Molecular characterization of CRMP5, a novel member of the collapsin response mediator protein family.

The CRMP (collapsin response mediator protein) family is thought to play key roles in growth cone guidance during neural development. The four members (CRMP1-4) identified to date have been demonstrated to form hetero-multimeric structures through mutual associations. In this study, we cloned a novel member of this family, which we call CRMP5, by the yeast two-hybrid method. This protein shares relatively low amino acid identity with the other CRMP members (49-50%) and also with dihydropyrimidinase (51%), whereas CRMP1-4 exhibit higher identity with each other (68-75%), suggesting that CRMP5 might be categorized into a third subfamily. The mouse CRMP5 gene was located at chromosome 5 B1. Northern blot and in situ hybridization analyses indicated that CRMP5 is expressed throughout the nervous system similarly to the other members (especially CRMP1 and CRMP4) with the expression peak in the first postnatal week. Association experiments using the yeast two-hybrid method and co-immunoprecipitation showed that CRMP5 interacts with dihydropyrimidinase and all the CRMPs including itself, except for CRMP1, although the expression profile almost overlaps with that of CRMP1 during development. These results suggest that CRMP complexes in the developing nervous system are classifiable into two populations that contain either CRMP1 or CRMP5. This indicates that different complexes may have distinct functions in shaping the neural networks.

It is now established that growing axons respond to a complex balance of guidance cues, attractive or repulsive factors acting at short range or long range. Well known guidance cues are collapsins/semaphorins, netrins, and ephrins (1). The semaphorins are a large group of axonal guidance molecules with at least 30 members (2). They serve as repulsive guidance cues, influencing growth cone guidance not only in a contact-dependent way but also from a distance (3)(4)(5). Using a COS cell expression cloning approach, a transmembrane protein called neuropilin-1 was identified as a collapsin-1/semaphorin-3A (Sema3A) 1 receptor (6,7). Both neuropilin-1 and Sema3A knockout mice exhibited defasciculation and spreading of cranial nerves over a large area (8,9). Dorsal root ganglion growth cones from the neuropilin-1 knockout mice did not collapse in response to Sema3A, proving that the receptor neuropilin is necessary for inducing collapse (8).
In contrast to the rapid progress in identification and characterization of axon guidance molecules and their receptors, much remains to be explored about the intracellular mechanism by which signals are transduced into the eventual response of the growth cone. Since chick CRMP62 (now designated CRMP2) was first isolated as a factor required for collapsin-1 (Sema3A)-mediated signaling (10), a new family of cytoplasmic proteins that may be involved in such signal transduction for semaphorins has been identified through the work of several groups (11). They are independently referred to as CRMP (collapsin response mediator protein; Refs. 10 and 12-15), TOAD (turned on after division; Refs. 16 and 17), Ulip (Unc-33-like phosphoprotein; Refs. 18 -20), or DRP (DHPaserelated protein; Ref. 21). To date, four different CRMP (TOAD/ Ulip/DRP) genes have been described in rats, mice, and humans, with exclusive expression in the developing nervous system (12,19); CRMP2 mRNA is detectable in lung (12), and CRMP4 is detectable in heart and adult testis (18,22).
We have been searching for topographically expressed molecules in the embryonic chick retina, using a cDNA subtraction technique and a novel cDNA display system called Restriction Landmark cDNA Scanning (23,24) to identify molecules that are involved in regional specificity in the retina, including those responsible for the topographic retinotectal projection. Among a number of molecules thus isolated, we found that CRMP3 was included in those asymmetrically expressed along the nasotemporal (anteroposterior) axis, although this expression pattern was transient during the retinal development.
In the present study, we performed a yeast two-hybrid screen of a mouse brain cDNA library using chick CRMP3 as bait to identify CRMP3-interacting molecules and found a novel CRMP. This molecule, which we refer to as CRMP5, was nearly equally divergent from other CRMP members and DHPase but exclusively expressed in the nervous system. CRMP5 inter-acted with DHPase and other CRMPs except CRMP1, suggesting that CRMP5 is involved in axon outgrowth and guidance together with the other CRMP members.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening-The cDNA fragment encoding the fulllength coding region of chick CRMP3 (GenBank TM accession number AF249294) was amplified by PCR with specific primers containing restriction sites using the isolated cCRMP3 clone as a template. After confirming the sequence identity, this fragment was inserted into the EcoRI and BamHI sites of the bait vectors, pBTM116 (kindly provided by Drs. P. Bartel and S. Fields) containing the LexA-coding sequence and pGBT9 (CLONTECH, Palo Alto, CA) containing GAL4 DNA-binding domain, to generate pLexA-cCRMP3 and pGAL4BD-cCRMP3, respectively. pLexA-cCRMP3 was then transformed into Saccharomyces cerevisiae strain L40, which harbors reporter genes HIS3 and LacZ under the control of upstream LexA-binding sites. The library screening was performed as described (25,26). Approximately 1.0 ϫ 10 6 clones were screened using a mouse 17-day embryo MATCHMAKER cDNA library (CLONTECH). Positive clones were selected with 1 mM 3-aminotriazole on plates lacking leucine, tryptophan, and histidine.
To confirm the positive interaction using another two-hybrid system, each positive clone thus obtained was again transformed into yeast strain SFY526 (CLONTECH) harboring reporter gene LacZ together with pGAL4BD-cCRMP3 and assayed for ␤-galactocidase activity according to the manufacturer's protocol (MATCHMAKER Library User Manual; CLONTECH). The nucleotide sequences of positive clones thus isolated were determined on both strands.
cDNA Cloning of the 5Ј-Portion of CRMP5-To obtain a longer 5Јflanking region, 5Ј-rapid amplification of cDNA ends (5Ј-RACE) was performed using total RNA prepared from embryonic day 17 (E17) mouse brains. Briefly, total RNA was reverse-transcribed with Ther-moScript reverse transcriptase (Life Technologies, Inc.) in the presence of 0.6 M trehalose (Wako, Osaka, Japan) (27). After two rounds of PCR, amplified DNAs were cloned into the pGEM-T Easy vector (Promega, Madison, WI) and subjected to sequence analysis.
Chromosome Preparation and Fluorescence in Situ Hybridization-The direct R-banding fluorescence in situ hybridization method was used for chromosomal assignment of the mouse CRMP5 gene. Preparation of R-banded chromosomes and fluorescence in situ hybridization were performed as described (28,29). Mitogen-stimulated splenocyte culture was synchronized by thymidine blockage, and the incorporation of 5-bromodeoxyuridine during the late replication stage was made for differential replication staining after the release from excessive thymidine. R-band staining was performed by exposure of chromosome slides to UV light after staining with Hoechst 33258. The chromosome slides were hardened at 65°C for 2 h and then denatured at 70°C in 70% formamide in 2ϫ SSC and dehydrated in a 70, 85, and 100% ethanol series at 4°C.
Genomic fragments covering the mouse CRMP5 gene were screened from mouse genomic library using SalI (linker site)-NcoI cDNA fragment (nucleotide residues Ϫ42 to ϩ249) as probe. One of the genomic fragments (22 kb long) was labeled by nick translation with biotin 16-dUTP (Roche Molecular Biochemicals) following the manufacturer's protocol. The labeled DNA fragment was ethanol-precipitated with 10 times volume of mouse Cot-1 DNA (Life Technologies, Inc.) and then denatured at 75°C in 100% formamide. After hybridization, the slides were washed with 50% formamide in 2ϫ SSC at 37°C for 20 min, 2ϫ SSC, and 1ϫ SSC for 20 min each at room temperature. They were incubated with Cy2-labeled streptavidin (Amersham Pharmacia Biotech) at a 1:500 dilution in 1% bovine serum albumin, 4ϫ SSC for 1 h at 37°C. The slides were washed with 4ϫ SSC, 0.1% Nonidet P-40 in 4ϫ SSC, and 4ϫ SSC for 10 min each on the shaker and then stained with 0.75 g/ml propidium iodide. Nikon filter set B-2A (see Fig. 2C) and UV-2A (data not shown) were used for observation. Kodak Ektachrome ASA100 films were used for microphotography.
Northern Blot Analysis-Total RNAs were prepared from various tissues of mice at postnatal day 0 (P0), and from whole embryos at E11, E13, and E15, whole heads at E17, and whole brains of postnatal mice at P0, P3, P7, and P14 by using ULTRASPEC TM RNA (Biotecx Laboratories, Houston, TX). As for staging of mice, the day on which a vaginal plug was detected was considered E0, and the day of birth was considered P0. Poly(A) ϩ RNA was isolated from E17 whole brains using Dynabeads Oligo(dT) 25 (DYNAL, Oslo, Norway). RNA separated on a 1% agarose gel containing 7.4% formaldehyde was transferred to a Hybond-N nylon membrane (Amersham Pharmacia Biotech). The blot was prehybridized at 68°C in a solution containing 5ϫ SSC, 0.1% SDS, 2% blocking solution (Roche Molecular Biochemicals), and 50% formamide. Hybridization was carried out overnight at 68°C in the same solution containing the DIG-11-UTP-labeled antisense riboprobe transcribed from the ApaI-SpeI cDNA fragment (nucleotide residues 1951-2828: noncoding probe) of CRMP5 (clone 1-8-7) (for nucleotide numbers, see Fig. 1) and detected using anti-DIG-alkaline phosphatase-conjugated antibody according to the manufacturer's instruction (Roche Molecular Biochemicals). To verify the amount of RNA loaded, the membrane was rehybridized with DIG-labeled cRNA probe for mouse glycelaldehyde-3-phosphate dehydrogenase.
Binding Assay Using the Two-hybrid System-To examine mutual associations among the CRMP family members and DHPase, DNA fragments covering the coding regions of mouse CRMP1-5 and DHPase were amplified by PCR using specific primers from the respective clones or RT-PCR (for CRMP2) to make restriction sites for cloning. After confirming the sequence, these fragments were inserted into the EcoRI and SalI sites of the pGBT9 bait vector. The CRMP1 and CRMP5 fragments were inserted also into the EcoRI and XhoI sites of the pACT2 prey vector (CLONTECH). The bait and prey constructs in various combinations were transformed into yeast strain Y190 (CLON-TECH) harboring reporter genes HIS3 and LacZ, and transformants were cultivated on plates lacking leucine and tryptophan. Assay was performed as described (26). For quantitative analysis of the mutual association, liquid ␤-galactosidase assay was performed using o-nitrophenyl ␤-D-galactopyranoside as substrate according to the manufacturer's protocol (Yeast Protocols Handbook; CLONTECH).
Immunoprecipitation and Western Blotting-Myc-CRMP5, FLAG-CRMP1-5, and FLAG-DHPase expression vectors were constructed in the following way: Oligo nucleotides encoding c-Myc or FLAG peptide including the first methionine were ligated to mammalian expression vector pcDNA3.1 (Invitrogen, Carsbad, CA), which were digested with NheI and EcoRI, to yield pcDNA-Myc and pcDNA-FLAG, respectively. Full-length CRMP1-5 and DHPase cDNA were cut out from the various pGBT9 bait vectors with EcoRI and SalI and inserted into pcDNA-Myc or pcDNA-FLAG digested with EcoRI and XhoI. Myc-CRMP5 and FLAG-CRMPs/FLAG-DHPase thus prepared were co-transfected into COS-7 cells by using LipofectAMINE Plus Reagent (Life Technologies, Inc.). At 48 h after transfection, cell lysates were prepared by sonication in RIPA buffer (50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1 mM dithiothreitol, 1% Triton X-100, 10 g/ml leupeptin (Sigma), 10 g/ml pepstatin A (Sigma), and 1 mM phenylmethylsulfonyl fluoride), followed by centrifugation at 12,000 ϫ g for 20 min. The lysates were incubated with agarose beads conjugated with 9E10 anti-Myc monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) for 3 h with agitation. The beads were collected by low speed centrifugation and washed four times by resuspension in 1 ml of RIPA buffer. Proteins eluted from the beads with sample buffer were subjected to SDS-polyacrylamide gel electrophoresis and analyzed by immunoblotting with M2 anti-FLAG monoclonal antibody (Sigma) or 9E10 anti-Myc monoclonal antibody (Santa Cruz Biotechnology). To check the expression levels of FLAG-CRMPs and FLAG-DHPase, cell lysates were also analyzed using M2 anti-FLAG monoclonal antibody.

RESULTS
Yeast Two-hybrid Screening for CRMP3-interacting Molecules-To know the signal transduction cascade of CRMPs upstream and downstream, we attempted to identify CRMP3interacting molecules by the yeast two-hybrid screening of a mouse embryo cDNA library using chick CRMP3 as bait (pLexA-cCRMP3). Sixty-one clones were isolated by surveying approximately 1.0 ϫ 10 6 transformants. Sequence analysis showed that CRMP1, CRMP3, CRMP4, and the mouse homologue of DHPase were included among them: six clones for FIG. 1. Nucleotide and deduced amino acid sequences for mouse CRMP5. The predicted amino acid sequence is shown by the one-letter codes below the nucleotide sequence. The original cDNA clone, 1-8-7, isolated by twohybrid screening, starts at position Ϫ42. The upstream sequence (Ϫ443 to Ϫ9) is derived from clone 4-1 obtained by 5Ј-RACE. A putative polyadenylation signal is shown in bold type. By searching the mouse EST data base, 12 overlapping sequences were found to match the 3Ј-untranslated region of CRMP5: GenBank TM accession numbers AA268030, AI509437, W40685, W54057, AA008915, W33434, W41424, AA032969, W78280, W77101, AA048239, and AA051700. An additional polyadenylation signal was found in the resultant contig, located about 400 bp downstream of the 3Ј end of the sequence shown here (data not shown). Consensus sequences for various phosphorylation sites are underlined: protein kinase A (double solid line), protein kinase C (solid line), casein kinase II (dashed line), and protein tyrosine kinases (bold line). Potential phosphorylation sites conserved between CRMP5 and CRMP1-4 are Ser-7, Thr-77, Thr-211, Ser-219, Ser-356, and Ser-549; conserved sites between CRMP5 and some of CRMP1-4 are Ser-108, Ser-252, Thr-458, Ser-466, Tyr-472, and Ser-538; conserved sites between CRMP5 and DHPase are Ser-108, Thr-211, Thr-229, and Ser-356; CRMP5unique sites are Thr-54, Thr-162, and Thr-330. The GenBank TM accession number for mouse CRMP5 is AF249295. CRMP1, one clone for CRMP3, 29 clones for CRMP4, and six clones for DHPase. This result supports the claim by Wang and Strittmatter (30) that CRMP isoforms associate with one another. The sequence of mouse DHPase, determined for the first time in the present study, shared 94 and 88% amino acid identity with rat and human DHPase, respectively (Gen-Bank TM accession number AF249296).
Besides the already identified CRMP isoforms described above, we found one clone, 1-8-7, encoding a novel protein sequence that shows significant but relatively low homology with CRMP1-4 and DHPase. Interaction of this novel CRMP isoform with chick CRMP3 was verified by the two-hybrid method using another bait construct pGAL4BD-cCRMP3 (data not shown). We here designate this new CRMP member as CRMP5. Moreover, two clones with no homology to these CRMP isoforms, which are supposed to be novel CRMP-interacting molecules, were also identified among them (to be reported elsewhere).
Molecular Cloning and Sequence Analysis of the Full-length CRMP5-The CRMP5 cDNA clone, 1-8-7, contained a putative open reading frame that was in frame with the GAL4 activation domain sequence and therefore expected to be expressed as a fusion protein. Because clone 1-8-7 included no in-frame stop codon located upstream of the first methionine (nucleotides 1-3 in Fig. 1), we performed 5Ј-RACE to identify further the upstream sequence. By reverse transcriptase reaction with trehalose (27), we obtained longer RACE-PCR products and found two in-frame stop codons in clone 4-1. Thus, we concluded that the initially isolated CRMP5 cDNA clone, 1-8-7, has the fulllength open reading frame. In Fig. 1, the contig nucleotide sequence composed of 1-8-7 and 4-1, together with the deduced amino acid sequence, is presented.
Our search of the mouse EST data base identified 12 overlapping mouse cDNAs matching to the 3Ј-noncoding region of 1-8-7. The resultant contig sequence had an additional poly(A) signal that is located about 400 bp downstream of the nucleotide sequence shown in Fig. 1 (data not shown; for GenBank TM accession numbers, see the legend). Therefore, two types of CRMP5 mRNA transcripts that are different only in the length of the 3Ј-noncoding region are probably generated.
We screened EST data bases for homologues of CRMP5 in other species. This analysis identified six human EST sequences (GenBank TM accession numbers N51749, AA350414, AA058664, AA351100, AA488145, and AI36969). They showed significant, although still partial, homology with mouse CRMP5; the N-terminal (corresponding to amino acid residues 1-184 of mCRMP5) and C-terminal (residues 482-564) amino acid sequences were predicted from the EST clones. Compared with mouse CRMP5, the deduced amino acid identity was 88% for the N-terminal part and 98% for the C-terminal part.
Amino Acid Sequence Comparison between CRMP5 and the Other CRMP Isoforms-The CRMP5 cDNA encoded a protein of 564 amino acids whose calculated molecular mass is 61,516 Da. It shares approximately 50% amino acid identity both with the members of the CRMP family (CRMP1-4; 49 -50%) and with DHPase (51%) through the entire length ( Fig. 2A). Because the already known CRMP family members exhibit approximately 70% identity with each other (68 -75%), CRMP5 might be categorized to another subfamily (12,19). As shown in Fig. 2B, a phylogenetic tree of CRMP isoforms indicated that CRMP5 is relatively close to DHPase and suggested that CRMP1-4 diverged first from a common ancestor with DH-Pase, followed after a lag by CRMP5.
Amino acid sequences of CRMP isoforms were aligned (Fig.  3). The N-terminal three-quarters is more conserved than the C-terminal region. When CRMP5 is compared with the other CRMPs, it is evident that some amino acid substitutions are identical or similar to those of DHPase, together with the gap at the N-terminal region. On the other hand, the sequence at aligned positions 54 -57 is absent, and the C-terminal region after 530 is present, similar to other CRMP members. It is notable that some regions (at aligned positions 279 -298, 313-339, 508 -536, and 572-576) were different from all the other CRMP isoforms including DHPase.
CRMP isoforms are known as phosphoproteins (18,19) and bear several consensus sequences for phosphorylation sites that are conserved among the members. CRMP5 contains 17 such potential phosphorylation sites (Fig. 1): a single potential site for tyrosine phosphorylation, two potential sites for protein kinase A, six sites for protein kinase C, and eight sites for FIG. 2. Similarities between mouse CRMP isoforms and chromosomal localization of mouse CRMP5 gene. A, amino acid identity between the indicated pairs of sequences. The percentage was calculated using Clustal W Software. B, an unrooted phylogenetic tree of CRMP isoforms. The relationship among CRMP isoforms was represented in a tree by Drawtree method (Phylip software package). C, chromosomal localization of the mouse CRMP5 gene. CRMP5 gene localization to mouse R-banded chromosomes was analyzed using a biotinylated genomic DNA fragment. The signals are localized to chromosome 5 B1 (arrows). The metaphase spreads were photographed with Nikon B-2A filter.
casein kinase II. Among them, three protein kinase C target sites are unique to CRMP5 (see legend to Fig. 1). This suggests that the function of CRMP5 could also be regulated by phosphorylation, specifically and/or commonly.
Recently, it has been proposed that the amidohydrolases including DHPase share the same active site architecture (31). Notably, four histidine residues (at aligned positions 77, 79, 202, and 258) and one aspartatic acid residue (at position 336) involved in metal binding are highly conserved (Fig. 3). Because CRMP5 has substitutions in two histidine residues, it is unlikely that it bears DHPase activity just as the other CRMP members do not (10,18,30).
Expression of CRMP5 mRNA-The CRMP family members are known to be expressed in the nervous system, whereas DHPase is present in liver and kidney (21). Because CRMP5 is almost equally divergent from the other CRMPs and DHPase, we next addressed the issue of whether CRMP5 expression is neural-tissue specific. Northern blot analysis using P0 mouse tissues (Fig. 4A), clearly showed that CRMP5 mRNA is present only in the brain but not in the other tissues. Next, the expression profile of CRMP5 mRNA during development was examined. CRMP5 mRNA was already expressed at E11, as early as CRMP4/Ulip1 (data not shown), peaking in the first postnatal week when neural network formation and its refinement culminates and subsequently declining to a lower level during the second week (Fig. 4B). By longer exposure, another minor hybridization band with an estimated mRNA size of 10.7 kb was detected. The hybridization band of about 4.2 kb is likely to be an artifact caused by overlap with 28 S ribosomal RNA. These results suggest that CRMP5 expression is neural-tissue specific and developmentally regulated as already reported for some CRMP members (10, 12, 14 -20). When poly(A) ϩ RNA was analyzed, the 5.0-kb band for CRMP5 turned out to be double bands of 4.8 and 5.2 kb (Fig. 4C). When the expression level of CRMP5 mRNA was compared with that of CRMP4 (18) by Northern blot hybridization, it was lower by approximately one order of magnitude, and this difference was retained during development up to adulthood (data not shown).
To elucidate the expression pattern of CRMP5, we conducted in situ hybridization on sections of E17 embryos with probes for CRMP5 and CRMP4. This analysis demonstrated that CRMP5 (Fig. 5, A-E) is expressed selectively in the nervous system with almost the same expression pattern as CRMP4 (Fig. 5, F-H). In our experiments, CRMP5 mRNA was not detected outside of the nervous system (Fig. 5A). In the brain at E17, CRMP5 and CRMP4 were restricted to postmitotic neural cells (Fig. 5, A-C  and F). Both CRMP5 and CRMP4 were prominent in the neocortex and also detected in the olfactory epithelium. In the retina, these CRMPs were strongly expressed in the postmitotic layer, which includes retinal ganglion cells, but undetected in the outer layer (Fig. 5, D and G). As shown in Fig. 5 (E and H), CRMP5 and CRMP4 displayed an almost identical expression pattern in the mantle layer of the spinal cord and dorsal root ganglion and sympathetic ganglia (data not shown). Overall, we conclude that CRMP5 is indeed a neural-specific CRMP.
Association Analysis between CRMP5 and the Other CRMP Isoforms-DHPase is well known to form tetramers by selfassociation (34), and the CRMPs are also shown to interact with one another and form tetrameric structures (30). Although CRMPs have no detectable DHPase activity, they are able to interact with the DHPase monomer weakly (30). Because CRMP5 was originally identified by its ability to interact with chick CRMP3 in our screening and has sequence homology with the other CRMP isoforms, it was expected that this protein may also interact with itself and with some other CRMP isoforms. Therefore, we tested for interaction with the other mouse isoforms using the two-hybrid system.
To further verify the interaction observed in the two-hybrid assay, co-immunoprecipitation experiments were conducted. Myc-tagged CRMP5 and FLAG-tagged CRMPs or DHPase were co-transfected into COS-7 cells, and the cell lysates were immunoprecipitated using agarose beads conjugated with 9E10 anti-Myc monoclonal antibody. FLAG-CRMP2, 3, 4, 5, and DHPase but not FLAG-CRMP1 were co-immunoprecipitated with Myc-CRMP5 (Fig. 7). Consistent with the two-hy- brid assay (Fig. 6C), associations between CRMP5 and CRMP2-4 were much tighter than that between CRMP5 itself, but that between CRMP5 and DHPase was significantly high for some unknown reason. When Myc-CRMP5 was not cotransfected, FLAG-tagged CRMPs were not immunoprecipitated with anti-Myc agarose. In these experiments, amounts of Myc-CRMP5 bound to anti-Myc agarose were nearly equal among the cell lysates (middle panel), and FLAG-CRMPs and -DHPase expression levels in the cell lysates were also comparable (lower panel). These data suggest that, contrary to the view that all CRMP members interact, association preferences between the family members do exist (30). DISCUSSION CRMPs are a family of cytoplasmic proteins predominantly expressed in the developing nervous system that have attracted attention because they might be key molecular components in shaping neural networks, although conclusive evidence about their functions remains to be obtained (11,35). To date, four CRMP members have been identified from several animal species and variously designated as TOAD/Ulip/DRP. Although individual members show somewhat different expression patterns in the nervous system (12), CRMPs are, as a whole, expressed in almost all the developing neurons peaking in the first postnatal week with a marked decrease in adulthood, except for CRMP2 and partly CRMP3.
While searching for topographically expressed molecules in the embryonic chick retina using the Restriction Landmark cDNA Scanning technique (24), we found that CRMP3 was expressed asymmetrically along the nasotemporal axis in the retina, although this expression pattern was transient during development. Analysis of the expression patterns of chick CRMP1-4 revealed that cCRMP1, cCRMP2, and cCRMP4 were expressed in a relatively similar pattern, whereas cCRMP3 was differentially expressed in the developing chick nervous system, 2 intimating its unique function in the process of neural network formation.
We have therefore attempted to screen CRMP3-interacting molecules using the yeast two-hybrid system to reveal the signal transduction cascade of CRMP3 upstream and downstream. In this study, we identified a novel CRMP, CRMP5. In previous studies using the yeast two-hybrid system with CRMP1-3 as bait, in vitro binding assay, and also purification of native CRMPs, CRMPs were suggested to form tetrameric structures as in the case of DHPase (30,34). Therefore, it was expected that CRMP isoforms would be included in our positive clones.
Mouse CRMP5 cDNA encoded a protein with 564 amino acids nearly identical in size to the other CRMP members. Amino acid sequence identity between CRMP5 and the four other CRMPs was significantly low (ϳ50%) compared with that between CRMP1-4 (68 -75%) and nearly equally divergent from that of DHPase (51%). A phylogenetic tree also showed that CRMP5 is significantly distant from CRMP1-4 and close to DHPase, suggesting that CRMP5 might be classified to another CRMP subfamily. It was shown that CRMP1-3 can associate with DHPase, but the interaction is very weak compared with that between DHPases (30). In this context, our finding that CRMP5 shows significant interaction with DH-Pase (see below) is interesting, although why it interacts with DHPase is not clear at present. Here, it should be noted that interaction between the two would not occur in vivo because of the difference in expressed tissues.
Recent studies 1 on the genomic structure of human CRMP1 and CRMP2 revealed that these genes consist of 14 exons and that the exon-intron organizations are completely conserved (36,37). On the other hand, the human DHPase gene is composed of 10 exons, and only five introns are shared by CRMP1, CRMP2, and DHPase (36 -38). We have cloned a fragment of the mouse CRMP5 gene. We found no intron insertions at the positions that correspond to intron 1 and intron 2 seen in the CRMP1 and CRMP2 genes but found an intron at the position that corresponds to intron 1 in the human DHPase gene together with an unique intron in the 5Ј-noncoding region. 3 This finding again supports the view that CRMP5 might be classi-fied to a subfamily distinct from the four other CRMP members. However, our low stringency screening of P1 mouse brain cDNA library and search for mouse and human EST data bases for CRMP5-related molecules yielded no additional cognates, suggesting the absence of related subfamily members.
Northern blot analysis detected three bands. The main 4.8and 5.2-kb bands, probably derived from the difference in poly(A) addition sites, represent CRMP5 mRNA. The minor 10.7-kb band presumably corresponds to a heterogeneous RNA that contains intronic sequences. However, the possibility that another related protein is coded by this mRNA species is not excluded. Northern blot analysis of CRMP5 mRNA in various mouse tissues indicated that CRMP5 mRNA were expressed only in brain but not in non-neural tissues like CRMP2 (in lung) or CRMP4 (in heart) (12,18). The CRMP5 expression profile during development resembled the CRMP1 and CRMP4 profiles, because the expression level peaked in the first postnatal week and markedly decreased in adulthood (12,19). In contrast, CRMP2 continues to be expressed in large amounts in adulthood, as does CRMP3 but only in the cerebellum (12,19).
In situ hybridization analyses showed that CRMP5 is expressed exclusively in the nervous system, with almost the same expression pattern as CRMP4 in the brain. It was also similar to CRMP2 expression analyzed by immunohistochemistry on mouse brain sections at E16.5 (14). In the spinal cord, the expression pattern of CRMP5 resembled those of CRMP1, CRMP2, and CRMP4 ( Fig. 5 and Ref. 12). These CRMP mem-3 I. Watakabe and M. Noda, unpublished observation.

FIG. 6.
Interaction between mouse CRMP1-5 and DHPase in the yeast two-hybrid system. Various CRMP-expressing bait constructs were tested for interaction with the prey, CRMP5 (A) and CRMP1 (B) by induction of reporter gene, HIS3. A and B, individual yeast transformants were streaked on plates lacking leucine, tryptophan, and histidine with 50 mM 3-aminotriazole to suppress leaky growth. Plates were incubated at 30°C for 3 days. Sections 1, CRMP1; sections 2, CRMP2; sections 3, CRMP3; sections 4, CRMP4; sections 5, CRMP5; sections 6, DHPase; sections 7, empty pGBT9 vector as a negative control. C, ␤-galactosidase activity was measured in triplicate using o-nitrophenyl ␤-D-galactopyranoside as substrate. Colonies were selected on the synthetic plate lacking leucine and tryptophan and cultured in the selective liquid medium. Data are the means from three independent colonies with standard errors (bars). The standard errors were less than 10% for Ͼ1 values. bers were expressed throughout the mantle layer in the developing spinal cord, whereas CRMP3 was localized to the dorsal matrix layer (12). Taken together with the developmental profiles obtained by Northern hybridization, the results suggested that expressional regulation of CRMP1, CRMP4, and CRMP5 is common and that of CRMP2 and CRMP3 is rather unique. It is thus conceivable that CRMP5 functions in cooperation with other CRMPs in the nervous system during development.
CRMP5 could be assigned to another family as mentioned above. However, we concluded that CRMP5 is a CRMP for the following reasons: it is expressed exclusively in the nervous system; it associates with other CRMPs (see below); and it probably lacks DHPase activity as do CRMP1-4 because a zinc ion binding site that is requisite for enzyme activity is not conserved. Here, it should be noted that CRMP isoforms more closely resemble each other in the N-terminal three-quarters, through which mutual association occurs (30). However, when human and mouse CRMP5 sequences were compared, the Cterminal portion was found to be more conserved than the N-terminal region, suggesting a specific role for the C-terminal region of CRMP5.
Our yeast two-hybrid analysis and immunoprecipitation analysis using COS-7 cells revealed that the CRMPs form multimeric structures, in which CRMP5 interacts with CRMP2, CRMP3, and CRMP4, but, interestingly, not with CRMP1. CRMP5 also interacts with itself, but this interaction was very weak compared with that between CRMP5 and CRMP2-4. There appears to be some preferences in the association that favor heterophilic oligomerization over homophilic oligomerization between the family members, except for DH-Pase (see also Ref. 30). This suggests that multiple combinations of CRMP hetero-multimeric complexes might exist even in a single neuron, which may serve distinct functions in the formation of neural networks. Our results suggest that CRMP oligomers are classifiable into two populations: those that contain CRMP1 and those that contain CRMP5. Unraveling the roles common to all the CRMPs regardless of composition of complex and/or characteristic of each member or particular combination of CRMPs through identification of the CRMPcomplex-binding partner(s) should lead to a better understanding of the basic principle for establishing complicated but extremely precise patterns of neuronal connectivity.
Finally, it should be pointed out that almost all the developing output neurons (projection neurons) including semaphorininsensitive cells apparently express several combinations of CRMPs. For example, retinal ganglion cells, which do not respond to semaphorins (43), express all the CRMPs. This suggests that their roles are not restricted to the semaphorininitiated signal transduction cascade in growth cones as was first proposed (10) but extended to more general processes underlying neural network formation in the entire nervous system. Of note, Sema3A stimulated anterograde and retrograde axonal transport of organellas in mouse dorsal root ganglion neurons (44). Furthermore, CRMP2 is associated with the neurofibrillary tangles observed in Alzheimer's patients (13), and Sema3A and CRMP2 are synchronously induced in apoptosis-destined neurons (45). CRMP3 is recognized by anti-CV2 autoantibodies which are present in patients with paraneoplastic neurological diseases (39). These results suggest that CRMPs are also involved in neuritic degeneration under not only developmental but also morbid conditions. Further study to elucidate the signaling cascade located upstream and downstream of CRMPs will be required.