Serum peptides as candidate biomarkers for relapsing polychondritis

Abstract

Introduction: Relapsing polychondritis (RP) is an intractable disease characterized by recurrent inflammation of cartilaginous tissue throughout the body. It is difficult to accurately diagnose RP, and no useful biomarkers have yet been identified.

Objectives: We analyzed serum peptide profiles to identify novel candidate biomarkers for RP.

Methods: Thirty-seven patients with RP, 42 patients with rheumatoid arthritis (RA), and 35 healthy control (HC) subjects were divided into training and testing sets. Seven patients demonstrating granulomatosis with polyangiitis (GPA) were used for validation. The ion intensity of serum peptides was comprehensively measured by matrix-assisted laser desorption/ionization time-of-flight/time-of-flight mass spectrometry and applied to a supervised multivariate analysis. Peptides of interest were analyzed by liquid chromatography-tandem mass spectrometry.

Results: In the training set, models developed based on 11 (RP/HC-11P model), 9 (RP/RA-9P model), and 14 (RP/nonRP-14P model) peptides, out of 160 peptides detected were able to completely discriminate the RP group from the HC, RA, and nonRP (HC + RA) groups. Almost all of the 15 identified discriminatory peptides comprising these models were fragments of proteins associated with coagulation. Four models, each consisting of 4 out of 10 identified peptides of the RP/nonRP-14P model (models RP/nonRP-4P-2, -10, -11, and -38), provided ≥ 70.0 % sensitivity and specificity when applied to the validation set (the testing set and the GPA group) (AUROC, 0.779-0.815). Notably, the RP/nonRP-4P-2 model provided 83.3 % sensitivity and 71.7 % specificity in the validation set (AUROC, 0.802).

Conclusions: Serum peptides are useful as candidate biomarkers for discriminating RP and may be involved in the pathophysiology of RP.

Keywords: Biological markers; Mass spectrometry; Relapsing polychondritis; Serum peptides.

Fig. 1

Score scatter plots from PCA using serum peptide ion intensity. Results of the RP group (A), the HC group (B), the RA group (C), and all three groups (D) in the training set. The two individuals in the HC group and one patient in the RA group who were identified as outliers are specifically indicated by their ID numbers.

Fig. 2

Serum peptide profiles of the RP, HC, and RA groups in the training set. One hundred sixty serum peptide ion peaks were detected from 19 RP patients, 17 HC subjects, and 21 RA patients in the training set. The average ion intensity of the peptides in the RP, HC, and RA group is shown.

Fig. 3

Models to discriminate between the RP and nonRP groups. Two models were generated using the peptide ion intensity of the 19 RP patients and the non-RP subjects (HC subjects, n = 17; and RA patients, n = 21) in the training set. First, the ion intensity values of the 160 peptides were subjected to an OPLS-DA (RP/nonRP-160P model, A-D). A, T he score scatter plot of the RP/nonRP-160P model. The x axis indicates the first principal component for the discrimination. B, The loading scatter plot of the RP/nonRP-160P model. Parameters localized further away from the center of the x-axis contribute more to the discrimination. C, The S-plot of the RP/nonRP-160P model. The magnitude (x axis) and reliability (y axis) of the peptides are visualized. D, VIP of the RP/nonRP-160P model. Thirty-six of the 160 peptides relatively highly contributed to the generation of this model (VIP >1.0). The second model between the RP and nonRP groups was generated using a minimum of 14 peptides (RP/nonRP-14P model, E-H). The score scatter plot (E), loading scatter plot (F), S-plot (G), and VIP (H) of the RP/nonRP-14P model are shown. The numbers in brackets in H indicate the order of the VIP scores of the peptides in the RP/nonRP-160P model (D).

Fig. 4

Validation of the RP/nonRP-4P models. In each model, the prediction values of individual patients and subjects were calculated as the value of an ideal RP patient as 1.0 and as that of an ideal non-RP patient or subject as 0.0 in the training set. The prediction values of the patients and subjects in the validation set were calculated using the formula obtained by the above calculation of the training set. Dotted lines show the cutoff points (See Table 4). In the RP group, ratios of the cases discriminated as RP are shown. In the RA, GPA, and HC groups, ratios of the patients and subjects discriminated as non-RP are shown. The significance of the differences between the RP group and the other groups is shown (*p < 0.05).