Identification of anti-SF3B1 autoantibody as a diagnostic marker in patients with hepatocellular carcinoma

The mouse B cell clones secreting a monoclonal autoantibody reactive to human hepatoma cells were selected from a B cell hybridoma pool constructed using hepatitis B virus X (HBx)-transgenic (tg) mice, as described previously [28]. The cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA). All cell lines originated from humans, except for HT22, which is a mouse hippocampal neuronal cell line. The cells were cultured in Dulbecco’s modified Eagle medium or RPMI-1640 (Thermo Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich, St Louis, MO, USA). Mouse liver or HCC tissues used for western blotting or immunohistochemistry were obtained from Hras12V- or HBx-tg mice provided by Disease Model Research Laboratory of KRIBB [15, 29]. Exosome-containing conditioned media were prepared from 70 to 80% confluent cancer cells, which were cultured for 48 h with exosome-free fetal bovine serum. Human HCC serum samples for this study were provided by the Ajou Human Bio-Resource Bank (AHBB; Suwon, Korea), a member of the National Biobank of Korea (Additional file 1: Table S1). Normal human serum samples (Non-HCC) were provided by the Korean Red Cross (Seoul, Korea). Serum samples were kept at − 70 °C until use.

XC24, an autoantibody secreted from one of our tumor model mouse B-cell clones, was analyzed in this study. The isotype of XC24 autoantibody was determined using a Pierce Rapid ELISA Mouse mAb Isotyping Kit (Thermo) as immunoglobulin (Ig) M. Identified XC24 autoantibody was purified from ascites fluid or hybridoma cell culture media using protein L-agarose (Thermo).

To examine the reactivity of XC24 autoantibody to human cancer cells, flow cytometric analysis and immunofluorescence microscopy were performed. For flow cytometric analysis, the suspended cells were fixed and permeabilized with BD cytoperm/cytofix solution (BD Bioscience, Franklin Lakes, NJ, USA), followed by an incubation with primary antibody solution and with goat anti-mouse IgG F(ab′)₂-FITC. The stained cells were analyzed by FACScalibur (BD) and obtained data were analyzed using CellQuest software (BD). When determining whether the autoantibody-mimotope phage can compete with target cellular antigen for antibody binding, primary antibodies were pre-incubated with each phage at room temperature for 60 min and then used for staining.

To detect the XC24 antibody-specific antigen, western blot analysis of tumor cell lysates was performed. Cell lysates were prepared with NP40 lysis buffer [PBS (10 mM sodium phosphate and 150 mM NaCl, pH 7.4) containing 1% NP-40 and protease inhibitor cocktail (Sigma-Aldrich)]. For the proteomic analysis of the XC24 antigen, HepG2 cell lysates prepared with NP40 lysis buffer were immunoprecipitated with XC24 antibody-conjugated beads. Antibody-conjugated beads were prepared using immunoprecipitation kit (Thermo). The precipitated beads were resolved by 10% SDS-PAGE and analyzed by western blotting or Coomassie Blue staining. XC24 reactive antigen bands were then excised from SDS-PAGE gel and in gel-digested with trypsin (Trypsin Gold, Mass Spectrometry Grade; Promega, Fitchburg, Wisconsin, USA). The peptide extracts of in-gel digestion were analyzed by nano-liquid chromatography electrospray ionization–tandem mass spectrometry (nano-LC–ESI–MS/MS), as previously described [28].

To confirm the result of mass spectrometric analysis, HepG2 cells were transfected with AccuTarget™ siRNA against human SF3B1 or HYOU1 (Bioneer Corporation, Daejeon, Korea; si-SF3B1: sense 5′-CGA AGA UCG CCA AGA CUC A(dTdT)-3′, antisense 5′-UGA GUC UUG GCG AUC UUC G(dTdT)-3′; si-HYOU1: sense 5′-CAG AGA UGG ACC AGA UCU U(dTdT)-3′, antisense 5′- AAG AUC UGG UCC AUC UCU G(dTdT)-3′) using Lipofectamine RNAiMAX reagent (Thermo), and then reverse transcription-polymerase chain reaction (RT-PCR) and western blot analysis were performed 72 h after transfection. AccuTarget™ Negative Contol siRNA (Bioneer) was used as negative control. Total RNA was extracted from cells using Qiagen RNA extraction kit (Qiagen, Hilden, Germany) and the first-strand cDNA was synthesized using Superscript III (Thermo). RT-PCR was performed using the following primer pairs (Bioneer; SF3B1: forward 5′-CTC GAG ATG GCT GCA TTG CGT CAG ATT AC-3′, reverse 5′-CTC GAG CAT TAC TGA GTC CTC TGT AGT-3′; GAPDH forward 5′-AGA GAC TGG AGC CAT TAC TTC-3′, reverse 5′-CAA CCT CAG CAG ACT GTG TG-3′). To further confirm whether XC24 antigen is SF3B1, the immunoprecipitated complex with anti-SF3B1 antibody (Bethyl Laboratories, Montgomery, TX, USA) was probed with XC24 antibody or vice versa.

For the immunofluorescence microscopy analysis, cells plated on 18 × 18-mm glass coverslips in 6-well plates were treated with BD cytofix/cytoperm solution to fix and permeabilize the cells. Then, cells were incubated with antibodies (XC24 or anti-SF3B1 antibody) diluted in BD cytoperm/wash solution (5 μg/mL) and with FITC-conjugated secondary reagents. The immuno-stained coverslips were mounted with Vectashield medium containing DAPI (Vector Laboratories, Burlingame, CA, USA) and analyzed using a Zeiss LSM510 Meta microscope (Carl Zeiss MicroImaging, Jena, Germany).

Subcellular localization of SF3B1 or XC24 antigen was determined by western blotting on subcellular fractionated cell lysates. Cells were washed twice with PBS and lysed using NE-PER nuclear and cytoplasmic extraction reagents (Thermo). Membranes were probed with one of the following antibodies: anti-Lamin B1 (nuclear marker; Santa Cruz Biotechnology Inc. Santa Cruz, CA, USA), anti-GAPDH (cytoplasm marker, Santa Cruz), and anti-SF3B1 antibody or XC24 monoclonal autoantibody.

To evaluate the expression of SF3B1 in HCC tissues, cell lysates prepared using RIPA buffer as previously described [28] or tissue lysates prepared by homogenization and sonication were analyzed by western blot analysis. Exosomes collected from the cell culture media by ultracentrifugation [30] was resolved in RIPA buffer and analyzed by western blotting. The protein concentration was determined by the Bradford method (Bio-Rad, Hercules, CA, USA), and equal amounts of protein (50 μg) were resolved by SDS-PAGE and transferred onto a PVDF membrane (Millipore, MA, USA). The membranes were probed with XC24 monoclonal antibody or anti-SF3B1 antibody; β-actin was probed as a loading control (Abcam) and an exosome marker ALIX was detected as with anti-ALIX antibody (Millipore). Calexin, an endoplasmic reticulum marker, was probed (Abcam) as cell contamination marker. Positive bands were detected by horseradish peroxidase (HRP)-linked anti-mouse IgG/M/A antibody (Abcam) or other corresponding secondary reagents, followed by enhanced chemiluminescence reagents (GE Healthcare Life Sciences, Pittsburgh, PA, USA). Band intensity was quantified using Image J (NIH, USA) and the relative intensity compared to β-actin was calculated.

For the staining of mouse tissue, 4-μm-thick sections were cut from the formaldehyde-fixed and paraffin-embedded tissue specimens and mounted on charged glass slides as described previously [15]. Tissue section samples were incubated overnight at 4 °C with anti-SF3B1 antibody. The sections were then incubated with HRP-conjugated secondary antibody for 30 min at room temperature, and 3,3′-diaminobenzidine (DAB) substrate chromogen solution was applied. Finally, the sections were counterstained with hematoxylin, dehydrated using graded alcohol and xylene series, and mounted with Permount (VWR International, Strasbourg, France). The photomicrographs were acquired at 200× or 400× magnification. DAB intensity was quantified using Image J (NIH) and plotted.

We screened the conformational epitopes against XC24 antibody with the cyclic peptide library Ph.D.™-C7C [New England Biolabs (NEB), Ipswich, MA, USA] to use them as detection antigens of SF3B1 autoantibody instead of using a recombinant SF3B1 protein. Screening of conformational epitopes was already proven to be successful in the study of anti-FASN and anti-CK8/18 autoantibody biomarkers [28, 31]; Panning was repeated four times, and sequencing of selected mimotope phages was performed following the manufacturer’s instructions. To confirm the specificity of selected epitopes displayed on M13 phages, phage ELISA was performed [28]. Briefly, the ELISA plate (MaxiSorp™; Thermo) was coated with a selected phage (10¹⁰ pfu/well) in 0.1 M sodium carbonate buffer (pH 8.6). After the wells were blocked with protein-free blocking buffer (Thermo), primary antibody solution (100 ng purified XC24 or XC20 monoclonal antibody in 100 μl blocking buffer) was added and incubated at room temperature for 2 h. HRP-linked anti-mouse IgG/M/A antibody (Sigma-Adrich; 1:2000 diluted in blocking buffer) was used as a secondary reagent. Tetramethylbenzidine (TMB)-based substrate solution (Thermo) was used for color development. When evaluating the effect of cyclic conformation on peptide antigenicity, cyclic phages were reduced and alkylated and then coated onto the ELISA plates described previously [28]. For the competitive FACS analysis to confirm the specificity of selected epitopes, the primary antibody XC24 or XC20 was pre-incubated with the indicated epitope-displaying phages (indicated pfu/100 μl reaction) and used for cell staining. Competitive western blotting was also performed. The HepG2 or other cell lysates were resolved by SDS-PAGE and western blotted with primary antibody XC24 or XC20, which was pre-incubated with epitope-displaying phages (10¹¹ pfu of indicated phages/100 μl reaction).

For the preparation of epitope-display coating antigens for human serum ELISA, we constructed XC24p11 epitope-fused streptavidin expression vector. The streptavidin-coding region was PCR-amplified from Streptomyces avidinii (ATCC 27419) and cloned into a NotI/XhoI-digested pET28a (+) vector. The DNA coding the cyclic peptide XC24p11 epitope sequence (-CDATPPRLC-) with restriction enzyme sites NdeI and NotI was synthetized (Bioneer; foward 5′-tatg ggt ggt gcg TGC GAC GCG ACC CCG CCG CGT CTG TGC ggt gga ggt tcg gcc-3′, reverse 5′-ggc cgc cga acc tcc acc GCA CAG ACG CGG CGG GGT CGC GTC GCA cgc acc acc ca-3′; underlined sequences correspond to XC24p11 epitope sequence) and cloned into the streptavidin-cloned pET28a (+) vector. The cyclic epitope gene was separated from the N-terminal His-tag and streptavidin gene by a linker encoding the amino acids GSGSA. DsbA, a bacterial thiol disulfide oxidoreductase (TDOR), was also cloned into the cyclic epitope-streptavidin cloned-pET28a (+) vector to catalyze the intrachain disulfide formation of the cyclic epitope. In addition, for the effective disulfide bond formation, SHuffle^® T7 (NEB) was used as a host cell, which is an E. coli K12 strain suitable for T7 protein expression with enhanced capacity to correctly fold proteins with multiple disulfide bonds in the cytoplasm.

Transformants of E. coli strain SHuffle^® T7 with XC24p11 epitope-fused streptavidin expression vector were grown overnight at 30 °C in 2× YT broth containing kanamycin (50 μg/ml). The culture was diluted 100-fold into fresh medium and grown in a shaking incubator at 30 °C. When the culture attained an absorbance at 600 nm of 4, isopropyl-β-d-thio-galactopyranoside (Sigma-Aldrich) was added to a final concentration of 1 mM, and incubation was continued overnight at 25 °C. Cells were harvested at 24–26 h post-inoculation by centrifugation, washed with TBS [10 mM Tris–HCl and 150 mM NaCl (pH 7.4)], and pelleted by centrifugation. The cells were resuspended in an ice-cold solution of 5% glycerol, 50 mM Tris–HCl (pH 7.4), and 2.0 mg/ml lysozyme, placed on ice for 30 min, and the cell suspension was sonicated to obtain the cell lysate. The lysate solution was centrifuged (10,000×g for 90 min, 4 °C) to pellet the cellular debris. The supernatant was decanted and filtered through a 0.2-μm filter. The lysate solution, to which NaCl and imidazole were added to a final concentration of 500 and 10 mM, respectively, was then applied over a Talon affinity column (Takara Bio, Mountain View, CA, USA) equilibrated in 50 mM Tris–HCl (pH 7.4) containing 0.25 M NaCl, 10 mM imidazole, and 5% glycerol. The column was washed with 10 column volumes of equilibration buffer and then eluted with an imidazole gradient up to 0.5 M in equilibrium buffer. The eluate fractions containing streptavidin antigen were pooled and concentrated using Amicon^® Ultra Centrifugal Filters (Merck Millipore, Darmstadt, Germany) down to 1 ml. The concentrate was applied to a HiLoad 16/600 Superdex 200 prep grade column (GE) equilibrated with PBS containing 1 mM β-mercaptoethanol. Each fraction obtained by size-exclusion chromatography was analyzed by SDS-PAGE and western blotting, and fractions containing tetrameric epitope-fused streptavidin were pooled and used for further analysis and human serum ELISA.

The ELISA plate, MaxiSorp-, or biotin-coated plates (Thermo) were coated with the indicated amounts of epitope-fused streptavidin in 0.1 M sodium carbonate buffer (pH 8.6) overnight at 4 °C. After the wells were blocked with protein-free blocking buffer, XC24 primary antibody solution (containing the indicated amount of purified monoclonal antibody in 100 μl blocking buffer) was added and incubated at room temperature for 2 h. HRP-linked anti-mouse IgG/M/A antibody (Sigma-Aldrich; 1:2000 diluted in blocking buffer) was used as a secondary reagent. TMB solution was used for color development. For the detection of reactivity of patient sera to XC24p11-streptavidin, the MaxiSorp plates were coated with XC24p11 epitope-fused streptavidin at 500 ng/well, and after blocking with protein-free blocking buffer as described above, the plates were treated with albumin-depleted human sera (1:1000 diluted in blocking buffer) and detected by HRP-conjugated anti-human IgG/M/A antibody (1:2000 diluted in blocking buffer). Albumin depletion of the human serum was performed using Affi-Gel^® Blue Gel following the manufacturer’s instructions (Bio-Rad). Empty-streptavidin (Eph) without a peptide epitope insert was used as control coating antigen. Alpha-fetoprotein (AFP) levels in human sera were evaluated with Human alpha-Fetoprotein Quantikine ELISA Kit (R&D systems, Minneapolis, MN, USA).

To identify the TA autoantibody biomarkers, each autoantibody produced by HBx-Tg mice B cell hybridoma clones was purified and reactivity to human cancer cells were analyzed. XC24, one of such TA autoantibodies (Fig. 1a), showed specific responses to HepG2 and other cancer (Hep3B, HeLa, LNCaP-LN3, MCF7, HT29, A549, H460) or non-cancer cell lines (Chang, HT22) (Fig. 1b). XC24 also could be applied to western blot analysis, which showed its target antigen is a protein with molecular weight of about 170 kDa (Fig. 1c). To identify an XC24 antibody-specific antigen, HepG2 cell lysates were immunoprecipitated with XC24-conjugated beads and resolved by SDS-PAGE. The protein bands corresponding to XC24 antigen, which was confirmed by western blotting, was then excised and in-gel digested with trypsin (Fig. 1d). The peptide extracts of in-gel digestion were analyzed by nano-liquid chromatography electrospray ionization-tandem mass spectrometry (nano-LC–ESI–MS/MS). Several protein candidates were identified, including splicing factor 3b subunit 1 (SF3B1) and hypoxia up-regulated protein 1 (HYOU1) (Table 1), among which the highest Mascot Score was shown by SF3B1. To confirm the result of mass spectrometric analysis, tumor cells were transfected with small interfering RNAs (siRNA) against SF3B1 or HYOU1, and total cell lysates were analyzed by western blotting. As shown in Fig. 1e, the protein bands stained with XC24 autoantibody in HepG2 or MCF7 cell lysate was disappeared when SF3B1 was knocked down by siRNA. The immunoprecipitated complex with anti-SF3B1 antibody was also probed with XC24 antibody or vice versa to confirm that XC24 antigen is SF3B1 (Fig. 1f, Additional file 2: Fig. S1). SF3B1 is a component of the pre-mRNA splicing multi-protein complex SF3B, which functions mainly in nucleus [32]. To confirm the XC24 autoantibody-specific antigen as SF3B1, the intracellular staining with XC24 antibody was compared with anti-SF3B1 antibody staining. As shown in Fig. 1g, commercial antibody staining against SF3B1 showed nuclear as well as cytoplasmic staining. XC24 autoantibody staining was also localized in nucleus as well as cytoplasm, but mainly in cytoplasm and plasma membrane. Occasionally, intracellular cell staining with antibodies may be influenced by its accessibility to subcellular compartments. To confirm again XC24 antigen as SF3B1, cellular fractionation and western blot analysis was performed. When nuclear envelope was disrupted XC24 antibody also detect nuclear fraction more than cytoplasmic fraction as commercial anti-SF3B1 antibody (Fig. 1h, Additional file 3: Fig. S2). Collectively, XC24 antibody was revealed as a tumor-associated antibody against SF3B1.

Northern blot analysis and whole mount in situ hybridization revealed SF3B1 to be ubiquitously expressed in a variety of adult tissues and mid-gestation embryos [32]. Recently, SF3B1 mutations have been described in several malignancies including HCC [33‐36], which cause aberrantly spliced transcripts, thus providing a link between SF3B1 mutations and disease [37]. Considering its ubiquitous expression in human tissues, SF3B1 must be recognized as self in the healthy human body; however, a certain alteration in the antigenic presentation of SF3B1, such as its expression level, localization, mutation or their combinations, can induce a specific autoantibody during tumorigenesis. We examined first, therefore, the SF3B1 expression in liver tissues. Western blot analysis showed that SF3B1 expression was significantly increased by about 2- or 3-fold (p < 0.05) in HCC tissues of H-ras12V transgenic (Hras12V-Tg) mice, another mouse model of liver cancer [29], compared to normal liver tissues (Fig. 2a). The elevation of SF3B1 was further confirmed by immunohistochemistry analysis. In the tumor region of the liver tissue of HBx-Tg mice (especially in large tumor region), SF3B1 was increased compared to that of the non-tumor region (Fig. 2b, Additional file 4: Fig. S3). Public gene expression database analysis using GENT analysis (Gene Expression across Normal and Tumor tissue; http://​medicalgenome.​kribb.​re.​kr/​GENT/​) also showed the elevation of SF3B1 in human liver tumor tissues compared to normal tissues (Fig. 2c). Collectively, SF3B1 expression is increased in liver cancer, which may be the primary cause of induction of abnormal autoimmune-response.

For the induction of TA autoantibodies, the elevated intracellular antigen should be released to the extracellular region, in which antigen can stimulate immune system. Cell death or necrosis accompanied by tumorigenesis has been assumed to be a predominant condition that exposes intracellular components to extracellular space [38]. Recent studies on TA exosomes, however, showed that intracellular antigens can be exposed to the extracellular region even without cancer cell necrosis. SF3B1 had already been confirmed as a component of TA exosomes derived from various cancers such as colorectal and ovarian cancer [39, 40]. Therefore, we examined whether SF3B1 is secreted as an exosomal component from HCC by analyzing exosomes derived from HCC cancer cell lines. Exosomes from the HepG2 or Huh7 cancer cells were collected by ultracentrifugation and probed with XC24 as well as with commercial anti-SF3B1 antibody (Fig. 2d). Antigen bands stained with XC24 or anti-SF3B1 antibody are overlapped at the same protein band of molecular weight of about 170 kDa. Several protein bands less than molecular weight of 170 kDa were also detected with anti-SF3B1 or XC24 antibody at the same locations (Fig. 2d; indicated by red arrow head), which may be the degradation fragments of SF3B1. Interestingly, an excess amount of low molecular weight protein (about 30 kDa; indicated by asterisk) was detected by XC24 autoantibody, which also may be a fragment derived from SF3B1, but not detectable with commercial anti-SF3B1 antibody.

The results described above implicates that SF3B1 is increased in HCC and can induce a specific autoantibody as it is secreted from cancer cells as an exosomal component. Since the characteristics of human HCC cells are similar to those of HCC model mouse [15‐27], human SF3B1-specific autoantibody is also expected to be induced in human HCC patients. Thus, we had screened the conformational epitopes of XC24 antibody from a cyclic peptide library to use them as detection antigen of anti-SF3B1 autoantibody. Screening conformational epitopes using a cyclic peptide display phage library was already proven to be successful in our previous studies [28, 31]; hence, the same screening method was applied for anti-SF3B1 autoantibody. Bio-panning of M13 phage library displaying 10¹¹ to 10¹³ cyclic peptides with XC24 autoantibody was repeated four times and specific binding phages were amplified about 1000 times (Fig. 3a). The epitope sequences of amplified phages from about 20 phage plaques were determined and 10 different epitopes were obtained (Table 2). In the ELISA using these cyclic peptide epitope-displaying M13 phages as the coating antigens, XC24 antibody showed different titers for each phage display epitope, and the reactivity against XC24p1, XC24p4, XC24p5, XC24p8, XC24p10, and XC24p11 epitopes was shown to be high (Fig. 3b). These epitopes did not show any reactivity against another autoantibody, XC20, indicating that the selected cyclic peptides display XC24 antibody-specific epitopes. All of selected epitope sequences showed a consensus sequence motif of xDxTPxx type (D: Aspartic acid; T: Threonine; P: Proline; x: diverse amino acids except cysteine), although this sequence motif of conformational epitope was not found in the protein primary sequence of SF3B1. To confirm whether the specificity of cyclic peptide epitopes against XC24 antibody depends on their cyclic structures, the antigenic cyclic peptide-displaying phages were reduced with dithiothreitol and alkylated, and were then subjected to ELISA. As shown in Fig. 3c, the reactivity of antigenic phages to the XC24 antibody disappeared when the disulfide bonds were reduced and the cyclic structures were linearized, confirming that the cyclic form of these epitopes is important for their binding to XC24 autoantibody. Competitive FACS analysis was performed to confirm how well the antigenic epitopes simulate the actual intracellular antigenic structures. After HepG2 cells were fixed and permeabilized, maintaining most of the structural features of antigenic determinants, they were probed with XC24 or XC20 antibodies and specific bindings of each antibody were detected (Fig. 3d). However, by adding antigenic epitope-displaying phages, such as XC24p1, XC24p8, or XC24p11, the XC24 antibody-specific binding was inhibited to almost zero. These phages did not affect the XC20 antibody binding to HepG2 cells; rather, they enhanced the antibody binding, maybe by increasing non-specific reactions. The antigenic specificity of these epitope-displaying phages was also confirmed by competitive western blot analysis (Fig. 3e). The XC24 antibody response to the HepG2, MCF7 or HT29 cell lysate on western blot was completely inhibited by XC24-specific epitope phages (XC24p1, XC24p8, and XC24p11), but not by epitope-free M13 phage (Eph). Collectively, we confirmed that the cyclic peptide epitopes including XC24p11, which sufficiently mimics the conformational characteristics of the XC24 autoantibody epitope, can be used as a detection antigen of XC24 autoantibody instead of cellular SF3B1 with a molecular weight of 155 kDa. In order to establish practical diagnostic tests, it is necessary to mass-produce the coated antigens. For this purpose, we fused the cyclic peptide epitope sequences to the amino-terminus of matured streptavidin (without the secretion leader sequence) and expressed as a streptavidin tetramer in E. coli (Fig. 4a). Among X24-reactive epitopes, XC24p11 was expressed on streptavidin carrier, because of protein expression efficiency. DsbA, a bacterial thiol disulfide oxidoreductase (TDOR), was also cloned into an antigen-expressing pET28a vector and expressed simultaneously with cyclic peptide display streptavidin to induce the correct disulfide bond necessary for formation of the cyclic peptide structure. In addition, SHuffle^® T7 was used as a host cell for effective disulfide bond formation. The expressed epitope-displaying streptavidins were purified using His-tag affinity chromatography and Superdex 200 size-exclusion chromatography. The streptavidin tetramer was confirmed by its size of about 80 kDa in the elution profile of Superdex 200 size-exclusion chromatography and SDS-PAGE (Fig. 4b). For the detection of human serum autoantibody, the conditions of each step in ELISA were optimized. The antibody binding to streptavidin antigens was compared on MaxiSorp plate and biotin-coated plate to select a solid phase suitable for coating epitope-fused streptavidin antigens (Fig. 4c). XC24 antibody binding to the XC24p11-streptavidin antigen was saturated at about 400 ng/well in both type of coating plates. However, antibody binding was diminished when the amount of coating antigen in the biotin-coated plate was increased. The decrease of antibody binding was particularly noticeable when the amount of primary antibody was low. When considering these properties of the streptavidin antigen on coating plates and the sensitivity of ELISA, the MaxiSorp plate seems to be more suitable for the human serum autoantibody ELISA, which must detect the autoantibody more stably even at low concentrations. Protein-free blocking buffer was used for plate blocking and human serum was pretreated with albumin depletion resin to remove albumin, which is a typical source of reducing potential in blood, and diluted 50-fold in protein-free blocking buffer.

Using these conditions, ELISA for the detection of human serum autoantibody against SF3B1 was performed. A total of 102 serum samples from HCC patients (Additional file 1: Table S1) and 85 serum samples of healthy control subjects were analyzed, and the reactivity to XC24p11 represented by the difference of antibody reaction between XC24p11-streptavidin (XC24p11-STA) and epitope-free streptavidin (Eph-STA). As shown in Fig. 5a, the reactivity of HCC patient sera against XC24p11 was significantly different from that of the non-HCC control sera, with an AUC value of 0.87 [95% confidence interval (CI) 0.8219–0.9243, p < 0.0001]. The sensitivity of this ELISA was 73.53% and the specificity was 91.76% when the cutoff value (CV) was 0.09194. AFP, a conventional HCC biomarker, was also determined in the same sample set for comparison (Fig. 5a). The CV for the AFP biomarker of 40 ng/ml and ultrasonography were the criteria for evaluation of HCC currently used by clinicians today. AFP detection for HCC diagnosis showed a sensitivity of 52.94% and specificity of 100% when the CV was 40 ng/ml with an AUC value of 0.92 (95% CI 0.8781–0.9637, p < 0.0001). The responses to XC24p11 were analyzed also depending on tumor stage, size, and viral infection status (Fig. 5b). Anti-SF3B1 antibodies were detected in patient’s sera even at the early tumor stage (T1) or in small size tumor (size < 2 cm). Viral infection, which is a major cause of HCC, did not influence anti-SF3B1 autoantibody level. In all subgroups of HCC patients, the portion of the anti-XC24p11 response above the CV was about 70–80%. By contrast, in non-HCC samples, the portion of the anti-XC24p11 response above the CV was about 10%.

For decades, AFP has been considered as a gold standard in the diagnosis of liver cancer with high reliability based on long accumulated clinical experiences, even though its diagnostic sensitivity to HCC is low at about 50% [13]. Thus, a novel HCC diagnosis method may be easily acceptable when it complements the defect of AFP test. In this context, we examined the use of the anti-SF3B1 autoantibody test in providing additional information after AFP-based HCC diagnosis. As shown in Fig. 6a, the detection of anti-SF3B1 antibody showed no correlation with AFP level (Pearson’s r = 0.00727). The anti-SF3B1 antibody response pattern of HCC patients was not significantly distinguished between in AFP-low or AFP-high groups [ns (p > 0.05); Fig. 6b]. However, the simultaneous detection of AFP and anti-SF3B1 antibody in serum allowed the subjects to be grouped in detail (Fig. 6c). When the diagnostic values of anti-XC24p11 autoantibody or AFP were indicated as positive (+) or negative (−) according to the value above or below the CV, subjects were classified into 3 groups [AFP(+)/XC24(+), AFP(+)/XC24(−) or AFP(−)/XC24(+), AFP(−)/XC24(−)]. Most of Non-HCC subjects showed AFP(−)/XC24(−) pattern, whereas most of HCC patients showed AFP(+)/XC24(+), AFP(+)/XC24(−) or AFP(−)/XC24(+) pattern. ROC curve analysis of these results showed that although AFP was not detected in half of patients with HCC, the simultaneous detection of AFP and anti-SF3B1 antibodies can discriminate HCC up to 87.25% with AUC value of 0.9081 (95% CI 0.8635–0.9528, p < 0.0001), which is also superior to anti-SF3B1 antibody biomarker detection only (73.53%; Fig. 6d).