N-glycosylation of serum proteins for the assessment of patients with IgD multiple myeloma

Representative serum samples from MM patients recruited during a 2-year period (March 2010 to April 2012) at Shanghai Jingan District Zhabei Central Hospital were studied, excluding those with the IgG or IgA subtype. Additionally, serum samples from healthy donors who were deemed free of diseases after a physical examination were collected. The 61 MM patients consisted of 20 IgD MM patients and 41 LCMM patients, and 42 CTRs were evaluated. All MM patients were defined based on the International Myeloma Working Group (IMWG) criteria 2003 [Clonal bone marrow plasma cells ≥10%, presence of serum or urinary monoclonal protein (except in patients with true non-secretory MM), and evidence of end-organ damage that can be attributed to the underlying plasma cell proliferative disorder] and were staged according to the international staging system (ISS) and Durie–Salmon staging system after a confirmed diagnosis [22, 23]. Both newly diagnosed MM patients and confirmed MM patients who were undergoing routine chemotherapy were included in the study. MM patients who had serious infectious diseases, acute or chronic inflammatory diseases, history of another malignant cancer besides MM, and drug abuse were excluded. The serum samples from newly diagnosed and treated patients were obtained before chemotherapy initiation and before the next chemotherapy session, respectively. Based on a computerized database and medical records of all patients seen at the hospital, the progression-free survival (PFS) and overall survival (OS) of 20 patients with IgD MM were monitored. Blood was collected using a standard protocol and serum samples were separated by centrifugation at 3000 rpm for 10 min, followed by storage at −80 °C until analysis. Nine patients were newly diagnosed and 11 were already treated in the IgD MM group. All the patients were already symptomatic or became symptomatic during follow-up. Next, the symptomatic patients were treated with conventional modalities. The median follow-up for IgD MM patients was 31.5 months.

The haematological index of the haemoglobin (Hb) level was determined using an automatic cell counter and associated reagent (Sysmex XZ-2100D Cell Counter; Sysmex, Kobe, Japan; Sysmex diagnostic reagents). Biochemical indexes, such as total protein (TP), albumin (ALB), urea (BUN) and creatinine (CREA) were measured using an automatic chemical analyser and associated reagent (ADVIA 2400 Analyzer; Siemens AG, Munich, Germany; Siemens diagnostic reagents). SPE and IFE were run using a semi-auto electrophoresis system (Sebia HYDRASYS2 electrophoresis system; Sebia, Tours, France). Serum and urine light chain levels were analysed on a Dade Behring BNII nephelometer (Siemens AG; Siemens diagnostic reagents).

Clinical features, such as bone lesion and extramedullary infiltration were documented from the medical records of patients with IgD MM.

PFS was calculated from the start of the first treatment to disease progression or death from any cause, or the date on which the patient was last known to be in remission. OS was calculated from the start of first treatment to the date of death or the date on which the patient was last known to be alive. For the analysis of the correlation between N-glycan abundance and survival, PFS or OS, the patients were divided into two groups: one group had a specific N-glycan structure level less than the median level, and the other group had a specific N-glycan structure level above the median level.

The N-glycans present on the serum proteins were analysed using DSA-FACE technology as described previously [12, 24]. Briefly, the N-linked glycans were denatured and released from serum glycoproteins by adding the peptide N-glycosidase-F (PNGaseF) (New England Biolabs, Boston, MA) in 2 μl of serum. Thereafter, N-glycans were labelled with APTS (8-aminonaphtalene-1, 3, 6-trisulphonic acid) (Invitrogen, Carlsbad, CA). Sialic acid was removed using Arthrobacter ureafaciens sialidase (Roche Bioscience, Palo Alto, CA) and the processed samples were analysed using a capillary electrophoresis-based ABI3500 Genetic Analyzer (Applied Biosystems, Foster city, CA). Twelve obvious N-glycan peaks detected in all serum samples were analysed using the GeneMapper v4.1 software (Applied Biosystems). The abundance of each N-glycan peak was described by normalizing its height to the sum of the heights of all 12 peaks. Before applying DSA-FACE method to discover N-glycan biomarkers for IgD myeloma, we have evaluated the feasibility of this method, such as test reproducibility. The coefficient of variation (CV) value of run-to-run was less than 15% for each N-glycan marker. In addition, a pooled serum was aliquoted as the standard sample, and it was analyzed in each experiment to ensure the stability of the system and the reliability of the results (Additional file 1: Table S1).

All quantitative variables are expressed as means ± standard deviation (SD). Quantitative variables were compared using Student’s t test, ANOVA or nonparametric tests. After one-way ANOVA, the LSD (least significance difference) test was applied for pair-wise comparison between the three groups. Pearson’s coefficients of correlation and their associated probabilities (P) were used to evaluate the relationship between the peak values and other independent parameters. The diagnostic or differentially diagnostic performance of a single marker was evaluated by receiver operating characteristic (ROC) curve analysis and the area under the curve (AUC). The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated using cut-off values optimally selected based on the ROC curves. Survival curves were estimated according to the Kaplan-Meier method. Differences between survival curves were tested for statistical significance using the two-sided log-rank test. All reported P values were 2-tailed, and P values <0.05 were evaluated as statistically significant. Statistical analyses were performed using SPSS16.0 for Windows software (SPSS, Chicago, IL).

Twenty patients with IgD MM were identified, and the patient characteristics are presented in Table 1.

As mentioned above, 12 N-glycan peaks were detected in all the serum samples (Fig. 1). The corresponding structure of each N-glycan peak was previously determined by Liu et al. and Bunz SC et al. [25‐27]. The average abundance of the 12 N-glycan peaks in the IgD MM, LCMM and healthy CTR groups is shown in Table 2.

The levels of three complex-type biantennary N-glycans, the agalactosylated NGA2F, α(1,6)-arm monogalactosylated NG1(6)A2F and α(1,3)-arm monogalactosylated NG1(3)A2F (identified as peaks 1, 3 and 4, respectively) were significantly lower in the IgD MM patients than in the CTRs. Moreover, the levels of bigalactosylated biantennary glycan NA2 (peak5), together with all the triantennary and tetra-antennary glycans NA3, NA3Fb, NA3F2, NA4 and NA4Fb (defined as peaks 8, 9, 10, 11 and 12, respectively), were significantly higher in the IgD MM patients than in the CTRs.

Next, we compared the IgD MM and LCMM groups because their serum N-glycomic profiles were similar to those of the CTRs. The 2 groups showed low levels of partial biantennary structures and high levels of NA2, triantennary and tetra-antennary glycan structures. The increase in multi-antennary peaks may affect the decreased concentration of IgG present in high abundance in the human circulation and the peaks represent predominantly biantennary complex-type structures. Notably, the significant decrease in NG1(3)A2F levels and prominent increase in NA2FB (peak7) levels were more evident in patients with IgD MM than in those with LCMM. In addition, none of the N-glycan peaks were significantly different between the newly diagnosed and treated IgD MM patients (Additional file 2: Table S2), and there were no differences between the kappa and lambda chain subjects with IgD MM because of the predominance of lambda chain (accounting for 95%).

Next, we evaluated the association between the levels of serum N-glycomic structures and different tumour stages. For this purpose, 4 ISS stage I, 5 ISS stage II and 11 ISS stage III IgD MM patients were recruited; additionally, all the patients with IgD MM were diagnosed with Durie Salmon stage III (including 16 IIIA and 4 IIIB patients). NG1(3)A2F levels differed significantly between ISS stage I and ISS stage III patients (P = 0.011, Fig. 2a). Moreover, the level of NA3F2 was significantly higher in patients with Durie Salmon stage IIIB than in those with stage IIIA (P = 0.036, Fig. 2b). Therefore, the lower trend of the NG1(3)A2F level was related to a higher ISS stage, and a higher level of NA3F2 was associated with a greater probability of Durie Salom stage IIIA to IIIB.

Until now, SPE and IFE were the preferred methods of detection of monoclonal proteins in IgD MM patients. Therefore, we compared the levels of N-glycan peaks between the negative and positive groups. We found that, for SPE and IFE, the levels of NA2FB were significantly higher in the positive group compared with the negative group (P = 0.001 and P = 0.036, respectively, Fig. 3). Moreover, other laboratory parameters, namely, haemoglobulin (Hb), platelets (PLT), serum/urine light chain ratio, urea (BUN), creatinine (CREA), total protein (TP) and albumin (ALB) were collected. The decreased N-glycan peaks in the IgD MM patients presented a negative association with the serum or urine light chain ratios and a positive association with Hb, such as NGA2F versus the urine light chain ratio; NG1(6)A2F or NG1(3)A2F versus Hb and the serum/urine light chain ratio (Table 3). Additionally, increased N-glycan peaks showed a negative correlation with TP and ALB, e.g., NA2 or NA4 versus TP and NA3F2 versus ALB (Table 3).

To further evaluate the potential clinical usefulness of these 9 N-glycan constructions as diagnostic serum markers, we employed ROC curve analysis to assess these N-glycan markers. As shown in Fig. 4, NG1(6)A2F and NG1(3)A2F presented the best diagnostic performance, with AUCs of 0.981 (95% CI 0.949–1.013) and 0.936 (95% CI 0.869–1.003). The optimal cut-off value for NG1(6)A2F was 4.819 (sensitivity 95.0%, specificity 95.2%, accuracy 95.1%), and that for NG1(3)A2F was 4.640 (sensitivity 95.0%, specificity 78.6%, accuracy 86.8%) (Table 4). In this study, 7 SPE negative and 2 IFE negative IgD MM patients could be diagnosed properly by N-glycan markers NG1(6)A2F and NG1(3)A2F (Additional file 3: Table S3).F.

Because most of the IgG MM and IgA MM cases can be identified by SPE and IFE and the N-glycomic profiles of serum proteins were quite similar between the IgD MM and LCMM patients (as described above), we intended to differentially diagnose IgD MM from LCMM only. The ratio of NA2FB/NG1(3)A2F (peak7/peak4) was selected as the differential diagnostic marker for NA2FB levels were significantly higher in the IgD MM group than in the LCMM group, while NG1(3)A2F levels were significantly lower in the former than in the latter. NA2FB/NG1(3)A2F showed comparably differential diagnostic performance with an AUC of 0.744 (95% CI 0.605–0.883, Fig. 5). The optimal cut-off value for NA2FB/NG1(3)A2F was 0.953 (sensitivity 95.3%, specificity 50.0%, accuracy 72.7%) (Table 4).

In this study, one patient whose NA2FB/NG1(3)A2F level was 3.307 at the first-time diagnosis of LCMM was eventually confirmed with the subtype of IgD lambda 37 months later. Therefore, the increased level of NA2FB/NG1(3)A2F probably indicated the presence of IgD monoclonal protein, while IFE and SPE already showed a light chain-only or negative result.

To determine the correlation between the abundance of N-glycan markers and clinical outcome, we analysed the prognostic significance of N-glycan markers using Kaplan-Meier analysis. The median follow-up of the entire IgD MM patient cohort was 31.5 months. The comparison was executed between the above- and below-median concentrations of each N-glycan peak. Finally, as shown in Fig. 6, the progression free survival (PFS) time of IgD MM patients with above-median levels of NA2FB was significantly shorter than that of patients with below-median levels of NA2FB (P = 0.008). Additionally, the subjects with NA2FB levels above the median level exhibited a shorter overall survival (OS) time, and those with NA2FB levels below the median level showed a relatively longer OS time (30.7 and 41.6 months, respectively), although the difference was not statistically significant.