Human IgGFc Binding Protein (FcγBP) in Colonic Epithelial Cells Exhibits Mucin-like Structure*

Cloning a cDNA for human IgGFc binding protein (FcγBP) from human colonic epithelial cells reveals an mRNA and coding region of 17 and 16.2 kilobases, respectively. The predicted amino acid sequence contains 12 occurrences of a 400-amino acid cysteine-rich unit resembling that found in mucin. A motif (CGLCGN) in FcγBP is conserved in MUC2 and prepro-von Willebrand factor. The N-terminal 450-amino acid sequences are necessary and sufficient to confer IgG Fc binding activity. FcγBP mRNA is expressed only in placenta and colonic epithelial cells. These results suggest that FcγBP may play an important role in immune protection and inflammation in the intestines of primates.

Each antibody isotype has specific biological activities that are dependent on the Fc receptor (FcR) 1 it binds (1,2). For example, immunoglobulin G (IgG) complexes evoke numerous functions. They are phagocytosed by macrophages, activate the lytic potential of cytotoxic lymphocytes and monocytes, and regulate B lymphocyte activation via different receptors for IgGFc (Fc␥R) expressed on different cells (3,4). Soluble forms of FcR, called immunoglobulin-binding factors, may also be involved in regulatory functions (5).
Other types of immunoglobulin-binding proteins are involved in transport of immunoglobulins. For example, polyimmunoglobulin (Ig) receptors have been implicated in the transcytosis and secretion of polymeric IgA and IgM in mucosal epithelial cells (6,7). Further, neonatal FcR (FcRn) expressed on the apical cell surface binds IgG in milk for endocytosis and subsequent transcytosis to the basolateral cell surface (8 -12).
Recently, binding sites specific to the Fc portion of IgG mol-ecules have been reported in human intestinal goblet cells (13,14). Monoclonal antibodies K9 and K17, which block the binding to the Fc portion of IgG, react with antigens Ͼ200 kDa in immunoblot analysis using non-reduced colonocyte lysates. Under reducing conditions, K17 monoclonal antibody also reacts with proteins around 70 -80 kDa. The binding site is immunologically distinct from the known Fc␥R on leukocytes, is present in mucous granules, and appears to be secreted with mucus into the intestinal lumen (14). In active ulcerative colitis, a marked increase of IgGFc-binding sites in the endoplasmic reticulum and a decrease in mucous granules are observed, apparently reflecting increased synthesis and secretion. 2 Using monoclonal antibodies K9 and K17, we have cloned a full-length cDNA of about 17 kb, which encodes an IgG Fc binding protein (Fc␥BP). Expression of an Fc␥BP cDNA fragment (about 8 kb) confers IgG Fc-binding activity in both COS and CHO cells. Our studies reveal that Fc␥BP shares some structural features with mucin proteins. So far, the complete primary sequences of mucins MUC1 (15)(16)(17), MUC2 (18), and MUC7 (19) have been reported, but only partial sequences are available for most other mucins because of tandem repeats of sequences. Notably, the mucin-like protein we characterize here has the biological activity of IgG binding, suggesting that it has an important role in mucosal immune system and inflammation.

MATERIALS AND METHODS
Isolation and DNA Sequence Analysis of cDNA Clones-Human colonic or ileal epithelial cells were isolated as described previously (13). The epithelial cells were pelleted and resuspended in PBS(Ϫ) and centrifuged at 4°C for 10 min. Total RNA was prepared with the method of Chomczynski and Sacci (20). Poly(A) ϩ RNA for the cDNA library construction was prepared by affinity purification with oligo(dT)-latex beads (oligotex-dT30, Takara Shuzo Co.). cDNAs were synthesized from the poly(A) ϩ RNA using random primers and oligo(dT) [12][13][14][15][16][17][18] primer by Moloney murine leukemia virus reverse transcriptase, as described by the manufacturer (cDNA kit, Amersham Life Science, Inc.). Then, EcoRI-NotI-BamHI linkers (Takara Shuzo Co.) were ligated with the cDNAs. A random primed cDNA library for screening with monoclonal antibodies was fractionated in size greater than 500 bp, with an average of about 1.5 kb, and constructed into the EcoRI site of gt11 vector (Stratagene). The cDNA library was screened by standard procedures (21) with monoclonal antibody K9 or K17 directed against colonic Fc␥BP. Restriction endonuclease fragments (q and a/b as shown in Fig. 1B) from the isolated cDNA clones were used as probes to isolate longer cDNA clones from the gt10 library constructed as follows. Poly(A) ϩ RNA prepared from human colonic epithelial cells was used for synthesis of random primed or oligo(dT)primed cDNA as described above. Then the cDNA was size-selected on agarose gels (Ͼ3 kb) and used to construct gt10 libraries for the * A part of this work was performed as a part of the R&D Project of Industrial Science and Technology Frontier Program supported by the New Energy and Industrial Technology Development Organization. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ and GenBank TM  screening with DNA probes. Subsequent screening was carried out with the unamplified library. 2 ϫ 10 4 phage clones were plated on Luria-Bertani agar dish (10 ϫ 13 cm), and the plaques were transferred onto Biodyne A nylon filter (Nippon Genetics Co.). The filters were hybridized with probe q and probe a/b in a solution of 50% formamide, 5 ϫ SSCP, 1 ϫ Denhardt's solution, 0.5% SDS, 100 g/ml denatured salmon sperm DNA at 42°C overnight. Filters were washed three times with 0.2 ϫ SSC containing 0.2% SDS at 65°C for 40 min and exposed to x-ray film at Ϫ80°C overnight. For DNA sequencing, a nest of insert unidirectionally deleted from each end was generated by exonuclease III and mung bean nuclease digestions (Kilo sequence deletion kit, Takara Shuzo Co.) as described in the manufacturer protocol. DNA sequencing of double-stranded plasmid DNAs and single-stranded DNAs rescued with VCSM13 helper phage was performed with standard dye primerlabeled cycle-sequencing techniques using an Applied Biosytems 373A DNA sequencer, and, in some experiments, internal sequencing primers were synthesized with a DNA synthesizer Model 394 (Applied Biosystems) for sequencing. Gene Works 2.3 (IntelliGenetics, Inc.) Macintosh software was used for the DNA sequencing analysis.
Northern Blot Analysis-20 g of total RNA prepared from colonic epithelial cells or from the human HT-29 -18-N2 (designated as N2) cells was subjected to denaturing formaldehyde-agarose gel electrophoresis, transferred, and fixed onto a nylon membrane. The blots were prehybridized in the same solution as described in the screening method. Hybridization was performed with probe a, q, or y in the same solution at 42°C overnight. The blots were washed 3 times with 0.2 ϫ SSC and 0.2% SDS at 65°C and exposed to x-ray film.
Zoo Blot Analysis-High-molecular-weight DNA was isolated from human colon epithelial cells by the methods of Nelson and Krawetz (22). Animal genomic DNAs were purchased from CLONTECH. Each DNA (5 g) was digested with restriction enzyme EcoRI, electrophoresed in 0.7% agarose gel, and transferred to nylon membrane under alkali conditions. The filter was hybridized with a 32 P-labeled 700-bp fragment of probe a.
Screening of 5Ј-Flanking Region of Human Genomic DNA for Fc␥BP-EMBL3 SP6/T7 human leukocyte genomic library (CLON-TECH) was screened by plaque hybridization using a 32 P-dCTP random primer-labeled 908-bp BamHI fragment of clone NZ4 and a 32 P-5Ј end-labeled synthetic 33-mer nucleotide RPS1 (GCTCCAGCCCAGAG-TATCCACCAGCTCCATAGG, which is complementary to nucleotide sequences at the position of ϩ17 to ϩ49 of the NZ4 cDNA) as a hybridization probe. Hybridization was performed at 42°C in a hybridization solution containing 5 ϫ SSPE, 50 (for pNZ4 probe) or 20% (for RPS1 probe) formamide, 2.5 ϫ Denhardt's solution, 0.1% SDS, and 100 g/ml heat-denatured salmon testis DNA. Filters were washed five times for 15 min in 0.2 ϫ SSC with 0.2% SDS at 65°C (for NZ4 probe) or 0.5 ϫ SSC with 0.1% SDS at 55°C (for RPS1 probe). Phages yielding positive signals on duplicate filters were purified, and the fragments digested with restriction enzymes, including XhoI and SacI, were subcloned into pBluescriptSK(ϩ).
Primer Extension Analysis and S1 Nuclease Mapping-Antisence 50-mer (5Ј-GCTGATAGTTCTGCAGGAAGGCTGTGAGGAATTCC TCTCTGGCCAGTGTTC) and 33-mer (5Ј-GCTCCAGCCCAGAGTATC-CACCAGCTCCATAGG-3Ј) oligonucleotides, respectively complementary to nucleotides ϩ95 to ϩ144 and ϩ17 to ϩ49, were synthesized and purified by using an OPC cartridge (Applied Biosystems) followed by polyacrylamide gel electrophoresis. The 50-mer oligonucleotide primer was end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase (Takara Shuzo Co.) and was used for the extension of cDNA from 2.5 mg of poly(A) ϩ RNA, which was isolated from either N2 cells or human colon epithelium cells using Moloney murine leukemia virus-reverse transcriptase (Toyobo Co.). The extension products were analyzed on a denaturing 6% polyacrylamide gel. A universal primer was used to prime a sequencing ladder from single strand M13mp18 template as size markers. For preparation of S1 probe, the end-labeled 33-mer oligonucleotide was annealed to single-stranded DNA that was derived from XhoI/EcoRI fragment of GHFc and contained the 5Ј-flanking region of Fc␥BP gene, extended with BcaBest polymerase (Takara Shuzo Co.) at 60°C for 10 min, and digested with BamHI. The resultant double-stranded S1 probe was purified by polyacrylamide gel electrophoresis. The 50,000 cpm of the S1 probe was hybridized at 45°C with 50 g of total RNA isolated from colon epithelium cells, and the mixture was subjected to S1 nuclease digestion.
PCR/RFLP Analysis for Fc␥BP Gene-Polymorphic SmaI restriction site of the Fc␥BP gene was analyzed in four cases of normal part of the colon epithelium from patients harboring colon cancer and in 6 cases of normal blood lymphocytes. The PCR reaction mixture contained 0.5-1 g of DNA, 200 mM each dNTP, 50 mM KCl, 20 mM Tris, pH 8.4, 2 mg/ml bovine serum albumin, 1.5 mM MgCl 2 , 0.5 units Taq polymerase (Perkin-Elmer Cetus), and 20 pmol of each primer set in a total volume of 50 ml. The following oligonucleotide primers for PCR were synthesized: BC1 (forward 5Ј-ACCACTCCTTCGATGGCC-), GS4R (reverse 5Ј-TG-GTGCCGAGGGCAGCCACG-3Ј), GS1 (forward 5Ј-ACCTGTAACTAT-GTGCTGGC-3Ј), and GS3R (reverse 5Ј-ACAGCAGGGTTGCCCCGG-3Ј). Reverse-transcription PCR (RT-PCR) was also carried out. Total RNA (0.6 g) from N2 cells was used as starting material. The first strand synthesis was performed at 60°C for 30 min using avian myeloblastosis virus reverse transcriptase with primer GS3R or GS4R followed by inactivation of the enzyme at 95°C for 5 min, and then each forward primer was added to the mixture. The thermal cycle profile for both PCR and RT-PCR was as follows: 95°C for 1 min, 60°C for 1.5 min, 72°C for 2.5 min for 30 cycles. Each 7 ml of the amplification product was digested with 50 units of SmaI in a total volume of 10 ml at 30°C for at least 12 h. The digested-amplification products were electrophoretically separated on 2% agarose gels and visualized by ethidium bromide staining under UV-light. Quantitative determination of products was carried out using scan analysis software with photographs of the gels.
Plasmids for cDNA Expression-pcDL-SR␣296 vector (23) was kindly provided by Dr. Takebe of the National Institute of Health (Japan). pMSXND vector (24) contains metallothionein promoter for foreign cDNA expression and dhfr gene for selection of clones and gene amplifying.
Construction of Fc␥BPf Expression Plasmids-Plasmids to express about half of full-length cDNA were constructed as follows. Five cDNA fragments (see Fig. 1) were used for construction of expression cDNA. The fragment of nucleotide 1 (5Ј end) to nucleotide 313 (BglII) was derived from clone NZ4, nucleotide 314 (BglII) to nucleotide 1244 (BstXI) was from C72, nucleotide 1245 (BstXI) to nucleotide 1664 (Hin-cII) was from Y11, nucleotide 1665 (HincII) to nucleotide 4442 (BamHI) was from X1, and nucleotide 4443 (BamHI) to nucleotide 7784 (3Ј end of the clone) was from V11. The fragments were ligated to each other, and the ligated 7.9-kb cDNA fragment was inserted at the cloning site of pcDL-SR␣296, which was under control of the SV40 promoter, to yield pSR-NV11.
Fc␥BP cDNA Deletion Derivatives and Their Expression-The deletion constructs were prepared as follows. (i) The 5Ј end-BglII fragment of NZ4 clone was ligated to the BglII site of C72 clone to produce NZ. (ii) NZC was ligated to the ϳ420-bp fragment of Y11 clone at the BstXI site to yield NZCY. (iii) The XhoI (cloning site on a vector)-BstXI fragment from NZC was ligated to the BstXI-HincII fragment from Y11, and the resultant 1.7-kb fragment was inserted in the HincII-digested X1 clone to produce NX. (iv) NV11 was recovered from the plasmid by digestion with NotI and inserted in pUC119, of which the HincII site changed to NotI site to generate pUC-NV11. pUC-NV11 was digested with HincII, BssHII, Tth111I, or SplI and SpeI, respectively, and each digest containing vector sequences was self-ligated to produce ÇHinc, ÇBssH, ÇTth, or ÇSpl, respectively. (v) ÇTth, NX, and ÇHinc were digested with BssHII, and then each digest containing vector sequences was self-ligated, respectively, to generate ÇBssH/Tth, NX, ÇBssH, and ÇHinc/BssH. (vi) NotI-digested insert of V11 was blunted by Klenow fragment and was inserted into the blunted SpeI site of NZC to produce NZCV11.
For expression, the original initiation site of each construct was alive except for X1. To express clone X1, an oligonucleotide adapter containing the initiation site sequences originated from Fc␥BP cDNA (or NV11), intervened between HindIII-and EcoRI-site sequences, was designed as follows (5Ј-AAGCTTCTGCAGCCATGGGGGATCC-3Ј). The oligonucleotide was inserted into HindIII-EcoRI site of pBLS-STOP to make transcription start followed by cDNA sequences. The insertion sequences of clone X1 digested with EcoRI were ligated into EcoRI site of pBLS-STOP to produce X1. Stop codon sequences were also designed as follows. The two adapter sequences were synthesized (upper strand 5Ј-CTAGTTAGTTAGTTAGGGTACCGC-3Ј and lower strand 5Ј-GGC-CGCGGTACCCTAACTAACTAA-3Ј). This double strand provided stop codon in all frames. It was inserted between SpeI and NotI site of pBluescript II SK(ϩ), followed by modification of cloning site of XhoI to XbaI site to yield pBLS-STOP vector. All constructs encoding Fc␥BP derivatives were cloned into pBLS-STOP, and the insertion sequence containing stop codons was subcloned into pcDL-SR␣296 for protein expression. The expression vectors containing these deletion constructs were transfected transiently in COS7 cells, and produced proteins were stained with Fc fragment of human IgG or monoclonal antibodies K9 and K17.
Transfection-For transient expression, COS7 cells were grown in RPMI 1640 medium containing 10% fetal bovine serum (FBS), 10,000 IU/liter penicillin, and 100 mg/l streptomycin. COS7 cells were plated at 1 ϫ 10 5 cells/35-mm tissue culture dish a day before transfection. Then the cells were transfected with pSRNV11 by lipofection procedure using 10 g of DNA and 5 l of Transfectam (Promega) per 500 l of RPMI 1640 medium. After 6 h, medium was changed and cultured for a further 48 h. For permanent expression, CHO cells were transfected with a pMSXND vector bearing Fc␥BP cDNA fragment, NV11, or the deleted cDNA by lipofection. Cells were passaged the day before transfections, and the subconfluent 2 ϫ 10 5 cells were treated with CsClpurified plasmids in the same way as COS cells. The cell medium was replaced with F-12 medium supplemented with nucleotides and 10% FBS and incubated at 37°C for 2 days. Then the cells were selected in ␣-minimum Eagle's medium without nucleotides containing 1 mg/ml G418, 10% FBS, and antibiotics. After culturing under these conditions for 14 days, the clonal cell lines were obtained by limiting dilution and were examined for Fc␥BP expression by immunostaining. Then the clonal cell lines were propagated with increasing concentrations of methotrexate, starting with 0.02-6.4 M. The gene-amplified cells were confirmed by a higher amount of expression of Fc␥BP, and clonal cells were obtained by limiting dilution.

Cloning and Sequencing of Full-length cDNA for Human
Fc␥BP-Random primed cDNA libraries (Ͼ600 bp) were constructed from human colonic epithelial cells using gt11 as a vector. The gt11 human epithelial cell library was first screened with two independent monoclonal antibodies against Fc␥BP, K9, and K17 that block the binding of IgG Fc fragment (14). The screening of 1 ϫ 10 6 recombinants with monoclonal antibody K9 led to the isolation of a cDNA clone 618 bp long (Fig. 1B) in the region defined by nucleotide position 13788 -14405 (probe q). Furthermore, the screening of 6 ϫ 10 5 recombinants with monoclonal antibody K17 led to the isolation of 7 clones, one of which was about 1300 bp long and could be digested with BamHI into 2 fragments (a and b). These adjacent fragment probes (a/b) were corresponding to nucleotide positions 7368 -8045/8046 -8697 or 10970 -11649/11650 -12300, respectively. The nucleotide sequences of probes q and a/b revealed that they are independent cDNA clones. However, both probes hydridize to a single band of Ͼ16 kb on a Northern blot (described below). These results suggest that the two clones are derived from the same mRNA for Fc␥BP. Three probes, q, a, and b, were used to determine the full-length cDNA sequence. Out of more than 70 clones isolated, 10 clones (T5, A43, A8, A31, A40, A53, V11, X1, Y1, and C72) were shown to cover most of full-length cDNA for Fc␥BP except for 5Ј-terminal sequence (Fig. 1C).
Sequence analysis of these clones showed that the Fc␥BP cDNA is composed of three homologous units that are tandemly repeated (see Fig. 1B). Probe q hybridizes with 3 regions, as does probe a/b. Each unit shares more than 95% homology with one another. Although a cDNA clone representing the 3Ј-terminal region, T5, was isolated from oligo(dT)-primed cDNA library, no 5Ј-terminal clones (i.e. extending from C72) were isolated from human colonic libraries. Thus, we constructed a cDNA library of human N2 cell line, which can differentiate into goblet cells and express Fc␥BP (25,26). Using a 5Ј-terminal probe derived from the C72 clone, we obtained a clone, NZ4, containing the 5Ј-terminal ATG. Sequences flanking the first ATG (GCC(A/G)CCATGG) in NZ4 clone are consistent with those described by Kozak (27) for an initiation codon. The complete nucleotide sequence data has been submitted to the DDBJ, EBI, and GenBank TM Data Banks; a map of the major restriction enzyme sites is shown in Fig. 1A.
Predicted Amino Acid Sequence and Sequence Homology of Fc␥BP with MUC2 and von Willebrand Factor-A full-length sequence of the predicted amino acids is shown in Fig. 2A. A dot matrix plot of the entire predicted amino acid sequence against itself revealed 12 repeated domains flanked by unique N-terminal (450 amino acids) and C-terminal (160 amino acids) domains (data not shown). These 12 repeated domains (r1-r12 in Fig. 2B) were classified into 5 types that shared 30 -40% homology. The R1 unit (r3-r5) corresponds to the first 5Ј-terminal A-B-C-Q region in Fig. 1. The repeated domains r1-r12 are each composed of about 400 amino acids.
As some mucin-related proteins have been reported, the predicted amino acid sequence of human Fc␥BP was compared with those amino acid sequences, and a broader search for similarities to other protein sequences was carried out using Gene Works (IntelliGenetics, Inc.) loading GenBank TM Release 8.7. These analyses demonstrate that the Fc␥BP sequence has significant similarity to portions of MUC2 and prepro-von Willebrand factor (vWF) but that it does not have homology to either Fc receptors or IgG-like domain. A close examination reveals that each repeated domain of Fc␥BP was homologous with four MUC2 D domains and to four prepro-vWF D2 domains (about 30% homology). Interestingly, as shown in Fig. 3, Fc␥BP, vWF, and MUC2 all have the conserved amino acid motif CGLCGN. These sequences are also characteristic of thioredoxin (28) and protein disulfide isomerase (29). Another feature of this protein is that it is high in cysteine content. As shown in Fig. 2B, 8.1% of the amino acid residues are cysteines (cf. 5.7% for human serum albumin (hSA)). A total content of the predicted serine/threonine residues for O-linked glycosylation is 12.3% (Fig. 2B, cf. 9.3% for hSA). Further, the content of hydrophobic and neutral amino acids is 78.6% (cf. 63.8% for hSA).
Size Determination of Native mRNA for Human Fc␥BP-Northern blot analysis was carried out to detect the expressed mRNA for Fc␥BP from colonic epithelial cells. Probes q, a, or y (an 800-bp probe derived from the 5Ј-terminal region of clone Y1) were used. A single band of larger than 15 kb with a smeared front was detected with each probe, as shown in Fig.  4A. Probe y gave the same Northern blot pattern as with probes q and a, indicating that these three probes hybridize to the same mRNA. The same result was obtained using a part of clone NZ4 as a probe (data not shown).
As described above, however, mRNA size is critical for determining the number of 4.5-kb units (A-B-C-Q). To determine more precisely the size of the Fc␥BP mRNA, we compared it with mRNAs for ryanodine receptor (15.2 kb) and for dystrophin (14 kb), which are among the longest mRNAs reported. We prepared the respective probe for them by PCR method, using poly(A) ϩ RNA from human cytoskeletal muscle. Since Fc␥BP has been shown to be present in human N2 cells by immunohistochemical staining with monoclonal antibodies K9 and K17 (26), we carefully prepared the mRNA from the N2 cells. As shown in Fig. 4B, Northern analysis using probe a showed that the size of our mRNA was estimated to be 17 kb by relative mobility of two standard mRNAs. These results confirmed the fact that a 4.5-kb-long unit was repeated 3 times, in addition to 5Ј-terminal and 3Ј-terminal unique sequences.
Transcription Start Sites-To confirm that the 5Ј-terminal cDNA sequence of human Fc␥BP is identical to genomic DNA, approximately 2 ϫ 10 6 recombinant phages from a human leukocyte genomic library were screened using the 5Ј-terminal fragment of clone NZ4 as a probe. One independent clone, GHFc1, and two overlapping clones, GHFc2 and GHFc3, were identified. A 1908-bp SacI/EcoRI fragment at the 5Ј-flanking region from GHFc1 as well as the exon 2/intron boundary regions from GHFc2 and GHFc3 were subcloned into pBluescript and sequenced using exonuclease III-generated deletion templates. Fig. 5 shows the sequence of the first and second exons of genomic DNA, which completely coincides with the corresponding region of cDNA. The exon/intron splice sites were confirmed to be GT-AG border element consensus sequences. The first exon contains the putative ATG initiation codon suitable for Kozak's rule (27), as described above. Examination of the 5Ј untranslated region reveals a TGA in-frame stop codon at Ϫ78 position of the gene. However, no TATA, CAAT, or other promoter/enhancer motifs were detected within 2 kb further upstream (data not shown).
To identify the transcription start sites, primer extension and S1 mapping were also performed. Primer extension analysis using an oligonucleotide complementary to the sequence double-underlined in Fig. 5 reveals multiple start sites (see Fig.  6A). Three major bands are located at ϩ27, ϩ28, and ϩ30 nucleotides downstream. The upper broad band is located at Ϫ5 to ϩ25 in a boundary region of 5Ј terminus for cDNA clone NZ4. Similar results are observed using human colon epithelial cells, except for additional faint bands. Clusters of multiple start sites are common for genes lacking a TATA box (30).
As shown in Fig. 6B, the end-labeled 205-bp fragment hybridizes with total RNA from human colon epithelial cells protected from S1 nuclease digestion, and several hybrids are detected. These products also reveal multiple start sites, and the longest digestion showed the predominant start site to be an adenosine residue that is only 9-bp upstream from the deduced ATG translation initiation site. Thus, the same transcription start sites are suggested by the broad extension band seen in primer extension analysis.
Distribution of SmaI RFLP-To detect sequence variants and to study whether those variants are allele-specific or repeated region-specific in each allele, RFLP analysis with SmaI was investigated for 10 individuals and cultured N2 cells. We designed two distinct primer sets, as described under "Materials and Methods." The PCR-amplified products are 186 bp (primers BC1 and GS4R) as shown in Fig. 7 and 107 bp (primers GS1 and GS3R) (data not shown), and the complete digestion of these products with SmaI was followed by electrophoresis. The digestion with SmaI results in fragmentation into 113 and 73 from 186 bp and into 72 and 35 from 107 bp, respectively. In the case of sample 4, only 113-and 73-bp SmaIdigested bands were observed, but 9 other cases showed the additional band (186 or 107 bp) non-digested with SmaI. Using primers GS1 and GS3R, consistent results were obtained by detecting a set of 72-and 35-bp fragments after SmaI digestion.
Reflecting the sensitivity to SmaI digestion, we determined two alternative sequences. One was present in clones of A53, A8, and others (ACT-GGC-TGC-CCC-GGG-GGT), which are sensitive to SmaI, and another sequence (ACT-GGC-TGC-CTG-GGG-GGT) present in clones of V11 and others, which are resistant to SmaI. This difference in sequence results in chang- ing one amino acid residue of proline to leucine. As Fc␥BP gene has three large repeats, six large repeats are involved in a diploid genome. Although theoretical ratios of SmaI-digested bands to undigested bands are 6:0, 5:1, 4:2, 3:3, 2:4, 1:5, or 0:6, no case in which all 6 repeated regions are SmaI-resistant has yet been observed. RT-PCR of mRNA from N2 cells reveals the message to be all SmaI-sensitive (Fig. 7, sample RT).
Expression Pattern in Human Tissues and Other Species-Human Fc␥BP mRNA expression in various human tissues was examined. 17-kb mRNA is expressed in the placenta and colonic epithelial cells, but no expression was detected in heart, brain, lung, liver, skeletal muscle, or kidney (Fig. 8). Since IgGFc binding activity has not yet been detected in non-human mammals such as mice and rabbits (13), we investigated the species specificity of Fc␥BP gene. The gene for Fc␥BP was detected in human and monkey but not in mouse, rabbit, rat, dog, bovine, or porcine by zoo blot analysis (data not shown) even when hybridization was carried out under several diffferent conditions (data not shown).
Expression of Active Recombinant Molecules-The large size of the intact cDNA for human Fc␥BP precluded expression in a standard expression vector. Therefore, COS cells were transfected with a cDNA fragment containing the H domain and only one A-B-C repeat unit. About 30% of transfected cells produced proteins that reacted with monoclonal antibodies K9 or K17 against Fc␥BP, compared with nontransfected cells (Fig. 9, panels a, b, and c). We then examined whether transfected COS cells had IgG binding activity using IgG conjugated with horseradish peroxidase (HRP). The fraction of transfected COS cells bound to HRP-IgG was almost the same as that stained with monoclonal antibodies (Fig. 9, panel d). Furthermore, a similar fraction of transfected COS cells bound IgG Fc fragment, whereas IgG F(abЈ) 2 failed to bind to the cells. We also examined the binding of IgM, IgA, and secreted IgA to the transfected cells. None bound to the cells (data not shown). These results show that the NV11 cDNA fragment for human Fc␥BP is sufficient to express a biologically active protein fragment (designated as Fc␥BPf) that can bind IgG Fc but not IgG F(abЈ) 2 .

Inhibition of IgG Binding by Heat-aggregated IgG Using
Fc␥BPf-Expressed CHO Cells-We isolated a CHO stable transfectant clone expressing Fc␥BPf. More than 90% of cells produced the protein product. Therefore, we could determine the binding activity quantitatively using HRP-conjugated human IgG. Fig. 10 shows the competitive inhibition of HRPlabeled IgG binding in the presence of the various concentration of human monomeric IgG or heat-aggregated IgG. Monomeric IgG can specifically inhibit the binding of HRPconjugated IgG, and heat-aggregated IgG shows 10 times stronger inhibition than monomeric IgG. These results suggest that polymeric IgG, like the heat-aggregated form, has a higher affinity to Fc␥BPf than monomeric IgG.
Functional Analysis of Fc␥BPf cDNA-Subsequently, we prepared nested deletions of NV11 cDNA to identify regions of cDNA essential for biological activity (Fig. 11). Monoclonal antibody reactivity and IgG binding activity of the protein fragment produced in the transfectants are also summarized in Fig. 11. Notably, clones deleted within the H region (ÇHinc, ÇHinc/BssH, or X1) can express proteins with monoclonal antibody reactivity but no IgG binding activity. These results suggest that the subregion from r1 through r5 (Fig. 11) is responsible for IgG binding and that the H region is essential for expression of functional molecules. Although about half the r5 domain is deleted in clone ÇSpl, its reactivity with monoclonal antibody K9 is retained. In clones ÇTth and ÇBssH/Tth, K17 reactivity is attributable to a part of the r6 domain, which is highly homologous to the r3 domain.
Prediction of IgG Binding Sites-Domains r1-r5 show 30 -40% amino acid sequence homology with one another. We tried to determine whether each domain is responsible for IgG binding. COS cells transfected with the deleted cDNAs were treated with HRP-IgG for IgG binding assay and also incubated with the HRP-IgG in the presence of excess unlabeled competitors, monoclonal antibodies K9 and/or K17. When some r domains were deleted (e.g. clones NX and NZCY in Fig. 11), the strength of IgG binding was proportional to the number of intact r domains; a clone completely lacking r domains (such as NZC) showed no IgG binding activity. Since monoclonal antibodies Capital letters indicate exon sequences, and small letters represent intron or 5Ј untranslated regions. Underlined triplet denotes in-frame TGA stop codon. Doubleunderlined sequences show primers for primer extension and S1 mapping. The putative transcriptional start site is in bold below a superscript Ͼ. The dashed line shows the omitted sequences of an intron.
K9 and K17 interact with the r3 and r5 domains, respectively, K9 and K17 competitively inhibited the IgG binding activity of clones expressing Fc␥BP containing r3 and/or r5 domains (Fig.  11). These findings suggest that at least r1, r3, and r5 domains are involved in the IgG binding capacity of Fc␥BPf.

DISCUSSION
Predicted Characteristics of Fc␥BP-Our cDNA sequence reveals that full-length Fc␥BP comprises 5405 amino acid residues (Fig. 2). Based on its structural features, this protein can be divided into three major domains. The largest central domain is composed of 12 tandem repeats of about 400 amino acids each (Fig. 2B). Each repeat contains about 8% cysteine residues. Repeats are homologous (about 30 -40%) to each other. The lability of IgG binding activity in periodic acid and hydrogen peroxide treatments (13) and broad bands observed on SDS-PAGE gels suggested that Fc␥BP is a highly glycosylated protein. The presence in the predicted amino acid sequences of many N-and O-linked glycosylation sites (15.5%) is consistent with this. The structure of Fc␥BP is related to the mucin-like protein MUC2. Taking together its intracellular transport and localization in goblet cells (14), this fact indicates that Fc␥BP may be a component of mucus.
Gel-forming mucin is thought to be a giant molecule formed by several to tens of molecules bound by inter-cystine bonds (31). The tandem repeat domain, for example, composed of repeats of 17-amino acid units for MUC3 (32), 16 for MUC4 (33), and 169 for MUC6 (34) is characteristic of conventional mucin. However, no such domain containing short repeat units was detected in the molecule of Fc␥BP. Nevertheless, Fc␥BP should be classified as one of the mucins because of the fact that (i) it has a high molecular weight (Ͼ200 kDa) with S-S linkages (13), (ii) it is secreted with mucus from goblet cells into the intestinal tract (14), (iii) it may be glycosylated (13), and (iv) it contains several cysteine-rich domains (Figs. 2 and 3).
Thus, Fc␥BP appears to be a mucin-like protein and to be involved in the maintenance of the mucosal structure as a gel-like component of the mucosa.
Gene Polymorphism-We analyzed cDNA libraries prepared from human tissue and the N2 cell line, and detected several different types of Fc␥BP cDNA, as assayed by SmaI sensitivity. Since a difference was specifically seen in the SmaI site of highly homologous 3.6-kb DNA repeats, we focused on the analysis of polymorphisms of this site. This analysis revealed several particular digestion patterns dependent on the number of SmaI sites, including polymorphisms of each allele in the Fc␥BP gene (Fig. 7). This finding also suggested the expression of Fc␥BP gene in different alleles in individuals. No major variations in the Fc␥BP gene, such as variations in the number of homologous repeats or in splicing, have been detected to date (Fig. 4). While the amino acid residue of the SmaI site was changed from proline to leucine or reverse, it is unknown whether polymorphism of the Fc␥BP open reading frame caused by point mutation alters the Fc binding activity of Fc␥BP related to colorectal diseases, as is the case for familial intestinal polyposis and polymorphisms of its causative gene (adenomatous polyposis coli gene) (35). The physiological function of gene polymorphisms in Fc␥BP remains to be elucidated.
Role of Subdomain-The protein Fc␥BPf produced by the expression of an approximately 8-kb fragment including the 5Ј terminus of Fc␥BP cDNA (NV11 clone) contained an H domain and 6 r domains (Fig. 11). To identify the domain possessing Fc binding activity, we assessed activity using a series of deletions in cDNA. When some of the r domains were deleted, IgG binding tended to become weaker in proportion to the length of the FIG. 6. Determination of mRNA initiation site for Fc␥BP transcript by primer extension analysis and S1 mapping analysis. A, the antisence 50-mer (5Ј-GCTGATAGTTCTGCAGGAAGGCTGTGAG-GAATTCCTCTCTGGCCAGTGTTC-3Ј) oligonucleotide complementary to nucleotides ϩ95-144 was synthesized and purified. The 50-mer oligonucleotide was used in primer extension after labeling the 5Ј end. Universal primers were used for the sequence ladders with the singlestranded M13mp18 DNA template as size marker (lanes T, C, G, and A). The numbers from the primer and of the transcription start site are indicated on the left side and in parentheses, respectively. Lanes HCE and N2 indicate the cDNA reverse-transcribed from RNAs from human colon epithelium cells and N2 cells, respectively. Band locations are shown by arrows and a solid line on the right side. B, the end-labeled 33-mer (5Ј-GCTCCAGCCCAGCCCAGAGTATCCACCAGCTCCATAG-G-3Ј) oligonucleotide was used as a primer for the sequence ladders (lanes T, C, G, and A). Single-stranded DNA complementary to nucleotide Ϫ150 to ϩ50 was labeled and used as template in S1 nuclease mapping (lane S1). The arrow (ϩ1) denotes the transcription start site indicated in Fig. 5. remaining r domains. The deletion experiments also showed that r5 domain is critical for IgG binding. Competition experiments using inhibitory monoclonal antibodies suggest r3 and r5 domains are involved in the binding of Fc although the r1 domain of clone NZCY still shows significant binding activity. Staining with monoclonal antibodies revealed that K9 recognizes the r5 domain, as well as r11, and that K17 recognizes the r3 domain, as well as r6 (Fig. 11). Our results further suggest that both the r3 and r5 domains possess independent Fc binding sites as well as r1 domain. Therefore, we speculate that at least three IgG molecules can bind to a single Fc␥BPf protein molecule. Furthermore, the inhibition of IgG binding by heataggregated form implies that the polymeric form such as immune complexes may be a better ligand for Fc␥BP.
Considering the lower cysteine content of H domain and its unique amino acid sequences, it is likely that it may differ in nature from r domains. Deletion of the H domain results in loss of Fc binding activity (Fig. 11). This suggests that the H domain plays an important role in maturation of biologically active protein products. We speculate that H domains are involved either in the processing of the Fc␥BP polypeptide into an active form with Fc binding activity or in the intracellular localization for suitable protein processing, such as Golgi apparatus or mucus granules. These speculations seem to be reasonable because the following facts were observed. Although the predicted molecular weight of the whole Fc␥BP (about 5,000 amino acids) is more than 500 kDa, Kobayashi et al. (14) detected a more than 200 kDa band by non-reduced electrophoresis and 70 -80 kDa bands under reduced conditions. Thus, the Fc␥BP may be processed by protease after translation. So far, our Fc␥BPf expressed in COS and CHO cells is also recovered as broad bands of 50 -80 kDa under reduced conditions by Western blot analysis. 3 Physiological Functions of Fc␥BP-Our previous study using monoclonal antibodies showed the presence of Fc␥BP in the mucosa of the large and small intestines (13,14). In the present study, Fc␥BP mRNA is detected not only in the colorectal epithelial cells but also in the placenta (Fig. 8). These findings imply that Fc␥BP may be distributed in the systemic mucosa, as are the secreted mucins MUC2, MUC3, and MUC4. Fc␥BP was detected only in humans and monkeys by Southern blotting. Therefore, Fc␥BP may play a role in the mucosal immune system of primates. IgA and IgM are transported within epithelial cells via polymeric Ig receptors, and they eliminate toxic antigens from the lamina propria. Analysis of the amino acid sequences of Fc␥BP, however, reveals no membrane-penetrating domain or signal peptide sequences (Fig. 2). Immunohistochemical studies reveal that Fc␥BP is transported from Golgi apparatus to the mucous follicles (14) although it is unknown whether protein localization is dependent on binding to IgG. It therefore seems unlikely that Fc␥BP serves as a transporterlike poly-Ig receptor or incorporates IgG-like FcRn. However, similar to poly-Ig receptor, a portion of which serves as a secretory component and contributes to multivalent IgA formation and its stabilization, it is likely that Fc␥BP enhances antigen trapping through its promotion of multivalent IgG  (d-f). cDNA fragment, NV11, was constructed using clones NZ4, C72, Y1, X1, and V11 as described under "Materials and Methods." The NV11 fragment was inserted into mammalian expression vector pcDL-SR␣ 296. The expression vector was transfected transiently into COS7 cells and stained with the indicated antibodies followed by HRP-conjugated anti-mouse IgG F(abЈ) 2 fragment to detect protein product. a, monoclonal antibody K9; b, monoclonal antibody K17; and c, without antibody. The IgG binding activity was detected by the combination of the indicated immunoglobulin and the second HRPconjugated anti-human IgG F(abЈ) 2 . d, human whole IgG; e, IgG Fc fragment; f, IgG F(abЈ) 2 fragment.
FIG. 10. Competitive inhibition of IgG binding to Fc␥BPf by monomeric IgG and heat-aggregated IgG. The CHO cells stably transformed by NV11 expression vector were cultured and fixed to 96-well micro test plate. HRP-labeled human IgG (5 g/ml) was added to the cells with indicated concentrations of monomeric IgG (open circle) or heat-aggregated IgG (closed circle). After washing, the amount of bound labeled IgG was determined under the standard assay conditions, and the supernatant reaction mixtures were transferred for colorimetric determination at 495 nm with a plate reader. formation or protects IgG from degradation by bacterial proteases. Further, secreted into the mucus, native Fc␥BP may prevent the invasion of antigens into the mucosa by efficiently trapping antigen-IgG complexes through its binding to more than nine of the aggregated IgG complexes. Further studies are being conducted to reveal where Fc␥BP is able to bind IgG molecules during protein processing.
Fc␥BPf expressed from an 8-kb cDNA are cleaved into several polypeptides cross-linked by disulfide bonds. 3 Intra-and inter-disulfide bonds are also suggested by the conserved motif (CGLCGN) in all the repeated domains (Fig. 3). Interestingly, such vicinal cysteines are conserved in all members of thioredoxins, as well as vWF and MUC2. Thioredoxins are involved in a wide variety of biochemical systems, and the vicinal cysteines (CGPC for thioredoxins) are essential for redox functions in E. coli. In mammalian cells, thioredoxin functions as an endogenous glucocorticoid receptor-activation factor (36). Moreover, thioredoxin-like domains have been found in several proteins of higher molecular weight, such as protein disulfide isomerase (29) and phosphoinositide-specific phospholipase C (37). Another interesting function of eukaryotic thioredoxin is its stimulation of interleukin 2 receptor expression in human T-cell leukemic virus (HTLV)-1 transformed T cells, an activity originally described as adult T-cell leukemia-derived factor (ADF) (38,39). Recently, recombinant ADF prevented the cytotoxicity caused by H 2 O 2 (40). These results suggest that high cysteine content, as well as the conserved sequences in human Fc␥BP molecules, serves as an anti-oxidant in mucus. This speculation led to protective elimination of antigen-immune complex from the intestinal tract.
Regarding the vicinal cysteines, another potentially interesting parallel is the possibility that Fc␥BP, like pro-vWF (41), may have the ability to catalyze its own oligomerization. Pro-vWF autocatalysis occurs in vitro at low pH and is dependent upon the integrity of two sets of vicinal cysteine residues, both with the sequence CGLC (42). These tetrapeptides apparently conduct thiol oxidation through a mechanism involving the formation of a strained 14-member ring joined at the sulfur atoms of the two cysteine residues. These sequences are conserved in each subdomain of Fc␥BP. These vicinal cysteines may be important in processing to an active form as well as H domain, and in formation of structural networks by mucin proteins. FIG. 11. Schematic diagram of deletion constructs and the determination of the monoclonal antibody reactivity of protein products and their Fc binding activities. A series of deletion derivatives generated from NV11 were produced as shown in the left panel and were then constructed into expression vector pcDL-SR␣ 296. The constructs were transfected transiently into COS7 cells, and the protein products were detected by the monoclonal antibodies K9 and K17. The binding activities to the Fc fragments of IgG were determined after the cells were fixed with ethanol. The binding activity was evaluated by the staining intensity of each cell.