A Novel Cytoplasmic Protein with RNA-binding Motifs Is an Autoantigen in Human Hepatocellular Carcinoma

HCC sera were obtained from 95 subjects included in an epidemiological study previously described and were from Henan Province, People's Republic of China (31). Sera from 77 patients with liver diseases (26 asymptomatic HBsAg carriers, 31 patients with acute hepatitis, and 20 patients with chronic hepatitis and liver cirrhosis), and 30 normal human sera, all from the same province, were available for these studies. All of the above sera came from the Sanitary and Anti-Epidemic Station (Henan Province). 40 normal human sera from the San Diego, CA area were also included as controls. Human prototype sera containing autoantibodies to previously identified intracellular antigens were from patients with systemic autoimmune diseases (2) and obtained from the serum bank of the Autoimmune Diseases Center of The Scripps Research Institute (La Jolla, CA).

MOLT-4 (human acute lymphoblastic leukemia), T24 (human transitional cell bladder carcinoma), HEp-2 (human epidermoid laryngeal carcinoma), HeLa (human epitheloid cervical carcinoma), HepG2 (human hepatocellular carcinoma), A549 (human lung carcinoma), and 3T3 (mouse fibroblast) cell lines were obtained from the American Type Culture Collection and cultured following the specific protocol for each cell line. Cells grown in monolayers were solubilized directly in Laemmli's sample buffer containing protease inhibitors (Boehringer Mannheim). Solubilized lysates were briefly sonicated before electrophoresis on SDS–polyacrylamide gels.

Initial identification of autoantibodies in sera was performed using methanol- and acetone-fixed commercial HEp-2 cell slides (Bion Enterprises, Ltd.). The findings were usually confirmed in other experiments using T24, HepG2, and 3T3 cells that were grown on coverslips, fixed for 5 min at −20°C in 100% methanol, and permeabilized for 3 min at −20°C in 100% acetone. As a second antibody, FITC-conjugated goat anti–human IgG (Caltag Laboratories) was applied. A titer of >1:40 dilution was interpreted as positive.

Western blotting was performed essentially as described by Chan and Pollard (32). Cell extracts were electrophoresed on SDS-PAGE and transferred to nitrocellulose paper. After preblocking with PBS containing 0.5% Tween-20 and 5% nonfat milk for 30 min at room temperature, the nitrocellulose strips were incubated for 60 min at room temperature with a 1:100 dilution of serum. As secondary antibody, horseradish peroxidase–conjugated goat anti–human IgG (Caltag Laboratories) was applied (1:2,000 dilution). The detection of immunoreactive bands was performed with an ECL kit (Amersham Corp.) according to the manufacturer's instructions and followed by autoradiography.

T24 and HeLa cells were cultured and radiolabeled with [³⁵S]methionine. For preparation of T24 and HeLa cell extracts, cells were collected by centrifugation, combined with two times packed cell volume buffer A (10 mM Tris-HCl, pH 7.5; 150 mM NaCl, 1.5 mM MgCl₂, and 0.5% NP-40), and held on ice for 10 min to allow cell lysis. The supernatant obtained by centrifugation at 10,000 g for 10 min at 4°C was used as antigen preparation in immunoprecipitation studies. Before immunoprecipitation, labeled cell extracts were precleared by adding 100 μl 10% protein A–Sepharose stock/ml extract, mixed for 5 min on ice, and centrifuged to collect supernatant. Typically, 100 μl 10% protein A–Sepharose, 500 μl buffer B (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 5 mM EDTA; 0.5% NP-40; 0.5% deoxycholic acid; 0.1% SDS; and 0.02% sodium azide) containing BSA at 10 μl (stock: 10 mg/ml), 40 μl labeled cell extract, 10 μl serum, and 10 μl protease inhibitor (Boehringer Mannheim) was added to a standard immunoprecipitation reaction. After incubation for 1 h, the immunoprecipitated beads were washed five times with 1 ml buffer B. Finally, the beads were eluted with an equal volume of 2× Laemmli's sample buffer and analyzed in SDS-PAGE followed by autoradiography.

HCC serum YZ was diluted 1:100 and used for screening a T24 λzap cDNA expression library. Before screening of the cDNA library, the serum was extensively absorbed against bacteria infected with wild-type λzap phage (33). The preabsorbed serum was used to immunoscreen 3.0 × 10⁵ recombinant plaques using ¹²⁵I-labeled goat anti–human IgG as the secondary detecting reagent. Screening was carried out on duplicate filters, and one double-positive clone, JY1, was isolated and subcloned in vivo into pBK-CMV plasmid using ExAssist helper phage (Stratagene Inc.) as recommended in the manufacturer's instructions. The clone JY1 was amplified, purified, and used for sequence analysis. cDNA insert was analyzed by restriction mapping and sequencing. The clone JY1 was a partial sequence, and RACE methodology was used to obtain overlapping 5′ clones using the Marathon-Ready cDNA from human colorectal adenocarcinoma SW480 cell line (Clontech). Nucleotide sequence was determined in both strands using a semiautomated sequencer from Applied Biosystems (model 373). Oligonucleotide primers were synthesized with a DNA synthesizer (Applied Biosystems; model 394). DNA and protein sequences were analyzed by the Genetics Computer Group Sequence Analysis Software Package for UNIX computers (version 7.4; 34). Alignment of protein sequences was achieved with a Multiple Alignment Program (http://dot.imgen.bcm.tmc.edu:9331/multi-align/multi-align.html; reference 35).

The ORF of p62 was reamplified and confirmed by RT-PCR using T24 cell mRNA as template. One set of sense and antisense primers was designed, and their positions with respect to the p62 full length cDNA are indicated (see Fig. 2 A): rt3 sense, 5′-TTGAATTCGCCATGGTGAACAAGCTTTACATCGGGAACC-3′ and rt4 antisense, 5′-TTTATGTCGACGGTGTTGGAAGGGCTACATT-3′, incorporating an EcoRI and SalI site, respectively. RT-PCR was performed using the one-tube method as described by Pfeffer et al. (36). In brief, 1 μl T24 mRNA (0.5 μg/μl), 10 μM primer (1 μl each), 1.25 U Taq polymerase (GIBCO BRL), 100 U SuperScript II RNase H–reverse transcriptase (GIBCO BRL), 20 U RNase inhibitor (Promega Corp.), 0.25 μl of 10 μM dNTPs, and 2.5 μl 10× PCR buffer containing 500 mM KCl; 100 mM Tris-HCl, pH 8.3; 15 mM MgCl₂; and 0.1% gelatin were added to a final total volume of 25 μl, and the PCR steps were programmed using a thermocycler (Eppendorf). The reactions were performed at 50°C for 1 h and followed by 30 cycles at 57°C for 10 s, 72°C for 2 min, and 94°C for 10 s. RT-PCR products were analyzed by agarose gel electrophoresis.

For increased expression and purification of recombinant protein, p62 cDNA derived from RT-PCR was subcloned into the EcoRI and SalI sites of pET28a vector, which provides the NH₂-terminal fusion protein with 6× histidine and T7 epitope tags. The recombinant protein was overexpressed in Escherichia coli BL21 (DE3) and purified using nickel column chromatography. The protocol used for the high-level expression and purification of 6× histidine–tagged proteins was performed as described (Qiagen, Inc.). Elution buffer (8 M urea, 0.1 M NaH₂PO₄, and 0.01 M Tris, pH 4.5) was used to elute the recombinant protein.

The p62 cDNA was transcribed and translated in vitro using TnT-coupled reticulocyte lysate system (Promega Corp.) in the presence of [³⁵S]methionine (ICN) as described (Promega Biotec). Labeled products were used as substrates for immunoprecipitation analysis.

Recombinant protein was electrophoresed on 15% SDS-PAGE and transferred to nitrocellulose membranes. The membranes were cut into strips and the recombinant protein bands confirmed by Western blotting. Nitrocellulose strips were incubated with diluted serum at 1:100, and unbound antibodies were removed by washing with PBS containing 0.5% Tween-20 before elution of bound antibodies with 0.5 ml elution buffer (200 mM KH₂PO₄, 150 mM NaCl, and 0.1% BSA, pH 2.5). Affinity-purified antibodies were immediately neutralized by the addition of 1 M Tris-HCl, pH 8.7. The antibodies were concentrated with Centricon-30 microconcentrators (Amicon Corp.), and different dilutions (1:5, 1:25, and 1:50) were used for immunofluorescence assay and Western blotting analysis.

Four female New Zealand White rabbits were immunized by subcutaneous injections of 0.5 mg of p62 recombinant protein in complete Freund's adjuvant. Rabbits were boosted two times with 0.5 mg p62 recombinant protein in incomplete Freund's adjuvant at 1-mo intervals, and blood was collected 10 d after the last booster injection.

Nylon membranes blotted with poly A⁺ RNA isolated from multiple human tissues and several human cancer cell lines were obtained from Clontech. An antisense riboprobe was generated from a 480-bp fragment corresponding to the NH₂-terminal domain of p62 (see Fig. 2 A) and labeled with [α-³²P]UTP (Clontech) as described. In brief, the membranes were hybridized with ³²P-labeled p62 riboprobe for 2 h at 74°C, washed in 2× SSC and 0.1% SDS at 74°C for 20 min and in 0.1× SSC and 0.1% SDS at 65°C for 20 min, and exposed to x-ray film for 4 h at −70°C. A 2.0-kb human β-actin cDNA provided by Clontech was used as control probe.

Standard protocol for ELISA was used as described by Rubin (37). Purified p62 recombinant proteins were diluted in PBS to a final concentration of 1 μg/ml for coating Immulon 2 microtiter plates (Dynatech Laboratories). Human sera diluted 1:100 were incubated in the antigen-coated wells. Horseradish peroxidase–conjugated goat anti–human IgG (Caltag Laboratories) and the substrate 2.2′-azinobis (3-ethylbenzthiazoline sulfonic acid; Boehringer Mannheim) were used as detecting reagents. Each sample was tested in duplicate, and the average OD at 490 nm was used for data analysis. The cutoff value designating positive reaction was the mean OD of normal sera + 3 SD.

Initial studies focused on the analysis of serum specimens from 95 HCC patients from Henan Province. Many of these sera contained antibodies reactive with a 62-kD cellular antigen that was present in low concentration in MOLT-4 cell extracts but in higher concentration in extracts from other cell lines. Fig. 1 shows Western blotting of three such sera against whole cell extracts from MOLT-4, T24, and HepG2. Lanes 1, 5, and 9 are normal human serum showing negative reactions. Lanes 2, 6, and 10 show the reactivities of HCC serum YZ. This serum showed strong reactivity with a 90-kD band in MOLT-4 cell extracts and weak reaction with a 62-kD band. With T24 cell extract, the 62-kD band was slightly stronger, but in HepG2 cell extract the 62-kD reactivity was as strong or stronger than the 90-kD reactivity. In addition, a good signal was observed at 50 kD. Serum YL from a different patient contained antibodies primarily against the 90-kD antigen and displayed very weak reactivity with the 62-kD antigen (lanes 3, 7, and 11). A third serum, CH, appeared to have antibodies primarily against the 62- and 50-kD antigens (lanes 4, 8, and 12). In addition, serum CH, as well as YZ, demonstrated weak reactivity with other antigens of higher and lower molecular sizes, but the predominant reactivities detected were against the 90-, 62-, and 50-kD proteins. These data demonstrated that different cell lines showed different levels of expression of these intracellular proteins, with T24 expressing higher levels of the 62-kD and HepG2 expressing higher levels of both 62- and 50-kD antigens than MOLT-4.

The above data suggested that in order to isolate cDNA clones encoding the 62-kD antigen, T24 or HepG2 cDNA expression libraries would be the preferred cell lines. HCC serum YZ was used to immunoscreen a T24 λzap cDNA expression library. One positive clone was isolated from 3.0 × 10⁵ recombinant plaques, purified to homogeneity, and subcloned in vivo into pBK-CMV. This clone, designated JY1, was a cDNA of 3.3 kb and was shown to have an apparent coding sequence with a termination codon and a long 3′ UTR with a poly A tail. 5′-RACE methodology was used to obtain overlapping 5′ clones using the Marathon-Ready cDNA repertoire derived from human colorectal adenocarcinoma SW480 cell line. Two overlapping independent clones were obtained and analyzed, with the longest clone, H27, containing a cDNA insert of 374 bp. An in-frame TGA stop codon at position 320 bp upstream of a methionine start codon was found. The presumptive full length cDNA is shown in Fig. 2 C. The cDNA has a 5′-UTR of 435 bp, an ORF of 1,668 bp, and a 3′-UTR of 1,564 bp. RT-PCR was performed using T24 cell line RNA as template with one pair of designed primers (rt3 and rt4, shown in Fig. 2 A). The RT-PCR products were of ∼1.7 kb, compatible with the size of the ORF in the isolated cDNA clone (Fig. 2 B). The ORF codes for 556 amino acids with a predicted molecular mass of 62 kD and a calculated pI of 8.59.

The 62-kD protein contained two types of RNA binding motifs, the consensus sequence RNA–binding domain (CS-RBD) and four hnRNP K homology (KH) domains (38). The CS-RBD domain was located in the NH₂-terminal region, and the four KH domains extended from the middle region to the COOH-terminal region (Fig. 2). Three proteins were found to have high degrees of homology to p62 at both nucleotide and protein levels. Table I compares percent similarity and identity of p62 with the other three proteins. KH domain–containing protein overexpressed in cancer (Koc), a putative oncogene (39), had 66.5% identity and 80.3% similarity to p62. Zipcode binding protein (ZBP1), a β-actin mRNA-binding protein in chicken (40), had a 70.5% identity and 83.9% similarity to p62, and B3 (X. laevis TFIIIA–binding protein), an oocyte factor that binds to a developmentally regulated cis element in the TFIIIA gene (41), showed 69.7% identity and 82.7% similarity to p62. Table I also shows that other KH domain–containing proteins such as FMR1 (42), hnRNP K (43), and hnRNP X (44) showed much lower levels of homology. The sequences of p62, Koc, ZBP1, and B3 are shown in Fig. 3 A, demonstrating the CS-RBD and four hnRNP K homology domains. In addition, a nine–amino acid sequence (VGAIIGKE/KG) of unknown function previously reported in ZBP1 (40) was also found in the first three KH domains of the other three proteins. A potential REV-like nuclear export signal also found in ZBP1 protein (40) was present in position 308–319 of p62. The sequence alignments of the four proteins are depicted in Fig. 3 A, and their domain structures are shown in Fig. 3 B.

The p62 recombinant protein was expressed using the pET28a vector in E. coli and purified using nickel affinity column chromatography. A 62-kD polypeptide was detected by Coomassie blue staining together with one prominent lower molecular mass product (Fig. 4 A, lane 1). HCC sera containing antibodies to the 62-kD cellular antigen reacted strongly with the 62-kD recombinant protein and also with the lower molecular weight product (Fig. 4 A, lane 3). In contrast, HCC sera containing only 90-kD autoantibodies such as serum YL and normal human serum (Fig. 4, lane 2) were nonreactive. In vitro–translated products of p62 cDNA showed that the polypeptide migrated at 62-kD (Fig. 4 B, lane 6) and was immunoprecipitated by human anti-p62 prototype serum (Fig. 4 B, lane 4) but not normal human serum (Fig. 4 B, lane 5). The in vitro–translated products (Fig. 4 C, lane 8) comigrated with the 62-kD antigen detected by Western blotting of HepG2 cell extract (Fig. 4 C, lane 8a).

To further confirm that the recombinant protein was identical to the 62-kD cellular protein, four female New Zealand White rabbits were immunized with p62 recombinant protein. Rabbit immune sera immunoprecipitated in vitro–translated products (Fig. 5 A, lane 3). In addition, Fig. 5 B shows that the cellular 62-kD protein was immunoprecipitated by immune rabbit serum (lane 10) in a fashion similar to human anti-p62 prototype sera (lanes 7 and 8), using extracts from [³⁵S]methionine–labeled HeLa cells as the antigen preparation. Immune rabbit sera were also reactive in Western blotting with 62-kD proteins from T24, HepG2, and A549 cells (data not shown). Fig. 5 B shows that other proteins were also immunoprecipitated, but many of these were also immunoprecipitated by preimmune rabbit serum and normal human sera and are presumed to be nonspecific precipitates.

It had been previously observed by immunofluorescent histochemistry that HCC sera with antibodies to the 62-kD antigen were negative for staining of the nucleus but positive for cytoplasmic staining. Immune rabbit antiserum affinity-purified from recombinant protein showed cytoplasmic staining, as depicted in Fig. 6 B. The pattern of human HCC prototype serum YZ is shown in Fig. 6 A. The fine details of cytoplasmic staining of human serum YZ and rabbit immune sera are somewhat different, with rabbit serum showing a coarser pattern of cytoplasmic staining than human serum. This could be related to the fact that human sera are polyclonal and might contain other autoantibodies. The relevant finding was that both human HCC sera with anti–p62-kD antigen reactivity and rabbit immune sera were reactive with antigens that were cytoplasmic in location.

Using a 480-bp fragment corresponding to the COOH-terminal domain of p62 as probe (Fig. 2 A), two major forms of p62 transcripts were detected in Northern blotting using commercially available poly A⁺ RNA from a number of human tissues and cell lines. Fig. 7 demonstrates that there was a 3.7-kb transcript (lower arrow) that was reactive with the probe, as well as a 5.2-kb transcript (upper arrow). The 3.7-kb transcript of p62 was found in heart and placenta, whereas brain, lung, liver, kidney and pancreas gave much lower signals or were negative for p62 expression. A signal for skeletal muscle migrated somewhat more slowly than the 3.7-kb transcript (see data showing the probe corresponding to the NH₂-terminal domain below). High levels of the 3.7-kb transcript were also detected in HeLa, K-562 (chronic myelogenous leukemia), SW480 (colorectal adenocarcinoma), A549 (lung carcinoma), and G361 (melanoma) cell lines, whereas low expression was observed in HL60 (promyelocytic leukemia), MOLT-4 (lymphoblastic leukemia), and Raji (Burkitt's lymphoma) cell lines. The presence of the 3.7-kb transcript in these cell lines showed good correlation with the presence of a 62-kD polypeptide antigen in extracts of these cell lines, whereas cell lines negative for the 3.7-kb transcript contained negligible or barely detectable amounts of 62-kD polypeptide. The data are not shown for all cell lines, but as can be seen in Fig. 1, MOLT-4 cell extracts contained low levels of p62 antigen and low levels of 3.7-kb transcript (Fig. 7 B). The Northern blot data just described were confirmed with an antisense riboprobe generated from a 687-bp fragment corresponding to the NH₂-terminal domain of p62, and the same membrane shown in Fig. 7 A was stripped and rehybridized with this probe. The 3.7- and 5.2-kb transcripts were specifically detected by this probe, a result similar to the observations shown in Fig. 7 A with the exception that the “slower” 3.7-kb band in skeletal muscle was not detected. From these observations, we propose that both the 3.7- and 5.2-kb transcripts represent cellular mRNAs encoding the 62-kD protein. The identity of the 5.2-kb transcript has not been determined, but it could represent an alternative mRNA for the 62-kD protein with extra 5′ upstream or 3′ untranslated sequences.

An enzyme-linked immunosorbent assay system was developed using p62 recombinant protein as antigen and a total of 242 sera from humans with different conditions were examined for reactivity. Table II shows that detectable antibodies to p62 antigen were present in 21.1% of patients in a group of patients with HCC from Henan Province. However, in sera from patients with conditions that are known to be precursor diseases to HCC, including asymptomatic HBsAg, acute hepatitis, and chronic hepatitis, no antibodies to p62 were detected. Normal human sera from patients who came from Henan Province or the San Diego area were also negative for p62 antibodies.

Study of human tumor antigens recognized by the autologous host has had a long history, and T cell–recognized epitopes on human tumor cells have been extensively characterized (30). However, antibody-defined tumor antigens have been receiving greater attention recently, and several centers have used a method called SEREX (serological analysis of recombinant cDNA expression libraries), libraries of cDNA constructs made from mRNA extracted from tumor tissue (29) and screening of such libraries with autologous serum. From these studies, several novel as well as previously defined tumor antigens have been identified (30). Our approach was not restricted to use of cDNA libraries from autologous tumors because our previous studies had shown that antibodies in HCC sera were reactive with antigens expressed by a variety of tissue culture cell lines (23–25). The autoantigens identified included some previously recognized in autoimmune diseases such as lupus and scleroderma (22, 23). However, an unusual feature in the HCC model was that novel antigen-antibody reactions were detected in some patients during transformation from chronic liver disease to malignancy. Therefore, use of such sera to isolate and identify antigens might reveal special signatures of cancer cells. A novel protein with alternative splicing factor motifs (25) and a cell cycle–related nuclear protein with WD-40 motifs (26) have been isolated. These novel autoantigens, including p62 reported here, are not restricted to cancer cells but are also expressed in normal cells, an observation which has also been reported for cancer antigens in melanoma. These antigens have been called differentiation antigens (28). At this point, it is unclear why some differentiation antigens become capable of provoking autoimmune responses, but possibilities which have been raised in the case of p53 include gene mutations, gene product overexpression, and unusual complexes with other cellular proteins such as heat shock proteins (27, 45).

In addition to the remarkably high percentage of similarity and identity between p62 on the one hand and ZBP1 and Koc on the other, there are several features of these proteins that are of interest. ZBP1 is a chicken protein of 68 kD which was identified through its property of binding to a conserved element in the 3′ untranslated region of β-actin mRNA (40). Rabbit antibodies raised against ZBP1 polypeptides were shown to bind to β-actin mRNA in the leading edge of the lamella in cells such as chicken embryo fibroblasts and 3T3 fibroblasts (40). Studies are in progress to determine whether recombinant p62 is capable of binding in vitro to the “zipcode” element of β-actin mRNA as is the case with ZBP1.

The Koc protein was isolated following analysis of differential gene expression in pancreatic cancer tissue (39). High transcript levels of Koc were found not only in pancreatic cancer tissue but also in soft tissue sarcoma, gastric cancer, and colon cancer. Differences in mRNA expression of Koc in different tissues has been reported (39). We have also observed different transcript levels of p62 in different tissues and, in addition, high levels of expression of p62 transcript in some cell lines (HeLa, K-562, SW480, A549, and G361) but low expression in others including the promyelocytic leukemia HL-60, lymphoblastic leukemia MOLT-4, and Burkitt's lymphoma Raji. It is perhaps of some interest that the two human members (p62 and Koc) of this putative family of RNA-binding proteins appear to be associated in some way with cancer. It has recently been reported that two different Koc-homologous genes were also identified by SEREX methodology (46), but sequence information concerning these genes and their relationship to cancer are not yet available.

Previously, we reported that several patients with HCC mount de novo immune responses to nuclear antigens at the time of conversion from chronic hepatitis or liver cirrhosis to HCC (24) and that the autoantibodies that were produced against intracellular antigens might be regarded as immune system reporters of abnormal intracellular molecular events. Novel proteins that have been detected include HCC1 (25), a putative member of the SR family of alternative splicing factors, and SG2NA, a protein highly expressed in the S and G2 phases of the cell cycle (26). p62 appears to be another such molecule, but unlike the antibody responses to HCC1 and SG2NA, which were observed in individual patients, the autoimmune response to p62 was detected in >20% of one group of HCC patients, suggesting that a shared stimulus might be inciting the immune responses. This shared stimulus could be environmental in nature, but this is as yet only conjecture. An important study that could not be performed at this time was the analysis of HCC tissues to determine whether there were abnormalities in the p62 gene or in its expression, as tissue specimens were not available in this retrospective study. This will be the focus of a prospective study in newly identified HCC patients with autoantibodies to p62.

This work was supported by National Institutes of Health grants CA56956 and AR 32063 and by services provided by the Stein DNA Core Laboratory of the Scripps Research Institute. This is publication 11759-MEM from The Scripps Research Institute.

CS-RBD

consensus sequence RNA– binding domain

HCC

hepatocellular carcinoma

KH

K homology

ORF

open reading frame

RACE

rapid amplification of cDNA ends

RT

reverse transcriptase