine as a putative neuroprotective therapy in secondary progressive MS from the UK. Though this clinical trial did not demonstrate a significant difference in clinical outcomes between patients taking lamotrigine (treated) and the placebo group (untreated) [18], we believed it was still beneficial to investigate the expression profiles of CSF protein biomarkers in subjects of this study to determine if the lamotrigine treatment has any observable effect at protein level. Based upon previous discovery results and an up-to-date literature survey [4, 19–21], we selected a total of 26 proteins as potential CSF protein biomarkers for SPMS and developed a mass spectrometric assay for relative quantitation.

By taking the advantage of high specificity and sensitivity, we have developed a multiplex SRM assay for detecting even low abundant proteins in CSF with relatively simple sample preparation. The platform we have developed here can be easily adapted to other protein biomarker studies. The sensitivity and throughput might be further enhanced using newly developed technologies, such as the dual stage electrodynamic ion funnel interface described by Hossain et al. [22]. Although our assay successfully quantified both high abundant and low abundant proteins in CSF, several proteins which have been previously reported present in CSF were not detected during our initial method development using pooled CSF, such as nidogen-2 (NID2) and nitric oxide synthase (NOS2). Low abundance might be one of the contributing factors, though the intrinsic biological properties of the pooled CSF and the instability of peptides/proteins are also possible. Improvements in sample preparation and optimization of instrument parameters could increase the appearance of low abundant proteins in future studies.

The human proteome is more complex, compared to the genome, considering that each protein can be present in different isoforms even at the same time. However, mass spectrometry-based targeted proteomics has the advantage of specifically quantifying individual protein isoform, which could potentially help us better understand the pathogenesis underlying various diseases. Although we did not include such studies here, the intrinsic specificity of SRM provides the merit to do so if necessary in the future.

Despite CSF’s unique value for neurodegenerative diseases, there were some concerns in the process of searching for protein biomarkers in CSF, such as low protein abundance, difficulties in collecting samples, spontaneous variation through the day, etc. It is essential that CSF collections follow standard operating procedures, including but not limited to the following: all lumbar punctures were carried out at approximately the same time of day, and sample processing (e.g. time to centrifugation, centrifugation speed and time) and storage was handled in a standard way. Our study was carried out among a limited number of patients since the purpose of our study was to validate the feasibility of using SRM for protein biomarker verification. In the future, validation studies on large cohorts of samples originating from healthy controls, relapse-remitting MS (RRMS), SPMS, primary-progressive MS (PPMS), NIND and other inflammatory neurological disease (OIND) controls will determine clinical value of this pilot combinatorial biomarker.

Using SRM to target a specific set of surrogate biomarker peptides, we established a highly sensitive and specific multiplex assay to simultaneously detect twenty-eight potential protein biomarkers which might be involved in the pathology of SPMS. We quantified the relative levels of each targeted protein using an internal standard horse apomyoglobin and compared their expression profiles among three different cohorts: SPMS, NINDS and healthy controls. ANOVA indicated one protein biomarker- showing significantly different expression between SPMS and HC, and three protein biomarkers showing significantly different expression between SPMS and NIND. It is highly possible that these four proteins play a role in the pathophysiology of SPMS and could potentially benefit the diagnosis, prognosis and/or the development of newer generation treatments. In addition, two proteins were expressed differently between NIND and healthy controls, which could potentially benefit NIND studies. In the lamotrigine trial, although the trial itself came negative with respect to the efficacy of the drug therapy, our analysis suggested that eleven proteins in the treated group had significantly lower expression compared to the untreated group. Since lamotrigine therapy had an effect of decreasing expression for these proteins, it is plausible that these proteins are correlated with other proteins. In fact, our principal component analysis (PCA) revealed that these 26 proteins were correlated, and could be represented by four principal components.Overall, we have established a straightforward mass spectrometry-based platform for CSF protein biomarker development and verification. This platform can be easily adapted to study a broad range of proteins, especially feasible for those of interest to neurodegenerative diseases.