Hydrophobicity and molecular mass-based separation method for autoantibody discovery from mammalian total cellular proteins

2.1 Schematic workflow of autoantibody biomarker discovery by TRIzol-based TCP extraction, reversed-phase HPLC and 2-D PAGE

Autoantibody discovery from patient serum or plasma requires an antibody–antigen reaction. In this study, antigens for TCP were obtained from cultured cell lines (Figure 1). This procedure could also be applied to tissue-derived samples. TRIzol-based protein extraction is a simple and scalable method that minimizes endogenous protease digestion (Futami et al., 2014). The TRIzol extraction procedure quantitatively extracts TCP with complete removal of nucleic acid contamination. After dissolving the extracted proteins in 6 M guanidium hydrochloride (GdnHCl), all cysteine residues were modified via the alkyl disulfide reaction, which enhanced protein solubility by introducing positive charges by TAPS-sulfonate (Futami et al., 2015; Futami et al., 2017; Kimura et al., 2020; Miyamoto et al., 2022; Murata et al., 2006). Highly water-soluble TCP was recovered by dialyzing against pure water without any visible precipitates. Reversed-phase HPLC was used to separate proteins according to their hydrophobicity. The fractionated samples were lyophilized and dissolved in SDS-sample buffer. These scalable fractionated samples allowed multiple homogeneous samples to be subjected to SDS-PAGE, enabling their long-term storage under freezing conditions. Notably, this procedure allows the preparation of homogenous multiple polyvinylidene difluoride (PVDF) blots using the above-fractionated samples on a quality-certified commercially available polyacrylamide gel. The subsequent autoantibody discovery procedure was the same as that generally employed in SERPA, including western blotting using patient serum or plasma, in-gel digestion of the targeted protein bands, LC–MS/MS analysis to identify the target antigens, and confirmation of autoantibody reactivity using recombinant proteins.

2.2 Quantitative extraction of TCP using TRIzol

The quality of TCP extracted from cultured cells by TRIzol was validated by comparing it with the total cellular lysate (TCL) obtained by directly lysing the cells in lysis buffer. The protein contents of both TCL and S-cationized TCP by TAPS-sulfonate (TAPS-TCP) were found to be comparable by Coomassie Brilliant Blue (CBB) staining (Figure 2a), indicating the quantitative extraction of total cellular protein using TRIzol. The western blotting band intensities of the internal controls, β-actin, and β-tubulin were also comparable; thus, both samples contained comparable intracellular proteins (Figure 2b). When the diluted plasma from patient #1 with lung cancer was reacted with the PVDF blots, clear bands of approximately 50 kDa (band-1) were detected. Identical reaction patterns with anti-p53 specific antibodies confirmed the presence of an autoantibody-reactive band. The immunoblotting patterns of patients 2, 3, and 4 were comparable in both TCL and TCP, indicating that TCP includes total cellular proteins without losing them during the TRIzol extraction and dialysis steps. Interestingly, the native TCL included a degraded band of approximately 45 kDa, which was not observed for TAPS-TCP. This change may reflect the endogenous protease activity (Figure 2b). The strong detergent TRIzol completely denatured the endogenous protease, whereas the native TCL may have degraded the p53 protein due to the remaining endogenous protease during lysate preparation.

2.3 Separation of TCP by reversed-phase HPLC

We successfully separated water-soluble TCP mixtures that were nucleic acid-free and fully denatured by hydrophobicity. TAPS-TCP, which increased the hydrophilicity by S-cationization (Futami et al., 2015; Futami et al., 2017; Miyamoto et al., 2022; Murata et al., 2006), shifted the chromatographic profile to an earlier elution than nonconjugated TCP (Figure 3a). Total cellular proteins were distributed from 30 to 54 min of elution, with highly discriminated protein bands. The reversible disulfide bonds in TAPS-TCP-conjugated 3-(trimethylammonium)propyl moieties result in proteins migrating as nonconjugated bands on SDS-PAGE under reducing conditions. The native conditions of TCL with nucleic acids also allowed its separation by reverse-phase HPLC (Figure 3b). Extensively degraded nucleic acid fractions were observed around pass-through fractions. Because the elution time for proteins was lesser than that for the nondenaturing detergent Nonidet® P40, almost all protein UV absorption was monitored (Figure 3b). The elution patterns of denatured TAPS-TCP and native TCL were quite different. SDS-PAGE analysis revealed broader protein bands (19 min, from 33 to 52 min, Figure 3a) for TAPS-TCP and a narrower distribution (15 min, 36 to 51 min, Figure 3b) for TCL. It is concerning that the separation of hydrophobic proteins by reversed-phase HPLC can sometimes result in a ghost peak because of residual protein sample adsorption. However, no peaks were observed in the second elution cycle (Figure S1).

This methodology has been effectively extended to hydrophobic integral membrane proteins (Figure S2). The integral membrane proteins of coxsackie and adenovirus receptor (CAR) (Ortiz-Zapater et al., 2017), endogenously expressed in HeLa cells, were solubilized as TAPS-TCP and subsequently separated by reversed-phase HPLC. This separation yielded distinct protein entities, visualized through HYD-MM 2-DE blots employing specific anti-CAR antibodies. Furthermore, a highly effective elution buffer designed for hydrophobic protein separation in a reversed-phase HPLC column, consisting of a mixture of acetonitrile and 2-propanol (2:1) (Tarr & Crabb, 1983), showed no detectable peaks during the second elution cycle. Thus, this methodology enables proteomics even for hydrophobic membrane proteins.

2.4 Demonstration of controllable resolution and validation of reproducibility for 2-D protein separation

We successfully validated the HYD-MM 2-DE system, which utilizes TAPS-TCP from NCI-H1975 cell culture, as it has a controllable resolution for protein separation (Figure 4a) and demonstrated high reproducibility in sample preparations (Figure 4b,c).

A total of 431 discriminated bands were identified as detectable proteins at elution times ranging from 30 to 57 min when fractionated for 1 min each (Figure 4a, middle). The band resolution was controlled by adjusting the fractionation time. Highly condensed bands, consisting of 217 bands in the elution time range of 41–49 min, could be expanded to 546 bands when the fractionation time was reduced from 60 to 20 s. This demonstrated that the resolution of the discriminated protein bands could be adjusted as required.

The reproducibility of the 2-D separation is also critical for proteomic analysis. The inter-injection (1 mg of protein each) reproducibility of TAPS-TCP on reversed-phase HPLC confirmed a similar chromatographic trace pattern between the first and second injections (Figure 4b). Thus, the repetitive separation of TAPS-TCP by HPLC resulted in homogenously fractionated proteins. This allows for scalable proteomic analysis that is favorable for a large set of autoantibody screening using identical 2-D separated PVDF blots and gels. Inter-sample reproducibility was also demonstrated using TAPS-TCP prepared from different cell culture lots. Although the HPLC chromatographic trace pattern showed slight changes depending on the preparation of the cell culture lots (Figure S3), the HYD-MM 2-DE profiles were comparable, thus establishing a highly reproducible procedure (Figure 4c).

2.5 Demonstration of SERPA using HYD-MM 2-DE

The optimized HYD-MM 2-DE gel of total cellular proteins from NCI-H1975 cells was used to identify target antigens that reacted with autoantibodies in the plasma of patient #5 with lung cancer using the SERPA method. As shown in Figure 5a,b, HYD-MM 2-DE gels and blots were prepared on commercial minigels. The immunoreactive bands at approximately 55 kDa on the 1-D gel separated into four distinct bands on the HYD-MM 2-DE gel (Figure 5b). Proteomic analysis of each antigen revealed vimentin (band-2), keratin-7 (band-3), enolase-1 (band-4), and keratin-8 (band-5) as the target proteins. The three specific antibodies from our stock exhibited specific immune reactions against the corresponding antigens (Figure 5c–e).

Another cell line, NCI-H358 derived TAPS-TCP, was used for autoantibody screening (Figure 6a). After excising the target band corresponding to the immunoreactive band on the PVDF blots and performing in-gel digestion, candidate target antigens were identified using MASCOT search analysis (Figure 6b). The critical criterion for selecting target antigens among these candidates is data mining to match autoimmune or cancer-related antigens. The recombinant protein expressed as a HisTag-fused protein facilitated the purification and detection of the target antigen, which was then checked for an immune reaction against the autoantibody-targeted antigen (Figure 6c). Using this procedure, we identified 26 autoantigens from 14 patients with lung cancer (Figures 2, 5 and 6, S4).

2.6 Autoantibody detection via indirect immunofluorescence cytochemistry

Visualization of autoantibodies present in the plasma of patients with lung cancer was accomplished using indirect immunofluorescence cytochemistry against lung cancer-derived cell lines (Figure 7). Autoantibody reactions against the cell lines varied between individuals, and the cells that exhibited the higher reactions were identified by 1-D gel western blotting (Figure S5). Figure 7a shows clear autoantibody reactions against each cell line visualized in paraformaldehyde-fixed and permeabilized cells. The topographic distribution of the immunofluorescence patterns also varied among individuals, suggesting diversity and individual differences in immune responses. Notably, healthy donor pooled (HDP) serum from 10 healthy donors showed no clear autoantibody reactions (Figure 7b). These results demonstrate that autoantibodies in patients with lung cancer react with lung cancer-derived cell-derived proteins, and proteomic analysis revealed some targeted antigens.

Three of the identified autoantigens were confirmed to exhibit specific binding in the knockdown experiments (Figure 8). Endogenous proteins in each cell were found to react with autoantibodies in the plasma on a 1-D gel. In contrast, the target bands were significantly downregulated in knockdown cells. Interestingly, the autoantibody reaction in indirect immunofluorescence cytochemistry showed no significant reduction in fluorescence intensity. These results also support the conclusion that the autoantibodies present in peripheral blood are composed of multiple autoantibodies directed against different autoantigens. Autoantibody levels were quantified using Luminex beads immobilized with antigens (Futami et al., 2015; Miyamoto et al., 2022) which confirmed the heterogeneity of autoantibody levels among individuals. As all identified antigens were selected based on their binding to autoantibodies on PVDF blots, it can be inferred that autoantibodies specifically recognize linear epitopes on the denatured form of antigens. The water-soluble and denatured forms of the antigens immobilized on the beads through S-cationization demonstrated robust reactivity with autoantibodies, making them suitable for quantitative assays.

4.1 Cell culture and patient plasma

Human non-small cell lung cancer-derived EBC-1, Lu65, Lu99, NCI-H358, NCI-H1299, and NCI-H1975 cells (ATCC) and human cervical epithelioid carcinoma (HeLa) cells (ATCC) were cultured in RPMI1640 medium (Wako Chemical, Osaka, Japan) supplemented with 10% FBS in a humidified incubator. HDP from 10 healthy donors (serum from five males and five females, ages 19–49 years) were obtained from Tennessee Blood Services (TN, USA). Plasma samples from patients with lung cancer who provided written informed consent were obtained from the Okayama University Hospital Biobank (Okadai Biobank, Japan), in collaboration with Dr. Kiura K. and Dr. Ohashi K. (Okayama University Hospital) for biomarker screening, which was approved by the ethics committee of Okayama University.

4.2 Preparation of native cell lysate and water-soluble nucleic acid-free denatured total cellular protein mixtures

Subconfluent lung cancer cells cultured in 10 cm dishes were used in this study. The TCL under native conditions was prepared by dissolving cells in lysis buffer (20 mM HEPES-NaOH buffer, pH 7.5, 50 mM NaCl, 2 mM MgSO₄, 1% Nonidet® P40) supplemented with a protease inhibitor cocktail (Nacalai Tesque, Kyoto, Japan) and ruptured for 12.5 min (250 W, cycle of 20 s ON and 20 s OFF) on ice using a sonicator (Bioruptor UCD-250, CosmoBio, Tokyo, Japan). To digest nucleic acids, 125 unit/mL of Benzonase (Merck Millipore, Burlington, MA, USA) were added and incubated for 30 min at 25°C. Nucleic acid-free, fully denatured, and water-soluble total cellular proteins were prepared using TRIzol (Thermo Fisher Scientific, Waltham, MA, USA) according to previously described methods (Futami et al., 2014). The recovered denatured protein precipitates were solubilized in 6 M GdnHCl containing 0.1 M Tris–HCl (pH 8.5) and reduced with 30 mM dithiothreitol for 2 h at 37°C. S-cationization of all Cys residues in total cellular proteins was prepared by adding 70 mM [3-(trimethylammonium)propyl] methanethiosulphonate (TAPS-sulfonate, Katayama Chemical, Osaka, Japan) and incubating for 1 h at 37°C (Futami et al., 2015; Futami et al., 2017; Kimura et al., 2020; Miyamoto et al., 2022; Murata et al., 2006). Total cellular protein, with or without S-cationization, was adjusted to a protein concentration of 3 mg/mL with 6 M GdnHCl and mixed with a 0.1 volume of acetic acid. After extensive dialysis against pure water at 4°C, water-soluble nucleic acid-free denatured TCP mixtures were recovered without visible precipitation.

4.3 HYD-MM 2-DE

All total cellular proteins from TCL, TCP, and S-cationized TCP (TAPS-TCP) were filtered using 0.45 μm pore size syringe filters (Merck Millipore) before applying for HPLC. TCL protein concentrations were determined using the Bradford protein assay (Bio-Rad Laboratories, Hercules, CA, USA) with BSA as a standard. TCP and TAPS-TCP were determined by measuring the absorbance at 280 nm at 1OD = 1 mg/mL. One milligram of each protein sample was separated on a reversed-phase HPLC column (COSMOSIL Protein-R, 4.6 mm I.D. × 150 mm, Nacalai Tesque) using an acetonitrile linear gradient elution procedure in the presence of 0.1% HCl. The eluted proteins were collected every 60 or 20 s using a fraction collector and lyophilized using a centrifugal evaporator (CVE-200D; Eyela, Tokyo, Japan). Dried samples were then dissolved in SDS-PAGE sample buffer with a reducing reagent (Nacalai Tesque) and subjected to SDS-PAGE analysis on 5%–20% gradient mini-gels (Wako Chemical) or on a wide gel (ATTO, Tokyo, Japan). The fractionated samples were applied at 1/8 volume for the mini gels and 1/4 volume for the wide gels.

4.4 Immunoblot analysis

The following monoclonal (mAb) and polyclonal (pAb) primary antibodies were used for immunoblotting: CAR mAb (D3W3G, Cell Signaling Technology), β-actin mAb (13E5, Cell Signaling Technology, Danvers, MA, USA), β-tubulin mAb (10G10, Wako Chemical), anti-His-tag mAb (OGHis, MBL, Tokyo, Japan), vimentin mAb (D21H3, Cell Signaling Technology), and HNRNPA2B1 pAb (Proteintech, Rosemont, IL, USA). Rabbit polyclonal antibodies against human p53 (Uniport: P04637), alpha-enolase (ENO1, Uniport: P06733), keratin-8 (KRT8, Uniport: P05787), and cancer/testis antigen 1 (NY-ESO-1, Uniport: P78358) were prepared at Cosmo Bio (Tokyo, Japan) by immunizing rabbits with full-length antigens purified using a previously described procedure (Futami et al., 2015; Miyamoto et al., 2022). Target bands were visualized using horseradish peroxidase-conjugated anti-mouse or anti-rabbit IgG antibodies (Cell Signaling Technology) with a chemiluminescent reagent (Western Lightning Plus-ECL; PerkinElmer, MA, USA) on an Amersham ImageQuant 800 (GE Healthcare, Chicago, IL, USA).

4.5 Proteomics

After two rounds of sample separation using SDS-PAGE, one gel was subjected to CBB staining and the other was transferred onto PVDF membranes (Immobilon-P, Merck Millipore) using the semi-dry transfer method. The PVDF blots were subsequently blocked using Blocking One (Nacalai Tesque), incubated with human plasma diluted 400-fold in Can Get Signal solution 1 (Toyobo, Osaka, Japan), and probed with horseradish peroxidase-conjugated anti-human IgG (MBL). Western blot results were visualized by adjusting the size of the CBB-stained gel, and the target band was excised under transmitted light. All CBB-stained gel fragments were destained, reduced, and alkylated before tryptic digestion of the proteins. In-gel digestion of samples was performed using 10 ng/μL trypsin-TPCK TREATED (Worthington). The digested protein samples were subjected to HPLC-Chip/QTOF analysis (G6520 and G4240; Agilent). Candidate peptide sequences were analyzed using the MASCOT software (Matrix Science Inc., MA, USA).

4.6 Identification of autoantigens

Putative biomarker proteins were identified using LC–MS/MS and MASCOT search analyses. Full-length and mature forms of Heterogeneous Nuclear Ribonucleoprotein R (HNRNPR), FK506-Binding Protein 4 (FKBP4), Ribosomal Protein L7a (RPL7A), Emerin (EMD), Heterogeneous Nuclear Ribonucleoprotein A2/B1 (HNRNPA2B1), T-Complex Protein 1 (TCP1), Eukaryotic Elongation Factor 2 Kinase (EEF2K), Heat Shock Protein 60 (HSPD1), Lysophosphatidylcholine Acyltransferase 1 (LPCAT1), Alpha-Enolase (ENO1), Keratin 7 (KRT7), Polypyrimidine Tract-Binding Protein 1 (PTBP1), Keratin 8 (KRT8), Peptidyl-Prolyl Cis-Trans Isomerase B (PPIB), Eukaryotic Elongation Factor 1 Gamma (EEF1G), Keratin 19 (KRT19), Plectin (PLEC), Eukaryotic Translation Initiation Factor 3 Subunit A (EIF3A), TAR DNA-Binding Protein 43 (TARDBP), Keratin 18 (KRT18), Vimentin (VIM), Lamin A/C (LMNA), NY-ESO-1/CT6.1, Heterogeneous Nuclear Ribonucleoprotein A1 (HNRNPA1), and FK506-Binding Protein 1A (FKBP1A) were cloned into pET28b vectors (Novagen) to express a His-tag (MGSSHHHHHHSSGLVPRGSH) at the N-terminus, and a StrepTagII (GPGWSHPQFEK) at the C-terminus. Recombinant proteins were expressed in Escherichia coli BL21(DE3) cells. Recombinant PPIB (34–216 aa), PLEC (175–400 aa), and EIF3A (8–848 aa) proteins with His-tags and StrepTag II corresponded to partial amino acid sequences. HNRNPR, FKBP4, HNRNPA2B1, HSPD1, PTBP1, PLEC, HNRNPA1, and FKBP1A were expressed as soluble fractions and purified using immobilized metal affinity chromatography (IMAC). RPL7A, EEF2K, LPCAT1, ENO1, KRT7, EEF1G, KRT19, EIF3A, TARDBP, KRT18, VIM, NY-ESO-1/CT6.1, and TCP1 were expressed as insoluble inclusion bodies and were solubilized using S-cationization techniques (Futami et al., 2015; Futami et al., 2017; Miyamoto et al., 2022; Murata et al., 2006). EMD, KRT8, and LMNA were refolded from their insoluble fractions without any chemical modifications. Purified recombinant autoantigens were identified as potential biomarker proteins based on positive western blot bands observed in diluted plasma from a patient with lung cancer.

4.7 Immunostaining of intracellular autoantigens

Subconfluent cells in a glass-based dish (Iwaki Glass, Shizuoka, Japan) were fixed with 4% paraformaldehyde phosphate buffer solution (Wako) and permeabilized with 0.1% Triton X-100 in PBS for 30 min. Intracellular autoantigens were reacted with 600-fold diluted human plasma in blocking buffer (1% BSA, 0.1% Tween20, in PBS) for 1 h at 25°C. Immunoreactive antigens and nuclei were stained with Alexa Fluor 488-conjugated AffiniPure Goat Anti-Human IgG (Jackson ImmunoResearch, West Grove, PA, USA, 1:400) and 4,6-diamidino-2-phenylindole (DAPI, Dojindo Laboratories, Kumamoto, Japan), respectively. Fluorescent images were acquired using a FV3000RS laser scanning microscope (Olympus, Tokyo, Japan) and an APX100 digital imaging system (EVIDENT, Tokyo, Japan).

4.8 Knockdown of endogenous specific autoantigens

Short hairpin RNAs (shRNAs) were constructed by ligating annealed oligo DNA into the pLKO.1 vector. For lentivirus preparation, 293T cells were transfected with 300 ng of pLKO.1-shRNA and packaging plasmids (300 ng of pMDLg/pRRE, 150 ng of pMD2g, and 150 ng of pRSVRev) using the FuGENE 6 transfection reagent (Promega, Madison, WI, USA) according to the manufacturer's instructions. Cells were infected using the filtered culture supernatant from 293T cells in the presence of 8 μg/mL polybrene. The target sequences of the shRNAs were 5′-GCTAACTACCAAGACACTATT-3′ for VIM, 5′-AGCTTCAGGTTATCGAAATAA-3′ for HNRNPA2B1, 5′-TCACTGTGTCCGGCAACATAC-3′ for NY-ESO-1, and 5′-CCTAAGGTTAAGTCGCCCTCG-3′ for the scrambled control.

4.9 Quantification of autoantibodies by MUSCAT-assay

Autoantibodies in patient plasma were diluted to 1/200 and quantified using water-soluble full-length antigen-immobilized Luminex beads. The detailed procedures have been described previously (Futami et al., 2015; Miyamoto et al., 2022).