Biomarkers of Myeloma-Associated Lytic Bone Disease Using SELDI-TOF MS.

Abstract MGUS is a pre-malignant state, with a 1% annual risk of progression to overt myeloma (Kyle, NEJM346:564, 2004). Progression of MGUS to myeloma is preceded and can be predicted by increased rates of bone turnover (Betaille, JCI88:62, 1991). Progress to overt myeloma is associated with lytic bone disease in about 80% of patients. A simple, reliable test that identifies MGUS patients at risk of progression to symptomatic myeloma would elucidate biological mechanisms that govern the process and identify targets for early intervention. The intimate association of progression with changes in bone biology suggested that early recognition of changes in bone turnover would serve such a purpose. Therefore, we sought to identify protein biomarkers of bone disease in the sera of patients with myeloma that can be used as early predictors of progression. Given the limited reliability of individual markers of bone turnover to identify changes in bone metabolism, we elected to adopt a global proteomics approach. Sera from 62 untreated myeloma patients were collected and stored at −80°C for future analysis. 35 serum samples were from myeloma patients with 1–26 lytic bone lesions as identified on X-ray skeletal surveys and 27 serum samples were from patients without lytic lesions. The sera were analyzed by surface-enhanced laser desorption and ionization-time of flight mass spectroscopy (SELDI-TOF MS) using Ciphergen’s ProteinChip Biology System II (PBS II), to identify protein patterns associated with lytic bone disease. Each sample was applied in 4 replicates to randomly assigned spots on 12 IMAC30 ProteinChips placed in a bioprocessor and activated with Cu++ ions using a Biomek2000 robot. The chips were read on a PBS II reader. The mass spectra of proteins, generated using an average of 66 laser shots, were calibrated using peptide and protein standards and normalized to total ion current using CiphergenExpress 2.0 software. To develop a classification model that separates non-treated patients with focal lesions from those without focal lesions, we identified among the low molecular weight (1500–25000 kDa) proteins a set of 17 protein peaks that were differentially expressed between the two groups at a significance level of p<0.005. We then randomly selected 80% of patients for developing a model based on these peaks using a stepwise logistic regression method; 198 calibrated and normalized spectra from 50 randomly selected patients,28 with lytic bone lesions and 22 without lytic lesions (2 samples had only 3 replicates) were used for model development. The model used 10 protein peaks to classify the patients, with a receiver operating characteristic (ROC) area of 0.897. Finally the model was challenged by the set-aside test set of 50 spectra from 12 randomly selected patients (one patient had 6 replicates). The model exhibits high sensitivity and specificity, and its overall prediction accuracy is around 90%. Studies are underway to identify the biomarkers and to validate the utility of this model to predicting progression of MGUS to overt myeloma.