The local ethical review boards of Chiba University, Graduate School of Medicine, Toho University, Faculty of Medicine, Toho University Omori Medical Center, and Port Square Kashiwado Clinic, and the review boards of the participating hospitals approved this study. All experimental procedures adhered to the guidelines of the Declaration of Helsinki (2013). Serum samples were collected from the donors after obtaining written informed consent.

All serum samples were centrifuged at 3,000 g for 10 min. The supernatants were stored at –80°C to avoid repeated freezing/thawing.

Serum samples from 127 patients with AIS were obtained from Chiba Prefectural Sawara Hospital within two weeks of disease onset (AIS cohort). The Sawara cohort comprised 665 specimens obtained from Chiba Prefectural Sawara Hospital, including 139 from HDs, 228 from patients with AIS, 44 with transient ischemic attack, 122 with deep and subcortical white matter hyperintensity, 17 with asymptomatic cerebral infarction, 59 with chronic-phase cerebral infarction, and 56 disease controls. Sera of 275 and 100 patients with DM and CVD, respectively, were obtained from Chiba University Hospital (DM and CVD cohorts, respectively); the CVD cases included those with AMI and unstable angina pectoris. Serum samples from 300 patients with CKD were obtained from Kumamoto University (CKD cohort) [14, 15]. Serum samples from 285 patients with EC, gastric cancer (GC), or CRC were collected immediately before surgery, radiotherapy, or chemotherapy at the Department of Surgery, Toho University Hospital (cancer cohort). Serum samples from HDs (HD cohort) were collected from three institutions: Chiba University Hospital, Port Square Kashiwado Clinic, and Chiba Prefectural Sawara Hospital. All HD serum samples were obtained from individuals without any abnormalities on cranial magnetic resonance imaging.

The first screening was performed using the ProtoArray Human Protein Microarrays (version 4.0; Thermo Fisher Scientific, Waltham, MA, USA), which contained 9,480 proteins, as described previously [16–18]. Thirty serum samples (10 from HDs and 10 from patients with atherosclerosis) were analyzed to identify antigens specifically recognized by serum IgG antibodies. The data were processed using Prospector software, which employs M-statistics (Thermo Fisher Scientific). To compare the two groups, a positivity threshold for each protein was determined using M-statistics (Supplementary material), which included background subtraction, normalization of signals, and analysis of differences between patients and HDs. The proportion of subjects in each group showing an immune response above this threshold was scored, and the significance of the difference between the two groups was assessed by calculating the P-value, as described previously [19].

Full-length cDNA of human UBE2E3 (accession number: NM_006357) was purchased from Open Biosystems (Huntsville, AL, USA) and recombined into pGEX-6P (Cytiva, Pittsburgh, PA, USA). The cDNA product was expressed by treating Escherichia coli (E. coli) KRX (Promega, Madison, WI, USA) cells containing the pGEX-6P-UBE2E3 and pMINOR with 0.5 mM isopropyl-β-D-thiogalactoside at 37°C for 3 h [20]. The cells were lysed by sonication in phosphate-buffered saline containing 2 mM dithiothreitol. The proteins were purified using Glutathione-Sepharose 4 Fast Flow medium (Cytiva) and a HiPrep 26/10 desalting column (Cytiva) and concentrated to 1.3 g/mL in phosphate-buffered saline containing 2 mM dithiothreitol. For comparison, antigenic differential screening-selected gene aberrant in neuroblastoma (DAN) protein and sclerostin domain-containing protein 1 (SOSTDC1) peptide were prepared. The region between positions 267 and 731 of the DAN cDNA (accession number: X66872.1) was inserted into the EcoRI/XhoI site of pGEX-2T (GE Healthcare Life Sciences, Pittsburgh, PA, USA), which produced a truncated DAN protein (amino acid residues 25–178) lacking its potential signal peptide sequence. The glutathione S-transferase (GST)-fused DAN protein was purified, as described previously [21]. The biotinylated peptide between amino acid residues 156 and 170 of SOSTDC1 (accession number: NM_015464) was synthesized and purified using high-performance liquid chromatography (HPLC), as described previously [16]. The SOSTDC1 peptide structure was biotin-KITVVTACKCKRYTR-COOH, with a purity of > 90%.

Serum antibody levels were examined using amplified luminescence proximity homogeneous assay-linked immunosorbent assay (AlphaLISA) in 384-well microtiter plates (white opaque OptiPlate™, Perkin Elmer, Waltham, MA, USA). AlphaLISA required two types of beads: a donor bead that binds to the antigen via a tag and an acceptor bead that binds the IgG antibody via a secondary antibody. IgG antibody binding to the antigen in solution brings the two beads close together, and excitation at 680 nm results in emission at 618 nm. No plate washing was required. Each well contained 2.5 µL of 1:100-diluted serum with 2.5 µL of GST, GST-UBE2E3, or GST-DAN proteins (10 µg/mL) or biotinylated SOSTDC1 peptide (400 ng/mL) in AlphaLISA buffer. The buffer was composed of 25 mM N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (pH 7.4), 0.1% casein, 0.5% Triton X-100, 1 mg/mL dextran-500, and 0.05% ProClin-300, following the manufacturer’s instructions (https://www.revvity.co.jp/content/elisa-alphalisa-immunoassay-conversion-made-easy). The reaction mixture was incubated at room temperature for 6–8 h. Anti-human IgG-conjugated acceptor beads (2.5 µL at 40 µg/mL) and either glutathione- or streptavidin-conjugated donor beads (2.5 µL at 40 µg/mL) were added, and the mixture was incubated at room temperature in the dark for 7–14 days. Chemical emissions at 607–623 nm (alpha photon counts), which indicated the antigen-antibody binding level, were measured using an EnSpire Alpha microplate reader (PerkinElmer), as described previously [7–9, 13, 16–18]. Specific reactions were determined by subtracting the alpha emission counts of the GST and buffer control from those of the GST-fused proteins and biotinylated peptides, respectively.

We employed the Mann–Whitney U test to examine differences between two groups and the Kruskal–Wallis test (Mann–Whitney U test with the Bonferroni correction) to evaluate differences among three or more groups. Spearman’s correlation analysis, logistic regression analysis, the chi-square test, and univariate and multivariate analyses were used to calculate the correlations. We assessed the predictive values of the putative disease markers using receiver operating characteristic (ROC) curve analysis. The sensitivity and specificity were calculated using the cutoff values of the Youden Index. All statistical analyses were performed using GraphPad Prism (version 5; GraphPad Software, Inc.). Patient survival was evaluated using the Kaplan–Meier method, and the results were compared using the log-rank test. The cutoff values were determined using X-tile software (version 3.6.1; Yale University, New Haven, CT) [22]. All tests were two-tailed, and P-values < 0.05 indicated statistically significant differences.

We used ProtoArray to identify antigens using IgG antibodies from the sera of patients with atherosclerosis. UBE2E3 (accession number: NM_006357) was a candidate antigen because of its high positive reactivity rate (6 of 10 serum samples) in patients with atherosclerosis and low positive reactivity rate (2 of 10 samples) in HDs. All ProtoArray results are available in the public Figshare database (https://figshare.com/articles/dataset/Results_of_protein_array_for_atherosclerosis/25906330).

We examined serum antibody levels in patients with AIS using recombinant UBE2E3 protein. Sera from HDs and patients with AIS (AIS cohort) were obtained from the Chiba Prefectural Sawara Hospital. The sample numbers (total, male, female) and the average age ± standard deviation (SD) are shown in Table 1. The AlphaLISA results revealed that anti-UBE1E3 antibody (UBE2E3-Ab) levels were significantly higher in patients with AIS than in HDs (Figure 1A). Cutoff values were determined using the average plus two SDs of the HD values. The positivity rates of UBE2E3-Abs for the HDs and patients with AIS were 3.9% and 12.6%, respectively (Table 1). ROC analysis revealed that the area under the ROC curve (AUC) for UBE2E3-Abs was 0.6806 [95% confidence interval (CI): 0.6156–0.7456] (Figure 2A). Using the cutoff value based on the Youden index, the sensitivity and specificity for UBE2E3-Abs were 64.6% and 64.8%, respectively.

Next, we examined the UBE2E3-Ab levels in patients with DM. Serum samples from HDs and patients with DM (DM cohort) were obtained from Chiba University Hospital. The UBE2E3-Ab levels were significantly higher in patients with DM than in HDs (Figure 1B). At the cutoff value (average plus two SDs of the HD values), the positivity rates of UBE2E3-Abs in HDs and patients with DM were 4.9% and 22.2%, respectively (Table 2). ROC analysis was performed to evaluate the diagnostic ability for patients with DM. The AUC for UBE2E3-Abs was 0.6716, yielding a sensitivity and specificity of 50.2% and 79.0%, respectively (Figure 2B).

We then examined UBE2E3-Ab levels in patients with CVD. Sera from HDs and patients with CVD (CVD cohort) were obtained from Chiba University Hospital. UBE2E3-Ab levels were significantly higher in patients with CVD than in HDs (Figure 1C). Using the cutoff values (average plus two SDs of the HD values), the UBE2E3-Ab positivity rates for HDs and patients with CVD were 5.1% and 8.0%, respectively (Table 3). The AUC value of UBE2E3-Abs was 0.6821 (95% CI: 0.6069–0.7573), with a sensitivity and specificity of 88.4% and 41.7%, respectively (Figure 2C).

Next, we analyzed the antibody levels in the sera of patients with CKD (CKD cohort), a condition strongly associated with atherosclerosis. CKD was divided into three subtypes: type 1, diabetic kidney disease; type 2, nephrosclerosis; and type 3, glomerulonephritis. Samples from patients with CKD were obtained from the Kumamoto cohort, and samples from HDs were obtained from Chiba University Hospital. Patients in all three CKD groups had significantly higher serum UBE2E3-Ab levels than the HDs (Figure 1D). The UBE2E3-Ab positivity rates in HDs and patients with types 1, 2, and 3 CKD were 2.4%, 17.2%, 15.6%, and 8.1%, respectively (Table 4). ROC analysis suggested that the AUCs of types 1, 2, and 3 CKD were 0.8531 (95% CI: 0.8005–0.9057) (Figure 2D), 0.8798 (95% CI: 0.8189–0.9408) (Figure 2E), and 0.7941 (95% CI: 0.7300–0.8583) (Figure 2F), respectively. Overall, CKD showed much higher AUC values than AIS and DM, irrespective of the CKD type (Figures 2A-F).

An increasing number of reports have indicated that atherosclerosis is closely associated with cancer [7–9], which is supported by shared biomarkers for both atherosclerosis and cancer. To investigate this further, we analyzed serum samples from patients with EC, GC, or CRC (cancer cohort) obtained from Toho University Hospital. UBE2E3-Ab levels were significantly higher in patients with EC and GC, but not with CRC, than in HDs (Figure 1E, Table 5). The AUCs for EC, GC, and CRC were 0.6669, 0.6825, and 0.6071, respectively (Figures 2G–I).

UBE2E3 suppresses the cellular senescence of bone marrow mesenchymal stem cells, leading to osteoporosis [23], possibly via bone morphogenetic proteins (BMPs) [24]. We previously identified serum antibodies against BMP antagonists as atherosclerosis markers, such as anti-DAN antibody (DAN-Ab) [21] and anti-SOSTDC1 antibody (SOSTDC1-Ab) [16]. DAN-Ab levels were elevated in patients with EC and GC, but not with CRC, compared to those in HDs (Figure 1F, Table 5). The AUC values of DAN-Abs for EC, GC, and CRC were 0.6907, 0.6422, and 0.6071, respectively (Figures 2J–L). In contrast, only patients with EC had higher SOSTDC1-Ab levels than those in HDs (Figure 1G, Table 5). The AUC values of SOSTDC1-Abs for EC, GC, and CRC were 0.634, 0.582, and 0.527, respectively (Figures 2M–O).

Next, we tested whether UBE2E3-Ab levels were associated with the postoperative survival of patients with EC, GC, and CRC. We divided the UBE2E3-Ab levels into positive and negative groups using the cutoff values obtained from the X-tile software [22] to determine the optimal cutoff values for the discrimination of survival rates. Although there were no significant differences in overall survival between the UBE2E3-Ab-positive and -negative EC groups (P = 0.1271), the UBE2E3-Ab-positive group showed a tendency toward a better prognosis (Figure 3A).

The patients with EC in the DAN-Ab-positive group also showed a more favorable prognosis than the DAN-Ab-negative group, although the difference was not statistically significant (P = 0.1322) (Figure 3B). The prognosis of the UBE2E3-Ab-positive/DAN-Ab-positive group was significantly more favorable than that of the UBE2E3-Ab-negative/DAN-Ab-negative group (P = 0.0179) (Figure 3C). Patients with EC in the SOSTDC1-Ab-positive group showed no difference in prognosis compared with the SOSTDC1-Ab-negative group (P = 0.4126) (Figure 3D). The prognosis of the UBE2E3-Ab-positive/SOSTDC1-Ab-positive group was more favorable than that of the UBE2E3-Ab-negative/SOSTDC1-Ab-negative group, but the difference was not significant (P = 0.1545) (Figure 3E). Thus, significant discrimination was observed only for the combination of UBE2E3-Abs and DAN-Abs.

In contrast, the UBE2E3-Ab-positive group had a significantly poorer prognosis than the UBE2E3-Ab-negative group in patients with GC (P = 0.0193) (Figure 4A). The DAN-Ab-positive group showed a more unfavorable prognosis than the DAN-Ab-negative group, but the difference was not significant (P = 0.0640) (Figure 4B). The difference in prognosis between the UBE2E3-Ab-positive/DAN-Ab-positive and UBE2E3-Ab-negative/DAN-Ab-negative groups was greater than that of each alone (P = 0.0098) (Figure 4C). Likewise, the SOSTDC1-Ab-positive group showed a somewhat poorer prognosis than the SOSTDC1-Ab-negative group (P = 0.1148) (Figure 4D), and the combined UBE2E3-Ab-positive/SOSTDC1-Ab-positive group exhibited a somewhat more unfavorable prognosis than the combined UBE2E3-Ab-negative/SOSTDC1-Ab-negative group (P = 0.0063) (Figure 4E). Thus, a more precise prediction of GC prognosis was achieved using a combination of UBE2E3-Abs and DAN-Abs or SOSTDC1-Abs.

The prognostic analysis of CRC showed results similar to those of GC but not of EC. The UBE2E3-Ab-positive group had a significantly worse prognosis than the UBE2E3-Ab-negative group in patients with CRC (P = 0.0212) (Figure 5A). The DAN-Ab-positive group tended to have a worse prognosis than the DAN-Ab-negative group (P = 0.1478) (Figure 5B). The prognosis of the combined UBE2E3-Ab-positive/DAN-Ab-positive group was worse than that of the combined UBE2E3-Ab-negative/DAN-Ab-negative group (P = 0.0362) (Figure 5C). The SOSTDC1-Ab-positive group had a significantly poorer prognosis than the SOSTDC1-Ab-negative group (P = 0.0290) (Figure 5D), and the combined UBE2E3-Ab-positive/SOSTDC1-Ab-positive group exhibited an even more unfavorable prognosis than the combined UBE2E3-Ab-negative/SOSTDC1-Ab-negative group (P = 0.0033) (Figure 5E). Thus, a more precise prediction of CRC prognosis was achieved using a combination of UBE2E3-Abs and DAN-Abs or SOSTDC1-Abs.

Spearman’s correlation analysis of the CKD cohort of 300 participants revealed significant associations between the plaque score, maximum intima-media thickness (max-IMT) [25, 26], and cardio-ankle vascular index (CAVI) (right and left) [27] (Table 6), which are key indicators of atherosclerosis. Aspartate aminotransferase (AST) levels were weakly associated, and standardized urea clearance (Kt/V), urea nitrogen levels, and high-density lipoprotein cholesterol (HDL-c) levels were inversely correlated with UBE2E3-Ab levels. Other patient data, including age, height, weight, and body mass index (BMI), showed no significant correlation with UBE2E3-Ab levels.

We performed a correlation analysis of UBE2E3-Ab levels in the Sawara cohort. UBE2E3-Ab levels were significantly correlated with carotid artery stenosis, including IMT (right and left) and max-IMT (Table S1), confirming the results from the CKD cohort (Table 6). UBE2E3-Ab levels were strongly associated with blood pressure (P = 0.0015) and smoking duration (P = 0.0020), which are major risk factors for atherosclerosis [28, 29]. Furthermore, UBE2E3 levels were positively correlated with age, alkaline phosphatase levels, thymol turbidity test (TTT) results, and white blood cell numbers and negatively correlated with height, weight, and chlorine levels. The correlation between UBE2E3-Ab levels and the IMT in the Sawara cohort also suggested that UBE2E3-Ab levels were associated with arterial stenosis or atherosclerosis. A chi-square test was conducted to compare the sex differences between the UBE2E3-Ab-positive and -negative groups. No significant correlation was observed between sex and UBE2D3-Ab positivity (P = 0.9317) (Table S2). Univariate and multivariate analyses were performed using the AIS cohort. The UBE2E3-Ab level was not a significant independent predictor in the multivariate analysis (P = 0.34) (Table S3).

The correlation of UBE2E3-Ab levels was also examined in 275 patients with DM (DM cohort) at Chiba University Hospital. UBE2E3-Ab levels were correlated with the blood pressure and estimated glomerular filtration rate (eGFR) but were inversely correlated with calcium levels and platelet numbers (Table S4). Thus, the results of the correlation analyses in the CKD, DM, and Sawara cohorts support an association between UBE2E3-Abs levels and the level of atherosclerotic progression.