AMIGO, a transmembrane protein implicated in axon tract development, defines a novel protein family with leucine-rich repeats

Development of the nervous system with billions of connections is one of the most complex and fascinating phenomena in nature. One key feature in this event is the guidance of the neuronal growth cones to their appropriate targets. A wide variety of soluble, matrix, and cell surface molecules have been found to be involved in axonal growth and in association of axons to form mature fiber tracts (for review see Tessier-Lavigne and Goodman, 1996; Drescher et al., 1997).

Currently, Ig superfamily proteins form the most diverse and studied class of molecules, and have been shown to participate in contact-dependent regulation of neurite outgrowth, axon guidance, and synaptic plasticity (for review see Schachner, 1997; Walsh and Doherty, 1997; Stoeckli and Landmesser, 1998; Van Vactor, 1998). Extracellular proteins containing leucine-rich repeats (LRRs)^* have also been shown to participate in axon guidance. For instance, Slit proteins containing LRR domains act as midline repellents for commissural axons through the Robo (Roundabout) receptor (Battye et al., 1999; Brose et al., 1999), and recently, Battye et al. (2001) showed that the interaction of Slits with their Robo receptors was due to LRRs found in Slits. Furthermore, Pusch et al. (2000) showed that the disease called X-linked congenital stationary night blindness (XLCSNB) maps to a gene that codes only the LRR-containing protein nyctalopin in retina. Recently, the receptor for axonal regeneration inhibitor Nogo (Chen et al., 2000) was found to be a GPI-linked cell surface protein in which the only recognizable motifs are the LRR domains (Fournier et al., 2001).

Amphoterin (also known as HMGB1) is a heparin-binding protein that was isolated from perinatal rat brain as a neurite outgrowth–promoting factor (Rauvala and Pihlaskari, 1987) enriched in the growth cones of neuronal cells. Amphoterin has been proposed to be an autocrine factor in invasive cell or growth cone migration due to binding to the cell surface receptors (receptor for advanced glycation end products [RAGE] and sulphated glycan epitopes) and to activation of proteolysis of ECM through binding of plasminogen and it's activators to amphoterin (for review see Rauvala et al., 2000; Muller et al., 2001).

To examine the role of amphoterin in cell motility, especially in neurite outgrowth, we searched for genes that are induced on amphoterin matrix by using mRNA differential display. A transcript that was prominently up-regulated on amphoterin-coated matrix was found to encode a novel protein. In this paper, we describe the cloning and functional characterization of this novel protein, named amphoterin-induced gene and ORF (AMIGO). Cloning of AMIGO gave us sequence data to clone two other related proteins (AMIGO2 and AMIGO3); together, these three proteins form a novel family of transmembrane proteins. The predicted amino acid sequences of the AMIGOs suggest that they are type I transmembrane proteins containing a signal sequence for secretion and a transmembrane domain. Interestingly, the extracellular part of the AMIGOs contains six LRRs flanked by cysteine-rich LRR NH₂- and COOH-terminal domains and by one Ig domain close to the transmembrane region. This twin motif structure defines the AMIGOs as members of both the Ig and the LRR superfamilies.

In this paper, we have characterized in more detail AMIGO, the member of the protein family that is highly expressed in the nervous system. We suggest that AMIGO mediates cell–cell interactions via a homophilic mechanism in the fiber tracts of the nervous system. In addition, we show that all members of the AMIGO family display homophilic binding, and that they also bind each other in a heterophilic fashion, suggesting a cooperative role during development and in adult animals.

Ordered differential display (ODD; Matz et al., 1997) was used to search for amphoterin-induced genes in neurons. Comparison of ODD from embryonic d 18 rat hippocampal neurons grown on amphoterin- and laminin-coated plates revealed a transcript that was expressed more on amphoterin (Fig. 1 A). This expression difference was also confirmed with RT-PCR (Fig. 1 B).

To see whether the induction is mediated through the amphoterin receptor RAGE (for review see Rauvala et al., 2000; Schmidt et al., 2000), we also ligated RAGE by using affinity-purified anti-RAGE antibodies. The anti-RAGE matrix promoted prominent neurite outgrowth in hippocampal neurons, whereas a matrix coated with control rabbit IgG was inactive. The anti-RAGE–induced neurite outgrowth was specifically blocked by the peptide used for immunization (5 μg/ml), whereas the control peptide from a different part of RAGE did not have any effect on anti-RAGE–induced neurite outgrowth (unpublished data).

Expression analysis of the transcript was performed using hippocampal neurons on amphoterin-, anti-RAGE–, and laminin-coated plates. We compared the expression to chromogranin B (ChrB), for which the expression has been shown to be induced through RAGE signaling (Huttunen et al., 2002). Interestingly, the expression patterns were quite similar for these two genes; both were induced on amphoterin and anti-RAGE matrix, whereas on laminin there was no induction (Fig. 1, C and D). Expression of the transcript identified by ODD is thus induced by RAGE (but apparently not by integrin) signaling.

The sequence of the partial transcript did not give homology to any previously cloned genes. By using the 5′ RACE method (Matz et al., 1999), the cDNA encoding the whole coding sequence was cloned (Fig. 2 A). We named this differentially expressed gene AMIGO. Hydrophobicity profile analysis (Nielsen et al., 1997; SignalP V2.0.b2 software) revealed that the protein sequence of AMIGO contains a putative signal sequence and a putative transmembrane region. The deduced extracellular part of the protein contains six LRRs and one Ig domain. The deduced cytosolic part of the protein does not contain any known domains.

The human and mouse homologues of AMIGO were also cloned with the 5′ RACE method by using data from the rat AMIGO sequence and from EST sequences. The sequences of the rat, mouse, and human AMIGO are available from GenBank/EMBL/DDBJ under accession nos. AY237729, AY237008, and AY237007, respectively. Identity at the amino acid level between the rat and mouse AMIGO is 95%, and the murine sequences are 89% identical to the human AMIGO. In the extracellular part, the most conserved motifs between the murine and human AMIGO are the NH₂-terminal cysteine-rich domain and the LRRs 1–3. Interestingly, the whole transmembrane domain and the cytoplasmic tail are 100% identical between the murine and human AMIGOs.

By using homology search, we detected ESTs that gave homology but were not identical when compared to AMIGO. By using these EST sequences, we cloned two other novel genes that we named for convenience as AMIGO2 and AMIGO3. The sequences of the mouse and human AMIGO2 are available from GenBank/EMBL/DDBJ under accession nos. AY237006 and AY237005, respectively. The sequences of the mouse and human AMIGO3 are available from GenBank/EMBL/DDBJ under accession nos. AY237004 and AY237003, respectively. The deduced amino acid sequences show that AMIGO2 and AMIGO3 have the same domain organization as AMIGO; they also contain a putative signal seguence for secretion and six LRRs flanked on both the NH₂- and COOH-terminal sides by cysteine-rich LRRNT and LRRCT domains. Like AMIGO, the deduced extracellular parts of AMIGO2 and AMIGO3 contain an Ig domain close to the transmembrane domain (for schematic picture of AMIGO, AMIGO2, and AMIGO3, see Fig. 2 B).

Similarity at the amino acid level between AMIGO to AMIGO2 is 48%, AMIGO to AMIGO3 is 50%, and AMIGO2 to AMIGO3 is 48%. The alignment for AMIGO, AMIGO2, and AMIGO3 shows that the most conserved regions between the three proteins are the LRRs, the transmembrane region, and some parts of the cytosolic tail (Fig. 2 A). The LRRs found in the AMIGOs can be described as a motif LX₂LXLX₂NX(L/I)X₂aX₄(F/L/I), in which “a” denotes an aliphatic residue and “X” denotes any amino acid; this motif resembles a typical LRR sequence often found in extracellular parts of animal proteins (Kajava, 1998).

We made BLAST searches by using the extracellular part of the AMIGO to see whether there were some homologous neuronal proteins. The best fits were found from Slit family of extracellular axon-guiding proteins (Whitford et al., 2002), and from the Nogo-66 receptor (Fournier et al., 2001; Fig. 2 C).

RT-PCR analysis of adult mouse tissues (Fig. 3 A) revealed that AMIGO is mainly expressed in the nervous tissues (cerebellum, cerebrum, and retina), although weak expression could be also detected in liver, kidney, small intestine, spleen, lung, and heart. AMIGO2 expression is most prominent in cerebellum, retina, liver, and lung. A lower AMIGO2 mRNA expression is also seen in cerebrum, kidney, small intestine, spleen, and testis. AMIGO3 mRNA expression could be detected in every tissue studied, showing no specific expression pattern compared with AMIGO or AMIGO2. Thus, it appears that AMIGO is essentially a nervous system–specific member of the protein family, and we focused on AMIGO in more detail in the present work.

The AMIGO mRNA expression was studied in more detail using in situ hybridization. The AMIGO antisense probe gave a strong signal in the developing and adult nervous tissues, whereas the sense probe did not give any clear signal (unpublished data). A clear AMIGO expression was already detected in the E13 rodent embryo; at this stage the highest expression level was found in the dorsal root ganglia and the trigeminal ganglion, with some expression in the central nervous system (Fig. 3 B, a and b). During later stages of development and in the adult, AMIGO was also prominently expressed in the brain, where the most intense signal was detected in the hippocampus (Fig. 3 B, c).

To investigate the expression of AMIGO at the protein level, polyclonal antisera were produced against an extracellular 10-amino acid peptide sequence that is found in AMIGO, but not in AMIGO2 or AMIGO3. The anti-peptide antibodies recognized the 75-kD AMIGO–Ig fusion protein produced in Drosophila S2 cells (Fig. 4 A, lanes 1 and 3). Western blotting of crude brain extracts revealed specific binding to a 65-kD polypeptide (Fig. 4 A, lanes 2 and 4). The molecular mass of the recognized polypeptide is close to the calculated molecular mass (56 kD) of AMIGO. Binding of the antibodies to both the fusion protein and the 65-kD polypeptide of brain were blocked by the synthetic peptide used as the immunogen (Fig. 4 A, lanes 3–6). Furthermore, binding of the antibodies to tissue sections was blocked by the synthetic peptide (Fig. 4 B).

Western blotting of AMIGO using crude brain extracts from different developmental stages was consistent with the in situ hybridization data. The expression appears to start in the brain somewhat later than in the peripheral nervous system, and increases clearly between E13 to E14 (Fig. 4 C). The expression is maintained high during the perinatal developmental stage, but is down-regulated during the postnatal stages P6 to P10. After this, the expression is again up-regulated and remains high in the adult brain (Fig. 4 C). Because the time period of the postnatal up-regulation of the AMIGO expression appears to coincide with the onset of myelination, we compared the expression of AMIGO to that of the myelin-specific marker α-CNPase. Indeed, the expressions of AMIGO and the CNPase display a parallel increase during postnatal development (Fig. 4 C). Thus, the AMIGO expression displays a dual character during brain development; the first expression peak occurs during the late embryonic and perinatal development, and the second increase in expression accompanies myelination.

Immunohistochemistry using the anti-peptide antibodies revealed specific staining only in the nervous system. Intensity of the immunostaining was in agreement with the expression data inferred from Western blotting (Fig. 4 C), and specificity of the immunostaining was also suggested by inhibition of antibody binding to AMIGO and to tissue sections by the peptide used as the immunogen (Fig. 4, A and B). In general, AMIGO was intensely stained in developing and mature fiber tracts (Fig. 5, A–F). During embryonic development, when the spinal ganglia abundantly express AMIGO mRNA (see Fig. 3 B, a and b), the immunostaining was observed in the fiber tracts connecting to the ganglia and the spinal cord, but not in the ganglia themselves (Fig. 5 E), suggesting that the AMIGO protein is transported to axonal processes. In cerebellum, the most intense staining was observed in fibers on both sides of the Purkinje cell layer; the characteristic structure formed by the basket cell axons around the Purkinje cell soma was clearly discerned by the AMIGO immunostaining (Fig. 5 B). Consistent with the Western blotting data, AMIGO immunostaining labeled most myelinated axon tracts in the adult (Fig. 5 A). An example is shown in Fig. 5 (C and D), demonstrating the similarity of the AMIGO and α-CNPase immunostaining around the hippocampus. However, the AMIGO expression is not restricted to myelinated tracts; for example, in hippocampus, nonmyelinated tracts in the stratum lucidum CA3 region, which were negative for α-CNPase (Fig. 5 D) and myelin basic protein (unpublished data), stained clearly for AMIGO (Fig. 5 C). In general, AMIGO staining was detected (both during development and in adult animal) in large-diameter neurites (axons) that were also stained by antibodies against the 145-kD neurofilament (unpublished data). As in the forebrain, myelinated axon tracts were also stained for AMIGO in cerebellum, pons, medulla, and spinal cord.

AMIGO was also clearly immunostained both in the cell soma and in fasciculated and nonfasciculated processes of cultured hippocampal neurons (Fig. 5 F). As expected from immunostaining of tissue sections, double-immunostaining (unpublished data) revealed colocalization with the 145-kD neurofilament and the β-tubulin (TuJ1), but not with MAP2. Thus, AMIGO is preferentially expressed in axonal rather than dendritic processes.

Identification of AMIGO from hippocampal neurons growing neurites on amphoterin, the occurrence in fiber tracts in vivo, and the domain structure with LRRs and Ig domains suggest that AMIGO might have a role in neurite extension. To get insight into the function of AMIGO, we tested if it is able to promote neurite outgrowth of hippocampal neurons. The extracellular part of the AMIGO was fused to human IgG Fc part, and this fusion protein was immobilized on microtiter wells and used as a substrate for hippocampal neurons. These experiments showed that the AMIGO–Ig fusion protein promotes attachment and neurite outgrowth of hippocampal neurons (Fig. 6, A and C), whereas on the human IgG Fc control, neurite outgrowth was very low or undetectable (Fig. 6, B and C). Neurite outgrowth induced by the immobilized AMIGO–Ig fusion protein was inhibited by the soluble AMIGO–Ig fusion in the culture medium (Fig. 6 D).

Because AMIGO immunostaining could be found in vitro in hippocampal fasciculating axons and in the axon tracts in vivo, AMIGO might participate in fasciculation of neurites. We addressed this question by using the ectodomain of AMIGO as Ig fusion protein in the culture medium. Hippocampal neurons were plated on poly-l-lysine–coated wells to promote neurite outgrowth and fasciculation. Microscopy of the cultures revealed that the growth pattern of neurites was dramatically changed in the presence of the soluble AMIGO. In the control cultures, neurites formed fascicles in 4 d, whereas in the presence of the soluble AMIGO, the processes were mainly nonfasciculated up to at least 5 d in culture (Fig. 7, A–C).

Fasciculation of axons is known to involve homophilic interactions, and this might be reason why soluble AMIGO perturbs fasciculation. Therefore, we tested in a coimmunoprecipitation assay whether AMIGO, AMIGO2, and AMIGO3 could bind themselves and whether the AMIGOs display heterophilic binding within the protein family. To examine homo- and heterophilic association, 293T cells were cotransfected with GFP-tagged AMIGOs, V5-tagged full-length AMIGOs, and the control protein V5-tagged RAGE (Fig. 8 A). Immunoprecipitation of the GFP-tagged AMIGOs from the cell lysates precipitated all V5-tagged AMIGOs, and correspondingly, the GFP-tagged AMIGOs were all precipitated with anti-V5 antibodies (Fig. 8 A). The control protein V5-tagged RAGE was not coprecipitated with the AMIGO-, AMIGO2-, or AMIGO3-GFP and vice versa. (Fig. 8 A).

As another approach to study homophilic binding of the AMIGOs, we added AMIGO–, AMIGO2–, and AMIGO3–Ig fusion protein to protein A–coated beads to get the protein oriented in a manner that occurs at the cell surface. AMIGO and AMIGO2 caused rapid aggregation of the beads (Fig. 8 B and Fig. 8 C, b and c), but for AMIGO3 the aggregation was milder (Fig. 8 B and Fig. 8 C, d). Addition of the IgG Fc part as the control protein did not induce any clear aggregation (Fig. 8 B and Fig. 8 C, a).

In this paper, we have identified a novel family of transmembrane proteins called AMIGO, AMIGO2, and AMIGO3. These three proteins show clear homology with each other; their length and domain organization are highly identical (Fig. 2 B). This domain relationship suggests a common evolutionary origin of the AMIGOs.

Based on genomic sequence data, these three proteins probably occur in the puffer fish Fugu rubripes (unpublished data). Interestingly, Drosophila has a protein family called kekkon with three members of transmembrane proteins, kek1, kek2 (Musacchio and Perrimon, 1996), and kek3 (Ashburner et al., 1999), which show homology in their extracellular parts to the AMIGOs. The extracellular parts of both the AMIGOs and the kek proteins contain six LRR domains flanked with cysteine-rich LRRNT and LRRCT domains and one Ig domain close to the transmembrane region. However, the cytoplasmic parts of the AMIGOs and kek proteins do not display homology with each other. The gene expression data of kek1 and kek2 (Musacchio and Perrimon, 1996) is reminiscent of the one seen for AMIGO and AMIGO2; they all are expressed in the central nervous system of the adult organism. These domain and expression similarities suggest that the AMIGOs and kek proteins may be derived from a common ancestral gene.

In their extracellular parts, the most homologous motifs between the AMIGOs are the LRRs 3–5. The best fit in BLAST searches shows homology with Slit family of extracellular axon-guiding proteins (Whitford et al., 2002), and a clear homology is also found with the Nogo-66 receptor, where the only recognizable motifs are the LRR domains (Fournier et al., 2001; Fig. 2 C). The similarity found in the LRRs in AMIGO, Slit1, and Nogo-66 receptor suggests an evolutionary origin of these proteins from a common ancestor. The clear conservation seen at the LRR area between the AMIGOs suggests that this region is important for interactions with extracellular ligand(s), and that they could also share the same binding partner(s).

In the literature, there are reports of other transmembrane proteins that contain LRRs and Ig domains in the extracellular part of the proteins. ISLR (Nagasawa et al., 1997), 5 LRRs and 1 Ig domain; Pal (Gomi et al., 2000), 5 LRRs and 1 Ig domain; LIG-1 (Suzuki et al., 1996), 15 LRRs and three Ig domains; and GAC1 (Almeida et al., 1998), 12 LRRs and 1 Ig domain. Common for all of these proteins and the AMIGOs is the order of how the LRRs and the Ig domains are organized; the LRRs are always more distal to the transmembrane region than the Ig domains. Interestingly, BLAST searches by using Ig domain sequences from AMIGOs give no distinct homology with other Ig domains of the Ig superfamily proteins, but the closest are the ones found in proteins containing both Ig and LRR domains (unpublished data).

Although the cytoplasmic moieties of the AMIGOs do not display any significant homology with previously identified transmembrane proteins, the alignment of the AMIGOs (Fig. 2 A) shows two conserved serine-rich regions; one close to the transmembrane domain and the other at the COOH terminus. The COOH-terminal serine-rich area of AMIGO and AMIGO2 has a consensus sequence for Casein kinase II serine/threonine kinase (Allende and Allende, 1995), which is ubiquitously expressed in brain, but AMIGO3, which is not expressed in the brain, does not have this consensus sequence. Recently, Watts et al. (1999) showed that the transmembrane form of TNF-α has a consensus sequence SXXS that is a substrate for Casein kinase I–dependent phosphorylation. Interestingly, all three AMIGOs have four possible Casein kinase I phosphorylation sites in these two conserved serine-rich areas. Future work will reveal whether these conserved serine residues have important functions in signaling events of the AMIGOs.

There are an increasing number of reports in the literature and the data banks on mammalian transmembrane proteins having both LRR and Ig domains, but unfortunately at present, almost all data only comprise the cloning and tissue expression of these proteins. Our data here give a functional insight into these twin motif transmembrane proteins, belonging to both the LRR and Ig superfamily, in a form of more detailed characterization of AMIGO.

Based on RT-PCR experiments, in situ hybridization, and immunohistochemistry, AMIGO is essentially a nervous system–specific protein. Interestingly, AMIGO expression is up-regulated at two clearly distinct stages during brain development; the first peak is found perinatally, and the second up-regulation occurs during or slightly before the up-regulation of the oligodendrocyte-specific marker α-CNPase.

The first expression peak of AMIGO is compatible with a role in growth of axonal connections. The expression of AMIGO in developing axon tracts both in vivo and in vitro and our neurite outgrowth experiments support this role. One cellular mechanism in the growth of axonal connections is fasciculation. Axons grow along each other by using pioneer axons as the substratum for the growth cones of the later ones. Interestingly, addition of AMIGO ectodomain in the culture medium inhibits the formation of neurite bundles, suggesting a role for AMIGO in fasciculation. Furthermore, AMIGO displays a homophilic binding mechanism that would explain its role in fasciculation. Homophilic adhesion molecules belonging to both the Ig superfamily and to the cadherin family have been shown to mediate neurite outgrowth and fasciculation during the nervous system development (for review see Kamiguchi and Lemmon, 1997; Martinek and Gaul, 1997). It is also noteworthy that the LRR sequences of the AMIGOs display homology to the Slit proteins and to the Nogo receptor (Fig. 2 C) that have been implicated in axon growth, regeneration, and guidance. Further studies are thus warranted to identify ligands of AMIGO that do not belong to the AMIGO family and to study possible association of AMIGO to other transmembrane proteins.

The second up-regulation of the AMIGO expression suggests a role in myelination. It seems reasonable that AMIGO would mediate cell–cell interactions also at this stage of development. However, further studies are clearly warranted to understand the role of AMIGO in myelinating axon tracts, like in the interactions of axons with oligodendrocytes and Schwann cells. Interestingly, our present results suggest that the members of the AMIGO family are able to bind each other in a heterophilic fashion in addition to homophilic binding detected in all family members (Fig. 8, A–C). It is tempting to speculate that AMIGO and AMIGO2, which both are expressed in the central nervous system (Fig. 3 A), have a cooperative role by binding each other from opposing cell membranes, for example, in contacts of axons with each other or with glial cells.

Furthermore, AMIGO expression remains high until adulthood. This suggests that AMIGO plays a role in regeneration and plasticity of the adult fiber tracts, the mechanisms of which commonly recapitulate mechanisms of fiber tract development. From this viewpoint, it is of interest that AMIGO expression is induced by ligation of the Ig superfamily protein RAGE by amphoterin and by anti-RAGE, whereas integrin ligation by laminin does not induce AMIGO expression (Fig. 1, B and C). In general, RAGE signaling plays a role in tissue injury and in regulation of cell motility, for example, in growth cone and cell migration (for review see Rauvala et al., 2000; Schmidt et al., 2000).

To get further insight into the functional roles of AMIGO during development and adulthood, we have recently targeted the gene in ES cells and are currently producing AMIGO-null mice (unpublished data). In addition to the in vivo approaches using gene targeting, it will be important to understand what molecular domains mediate homophilic and heterophilic binding of the AMIGO family, and whether the intracellular domains of the AMIGOs have signaling properties. Furthermore, future studies will reveal whether the members of the AMIGO family mediate analogous cell–cell interactions in nonneuronal and neuronal tissues characterized in the present paper for AMIGO in axonal tracts.

ODD was performed as described by Matz et al. (1997), comparing genes induced on amphoterin versus laminin matrix. Hippocampi were dissected from 18-d-old rat embryos and triturated with a Pasteur pipette in HBSS (without calcium and magnesium; GIBCO BRL) containing 1 mM sodium pyruvate and 10 mM Hepes, pH 7.4. After washing in HBSS, neurons were suspended in neurobasal medium (GIBCO BRL), 2% B27 supplement (GIBCO BRL), 25 μM l-glutamic acid (Sigma-Aldrich), and 1% l-glutamine (GIBCO BRL), and were then seeded at the density of 10⁶ cells on 35-mm plastic plates (Greiner Bio-One) coated with 10 μg/ml laminin (Sigma-Aldrich) or 10 μg/ml recombinant amphoterin. RNA was isolated by using the RNeasy mini kit (QIAGEN) 24 h after seeding, and was used for ODD.

The rat AMIGO cDNA 5′ end was amplified using the method of Matz et al. (1999), based on template-switching effect and step-out PCR, and the full-length cDNA was cloned from postnatal d 14 rat cerebrum using a reverse transcription reaction with a 5′ primer (5′-ACTGCTTCTCGCCTGGCCCGT-3′) and a 3′ primer (5′-GAACCTCCCCATCAGCCTATACTG-3′). The rat AMIGO sequence was used to find out human and mouse ESTs to get sequences for cloning of the human and mouse AMIGOs. The human AMIGO cDNA was cloned from the THP-1 cell line (TIB-202; American Type Culture Collection) using a reverse transcription reaction with a 5′ primer (5′-CAGAACATGCCCGGGTGAC-3′) and a 3′ primer (5′-GGACCAATTCCCTTGAGGTCAG-3′). The mouse AMIGO cDNA was cloned from adult mouse cerebrum using a reverse transcription reaction with a 5′ primer (5′-ACTGCTTCTCGCCTGGCCCGT-3′) and a 3′ primer (5′-AACCTCCCCATCAGCCTACGCTG-3′). The AMIGO sequences were used for homology search with BLAST to find other possible related sequences. The human AMIGO2 cDNA was cloned from the HT1080 cell line (CCL-121; American Type Culture Collection) as above: the 5′ primer was 5′-CTCAGAGGCGACCATAATGTC-3′, and the 3′ primer was 5′-TGTTTATTTTGCAGACCACACAC-3′. The mouse AMIGO2 cDNA was cloned from adult mouse cerebrum with a 5′ primer (5′-CTCAGAGGCGACCATAATGTC-3′) and a 3′ primer (5′-GCGATGCTGAAGGCTAAGATG-3′). The human AMIGO3 cDNA was cloned from the HEK293 cell line (CRL-1573; American Type Culture Collection) with a 5′ primer (5′-CAACCTGCACAGAGCTGCTCTGTAC-3′) and a 3′ primer (5′-GCACAGTGCTTCCCACCAGTATCTG-3′). The mouse AMIGO3 cDNA was cloned from adult mouse cerebellum with a 5′ primer (5′-AGAAGTAGGTGAGTCTTGGAGCT-3′) and a 3′ primer (5′-TGTTGTGCAGGTAGAGCCTG-3′).

Total RNA was reverse transcribed in a reaction containing 1 μg RNA, 0.25 mM dNTP-mix, 1 μg random nonamers, 20 U recombinant RNasin^® (Promega), 200 U MMLV-RT (Promega) with 1× MMLV reaction buffer supplied. 2 μl of the reverse transcription mixture was then used for polymerase chain reaction with gene-specific primers. For the mouse and rat AMIGO, the primers were a 5′ primer (5′-AGCAACATCCTCAGCTGCTC-3′) and a 3′ primer (5′-CTTCAGCTTGTTGGAGGACAG-3′). For mouse AMIGO2, primers were a 5′ primer (5′-GGCACTTTAGCTCCGTGATG-3′) and a 3′ primer (5′-GTCTCGTTTAACAGCCGCTG-3′). For the mouse AMIGO3, primers were a 5′ primer (5′-AGGTGTCAGAGTCCCGAGTG-3′) and a 3′ primer (5′-GTAGAGCAACACCAGCACCA-3′). For rat ChrB, the primers were a 5′ primer (5′-CTTCCCAGAGTACAACTATGAC-3′) and a 3′ primer (5′-GCTGAACTTTTCCGCTATCTTC-3′). For GAPDH control, primers were a 5′ primer (5′-CAACGACCCCTTCATTGACC-3′) and a 3′ primer (5′-AGTGATGGCATGGACTGTGG-3′).

The subsequent PCR reaction was performed in a PCR mix (2.5 μM dNTP, 10 mM Tris-HCL, pH 8.8, 150 mM KCL, 1.5 mM MgCl₂, and 0.1% Triton X-100) containing 0.2 μM 5′ primer and 3′ primer and 1 U DyNAzyme™ II DNA Polymerase (Finnzymes). The amplification products were separated on 1.5% agarose gel and stained with EtBr.

For in situ hybridization with radiolabeled probes, a 1.2-kb fragment from the mouse AMIGO cDNA was PCR amplified with a 5′ primer (5′-CCGCTCGAGCCGGCCGATCTGTGGTTAG-3′) and a 3′ primer (5′-CGGAATTCTCACACCACAATGGGTCTATCAGA-3′). The reaction product was then ligated into pGEM-T vector. In situ hybridization analysis was performed using single-stranded RNA probes on mouse fetal and adult paraffin-embedded tissue sections as described previously (Reponen et al., 1994).

A 1,180-bp BamHI fragment containing the entire extracellular coding region of the mouse AMIGO was amplified by PCR with a 5′ primer (5′-CGGGATCCTAGGGTGACTCTCTCCCAGATCC-3′) and a 3′ primer (5′-CGGGATCCGTTGAGGGTGTCATGGTGTCC-3′). The reaction product was then ligated into pRMHA3–3c-FC-cDNA. The AMIGO2 and AMIGO3 extracellular coding region without signal sequence was cloned into the modified pRMHA3 vector, which contains CD33 signal sequence and FC-cDNA from signal pIgplus vector. A 1,065-bp NheI/NotI fragment of the mouse AMIGO2 was amplified with a 5′ primer (5′-TACTAGCTAGCTGCCCCACTGCTTGCATCTG-3′) and a 3′ primer (5′-ATAGTTTAGCGGCCGCCTCGTGGGCGTGGGA-3′). A 1,080-bp NheI/NotI fragment of the mouse AMIGO2 was amplified with a 5′ primer (5′-CTATACTAGCTAGCGATGTTCTGCCTCCTGCTCC-3′) and a 3′ primer (5′-ATAGTTTAGCGGCCGCTGTGTTGAAAGTCTCTGGCTCG-3′).

The AMIGO, AMIGO2, and AMIGO3–Ig fusion protein plasmids were cotransfected with the hygromycin resistance plasmid p-COP-hyg into Drosophila S2 cells using the FuGENE™6 transfection reagent (Roche). After a 3-wk selection with 300 μg/ml hygromycin B (Calbiochem), stabile AMIGO–Ig fusion S2 cell pools were cultured in shake flasks, where the protein expression was induced with 500 μM CuSO₄. After culturing for 6 d, the AMIGO–Ig fusion proteins were isolated from the supernatant by using protein A–agarose (Upstate Biotechnology) according to the manufacturer's instructions.

Rabbit anti-AMIGO peptide antibodies were raised against the synthetic peptide YAMGETFNET (corresponding to amino acids 341–350 of the mouse AMIGO and 342–351 of the rat and human AMIGO). Binding of the antibodies to AMIGO was verified using the recombinant AMIGO–Ig fusion protein and crude brain extracts in Western blotting (see below). Because the antibodies bound more intensely and specifically to the rat AMIGO compared with AMIGO from other species (possibly due to species differences in the glycosylation site close to the peptide sequence used in immunization), rat samples were primarily used in immunochemical detections.

Rabbit anti-RAGE peptide antibodies were raised against the synthetic peptide NTGRTEAWKVLSPQG (corresponding to amino acids 54–68 of the mouse, rat, bovine, and human RAGE). The RAGE antibodies were affinity-purified as described previously (Rauvala et al., 1988) by using the same peptide used for immunization. The synthetic control peptide, which was used to validate the specificity of the anti-RAGE antibodies, have the sequence RPLNTAPIQPRVRE (corresponding to amino acids 217–231 of mouse and 216–230 of rat RAGE).

Brains of embryonic, postnatal, and adult rats were extracted to the final concentration of 83.3 mg tissue/ml SDS extraction buffer (62.5 mM Tris, 1.8% SDS, 7.75% glycerol, and 4.4% 2-mercaptoethanol, pH 6.8). After addition of the SDS buffer, the extracts were pressed several times through a needle. The extracts were boiled 2 × 5 min and centrifuged at 10,000 g for 10 min to remove nonsoluble material. Samples corresponding to the same wet weight of tissue were analyzed by Western blotting. Ponceau staining of the membrane confirmed uniform protein amounts.

Precast 4–15% gels (Bio-Rad Laboratories) were used for SDS-PAGE in Western blotting. Proteins were transferred to Hybond™ nitrocellulose membrane (Amersham Biosciences) by semi-dry blotting technique. Rabbit anti-AMIGO peptide antibody (1/1,000 dilution) and monoclonal anti-CNPase, clone 11–5B (1/1500; Sigma-Aldrich) were used as primary antibodies. HRP-conjugated goat anti–rabbit IgG (Bio-Rad Laboratories) and sheep anti–mouse IgG (Amersham Biosciences) were used as secondary antibodies. The antibody complexes were detected using ECL™ reagents (Amersham Biosciences).

Immunohistochemistry of AMIGO was performed using paraffin sections. In brief, adult rats were killed after CO₂ treatment by cervical dislocation, tissues were fixed using ice-cold PBS with 4% PFA, and the samples were then transferred in paraffin. Hydrated paraffin sections (4–10-μm thick) were incubated with 1% hydrogen peroxide/methanol solution for 20 min, and were washed again with PBS. The sections were blocked for 1 h with 5% skimmed milk powder in PBS. The sections were then incubated with the rabbit AMIGO peptide antiserum, which was diluted 1/200 in the blocking buffer at 4°C overnight. After washing with PBS, the sections were incubated with HRP-conjugated goat anti–rabbit antibodies (Bio-Rad Laboratories) at a dilution of 1:500 for 2 h at RT, washed with PBS, and incubated with aminoethyl carbazole (Sigma-Aldrich) as a chromogenic substrate. Immunofluorescence staining for in vitro cultured hippocampal neurons was performed by using FITC-conjugated goat anti–rabbit secondary antibodies (Jackson ImmunoResearch Laboratories).

Hippocampi were dissected from 18-d-old rat embryos into a calcium/magnesium-free trituration medium (HBSS with 1 mM sodium pyruvate and 10 mM Hepes, pH 7.4). Cells were dissociated by pipetting 25 times with a glass Pasteur pipette and were washed once with the calcium/magnesium-containing buffer (HBSS plus calcium and magnesium with 1 mM sodium pyruvate and 10 mM Hepes, pH 7.4). The cells were seeded at the density of 70,000 cells/cm² on 96-well polystyrene dishes coated by the test protein in Neurobasal medium with 2% B27 supplement (GIBCO BRL), 1% BSA, 0.5 mM l-glutamine, 25 μM l-glutamic acid, and 1× penicillin-streptomycin. The dishes were coated with the test protein (3.125–100 μg/ml) in PBS overnight at 4°C, washed three times with PBS, and blocked with 1% BSA in PBS for 1 h at RT before adding the cells. The cells were cultured for 24 h before counting the neurite outgrowth. For counting of neurite outgrowth, images were taken from living cells using randomly selected microscopic fields, and the extensions, which were twice the length of the cell soma, were considered as neurites. For quantification of neurite outgrowth, 15 images (275 × 225 μm) with a total of 750 cells were evaluated from every concentration of the test protein (AMIGO–Ig fusion or Fc control substrate) used for coating. The data were pooled from three independent experiments.

To test the effect of soluble AMIGO–Ig fusion protein, the dishes were coated with the AMIGO–Ig fusion protein (12.5 μg/ml in PBS) at 4°C overnight, washed three times with PBS, and blocked with 1% BSA in PBS for 1 h at RT. The cells were seeded at the density of 70,000 cells/cm² and cultured for 24 h before counting the neurite outgrowth. Counting was performed as above from three independent experiments. A total of 750 cells were evaluated for every concentration of the test protein (AMIGO–Ig fusion or the Fc control protein) used in solution.

Fasciculation of neurites was studied with hippocampal neurons prepared as above. The 96-well plates were coated with poly-l-lysine at 4°C overnight, washed three times with PBS, and blocked with 1% BSA in PBS for 1 h at RT. The cells were seeded at the density of 70,000 cells/cm² in the serum-free medium (see Neurite outgrowth assay) with either the AMIGO–Ig fusion protein or the Fc control protein in solution. The AMIGO–Ig fusion and the Fc control protein were tested at 3.25–25 μg/ml. The experiment was repeated independently three times, and pictures were taken from living cells after 4 d in culture. For quantification of neurite outgrowth, 12 randomly taken images (45 × 35 μm) were taken for every concentration of the AMIGO–Ig fusion and the Fc control protein used in solution. To evaluate inhibition of fasciculation, the total length of the processes, the diameter of which is < 2 μm (formed only from 1–3 neurites), was measured from the 12 images taken for every protein concentration tested.

Pictures for the neurite outgrowth and fasciculation experiments were taken with a digital camera (model DP10; Olympus). The measurements were performed using the Image-Pro^® image analysis software (Media Cybernetics, Inc.).

Coimmunoprecipitation experiments were performed using transiently transfected HEK293T cells. The constructs were transfected into the cells using FuGENE™6 (Roche) according to the manufacturer's instructions. The full-length AMIGO, AMIGO2, and AMIGO3 were cloned in frame with the pEGFP-N1 (CLONTECH Laboratories, Inc.) and pcDNA6-V5-His (Invitrogen) vectors. Full-length RAGE was cloned in frame with the pcDNA6-V5-His vector. After transfection, the cells were grown for 48 h before lysing in the RIPA buffer with 10 mg/ml PMSF and 60 μg/ml aprotinin (Sigma-Aldrich). Coimmunoprecipitation experiments were performed using rabbit anti-GFP antibody (sc-8334; Santa Cruz Biotechnology, Inc.) and mouse anti-V5 antibody (46–0705; Invitrogen) at the concentration of 1 μg/ml.

The aggregation assay was performed using 1-μm protein A Fluoresbrite™ carboxylated beads (Polysciences, Inc.). The 100-μg beads were first washed three times with PBS, 2% BSA, and 0.1% Tween 20 solution, and were then mixed and sonicated in water bath in 50 μl of the buffer mentioned above. The beads were divided to two aliquots, and the test and the control protein (10 μg each) were added into the beads in 25 μl of PBS, 2% BSA, and 0.1% Tween 20 solution (final volume 50 μl). The aggregation samples were incubated at RT. After addition of the protein, 1.5-μl samples were taken into 100 μl of PBS, 2% BSA, and 0.1% Tween 20 solution in 96-well plates at different time points. The aggregation was evaluated and filmed under the fluorescence microscope. Kinetics of bead aggregation was calculated from two independent experiments from eight fields containing 9,000 beads. The extent of bead aggregation is represented by the index N_t/N₀, where N_t and N₀ are the total number of particles at the incubation times t and 0 (Agarwala et al., 2001).

We thank Ms. Eeva-Liisa Saarikalle for excellent technical assistance. We thank Professor Irma Thesleff for helping to carry out the in situ hybridization analysis.

This study has been supported by the Finnish Academy and Technical Research Center of Finland (Programme of Molecular Neurobiolgy; to H. Rauvala), the Sigrid Juselius Foundation (to H. Rauvala), Finnish Cancer Organizations (to H. Rauvala), and the Research and Science Foundation of Farmos (to J. Kuja-Panula).