Identification of specific and common diagnostic antibody markers for gastrointestinal cancers by SEREX screening using testis cDNA phage library

The present study was planned to identify novel serum antibody markers for digestive organ cancers. We have used screening by phage expression cloning and identified novel fourteen antigens in this experiment. The presence of auto-antibodies against these antigens in serum specimens was confirmed by western blotting. As for auto-antibodies against fourteen antigens, AlphaLISA (amplified luminescence proximity homogeneous assay) assay was performed in the sera of gastrointestinal cancers patients to confirm the results. Serum antibody levels against these fourteen recombinant proteins as antigens between healthy donors (HD) and esophageal squamous cell carcinoma (ESCC) patients, gastric cancer (GC), or colon cancer (CC) were compared. The serum levels of all fourteen auto-antibodies were significantly higher in ESCC and GC than those of HD. Among those auto-antibodies, except ECSA2 and CCNL2, were also detected significantly higher levels in CC than those of HD. Receiver operating curve (ROC) revealed similar results except CCNL2 in CC. AUC values calculated by ROC were higher than 0.7 in auto-antibodies against TPI1, HOOK2, PUF60, PRDX4, HS3ST1, TUBA1B, TACSTD2, AKR1C3, BAMBI, DCAF15 in ESCC, auto-antibodies against TPI1, HOOK2, PUF60, PRDX4, TACSTD2, AKR1C3, BAMBI, DCAF15 in GC, and auto-antibodies against TPI1, HOOK2, PUF60 in CC. AUC of the combination of HOOK2 and anti-p53 antibodies in ESCC was observed to be as high as 0.8228. Higher serum antibody levels against ten antigens could be potential diagnostic tool for ESCC. Higher serum antibody levels against eight antigens could be potential diagnostic tool for GC, and serum antibody levels against three antigens could be potential diagnostic tool for CC.


INTRODUCTION
There is a specific common mechanism of expressing some molecules (e.g., CEA) (these have been confirmed to be expressed in embryonic stage) during gastrointestinal tumorigenesis. However, there are few diagnostic biomarkers that are able to specifically detect and discriminate various cancers at an early stage. Recent research in the field of cancer therapeutic strategy suggests that there are few biomarker candidates that are able to predict the therapeutic effects. Moreover, these could also function as companion diagnostic tools [1]. The recently developed serum anti-p53 antibodies test was useful for the detection of superficial ESCC [2]. As positive rate of serum anti-p53 antibody was around 30%, other serum antibodies were necessary for a combinational test to improve positive rate and diagnostic accuracy.
The serological identification of antigens by recombinant cDNA expression cloning (SEREX) is an effective screening method for identification of serum antibody-type tumor markers [3]. SEREX could be utilized for the immune-screening of cDNA libraries prepared from tumor specimens with either autologous or allogeneic sera. Furthermore, sequencing the isolated cDNA clones easily identified antigens. This made SEREX suitable for the large-scale screening of tumor antigens. SEREX has been applied to various human tumor types and has identified more than 1000 novel tumor antigens (SEREX antigens) [4]. We have previously performed large-scale SEREX screenings and identified numerous ESCC SEREX antigens. Furthermore, 21 antigens were reported to have included the tumor suppressor p53, the oncoproteins phosphatidylinositol 3-kinase, and stathmin [5][6][7][8][9][10][11][12][13]. Anti-p53 antibody marker has been utilized for not only ESCC but also gastric cancer (GC) as well as colon cancer (CC). Further, the antibody levels against other SEREX antigens might be responsible for GC and CC. Thus, we examined whether antibody markers against ESCC SEREX antigens were common among digestive organ cancers.

Auto-antibody levels showed increase in patients with digestive organ cancers
We examined the levels of serum auto-antibodies using the sera of HD and patients with ESCC, GC, and CC procured from Chiba University Hospital. HD subjects from Higashi Funabashi Hospital were selected as healthy volunteers. The results of AlphaLISA showed that the levels of TPI1-Abs, HOOK2-Abs, PUF60-Abs, PRDX4-Abs, HS3ST1-Abs, TACSTD2-Abs, TUBA1B-Abs, AKR1C3-Abs, BAMBI-Abs, DCAF15-Abs, PDE4DIP-Abs, ECSA1-Abs, ECSA2-Abs, and CNNL2-Abs were significantly higher in patients with ESCC, GC, or CC than in HD (Figures 2A-2F and Supplementary Figures 2A-2H). The cutoff value was determined as the average + 2SD of HD, to keep the 95% confidence interval. Receiver operating curve (ROC) analysis was carried out to evaluate the ability of these markers to detect ESCC, GC, and CC. The areas under the curve (AUCs) of TPI1, HOOK2, PUF60, PRDX4, HS3ST1, TUBA1B, TACSTD2, AKR1C3, BAMBI, and DCAF15 for ESCC were significantly larger than 0.700 ( Table 1). The highest AUC values were obtained for HOOK2 when compared with AUC values of ESCC.

Combined ROC curve increases AUC
The receiver operating combined curve (ROC) analysis was carried out to evaluate the facility of these markers to detect ESCC, GC, and CC ( Figure 4 and To purify the SEREX-identified proteins, the insertion sequences of the 14 pBluescript plasmids were ligated in-frame into GST-tagged expression vectors. We confirmed by sequence analysis that the recombinant pGEX-4T-3 plasmids were properly recombined and GST-tagged recombinant proteins were affinity-purified using glutathione-Sepharose. To confirm the recombinant proteins to be the GST-tagged one that react with autologous plasma, the proteins were lysed in a SDS sample buffer, incubated at 100°C for 3 min, Affinity-purified GST-tagged antigens were separated on 11% SDSpolyacrylamide gels followed by Western blot using anti-GST antibody. All samples were examined simultaneously, at the same time on the same membrane. www.oncotarget.com Supplementary Figure 3). The antibodies with CEA, only the AUC of DCAF15+CEA increased to 0.7223 in ESCC ( Figure 4A and Supplementary Figure 3A). However, the AUC of TPI1+CEA decreased to 0.7435 in GC ( Figure  4B and Supplementary Figure 3B). Further, in CC, HOOK2+CEA showed AUC of 0.8075 ( Figure 4C and Supplementary Figure 3C). The AUC of all the antibodies combined with anti-p53 antibody, HOOK2 and BAMBI showed 0.8228 in ESCC ( Figure 4D and Supplementary Figure 3D). There was no antibody indicating AUC larger than 0.800 in CC ( Figure 4E and Supplementary Figure  3E). The AUC was summarized in early or advanced stages cancers were showed in Table 2. Advanced stage cancers showed higher AUC than that of early stage cancers. Further, in ESCC and CC, the efficiency of early diagnosis was increased when combined with HOOK2 and anti-p53 antibody (AUC of ESCC: 0.7985, AUC of CC: 0.7669).

DISCUSSION
We have identified novel potential diagnostic markers for digestive organ cancers by SEREX screening. Those serum antibody markers were detected using purified GST-fusion proteins as antigen. Two hundred seventy-seven patients with various cancers were evaluated for the presence of various-Abs. Patients with confirmed digestive organ cancers demonstrated  significantly higher levels of the antibodies against most ESCC SEREX antigens. This suggested that ESCC, GC, and CC have a common carcinogenesis process. However, some of the antibody markers showed differential antibody levels, i.e., all 14 markers were significantly higher in patients with ESCC or GC in comparison to those in HDs. On the contrary, the levels of ECSA2-Ab and CCNL2-Ab were not significantly different between patients with CC and HDs (Figure 2 and Supplementary Figure 2). Similar results were also attained by ROC analysis. The AUCs were higher than 0.6 for most of the markers except ECSA2 and CCNL2 in comparison to CC (Table 1). Also, the AUCs higher than 0.7 were observed for ten markers versus ESCC, eight markers versus GC, and three markers versus CC. Thus, tumor nature of ESCC might be more similar to GC than to CC. It is conceivable that HOOK2-Ab, PUF60-Ab, and TPI1-Ab are common markers for digestive organ cancers. Furthermore, the combined ROC analysis of antibodies with anti-p53 antibody, and CEA showed elevation in AUCs of almost antibodies in various cancers. The combinations of anti-p53 antibody and HOOK2 markers were valuable for early detection of ESCC and CC (Figure 4 and Table 2). Therefore, the combination of antibodies with anti-p53 antibody and CEA is a potential approach for the diagnosis of digestive organ cancers. However, prospective multi-institutional studies comparing the sensitivity and specificity of this combinational detection approach are necessary. AlphaLISA is an excellent method for measuring antibody levels as compared to ELISA because it has low variations, stable background, and high specificity. It does not involve plate-washing steps, but instead involves mixing of antigens with antibodies in sera followed by the addition of donor and acceptor beads. For instance, Figure 2 showed highly reproducible results, including distributions, P values, and positive rates despite using different sets of sera from healthy donors and patients. The precise measurement offered by AlphaLISA might enable establishment of antibody markers, although most of the existing tumor diagnosis methods involved antigen markers, with the exception of the anti-p53 marker [2,15]. The measurement of antibodies was more sensitive in comparison to the measurement of the antigen levels owing to the stability of IgG proteins and their amplification by repeated exposures to antigenic proteins. Prior to development, highly-malignant tumors could induce necrosis, leading to exposure of intracellular antigenic proteins to plasma. Therefore, using combinational antibody detection approach might enable precise tumor diagnosis. In this study, we examined and proposed some of these candidate markers for the early diagnosis of ESCC, CC, and GC. As per our knowledge, no other studies have suggested such kind of candidate markers in digestive organ cancers. Our study would be an important approach for further selecting diagnostic marker candidates for digestive organ cancers.

Clinical samples
The present study was performed in accordance with "The Code of Ethics of the World Medical Association" (Declaration of Helsinki). The Local Ethical Review Board of the Chiba University, Graduate School of Medicine, and those of co-operating hospitals approved this work. Sera of patients with ESCC (n = 85), GC (n = 96), and CC (n = 97) were obtained from the Department of Frontier Surgery, Chiba University Hospital, Chiba, Japan (Supplementary Tables 1-5). Sera of health donors (HDs) (n = 96) were obtained from Higashi Funabashi Hospital. Written informed consent was obtained from all participants prior to this study. Each serum sample was centrifuged at 2,000 × g for 10 min and then the supernatant was stored at -80°C until use. Repeated thawing and freezing of samples were avoided.

Screening by expression cloning
Recombinant DNA studies were performed with the official permission of the Chiba University Graduate School of Medicine and were carried out in accordance with the rules of the Japanese government. We used a λZAP II phage cDNA library prepared from the mRNA of the T.Tn cells and a commercially available human fetal testis cDNA library (Uni-ZAP XR Premade Library, Stratagene, La Jolla, CA) to screen for clones that were immunoreactive against serum IgG from patients with ESCC as described in earlier studies [16,17]. Escherichia coli XL1-Blue MRF' was infected with λZAP II or Uni-ZAP XR phage and the expression of resident cDNA clones was induced after blotting the infected bacteria onto NitroBind nitrocellulose membranes (Osmonics, Minnetonka, MN) The above membranes had been treated with 10 mM isopropyl-β-D-thiogalactoside (IPTG, Wako Pure Chemicals, Osaka, Japan) for 30 min. The membranes with bacterial proteins were rinsed 3 times with TBS-T [20 mM Tris-HCl (pH 7.5), 0.15 M NaCl, and 0.05% Tween-20], and non-specific binding was blocked by incubation with 1% protease-free bovine serum albumin (Nacalai Tesque, Inc., Kyoto, Japan) in TBS-T for 1 h. The membranes were exposed to 1:2000-diluted sera of patients for 1 h. After three washes with TBS-T, the membranes were incubated for 1 h with 1:5000-diluted alkaline phosphatase-conjugated goat anti-human IgG (Jackson ImmunResearch Laboratories, West Grove, PA). Positive reactions were developed using 100 mM Tris-HCl (pH 9.5) containing 100 mM NaCl, 5 mM MgCl 2 , 0.15 mg/mL of 5-bromo-4-chloro-3-indolylphosphate, and 0.3 mg/mL of nitro blue tetrazolium (Wako Pure Chemicals). Positive clones were re-cloned twice until obtaining monoclonality as described in previous studies [16,18,19].
Monoclonal phage cDNA clones were converted to pBluescript phagemids by excision in vivo using the ExAssist helper phage (Stratagene). Plasmid pBluescript containing cDNA was obtained from the E. coli SOLR strain after transformation by the phagemid. The sequences of cDNA inserts were evaluated for homology with identified genes or proteins within the public sequence database (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Statistical analyses
All statistical analyses were carried out using the GraphPad Prism 5 (GraphPad Software, La Jolla, CA). Mann-Whitney U test was used to determine the significance of the differences between the two groups. The predictive values of markers for diseases were assessed by receiver operating curve (ROC) analysis and the cutoff values were set at the values that maximize the sums of the sensitivity and specificity. All tests were twotailed and a P value below 0.05 was considered significant. We calculated antibody group-specific Z-scores for these measures to facilitate the comparison across anti-p53 antibody, CEA and antibodies groups. Z-score analysis was performed after normalization to healthy donors mean values: Z-score = [(control mean) -(individual value)] / (control SD) [32,33].
The Combined ROC analysis was performed by adding each Z score.