Nineteen TIA patients from Chiba Rosai Hospital were randomly selected for SEREX immunoscreening. To compare antibody levels, sera were acquired from 92 patients with TIA, 464 patients with acute cerebral infarction (aCI), and 65 patients with old cerebral infarction (oCI) at Chiba Prefectural Sawara, Chiba Rosai, and Chiba Aoba Municipal Hospitals. Sera were also acquired from 128 DM and 128 AMI patients at Chiba and Kyoto University Hospitals, respectively. Sera were acquired from 192 patients with ESCC, 96 patients with GC, and 191 patients with CRC at Toho University Omori Hospital. All subjects with ESCC were followed up until January 2020 or their death. Sera of healthy donors (HDs) were provided by Port Square Kashiwado Clinic, Chiba Prefectural Sawara Hospital, and Toho University Omori Hospital. These subjects had no aberrant cranial resonance imaging. The sera were centrifuged at 3,000 × g for 10 min and the supernatants were stored at -80°C until use. Sample freeze-thaw was avoided.

An improved version of the aforementioned method was used to screen clones that were immunoreactive to the sera of patients with TIA (31). A human aortic endothelial cell cDNA expression library (Uni-ZAP XR Premade Library, Stratagene, La Jolla, CA, USA) was transfected into Escherichia coli (E. coli) XL1−Blue MRF′ (Stratagene). The resident cDNA clones were transferred onto nitrocellulose (NC) membranes pretreated with 10 mM isopropyl-β-D-thiogalactoside (IPTG; Wako Pure Chemicals Industries Ltd., Osaka, Japan) for 30 min. Membranes with bacterial proteins were washed thrice with TBS-T (20 mM Tris-HCl (pH 7.5), 0.15 M NaCl, and 0.05% (w/v) Tween 20). Then the membranes were incubated for 1 h in 1% (w/v) protease-free bovine serum albumin (BSA; Nacalai Tesque, Inc., Kyoto, Japan) in TBS-T to block nonspecific binding. The membranes were then incubated overnight with diluted sera (1:2,000) from the TIA patients. The membranes were washed thrice in TBS−T and incubated for 1 h in alkaline phosphatase−conjugated goat anti−human IgG (1:5,000; Jackson Immuno Research Laboratories, West Grove, PA, USA). Positive responses were identified by cultivating the membranes in a color development solution (100 mM Tris-HCl (pH 9.5), 100 mM NaCl, and 5 mM MgCl₂) containing 0.15 mg/mL 5-bromo-4-chloro-3-indolyl phosphate (Wako Pure Chemicals Industries Ltd.) and 0.3 mg/mL nitroblue tetrazolium (Wako Pure Chemicals Industries Ltd.). Cloning was performed twice on the positives until monoclonality.

ExAssist helper phage (Stratagene) and in vitro excision were used to convert the monoclonalized phage cDNA clones into pBluescript phagemids. Plasmid DNA was extracted from the E. coli SOLR strains transformed by the phagemids. The infused cDNAs were sequenced for homology using the public database provided by the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/Blast.cgi/).

To construct the glutathione-S-transferase (GST)-fused protein expression plasmids, the cDNA sequences were recombined into the pGEX-4T vector (GE Healthcare Life Sciences, Pittsburgh, PA, USA) as previously described (32, 35, 37). The pBluescript plasmids associated with the cDNA inserts were digested with the restriction endonucleases EcoRI and XhoI and detached by agarose gel electrophoresis. GenElute Minus EtBr spin columns (Sigma-Aldrich Corp., St. Louis, MO, USA) were used to isolate the cDNA fragments which were ligated in frame to EcoRI- and XhoI-digested pGEX-4T-3 linearized vectors with ligation convenience kits (Nippon Gene, Toyama, Japan). The ligation mixtures were used to transform ECOS-competent E. coli BL-21 cells (Nippon Gene). Successful recombination was confirmed by DNA sequencing and protein expression analysis.

Escherichia coli BL-21 cells transformed with the pGEX-4T clone were cultured in 200 mL Luria-Bertani (LB) broth and treated with 1 mM IPTG for 3 h. The cells were collected in bacterial solution and lysed by sonication in BugBuster Master Mix (Novagen, San Diego, CA, USA). The lysates were then centrifuged at 13,000 × g and 4°C for 10 min. GST-tagged BRAT1 and GST-tagged WDR1 proteins were purified by glutathione-Sepharose column chromatography (GE Healthcare Life Sciences) and dialyzed as previously described (32, 36, 38).

Purified GST, GST-BRAT1, and GST-WDR1 proteins (0.3 μg) were separated by SDS-PAGE. After transfer and blocking, anti-GST (Rockland, Gilbertsville, PA, USA) or serum (1:5,000) from patients with TIA (#297) was used as a source of primary antibodies. The proteins were then incubated with HRP-conjugated secondary antibody (donkey anti-goat or anti-human IgG; Santa Cruz Biotechnology, Dallas, TX, USA) and detected as previously described (32, 39).

AlphaLISA was used to quantify the serum antibodies against the purified proteins. The α-luminescent photon counts represent the serum antibody levels (35, 40). The AlphaLISA assay was performed in 384-well microtiter plates (white opaque OptiPlate; PerkinElmer, Waltham, MA, USA). Each well contained 2.5 μL serum (1:100 dilution) and 2.5 μL GST or GST fusion protein (10 μg/mL) in AlphaLISA buffer (25 mM HEPES [pH 7.4], 0.1% (w/v) casein, 0.5% (w/v) Triton X-100, 1 mg/mL Dextran-500, and 0.05% (w/v) Proclin-300). The mixture was then incubated at 25°C for 8 h. Then 2.5 μL of 40 μg/mL anti-human IgG-conjugated acceptor beads and 2.5 μL of 40 μg/mL glutathione-conjugated donor beads were added and the mixture was incubated in the dark at 25°C for 7–21 d. Chemical emission was measured in an EnSpire Alpha microplate reader (PerkinElmer) as previously described. The reactions were calculated by subtracting the Alpha counts for the GST control from those for the GST fusion proteins.

The levels of serum squamous cell carcinoma antigen (SCC-Ag) (41) and p53 antibody (p53-Abs) (42) were evaluated as previously described. The serum SCC-Ag and p53-Abs cutoff values were 1.5 ng/mL and 1.3 IU/mL, respectively.

Tumor Immune Estimation Resource (TIMER2.0, http://timer.cistrome.org/) estimates the immune invasion levels of numerous cancers. The “Gene_DE” module permits users to compare gene expression levels between normal tissues and those of tumors associated with all 32 types of cancer listed in The Cancer Genome Atlas (TCGA) cohort (43, 44). C7orf27 (BRAT1) or WDR1 were inserted into the “Gene_DE” module of TIMER2.0 Web to analyze the differences in BRAT1 expression between normal tissues and the tumors listed in the TCGA cohort.

Gene Expression Profiling Interactive Analysis (GEPIA, http://gepia2.cancer-pku.cn/) analyzes gene expression in tumor and normal samples from the TCGA and Genotype-Tissue Expression (GTEx) databases (45, 46). GEPIA was used to analyze BRAT1 expression in various tumors in the TCGA cohort. Matching normal TCGA data served as the control.

The TNMplot (http://www.tnmplot.com) database contains 56,938 samples of normal, tumor, and metastatic tissues from gene chip studies, TCGA, Therapeutically Applicable Research to Generate Effective Treatments (TARGET), and GTEx (47). TNMplot was used to determine BRAT1 expression levels in different tumor tissues and compare them against the BRAT1 expression profiles of normal tissues.

UALCAN (http://ualcan.path.uab.edu) is an interactive web portal containing TCGA clinical data for 31 cancer types and RNA-seq. UALCAN has been used for in-depth TCGA gene expression data analysis (48). UALCAN can also analyze protein expression using data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC, http://ualcan.path.uab.edu/analysis-prot.html). Here, UALCAN was used to analyze protein expression.

Student’s t-test and Mann-Whitney U test were used to analyze differences between group pairs. Correlations among Alpha values and clinical data were calculated by multivariate logistic regression and Spearman’s correlation analyses. Differences in the distributions of two variables were calculated by Fisher’s exact test. The predictive values of the disease markers were assessed by ROC analysis. The antibody level cutoff values were set to maximize the sensitivity and specificity sums. Survival curves were plotted by the Kaplan-Meier method. Comparisons were made via the log-rank test. The Cox proportional hazards model was used to evaluate significant predictors. All tests were two-tailed. P < 0.05 was considered statistically significant. GraphPad Prism 6 (GraphPad Software, La Jolla, CA, USA) was used to perform all statistical analyses.

AS biomarkers were screened with SEREX ( Figures 1A–C ). Sera from the 19 patients in the TIA group were used for immunological screening. We identified certain clones and some of them were previously reported (32, 39). We focused on the antibody markers BRAT1 (accession No. NM_152743) and WDR1 (accession No. NM_017491). Full-length BRAT1 and WDR1 cDNAs were recombined into pGEX-4T-3 expression vectors. GST-labeled recombinant proteins were expressed in E. coli and purified by affinity chromatography with glutathione-Sepharose. Antigenic proteins were then purified from the precipitate fraction. The arrows in Figure 1 indicate that the GST-fusion proteins could be detected in equal amounts in the total extracts, precipitates (Ppts), dialysates, and flow-through, purified, and concentrated samples but not in the supernatant ( Figure 1D , middle).

Western blotting revealed antibodies against BRAT1 and WDR1 in the sera of patients with TIA. GST-BRAT1, GST-WDR1, and GST were recognized by anti-GST (αGST) antibody and detected as 58-kD, 59-kD, and 28-kD proteins, respectively ( Figure 1E ). The degradation products were marked with asterisks in Figure 1E . GST-BRAT1 and GST-WDR1 were recognized by serum IgG antibodies from a patient with TIA (TIA#297). There was no apparent reactivity against the serum IgG antibodies from the patients in the GST protein group ( Figure 1E , right).

We used AlphaLISA to measure serum BRAT1-Ab and WDR1-Ab in the HDs and patients with TIA, aCI, and oCI. Serum BRAT1-Abs and WDR1-Abs levels were significantly higher in the patients than the HDs ( Figures 2A, B ). Table 1A shows that there were 285, 92, 464, and 65 HDs and patients with TIA, aCI, and oCI, respectively. The distributions of men and women in these sample groups were 188/97, 55/37, 271/193, and 48/17, respectively. The average ages (± SD) in these treatment groups were 52.3 ± 11.7, 70.2 ± 11.6, 75.5 ± 11.5, and 73.3 ± 9.2 y, respectively. At the cutoff value, the positivity rates for BRAT1-Abs were 5.3, 14.1, 17.2, and 18.5% for the HDs and the patients with TIA, aCI, and oCI, respectively. The positivity rates for WDR1-Abs were 5.6, 14.1, 16.6, and 13.8% for the HDs and the patients with TIA, aCI, and oCI, respectively ( Table 1B ).

The abilities of BRAT1-Ab and WDR1-Ab to detect TIA, aCI, and oCI were evaluated by ROC analysis. In this manner, the efficacy of these markers at predicting TIA-related cardiovascular disease (CVD) was determined. The graphs in Figures 2C – H show the area under the curve (AUC), cutoff value, 95% CI, sensitivity, specificity, and P-value. The discriminant ability increases as the AUC value approaches unity (49). The BRAT1-Abs AUCs for TIA, aCI, and oCI were 0.68, 0.66, and 0.64, respectively ( Figures 2C – E ). Hence, BRAT1-Ab had good predictive efficacy for these diseases. The ROC analysis also disclosed that the WDR1-Abs AUCs were 0.67, 0.66, and 0.63 for TIA, aCI, and oCI, respectively ( Figures 2F – H ). Thus, BRAT1-Ab could effectively predict these conditions.

Spearman’s correlation analysis was performed to investigate the associations among the serum BRAT1-Ab and WDR1-Ab levels and the indices for the HDs and TIA, aCI, and oCI patients ( Table 2 ). Both BRAT1-Ab and WDR1-Ab were correlated with age (r = 0.2165, P < 0.0001; r = 0.2263, P< 0.0001), HT (r = 0.1298, P < 0.0001; r = 0.1279, P < 0.0001), IMT (r) (r = 0.1986, P < 0.0001; r = 0.1928, P < 0.0001), ALP (r = 0.1348, P < 0.0001; r = 0.09273, P = 0.007), CRP (r = 0.175, P < 0.0001; r = 0.1547, P < 0.0001), IMT (l) (r = 0.2032, P < 0.0001; r = 0.1996, P < 0.0001), and IMT_max (r = 0.2068, P < 0.0001; r = 0.2059, P < 0.0001). BRAT1-Ab and WDR1-Ab were also correlated with height, weight, AST, LDH, WBC, RDW, BP, and smoking but not with alcohol consumption. In contrast, the antibody levels were negatively correlated with A/G, CHE, ALB, T-CHO, and RBC. The AS-associated parameter TG was negatively associated with the level of WDR1-Ab but not that of BRAT1-Ab. The BRAT1-Ab and WDR1-Ab levels were not significantly elevated in patients with DM (P = 0.213 and 0.079, respectively). There was no apparent positive correlation between the antibody levels and HbA1c (r = 0.01993, P = 0.6547; r = -0.00131, P = 0.9766).

The foregoing results suggest associations among the BRAT1-Ab and WDR1-Ab levels and CVD including TIA, aCI and oCI. Moreover, BRAT1-Ab and WDR1-Ab are potential molecular markers for TIA, aCI, and oCI onset.

The serum antibody levels in the HDs and the AMI and DM patients were detected with AlphaLISA to verify the ability of BRAT1-Ab and WDR1-Ab to detect AS-associated diseases. Serum BRAT1-Ab and WDR1-Ab levels were significantly higher in AMI patients than HDs but not DM patients ( Figures 3A, B ). At the cutoff value, the BRAT1-Ab positivity rates were 3.1%, 10.2%, and 8.6% for the HDs, AMI, and DM, respectively. For WDR1-Ab, the positivity rates were 1.6%, 16.4% and 14.1% for the HDs, AMI, and DM, respectively ( Table 3 ). Hence, serum BRAT1-Ab and WDR1-Ab more effectively predict CVD than DM.

The ROC analysis parameters including AUC, 95% CI, cutoff value, sensitivity, specificity, and P value are shown in Figures 3C – F . The AUCs of BRAT1-Ab and WDR1-Ab for AMI were 0.64 (95% CI = 0.5649–0.7013) and 0.66 (95% CI = 0.5895–0.7224), respectively. The AUCs of BRAT1-Ab and WDR1-Ab for DM did not significantly increase to > 0.6 and were only 0.55 (95% CI = 0.4771–0.6185) and 0.54 (95% CI = 0.4604–0.6029), respectively.

The foregoing results indicate that serum BRAT1-Ab and WDR1-Ab effectively predict the onset of atherosclerosis-related diseases.

AS-associated diseases and cancers are major global causes of mortality and morbidity. They share common modifiable pathogenesis risk factors. Thus, the prophylactic strategies used against atherosclerotic vascular disease may also be efficacious against cancers (6, 50). We used the TIMER2.0 database to analyze the mRNA expression levels across all TCGA tumors and identify the differences in BRAT1 and WDR1 expression between tumors and adjacent normal tissues. Figure 4A shows that relative to adjacent normal, healthy tissues, BRAT1 was significantly upregulated in BLCA (bladder urothelial carcinoma), BRCA (breast invasive carcinoma), CHOL (cholangiocarcinoma), COAD (colon adenocarcinoma), ESCA (esophageal carcinoma), GBM (glioblastoma multiforme), HNSC (head and neck squamous cell carcinoma), KICH (kidney chromophobe), KIRC (kidney renal clear cell carcinoma), KIRP (kidney renal papillary cell carcinoma), LIHC (liver hepatocellular carcinoma), LUAD (lung adenocarcinoma), LUSC (lung squamous cell), PRAD (prostate adenocarcinoma), READ (rectum adenocarcinoma), STAD (stomach adenocarcinoma), and UCEC (uterine corpus endometrial carcinoma). BRAT1 was significantly upregulated in all five types of GI cancer (COAD, ESCA, LIHC, READ, and STAD). Contrastingly, WDR1 was significantly upregulated only in LIHC ( Figure 4B ). Therefore, the expression of BRAT1 but not WDR1 was positively correlated with GI cancers. As the statistical strategies and sample sizes differed between the TIMER2.0 and GEPIA databases, the latter was used to confirm BRAT1 upregulation in ESCA, STAD, READ, LIHC, PAAD, and COAD and their adjacent normal tissues. Matched TCGA normal data were used as controls. ( Figure 4C ). RNA-seq data generated by the TNMplot online tool showed that BRAT1 was upregulated in six different GI cancers compared with normal tissues ( Figure 4D ). We also measured BRAT1 protein expression in normal and LIHC, PAAD, and COAD tissues using Clinical Proteomic Tumor Analysis data ( Figure 4E ).

The preceding results acquired from multiple online databases indicated high BRAT1 expression in GI cancers and suggested that BRAT1 might play a crucial role in serum detection of these diseases.

AlphaLISA analysis was performed on HDs and patients with ESCC, GC, and CRC to establish whether serum BRAT1-Ab is a novel serum biomarker of these GI cancers ( Figure 5A and Table 4 ). Serum BRAT1-Abs levels were significantly higher for patients with ESCC, GC, and CRC than the HDs. Serum WDR1-Abs levels were markedly elevated in patients with ESCC and GC but not in those with CRC ( Supplementary Figure 1A ; Supplementary Table 1 ). At a cutoff value of the mean HD value plus 2 SD, the BRAT1-Abs positivity rates for the HDs and patients with ESCC, GC, and CRC were 6.3%, 17.9%, 15.6%, and 11.5%, respectively ( Table 4 ). The WDR1-Abs positivity rates for the HDs and patients with ESCC, GC, and CRC were 4.3%, 27.7%, 17.4%, and 16.0%, respectively ( Supplementary Table 1 ).

We then used ROC analysis to assess the abilities of these markers to detect GI cancers. Figures 5B – D show that the AUCs of BRAT1-Abs for ESCC, GC, and CRC were 0.68 (95% CI = 0.6203–0.7489), 0.64 (95% CI = 0.5676–0.7229), and 0.62 (95% CI = 0.5560–0.6924), respectively. The AUCs of WDR1-Abs for ESCC, GC, and CRC were 0.68 (95% CI = 0.6233–0.7479), 0.68 (95% CI = 0.6131–0.7651), and 0.59 (95% CI = 0.5306–0.6646), respectively ( Supplementary Figures 1B – D ). The cutoff value, sensitivity, specificity, and P value are shown under the curves. Significant increases > 0.6 were only observed for the AUCs of BRAT1-Ab vs. ESCC, GC, and CRC. In contrast, the AUC of WDR1-Ab vs. CRC was < 0.6.

The foregoing findings revealed that relative to WDR1-Ab, BRAT1-Ab is a superior predictor of GI cancers and the AS-associated biomarker BRAT1-Ab is a potential predictor of the onset of ESCC, GC, and CRC.

The AUC values were highest for ESCC ( Figure 6 ). Thus, 98 surgical ESCC cases were analyzed and used to validate the correlations between BRAT1-Ab and overall survival. We divided the BRAT1−Ab levels for ESCC into quartiles Q1 (n = 25), Q2 (n = 24), Q3 (n = 24), and Q4 (n = 25). There were no statistically significant differences in OS among groups ( Figure 6A ) according to a log-rank test (P = 0.12) ( Table 5 , left panel). However, the Q4 group presented with poor ESCC prognosis at 5–60 wks post-surgery ( Figure 6B ). The foregoing results show that the highest serum BRAT1-Ab levels (Q4) were associated with poor ESCC prognosis.

We applied various statistical methods to investigate the correlations among the serum BRAT1-Ab levels and the clinicopathological factors. In a Cox proportional hazards regression analysis, tumor depth, SCC-Ag, and BRAT1-Ab were the explanatory variables and gender, age, location, lymph node metastasis, p53-Abs level, WBC, neutrophils, lymphocytes, hemoglobin, platelets, CRP, and albumin were tested ( Table 5 ). A multivariate survival analysis only disclosed statistically significant correlations between the serum BRAT1-Ab levels and lymph node metastasis (P = 0.05), CRP level (P = 0.05), and tumor depth (P = 0.03). Hence, these parameters were designated as independent prognostic factors ( Table 5 , right panel).

Fisher’s exact probability test showed that the BRAT1-Ab levels were not associated with the aforementioned factors ( Table 6 , left panel). A logistic regression analysis demonstrated that platelet count was significantly correlated with high serum BRAT1-Ab levels ( Table 6 , right panel). A Mann-Whitney U test returned similar results for the associations between the foregoing clinicopathological factors and the median BRAT1−Ab levels in ESCC. Only platelet count and high serum BRAT1-Ab level were significantly correlated ( Table 7 ).