Autoantibody Landscape Revealed by Wet Protein Array: Sum of Autoantibody Levels Reflects Disease Status

Abstract

Autoantibodies are found in various pathological conditions such as autoimmune diseases, infectious diseases, and malignant tumors. However their clinical implications have not yet been fully elucidated. Herein, we conducted proteome-wide autoantibody screening and quantification with wet protein arrays consisting of proteins synthesized from proteome-wide human cDNA library (HuPEX) maintaining their three-dimensional structure. A total of 565 autoantibodies were identified from the sera of three representative inflammatory disorders (systemic sclerosis, psoriasis, and cutaneous arteritis). Each autoantibody level either positively or negatively correlated with serum levels of C-reactive protein, the best-recognized indicator of inflammation. In particular, we discovered total levels of a subset of autoantibodies correlates with the severity of clinical symptoms. From the sera of malignant melanoma, 488 autoantibodies were detected. Notably, patients with metastases had increased overall autoantibody production compared to those with tumors limiting to the primary site. Collectively, proteome-wide screening of autoantibodies using the in vitro proteome can reveal the "autoantibody landscape" of human subjects and may provide novel clinical biomarkers.

Keywords: autoantibody; autoimmunity; inflammatory disease; malignancy; protein array.

Figure 1

Schematic of the autoantibody screening pipeline. In the first step, proteins were synthesized in vitro from the proteome-wide human cDNA library (HuPEX). Promotors (P), Enhancers (E), and FLAG-GST tags were fused to open reading frames of the expression clones by Gateway LR reaction. After polymerase chain reaction amplification and in vitro transcription, translation was performed using the wheat germ cell-free synthesis system. In the second step, we conducted autoantibody screening by treating proteome-wide wet protein arrays with pooled serum. In the third step, autoantibody quantification was performed by treating focused wet protein arrays with individual sera. ORF, open reading frame; SSc, systemic sclerosis; Pso, psoriasis; CA, cutaneous arteritis; MM, malignant melanoma; Ctrl, healthy control.

Figure 2

Intra- and Inter-diversity of correlation between serum levels of C-reactive protein and autoantibody level. (A) Flow-chart of autoantibody selection. The number of entry cDNA clones was 12,946. As primary selection, proteome-wide autoantibody screening from pooled serum of patients with systemic sclerosis (SSc), psoriasis (Pso), or cutaneous arteritis (CA) detected 565 positivity. Their origin in primary screening is illustrated in a Venn diagram. As secondary selection, a search for association between the diseases by PubMed and for subcellular localization by the Cell Atlas was conducted. Finally, 178 proteins were spotted on the focused protein array. Their origin in primary screening is also shown in a Venn diagram. (B) Histograms of Spearman’s correlation coefficient between serum CRP and each autoantibody level in individual sera of patients with SSc, Pso, and CA. Autoantibodies that negatively or positively correlated with serum CRP with statistical significance are highlighted in cyan or in magenta, respectively. (C) UpSet plots that visualize the origin of autoantibodies in primary screening that significantly correlates with serum CRP.

Figure 3

Association between sum of autoantibody levels and clinical presentation in patients with systemic sclerosis. (A) Heatmap of 18 autoantibodies that positively correlate with serum levels of C-reactive protein (CRP) with statistical significance in individual sera of systemic sclerosis (SSc). The targets of autoantibodies are described with the gene symbols of their origin. Each case was labeled with the highest serum level among anti-topoisomerase I (ATA), anti-centromere (ACA), and anti-RNA polymerase III (RNAP) antibodies. SAL-P-CRP: sum of autoantibody levels that positively correlated with serum CRP with statistical significance. AU: arbitrary unit. (B) Violin plots of SAL-P-CRP in individual sera of SSc patients and healthy controls. Subpopulation analysis revealed that SAL-P-CRP was significantly elevated among SSc patients with interstitial lung disease (ILD) or diffuse cutaneous SSc (dcSSc) compared to limited cutaneous SSc (lcSSc). P values were calculated by Mann-Whitney U test. (C) Correlation between SAL-P-CRP and pulmonary function, serum ILD markers, skin thickness score, and serum CRP levels. r: Spearman’s correlation coefficient. P values were calculated by estimating r. %FVC: percentage of predicted forced lung vital capacity, %DLco: percentage of predicted lung diffusing capacity for carbon monoxide, KL-6: Krebs von den Lungen-6, SP-D: surfactant protein-D, mRSS: modified Rodnan’s total skin thickness score. (D) Receiver-operator characteristics curve of accuracy detecting ILD. Area under the curve (AUC) was the biggest for SAL-P-CRP, compared to those for ATA levels (TOP1) or for CRP levels. (E) Principal component (PC) analysis. Two-dimensional scattering isolated three representative subpopulations of SSc: ATA, ACA, and RNAP-positive groups. Each case was labeled with the highest serum level among ATA, ACA, and RNAP. PC1, the primary principal component. PC2, the secondary principal component. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 4

Serial measurements of pulmonary function and autoantibody level in a patient with interstitial lung disease associated with systemic sclerosis. Scales for pulmonary function and autoantibody level are shown at the left and right, respectively. The values of sum of autoantibody levels that positively correlated with serum CRP with statistical significance (SAL-P-CRP) are described as the mean and the standard error of the mean of technical triplicates. %FVC: percentage of predicted forced lung vital capacity, %DLco: percentage of predicted lung diffusing capacity for carbon monoxide, PSL, prednisolone; MMF, mycophenolate mofetil; AU, arbitrary unit.

Figure 5

Association between sum of autoantibody levels that correlates with forced lung vital capacity and clinical presentation in patients with systemic sclerosis. (A) Histograms of Spearman’s correlation coefficient between percentage of predicted forced lung vital capacity (%FVC) and each autoantibody level in individual sera of patients with systemic sclerosis (SSc). Autoantibodies that negatively or positively correlated with %FVC with statistical significance are highlighted in cyan or in magenta, respectively. Their targets are listed in the table with the gene symbols of their origin. AU, arbitrary unit. (B) Violin plots of sum of autoantibody levels that negatively correlated with %FVC with statistical significance (SAL-N-%FVC) in SSc patients and healthy controls. Subpopulation analysis revealed that SAL-N-%FVC was significantly elevated among SSc patients with interstitial lung disease (ILD), or diffuse cutaneous SSc (dcSSc) compared to limited cutaneous SSc (lcSSc). P values were calculated by Mann-Whitney U test. (C) Correlation between SAL-N-%FVC and pulmonary function, serum ILD markers, skin thickness score, and serum CRP levels. r: Spearman’s correlation coefficient. P values were calculated by estimating r. %DLco: percentage of predicted lung diffusing capacity for carbon monoxide, KL-6: Krebs von den Lungen-6, SP-D: surfactant protein-D, mRSS: modified Rodnan’s total skin thickness score. (D) Violin plots of sum of autoantibody levels that positively correlated with %FVC with statistical significance (SAL-P-%FVC) in SSc patients and healthy controls. Subpopulation analysis revealed that SAL-P-%FVC was significantly lower among SSc patients with ILD, or dcSSc compared to lcSSc. P values were calculated by Mann-Whitney U test. (E) Correlation between SAL-P-%FVC and pulmonary function, serum ILD markers, skin thickness score, and serum CRP levels. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 6

Association between sum of autoantibody levels and clinical presentation in psoriasis patients. (A) Heatmap of 30 autoantibodies that positively correlate with serum levels of CRP with statistical significance in individual sera of psoriasis (Pso). The targets of autoantibodies are described with the gene symbols of their origin. SAL-P-CRP: sum of autoantibody levels that positively correlated with serum C-reactive protein (CRP) levels. AU: arbitrary unit. (B) Violin plots of SAL-P-CRP in Pso patients and healthy controls. Subpopulation analysis revealed that SAL-P-CRP was significantly elevated among patients with psoriatic arthritis (PsA) compared to the others. P values were calculated by Mann-Whitney U test. (C) Correlation between SAL-P-CRP and psoriasis area and severity index (PASI) score, or serum CRP levels. r: Spearman’s correlation coefficient. P values were calculated by estimating r. (D) Receiver-operator characteristics curve of accuracy predicting PsA. Area under the curve (AUC) was 0.76 (95% confidence interval: 0.58-0.94). *P < 0.05, **P < 0.01.

Figure 7

Serial measurements of pulmonary function and autoantibody level in a patient with psoriatic arthritis. Scales for psoriasis area and severity index (PASI) and erythrocyte sedimentation rate (ESR) are shown on the left, and the scale for sum of autoantibody levels that positively correlated with serum C-reactive protein levels (SAL-P-CRP) is shown on the right. The values of SAL-P-CRP are described as the mean and the standard error of the mean of technical triplicates. IFX, infliximab; AU, arbitrary unit.

Figure 8

Association between sum of autoantibody levels and clinical presentation in cutaneous arteritis patients. (A) Heatmap of 2 autoantibodies that positively correlate with serum levels of C-reactive protein (CRP) with statistical significance in individual sera of cutaneous arteritis (CA). The targets of autoantibodies are described with the gene symbols of their origin. SAL-P-CRP: sum of autoantibody levels that positively correlated with serum CRP levels with statistical significance. AU, arbitrary unit. (B) Heatmap of 6 autoantibodies that negatively correlate with serum CRP with statistical significance in individual sera of CA. The targets of autoantibodies are described with the gene symbols of their origin. SAL-N-CRP: sum of autoantibody levels that negatively correlates with serum CRP levels with statistical significance. (C) Violin plots of SAL-P-CRP in CA patients and healthy controls. Subpopulation analysis revealed that SAL-P-CRP was significantly elevated among treatment non-responders than among responders. P values were calculated by Mann-Whitney U test. (D) Violin plots of SAL-N-CRP in CA patients and healthy controls. **P < 0.01.

Figure 9

Gross autoantibody product in localized and advanced malignant melanoma. The result of proteome-wide autoantibody detection from pooled serum of localized or advanced malignant melanoma (MM) patients is illustrated. Each row shows autoantibodies spotted on the focused wet protein array (WPA) and their signal strength on the proteome-wide WPA. Semi-quantification of signals on proteome-wide WPA was conducted as below: higher than positive control’s signal strength: 6, higher than 1/2 positive control’s signal strength but lower than positive control’s signal strength: 5, higher than 1/4 but lower than 1/2 of positive control’s signal strength: 4, higher than 1/8 but lower than 1/4 of positive control’s signal strength: 3, higher than 1/16 but lower than 1/8 of positive control’s signal strength: 2, higher than negative control’s signal strength but lower than 1/16 of positive control’s signal strength: 1. P value was calculated by Wilcoxon signed-rank test. ****P < 0.0001.