Identification of a novel transmembrane semaphorin expressed on lymphocytes.

Semaphorin (also known as collapsin) members are thought to be involved in axon guidance during neural network formation. Here, we report the isolation of a novel member, mouse semaphorin G (M-sema G), which encodes a semaphorin domain followed by a single putative immunoglobulin-like domain, a transmembrane domain, and a cytoplasmic domain. M-sema G is most closely related to M-sema F, which we previously reported, and semB and semC. These four members appear to constitute a transmembrane type subfamily in mouse semaphorins. In contrast to the predominant expression of M-sema F mRNAs in the nervous tissues, M-sema G mRNAs are strongly expressed in lymphoid tissues, especially in the thymus, as well as in the nervous tissues. The mRNAs are also detected in various cell lines from hematopoietic cells. By generating specific antibodies, we confirmed the strong expression of M-Sema G proteins on the surface of lymphocytes. These results provide the first evidence that semaphorin is expressed on lymphocytes and suggest that semaphorins may play an important role in the immune system, as well as in the nervous system.

During development, neuronal axons navigate to their appropriate targets. Many factors are involved in the guidance of neural growth cones to their final targets. However, the molecular nature of the mechanism of this navigation remains largely obscure. Adhesion molecules are among the main candidates for signals controlling the direction of growth cone extension. Additional molecules have recently been identified that act as attractive and repulsive cues effective in modulating the routes of neurite outgrowth (1)(2)(3)(4)(5). Collapsin was identified and purified from adult chick brain using a growth cone-collapsing assay of dorsal root ganglia. A medium of collapsintransfected COS-7 cells did not induce the collapse of retinal growth cone but did induce the collapse of dorsal root ganglia growth cones (2). A domain of about 500 amino acids of collapsin reveals high similarity to grasshopper Fasciclin IV, later renamed G-Sema I, 1 which is required for the proper guidance and fasciculation of the Ti1 growth cones in the limb bud of the grasshopper embryo (6). This domain is called the sema domain, and several molecules with this domain constitute the semaphorin family. Kolodkin et al. (7) have identified two members from Drosophila melanogaster, D-sema I and D-sema II. Sema IIIs identified from human and mouse are most likely to be mammalian homologues of chick collapsin. Puschel et al. (5) have isolated five semaphorin members from the mouse and predicted that all members are secreted proteins. All members share the motifs characteristic of sema domains, including conserved cysteine residues, potential N-glycosylation sites, and immunoglobulin (IgG)-like domains. By the similarity of the sema domains, they divided mouse semaphorins into two groups, group III and group IV. Group III includes sema III (also called semD), semA, and semE, and group IV includes semB and semC (5). We recently isolated an additional mouse semaphorin, M-sema F, which is predicted to encode a transmembrane domain by a hydrophobicity analysis (5,8).
In the present study, we isolated a novel member that is closely related to M-sema F, semB, and semC named M-sema G, which also encodes a transmembrane domain. Surprisingly, M-sema G mRNAs are very strongly expressed in lymphoid tissues, especially in the thymus, as well as in the brain. Generation of antibodies to M-Sema G confirmed strong expression of the protein on lymhocytes. These results suggest that M-sema G may play an important role in the immune system, as well as in the nervous system.
Isolation of cDNA Clones-An adult mouse brain library primed with oligo(dT) (Stratagene) was screened using probes radiolabeled by random priming (Pharmacia Biotech, Inc., Uppsala, Sweden). Probes were * This research was supported by Grant-in-aid 20238702 from the Ministry of Education for Scientific Research, Grant-in-aid 08044282 from the Ministry of Education, the Monbusho International Scientific Research Program, and a grant from the Ichiro Kanehara Foundation. 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 GenBank TM /EBI Data Bank with accession number(s) U69535.
DNA Sequencing and Analysis-Sequencing was carried out by a dideoxy chain termination method, using a Taq dye primer cycle sequencing kit or a Taq dye terminator cycle sequencing kit and ABI 373A DNA sequencer (Applied Biosystems). The final sequence was confirmed from both strands. Sequence analysis or comparison was done through BLAST on NCBI.
In Situ Hybridization-ddY bred mice were purchased from SLC (Hamamatsu, Japan). The day of detection of the vaginal plug was designated E0.5. Embryos at E14.5 were frozen on dry ice powder, sectioned on a cryostat at Ϫ15°C (15-m thickness) and processed by an in situ hybridization procedure described previously (8). Antisense and sense RNA probes were labeled with 35 S-UTP corresponding to the above-mentioned PCR fragment ligated to pCR-Script SK(ϩ) by T3 and T7 RNA polymerase.
Expression of a Fusion Protein of A18 -For the construction of two plasmids encoding GST-A18 and A18-His, the pGEX-4T2 vector (Pharmacia) and the pET21a vector (Novagen, Madison, WI) were digested with EcoRI and XhoI. The cDNA fragment flanked by EcoRI and XhoI of the presumed intracellular portion of A18 was prepared by PCR using primers A18 -5 (5Ј-CGGGGAATTCTCCTACAACTGCTACAA-GGG-3Ј) and A18 -3 (5Ј-AAATCTCGAGGTCCCCGTCAGCATCCGA-AT-3Ј). EcoRI and XhoI sites in the primers are underlined. This cDNA fragment was ligated to the vectors. Escherichia coli (XL1-blue) transformed with each plasmid was cultured at 37°C in 500 ml of Luria-Bertani medium with 500 g/ml ampicillin, and expression of the fusion proteins was induced by adding 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside when absorbance at 600 nm of the culture reached 0.40. After a 6-h incubation, cells were centrifuged, and each pellet was resuspended in 20 ml of 20 mM Tris-HCl (pH 7.4), 100 mM NaCl, 1 mM EDTA, 1% Triton X-100, and 1 mM phenylmethylsulfonyl fluoride. The bacteria were then lysed by sonication and centrifuged at 10,000 ϫ g for 30 min at 4°C. The supernatant was treated with glutathione-Sepharose beads (Pharmacia) and Ni-NTA resin (Qiagen, Hilden, Germany) following the protocols. The purified fusion proteins were stored at Ϫ80°C.
Preparation of Antibodies (Abs)-Two rabbits purchased from Kitayama Labeus (Kyoto, Japan) were immunized by an injection of A18-GST fusion protein emulsified with Freund's complete adjuvant followed by an injection every 2 weeks of the protein emulsified with Freund's incomplete adjuvant. Serum collected from the rabbits was purified with an A18-His protein affinity column.
RNA Analysis-Total RNA was prepared from various cell lines and murine organs by the acidic guanidium isothiocyanate-phenol-chloroform extraction method (9). Blood-rich organs (liver, lung, kidney, and brain) were dissected after perfusing 20 ml of phosphate-buffered saline (PBS) from the left ventricle to the right atrium to minimize the contamination of blood cells. RNA was extracted only when the color of organs turned pale after perfusion. An aliquot of each RNA sample was electrophoresed on a 1% agarose gel, and the intactness of the RNA was confirmed by visualizing 18S and 28S rRNA bands.
RNase protection assay was performed as described (10) with minor modifications. Briefly, we subcloned a cDNA fragment (nucleotides 1-260) into pBluescript SK and linearized the plasmid by restriction digestion. Radiolabeled antisense RNA probes were synthesized using [␣-32 P]UTP (DuPont NEN) and T3 RNA polymerase. After purification by 6% polyacrylamide gel electrophoresis, labeled RNA (1-2 ϫ 10 5 cpm) was hybridized to total RNA samples (5 g) in 15 l of the hybridization buffer containing 40 mM PIPES, pH 6.4, 400 mM NaCl, 80% formamide, and 1 mM EDTA. After hybridization for 12-20 h at 50°C, samples were digested with RNase A and T1 at 37°C for 60 min and separated on 6% polyacrylamide/7 M urea gel. Data were visualized by autoradiography.
Construction and Expression of the Influenza Virus Hemagglutinin (HA) Fusion Protein-To create a HA-tagged M-Sema G fusion protein, the annealed oligonucleotide corresponding to a tandem repeat of HA epitope (5Ј-GGCCGCTTACCCATACGATGTTCCGGATTACGCTTACC-CATATGATGTTCCGGATTACGCTTAA-3Ј and 5Ј-CTAGTTAAGCGT AATCCGGAACATCATATGGGTAAGCGTAATCCGGAACATCGTAT-GGGTAAGC-3Ј) was inserted into the NotI-XbaI sites of the mFas-Fc plasmid (11). The BamHI-NotI fragment of this plasmid, encoding a fusion protein of the extracellular portion of Fas and the Fc portion of human IgG 1 , was then replaced by a 2.6-kilobase BamHI-NotI fragment of M-Sema G that had been cloned into the EcoRI-XhoI sites of the pBluescript SK vector. The sequence of the M-Sema G-HA fusion plasmid encodes residues 1-860 of M-Sema G linked to a tandem repeat of HA by 26 amino acids (LEVDGIDKLDIEFLQPGGSTSSRAAA) from the restriction linker and pBluescript SK vector sequence.
The plasmid was transfected into COS-7 cells in a 100-mm tissue culture dish by DEAE-dextran methods (12). After 2 days of culture in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, cells were harvested, washed twice with PBS, pelleted by centrifugation, and frozen at Ϫ20°C before analysis by immunoblotting.
Surface Biotinylation and Immunoprecipitation-Cell surface biotinylation was performed as described previously (13). Cells were then solubilized with 0.5 ml of lysis buffer. After preclearing with protein A-Sepharose (Pharmacia), the lysate was immunoprecipitated with anti-M-Sema G serum prebound to protein A-Sepharose for 1 h at 4°C. After being washed with the lysis buffer, the eluates were subjected to SDS-PAGE and electroblotted onto polyvinylidene fluoride membranes. Membranes were blocked in PBS containing 5% bovine serum albumin, and biotinylated proteins were detected using streptavidin-biotinylated peroxidase complex (Amersham Corp.) and the ECL system.

Cloning and Structural
Features of M-Sema G cDNA-Using degenerative primers from conserved motifs of the semaphorin domain (2), cDNAs sharing homology with the collapsin and G-Sema I genes were amplified by PCR and used as probes to screen an adult mouse brain cDNA library. Of the cDNA clones isolated, the longest clone, A18, was analyzed. The nucleotide sequence and estimated amino acid sequence of A18 are shown in Fig. 1A. The longest open reading frame starting with an ATG codon matches the consensus sequence of a strong translation initiator (14) and ends with a TGA stop codon (Fig. 1A). There are in-frame stop codons preceding the first ATG. The cDNA predicts a protein of 861 amino acids. We named this protein mouse-semaphorin G (M-sema G). A hydrophobicity analysis of the predicted amino acid sequence displayed two stretches of hydrophobic sequence (Fig. 1B), one located at the N terminus that could function as a signal peptide and the other located at residues that could serve as a potential transmembrane-spanning domain. A schematic diagram of M-Sema G is shown in Fig. 1C. A single putative immunoglobulin-like domain of about 60 amino acids is conserved between all vertebrate semaphorins. The deduced amino acid sequence contains eight potential sites for N-glycosylation and 13 cysteine residues that are conserved among all vertebrate semaphorins. M-Sema G contains a highly acidic region at the C-terminal end that is predicted to be an intracellular domain.
The sema domain of M-sema G is 51% identical to that of M-sema F, 48% to that of semC, 44% to that of semB, and 40 -43% to those of semA, semD, and semE (5). The 295 amino acids within the sema domain are 77% identical to those of collapsin-4, which was isolated as a partial cDNA of about 0.9 kilobase from chicken; there is therefore no information on whether coll-4 contains a transmembrane domain (15). A phylogenetic tree was obtained for the complete amino acid sequences of mouse semaphorins (Fig. 1D). Two different groups can be distinguished; one contains semA, semD, and semE, and the other contains M-sema F, M-sema G, semC, and semB. M-sema G is closest to M-sema F and then to semC and semB.
Tissue Distribution of M-sema G mRNA-In situ hybridization histochemistry of frozen sections from mouse embryos (E14.5) distinctly detected M-sema G mRNA. The hybridization signals were found throughout the nervous tissues, with particularly strong expression in the cortical plate and dorsal root ganglia (Fig. 2). Surprisingly, among tissues other than nervous system, the thymus showed very strong expression for M-sema G mRNAs (Fig. 2). Moderate expression was also detected in the lung and kidney, whereas only background level expression was seen in the liver. The control sections hybridized with sense probes showed no distinct expression. So far there are no other semaphorin family members with strong expression in the immune tissues. For example, the expression of M-sema F mRNAs was very strong throughout the brain and spinal cord but weak in the thymus from E14.5 embryos.
The expression of M-sema G in various tissues from adult animals and cell lines was also studied by RNase protection assay. Radiolabeled antisense RNA, covering the 5Ј end of M-sema G, was used as a probe. As shown in Fig. 3, M-sema G was detectable in the adult brain, kidney, thymus, spleen, lung, heart, and bone marrow but was undetectable in liver. However, the expression level of M-sema G mRNA was much higher in the immune tissues, especially in the thymus, than in the other tissues. Although a high level of expression was observed by in situ hybridization analysis, the level detected in this assay was much weaker in the brain. This could be due to the maturation-associated decrease in the expression in the brain. In cell lines, the expression was detectable in T-lineage (EL4 and 2B4), B-lineage (LK, 70Z/3, and NS-1), myeloid (WEHI3 and J774), and mastocytoma (P815) cell lines but was undetectable in thymic and bone marrow-derived stroma (MRL104.8a and BMST) and fibroblast (BALB3T3) cell lines. Thus, M-sema G is a member of the semaphorin family, which is expressed predominantly in hematopoietic cells.
Protein Expression of M-Sema G-To check the tissue distribution of M-Sema G at the protein level, we generated a polyclonal Ab against M-Sema G by immunizing two rabbits with M-Sema G-GST fusion protein (see "Experimental Procedures"). To confirm whether immunized rabbit serum was directed against M-Sema G, an epitope-tagged form of M-Sema G was expressed in transfected cells. COS-7 cells transfected with the M-Sema G-HA fusion gene were lysed and analyzed by immunoblotting. As shown in Fig. 4A, HA-tagged M-Sema G immunoblotted with anti-HA monoclonal Ab migrated at about 125 kDa on SDS-PAGE under reducing conditions, although 100 kDa was expected for the core protein of mature M-Sema G-HA. Because there are eight potential sites for N-linked glycosylation in the deduced protein sequence of M-Sema G, the difference in the size could be explained by glycosylation at these sites. When HA-tagged M-Sema G on the same membrane (shown in Fig. 4A) was reprobed with anti-M-Sema G serum (Fig. 4B), the band migrating at the identical position appeared in the lysate from COS-7 cells transfected with M-Sema G-HA fusion gene, confirming the specificity of our polyclonal Ab.
Immunoblotting analysis using lysates from various organs revealed that M-Sema G was highly expressed in the thymus, in contrast to no expression in liver (Fig. 4C), which is consistent with the expression level of mRNA. Although low and faint Several other tissues, such as the olfactory epithelium (OE) and lung (Lu) showed significant expression of M-sema G mRNA, whereas the signals in the liver (Li) were seen only at background level. B, no distinct signals were detected in the sections hybridized with radiolabeled sense probes instead of antisense probes.

FIG. 3. RNase protection assay showing the expression of M-sema G in various tissues and cell lines.
For each RNA sample, an aliquot was electrophoresed on 1% agarose gel, and the intactness of the RNA was confirmed by visualizing 18S and 28S rRNA bands (data not shown). For blood-rich organs, RNA was extracted after perfusing the organ with PBS to minimize the contamination of blood cells. Five g of total RNA from the indicated organ or cell line was hybridized to the 32 P-labeled antisense RNA probe. The probe arises from the sequence coding the 5Ј end of M-sema G cDNA (nucleotides 1-260). The size of the undigested probe was 336 nucleotides, and the protection of M-Sema G transcripts yielded a 260-nucleotide fragment. Lane 1, 375 cpm of undigested probe; lanes 8 and 18, 500 cpm of undigested probe; lanes 7, 17, and 27, yeast tRNA that should not yield protection fragments. The cell lines used were EL4 and 2B4 (T lineage cell  expressions of M-Sema G were also observed in the brain and kidney, respectively, the size detected in the brain seemed to be smaller than that detected in the thymus. The reason for the different sizes of M-Sema G between thymus and brain is thus far unknown, but posttranslational modification of M-Sema G could vary with tissues. Cell Surface Expression of M-Sema G-Because the sequence of M-Sema G encoded a transmembrane domain, it could be expressed on the cell surface. To verify this possibility, thymocytes were surface biotinylated and cell surface proteins were immunoprecipitated with anti-M-Sema G Ab. As shown in Fig.  5, the protein that migrated at the same position as detected by immunoblotting analysis was immunoprecipitated with anti-M-Sema G Ab, demonstrating that M-Sema G is expressed on the surface of thymocytes. DISCUSSION The first member of the semaphorin family was identified by different methods in vertebrates and invertebrates. Sema I (known previously as FasIV) was cloned by using a monoclonal antibody recognizing special axonal tracts of the grasshopper embryos. Then several genes were cloned using a PCR method with degenerative primers for conservative amino acids, and a domain with about 500 amino acids was shown to be very conserved among invertebrates and vertebrates; it was named the sema domain (7). A member of the sema family was shown to be a human homologue of chick collapsin, which was independently identified as a protein collapsing a growth cone of dorsal root ganglion neurons (2). More members of the semaphorin family were cloned; semA, B, C, D, and E and M-sema F from mouse and coll-1, 2, 3, 4, and 5 from chick (5,15,8). Most of these members have been predicted to be secreted (5,15), and our hydrophobicity analysis suggests that M-Sema F and SemB and SemC of group IV are transmembrane proteins (8).
In the present study, we identified a novel member of the semaphorin family from mouse brain using a PCR method with degenerative primers; we named this semaphorin M-sema G. This cDNA deduces a protein that has a sema domain, an IgG-like domain, and a transmembrane domain. We also obtained evidence that this protein is a transmembrane type by a surface labeling analysis. This is the first evidence that transmembrane-type members of the semaphorin family exist in vertebrates and are different from invertebrate transmembrane-type members in that the latter have no IgG domain.
A BLAST search of a data base revealed that a fragment of chick coll-4 showed the highest homology, has 77% identical amino acids in the sema domain of M-sema G, and is thought to be a chick homologue of M-sema G. In addition, M-sema F, semC, and semB also showed high similarity to M-sema G, in that order, and are predicted to have a transmembrane domain from a hydrophobicity analysis. These results suggest that M-sema G and these members constitute a transmembranetype subfamily of semaphorins. However, the expression of M-sema G is unique among this subfamily, because the immune tissues, especially the thymus, displayed very strong expression for this member. Our results provide the first evidence that semaphorin is expressed on lymphocytes. It will be intriguing to elucidate the functional role of Msema G in lymphocytes. Our preliminary results demonstrated that the protein expression of M-sema G is much higher in thymocytes than in peripheral T and B cells. Moreover, in situ hybridization analysis using adult thymus revealed that Msema G is expressed predominantly in the thymic cortex, espe-  (open arrowhead). B, the membrane used in the experiment shown in A was stripped of bound Ab in stripping buffer, blocked, and reblotted with anti-M-Sema G serum. C, various murine organs were solubilized in lysis buffer. After the nuclear pellets were removed by centrifugation, lysates were mixed with 2 ϫ sample buffer, electrophoresed in 6% SDS-PAGE, and blotted with anti-M-Sema G serum. M-Sema G from thymus (arrowhead) migrated at a size similar to that from transfected COS-7 cells, whereas M-Sema G from the brain (arrow) is smaller than M-Sema G from the thymus.

FIG. 5. Cell surface expression of M-Sema G in thymocytes.
Thymocytes were surface biotinylated and lysed in lysis buffer. The precleared lysates were immunoprecipitated with normal rabbit serum or anti-M-Sema G serum prebound to protein A-Sepharose and subjected to SDS-PAGE analysis under reducing conditions. Proteins were blotted onto a polyvinylidene fluoride membrane and visualized as described under "Experimental Procedures." cially in the outer cortex. 2 The outer cortex contains thymic lymphoblasts, which give rise to more mature thymocyte populations, i.e. the small cortical thymocytes and the medullary thymocytes (16). Along with the progression of maturation, thymocytes move from the outer cortex to the medulla through the deep cortex. The mechanism responsible for the transport of thymocytes from the cortex toward the medulla is thus far unknown. Because semaphorins contribute to the guidance of axons on growth cones (17), the function of M-Sema G on thymocytes could be related to a maturation-dependent movement of immature thymocytes. Alternatively, because semaphorins have been shown to have repellent activity in axonal guidance of the nervous system, thymocytes may repel each other by expressing M-Sema G on the cell surface to interact efficiently with the thymic stromal cells that did not express M-Sema G (Fig. 3). The interaction between thymocytes and thymic stromal cells is considered to be a process essential for the proper differentiation of thymocytes. The precise distribution of M-Sema G on lymphocytes of various developmental stages must be determined to address these possibilities. To this end, we are currently generating a monoclonal Ab that can react with the extracellular portion of M-Sema G using M-Sema G-human IgG fusion proteins. Although the functions of M-Sema G in lymphocytes are currently unknown, our present results suggest that semaphorins are not a family of proteins whose functions are limited to the guidance of growth cones.
Semaphorins and their receptors may be used in a common pathway for communication within and between the nervous and immune systems. Identification of the natural ligand for semaphorins may contribute to a better understanding of the precise roles of these proteins in the formation of neural and immunological networks.