Alterations in a Panel of Serum Peptides Correlate with the Disease States of Multiple Myeloma

Background: Multiple myeloma (MM) is prone to misdiagnosis and missed diagnosis in early stage due to heterogeneity of clinical manifestations. Previously, we constructed the MM Supervised Neural Network Algorithm (SNN) diagnostic model based on four peptides through proteomic profiling. The current study is to validate the efficacy of the diagnostic model in a larger size of sample, to verify the identity of proteins the four peptides represent, and to characterize the expression patterns of these proteins in different disease states of MM.
Methods: Patients hospitalized in the Second Affiliated Hospital of Xi'an Jiaotong University from January 2010 to December 2011 were enrolled for the study, including 62 Newly diagnosed(ND) MM (age range 44-73, median age 56.5 years old, male/female 34/28), 38 complete remission and very good partial remission (CR&VGPR) MM (age range 44-71, median age 54.6 years old, male/female 21/17) and 43 refractory & relapsed (RR) MM (age range 45-73, median age 57.5 years old, male/female 27/16) after two courses of chemotherapy. 64 healthy controls (HC) (age range 45-70, median age 57.5 years old, male/female 34/30) were recruited from those who came to our hospital for physical examination. No abnormal symptoms or test results were found in HC. Weak cation exchange magnetic beads combined with MALDI-TOF-MS were to analyze serum peptide profiling of MM in different disease groups. We utilized a nano-LC/ESI-MS/MS system to perform peptide sequencing and peptide identification. Immunoblotting and ELISA were performed for validation. ANOVA test was used to analyze differences in different groups. Correlation analysis was performed using linear regression model. Statistical significance was defined as p<0.05 with assumption of equally deviation and two-sided distribution. For multiple comparisons among the four groups, significant level was adjusted to 0.0083(0.05/4(4-1)/2).
Results: 58/62 ND cases of MM were correctly assigned to MM group (93.55% sensitivity), while 59 HC were correctly identified as the non-diseased (92.19% specificity). The relative intensities of peptides with molecular weight of 2660.65, 2900.4 and 3315.96 Da were increased in the ND and RR groups when comparing with the CR&VGPR group and HC group(p=0.0000505; p<0.000001; p<0.000001), but there was no significant difference between the two groups for any of the peptides (p=0.14; p=0.25; p=0.21). No significant difference was found between the CR & VGPR group and HC group (p=0.52; p=0.24; p=0.15).The relative intensities of peptide with molecular weight of 7763.24 Da were reduced in the ND group and RR group when comparing with the CR&VGPR group and HC group (p<0.000001), and the two groups had no significant difference (p=0.37). No significant difference was observed between the HC group and CR&VGPR group (p=0.46). The four peptides were identified as fragments of fibrinogen alpha chain (FGA), dihydropyrimidinase- like 2, alpha-fetoprotein and PF4, respectively. Western blot results showed that levels of the FGA and dihydropyrimidinase-like 2 differ between the MM and normal cells (p=0.0034, p=0.0067). No or weak band representing PF4 was seen in ND and RR MM cases. Quantification of bands from western blot analysis showed a significantly decreased level of PF4 in MM-ND and RR samples, relative to age-matched MM-CR&VGPR and HC samples (p=0.0031). The contents of PF4 in 62 ND MM (0.67± 0.18μg/L) was significantly lower compared to 64 HC (1.43 ± 0.77μg/L) (p=1.86 E-6).The FGA contents of ND MM (276.78 ± 194.75μg/L) and HC (30.44 ± 22.15μg/L) differed significantly (p=1.82E-12). The serum VEGF contents of ND MM (669.69 ± 137.81ng/L) were significantly higher compared to HC (120.75 ± 36.96ng/L) (p=7.28E-29). No significant difference of PF4 content was found between thrombocytopenia group (0.65 ± 0.18μg/L) and platelet normal group (0.69 ± 0.17μg/L) in MM (p=0.55). Linear regression analysis showed no correlation between PF4 content and platelet count (correlation coefficient=0.18, p=0.17), whereas serum PF4 and serum VEGF were negatively correlated (correlation coefficient=-0.96, p=1.11E-14).
Conclusion: The MM SNN model has provided encouraging efficiency in discriminating MM from HCs. The close correlation of the four peptides with MM disease states pointed to their potential value in monitoring MM treatment response.