A Pilot Application of SELDI Serum Proteomics in Bone Marrow Failure Syndromes.

Abstract

Clinically, bone marrow failure syndromes show significant overlap. Newer analytic approaches may lead to improved diagnosis and better understanding of underlying pathophysiology. In contrast to traditional biomarker diagnostic principles, proteomic technologies such as surface-enhanced laser desorption/ionization (SELDI) method enable screening of protein patterns without knowledge of the identity of targets. The recent refinement of SELDI mass spectrometry has allowed for the early diagnosis of occult solid tumors. We explored the potential application of SELDI proteomics to the study of bone marrow failure states. An extensive sample library included 28 Aplastic anemia, 129 Myelodysplastic syndrome (MDS), 28 Paroxysmal Nocturnal Hemoglobinuria (PNH) patients and 36 control specimens. Within the MDS cohort, subgroups with RA (N=77), RARS (N=15) and RAEB/t (N=37) were studied. Analyses were also performed for patients with specific cytogenetic abnormalities, including those with 5q-syndrome (N=21), trisomy-8 (N=15), aberrations of chromosome 7 (N=15) and compared to MDS patients with normal cytogenetics (N=57). In preliminary experiments, H4 chips were selected for further study; spectra with >80 peaks were considered as informative. Initially, our analysis included learning sets in which 20 randomly selected samples from individual categories were compared to each other in order to identify specific peptide peaks. Subsequently, the results obtained in the learning set were confirmed using separate validation sample sets. A scoring system was applied for analysis in which signal/noise ratio was a sole criterion for peak recognition and assignment of non-parametric scores. Reproducibility was assured by repeating measurements of standard plasma sample and multiple determinations.When AA spectra were compared to controls, AA could be distinguished with 87% sensitivity and 85% specificity based on the presence of a single peak at M/Z value of 3.095kDa. When compared to RA, AA was discriminated with 95% specificity and 80% sensitivity (1.02kDa). Globally, we devised a decision tree (using 3 different peaks in 6 steps of separation) that allows for the plasma-based discrimination of MDS and controls. Within MDS group, RA could be further differentiated from a more advanced MDS with 95% specificity and 90% sensitivity. A single peak at M/Z value of 3.88kDa distinguished RA from RAEB/t. Similar learning data sets performed for RARS patient can be obtained. Not overlapping sample sets are used for validation of obtained results. Finally, samples with atypical spectra can be investigated for features that may prompt diagnostic reassignment; e.g., in patients with hypocellular MDS to AA, and PNH to AA/PNH syndrome. Most intriguing is the possibility that specific cytogenetic defects may result in proteomic patterns that may provide important clinical clues. A large sample collection allows for subanalyses based on karyotype: for example, data and result sets exist for patients with 5q-, trisomy-8 and other more common chromosomal changes.Our ongoing study demonstrates the feasibility of SELDI technology in detecting diagnostic protein signatures for individual bone marrow failure states. In the future, proteomic biomarker pattern-based analysis may allow for more precise definition and supplement the traditional diagnostic approaches.