Address correspondence and reprint requests to Dr. Pedro F. Esteban or Dr. Fernando de Castro, Grupo de Neurobiología del Desarrollo-GNDe, Hospital Nacional de Parapléjicos, Finca “La Peraleda”, s/n, E-45071-Toledo, Spain. E-mails: email@example.com; firstname.lastname@example.org
The protein anosmin-1, coded by the KAL1 gene responsible for the X-linked form of Kallmann syndrome (KS), exerts its biological effects mainly through the interaction with and signal modulation of fibroblast growth factor receptor 1 (FGFR1). We have previously shown the interaction of the third fibronectin-like type 3 (FnIII) domain and the N-terminal region of anosmin-1 with FGFR1. Here, we demonstrate that missense mutations reported in patients with KS, C172R and N267K did not alter or substantially reduce, respectively, the binding to FGFR1. These substitutions annulled the chemoattraction of the full-length protein over subventricular zone (SVZ) neuronal precursors (NPs), but they did not annul it in the N-terminal-truncated protein (A1Nt). We also show that although not essential for binding to FGFR1, the cysteine-rich (CR) region is necessary for anosmin-1 function and that FnIII.3 cannot substitute for FnIII.1 function. Truncated proteins recapitulating nonsense mutations found in KS patients did not show the chemotropic effect on SVZ NPs, suggesting that the presence behind FnIII.1 of any part of anosmin-1 produces an unstable protein incapable of action. We also identify the extracellular signal-regulated kinase 1/2 (ERK1/2) pathway as necessary for the chemotropic effect exerted by FGF2 and anosmin-1 on rat SVZ NPs.
extracellular signal-regulated kinases mitogen-activated protein kinases
foetal bovine serum
fibroblast growth factor receptor 1
Fibronectin-like type 3
surface plasmon resonance
Urokinase-type plasminogen activator
whey acidic protein-like
The protein anosmin-1 is coded by the KAL1 gene (Franco et al. 1991; Legouis et al. 1991) responsible for the X-linked form of Kallmann syndrome (KS) (Maestre de San Juan 1856; Kallmann et al. 1944). Anosmin-1 is an extracellular matrix (ECM) glycoprotein composed of several domains that include a cysteine-rich region (CR), a whey acidic protein-like (WAP) domain (similar to domains present in proteins with serine protease inhibitory activity), four consecutive FnIII domains and a histidine-rich C-terminal region with an abundance of basic residues (del Castillo et al. 1992; Lutz et al. 1993; Rugarli et al. 1996; Soussi-Yanicostas et al. 1996). Orthologues of KAL1 have been identified in invertebrate and vertebrate species, from the nematode worm C. elegans to rodents and primates. Despite the high degree of sequence identity among species shown by anosmin-1, no orthologue has been identified in mouse or rat. The KAL1 gene is located on the X chromosome (Xp22.3) adjacent to the pseudoautosomal region 1 (PAR1), a highly variable and unstable region (Perry et al. 2001; Church et al. 2009). From this section of the X chromosome, a 9-Mb block has been removed or translocated in a common ancestor of mouse and rat (Ross et al. 2005). Several of the genes located in the human PAR1 and the proximal Xp22.3 region have not been identified in the mouse genome or are located on autosomal chromosomes (Gläser et al. 1999).
Anosmin-1 participates in the adhesion, migration and differentiation of various cell types in the CNS; among others, anosmin-1 promotes the adhesion of neurons, neurite outgrowth, axonal guidance and branching of different CNS projection neurons (Soussi-Yanicostas et al. 1998), as well as having a role in the migration of different types of neuronal precursors (NPs), immortalized GnRH-producing neurons and oligodendrocyte precursors (Cariboni et al. 2004; Bribián et al. 2006, 2008; Hudson et al. 2006; Yanicostas et al. 2008; Hu et al. 2009; García-González et al. 2010; Murcia-Belmonte et al. 2010; Clemente et al. 2011). Recently, anosmin-1 has been shown to be implicated in the formation of the cranial neural crest in chicken (Endo et al. 2012).
The mechanisms of action by which anosmin-1 exerts its functions are not completely understood, and it has been proposed that this protein would interact with different components of the ECM and membrane receptors. Anosmin-1 would regulate in this way the activity of such proteins. In this sense, the five potential heparan sulphates (HS)-binding motifs described within anosmin-1 (Robertson et al. 2001) would interact with HS for the correct localization and binding of anosmin-1 to the ECM (Soussi-Yanicostas et al. 1996), which seems to be necessary for its biological activity (Bülow et al. 2002; Bülow and Hobert 2004; Cariboni et al. 2004; González-Martínez et al. 2004; Hudson et al. 2006). By means of co-immunoprecipitation (CoIP) and surface plasmon resonance (SPR), a direct interaction has been shown between urokinase-type plasminogen activator (uPA) and anosmin-1, which would regulate the proteolytic activity of uPA (Hu et al. 2004). Homophilic and heterophilic interactions between anosmin-1 and other ECM proteins, like fibronectin and laminin, have also been reported (Bribián et al. 2008; Murcia-Belmonte et al. 2010). As prokineticin 2 and its receptor Prokr2 are involved in the pathogenesis of KS, it has been hypothesized that their activity could be modulated by anosmin-1 (Dodé et al. 2006). It has also been shown that this protein interacts with growth factors and morphogens (FGF8, BMP5 and WNT3a) and modulates their expression and activity in the formation of the cranial neural crest (Endo et al. 2012).
Among all these mechanisms of action, the interaction with and the regulation of FGFR1 activity is the most studied. Interaction between these two proteins has been described by CoIP (González-Martínez et al. 2004; Bribián et al. 2006; Ayari and Soussi-Yanicostas 2007), and by SPR it has been determined that the N-terminal region of anosmin-1 (comprising the CR, WAP and FnIII.1 domains) interacts with some of the extracellular domains (D2–D3) of FGFR1 (Hu et al. 2009). By means of glutathione-S-transferase (GST) pull-downs, the interaction of the N-terminal region has been confirmed to the WAP and FnIII.1 domains and also that the FnIII.3 domain by itself interacts with FGFR1 (Murcia-Belmonte et al. 2010). Not surprisingly, missense mutations found in KS patients in FnIII.1 and FnIII.3 reduce or annul the binding to the receptor and turn the protein inactive, probably because of non-efficient binding to the receptor (Hu et al. 2009; Murcia-Belmonte et al. 2010). In this work, we show that the mutation C172R within the WAP domain found in KS patients (Oliveira et al. 2001) did not alter the binding capacity to FGFR1, contrary to the N267K substitution in FnIII.1 (Hardelin et al. 1993) that seemed to impede, almost completely, the binding. Both mutations annulled the chemoattractive effect on SVZ neuroblasts of anosmin-1 in the full-length protein, but not in the N-terminal-truncated proteins (comprising the CR-WAP-FnIII.1 domains, hereafter called A1Nt). The replacement of FnIII.1 with FnIII.3 in the A1Nt version of anosmin-1 (A1NtFnIII.3), annulled the chemoattraction on SVZ NPs, even when in the GST pull-down (GSTpd) the recombinant protein was still able to bind to FGFR1. This could indicate a specific binding site for FnIII.3 in FGFR1 and a likely distinctive function for the CR-WAP region. We also demonstrated the fundamental role of the CR domain in the function of the protein, albeit it is not necessary for the binding to FGFR1. Nonsense mutations found in KS patients that would give rise to truncated proteins larger that the functional A1Nt lacked the chemotropic ability exhibited by this truncated protein, a likely indication of a more unstable conformation. Finally, we identified the extracellular signal-regulated kinases mitogen-activated protein kinases (ERK1/2 MAPK) pathway as a key component of the downstream signalling of FGFR1, activation of which is necessary for FGF2 and anosmin-1 to exert their chemotropic effect on rat SVZ NPs.
Materials and methods
Chinese hamster ovary (CHO) cells were grown in Dulbecco's Modified Eagle's Medium (DMEM, GIBCO, Invitrogen, Life Technologies Corporation, Carlsbad, CA, USA) supplemented with 8% foetal bovine serum (FBS, GIBCO), 100 U/mL of penicillin and 100 μg/mL of streptomycin (GIBCO) at 37°C and 5% CO2. Cells were transfected using Fugene 6 or X-tremeGENE DNA Transfection Reagents (Roche Applied Science, Indianapolis, IN, USA) according to the protocol provided by the manufacturer.
GST pull-down assays
GST pull-down assays with the full-length FGFR1-HA protein and different GST fusion proteins carrying the WAP-FnIII.1 domains from the N-terminal region of anosmin-1 (K130–K285) and with the same region with the substitutions C172R and N267K were performed. Single point mutations were introduced using the QuikChange® II Site-Directed Mutagenesis Kit (Stratagene, Agilent Technologies, Inc., Santa Clara, CA, USA) or by PCR using primers with the desired mutation. A recombinant GST-WAP-FnIII.3 (K130–K180–G413–K525) was amplified from the expression plasmid carrying the A1NtFnIII.3 recombinant cDNA (see below, Chemotaxis assays: Plasmid generation). The presence of the mutations and the integrity of the rest of the plasmid were confirmed by sequencing.
CHO cells were transfected with the vector pEBG and all the different GST fusion plasmids generated. After 24–36 h, cell lysates were prepared in phosphate-buffered saline (PBS) lysis buffer (pH 7.4, 1 mM EDTA, 1% NP-40 and complete EDTA-free protease inhibitor, Roche). GST and the GST-fusion proteins were incubated overnight, at 4°C, in the presence of glutathione–Sepharose CL-4B beads (GE Healthcare Life Sciences, Uppsala, Sweden), with the same amount of CHO cell extracts containing the full-length C-terminal HA-tagged FGFR1 (FGFR1-HA) in a final volume of 1000–1500 μL. After incubation, the samples were washed five times with the same buffer, boiled for 5 min in 60 μL of 2X Laemmli's sample buffer (Sigma-Aldrich, St Louis, MO, USA) to elute the immobilized proteins, half of which was resolved by SDS–PAGE (10% polyacrylamide gels), transferred to nitrocellulose membranes (GE Healthcare) and immunoblotted for detection using an anti-HA monoclonal peroxidase-conjugated antibody (High Affinity 3F10; Roche) and an anti-GST peroxidase-conjugated antibody to detect GST and the GST-fusion proteins (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA, sc-138).
All the GST pull-down assays were repeated at least three times and a representative experiment is illustrated in the figures. The amount of FGFR1-HA pulled down by the different GST-fusion proteins was semiquantified using the Adobe Photoshop CS2 software (Adobe Systems Incorporated, San Jose, CA, USA). Digitalized pictures of the x-ray films were inverted and converted to greyscale to measure the area and the mean intensity of the pixels and calculate the absolute pixel intensity for each band. Data are presented as the relative amount of the different fusion proteins ± SEM, versus the control protein (GST-WAP-FnIII.1). The relative values were normalized against the relative amount of the corresponding GST fusion protein.
Expression plasmids carrying C-terminal HA-tagged versions of full-length anosmin-1 (A1), the N-terminal region of anosmin-1 comprising the CR, WAP and FnIII.1 domains (A1Nt, M1–A289) and full-length anosmin-1 with the E514K or F517L substitutions (A1514K and A1F517L) have been used before in our laboratory (Murcia-Belmonte et al. 2010). Plasmids carrying the C172R or N267K substitutions were constructed using the QuikChange® II Site-Directed Mutagenesis Kit (Stratagene), or by PCR, introducing in the primers the corresponding point mutation and cloning the appropriate DNA fragments (A1C172R, A1N267K, A1NtC172R and A1NtN267K). A recombinant plasmid with the C-terminal HA-tagged N-terminal region of anosmin-1 in which the FnIII.1 domain has been replaced with FnIII.3 (A1NtFnIII.3), as well as the plasmid lacking the CR domain (A1ΔCR) were constructed by cloning the appropriate PCR products and DNA fragments. Basically, the recombination was performed introducing an EcoRI site after the residue K180 to take advantage of the EcoRI site just before FnIII.3, which generated the recombinant fusion at K180–G413. The deletion of the CR region spans residues R46–L128 and introduces two new residues within a HindIII site (K46–L47). The protein resumes at position V129, the start point for the WAP domain, which is now the position V48 in the mutated protein. Finally, several C-terminal HA-tagged truncated versions of anosmin-1 were generated by cloning the appropriate PCR products and fragments of DNA, recapitulating different nonsense mutations, found in patients with Kallmann syndrome, in the FnIII.1 (A1Y328X) and FnIII.3 domains (A1Q421X, A1R457X, A1S478X and A1E514X). When necessary, the integrity of the plasmids was confirmed by sequencing.
The plasmids were used to transiently transfect CHO cells and the conditioned media (c.m.) from these cells, as well as c.m. from CHO control cells (CT) and from a stable CHO line expressing the full-length C-terminal HA-tagged anosmin-1 were used as a source of WT anosmin-1 and the different mutant proteins generated. The presence of the proteins in whole-cell lysates and c.m. was confirmed by western blot using and anti-HA peroxidase-conjugated antibody (High Affinity 3F10; Roche).
All animal manipulations were carried out in accordance with the Hospital Nacional de Parapléjicos Animal Review Board (SAPA001), Spanish (RD223/88) and European (86/609/ECC) regulations. P5 Wistar rat SVZ NPs were isolated as previously described and used in these experiments (García-González et al. 2010; Murcia-Belmonte et al. 2010). Briefly, SVZ coronal slices were prepared with a tissue chopper (McIllwain, The Mickle Laboratory Engineering Co. Ltd, Gomshall, Surrey, UK), the SVZs were carefully dissected and dissociated in DMEM containing 0.025% Trypsin (GIBCO) for 30 min at 37°C and 0.05% DNAse was added during the last 5 min of incubation. Trypsin was blocked by adding 10% FBS and the cells were collected by centrifugation (150 g; 5 min).
In each Boyden chamber (8-μm pore, 12 mm in diameter; Corning Inc., Lowell, MA, USA) 40 000 NPs were seeded in the upper compartment in culture medium: DMEM-F12 1 : 1; supplemented with 5% FBS, 5% horse serum, 100 U/mL of penicillin and 100 μg/mL of streptomycin from GIBCO; at 37°C, 5% CO2 and at 95% humidity, while in the lower compartment the same culture medium was supplemented for the different experimental groups as follows: (I) CT; (II) FGF2 20 ng/mL; (III); A1; (IV) A1C172R; (V) A1N267K; (VI) A1Nt; (VII) A1NtC172R; (VIII) A1NtN267K; (IX) A1NtFnIII.3; (X) A1ΔCR; (XI) A1Y328X; (XII) A1Q421X; (XIII) A1R457X; (XIV) A1S478X; and (XV) A1E514X. The cells were treated during the experiment with the MEK1/2 inhibitor U0126 (10 uM; Sigma) where indicated, and the rest of the cultures were exposed to an equal volume of the vehicle dimethyl sulfoxide (DMSO) (Sigma). After 20 h, the cultures were fixed with 4% paraformaldehyde (for 10–15 min, at 24°C), washed three times with PBS and the non-migratory cells on the upper membrane surface were removed with a cotton swab. The presence of transmigrated NPs on the opposite side of the membrane was evaluated by immunocytochemistry using anti-tubulin beta III (Tuj-1, 1 : 500; Covance, Princeton, NJ, USA) antibody and Hoechst. Appropriate fluorescence-labelled secondary antibodies were used in each case. After immunostaining, the Boyden filters were examined using an In Cell 1000 Analyzer (software In Cell 1000 Analyzer; GE-HealthCare) and 16 microphotographs from each membrane were taken randomly. To quantify chemotropism, the number of transmigrated NPs per field was counted (double-stained cells for Tuj-1 and Hoechst) using the software In Cell 1000 Analyzer Workstation (GE-HealthCare). The data were expressed as percentage of migrating NPs relative to control conditions ± SEM, considered as 100% (García-González et al. 2010; Murcia-Belmonte et al. 2010). Representative images of the different conditions and experiments are shown. For illustration purposes, a greyscale inverted image of the Tuj-1 channel is shown.
Dissociated cells from the SVZ (1 × 106 cells per well) were incubated for 24 h in poly-l-lysine-coated P12-well plates in culture medium: DMEM-F12 1 : 1; supplemented with 5% FBS, 5% horse serum, 100 U/mL of penicillin and 100 μg/mL of streptomycin from GIBCO; at 37°C, 5% CO2 and at 95% humidity. Cells were serum starved for 6 h in DMEM and treated with the FGFR blocker SU5402 (10 μM; Calbiochem-Merck, Merck KGaA, Darmstadt, Germany) where indicated (Fig. 5c, d) for 30 min prior to stimulation, while the rest of the wells were exposed to an equal volume of the vehicle DMSO (Sigma). Stimulation was carried out for 30 min with FGF2 (20 ng/mL) and c.m. from CHO control cells (CT) and from CHO cells expressing the full-length C-terminal HA-tagged anosmin-1. The FGFR blocker SU5402 and the DMSO were used where indicated during the stimulation (Fig. 5c, d). After 30 min, the cells were washed with cold PBS and lysed in lysis buffer (phosphate-buffered saline buffer pH 7.4, 1 mM EDTA, 1% NP-40, complete EDTA-free protease inhibitor and PhosSTOP Phosphatase Inhibitor Cocktail Tablets, Roche). An equivalent amount of lysate was boiled for 5 min in 2X Laemmli's sample buffer (Sigma) and resolved by SDS–PAGE (10% polyacrylamide gels), transferred to nitrocellulose membranes (Santa Cruz Biotechnology) and immunoblotted for detection with anti-ERK (Santa Cruz Biotechnology, sc-93) and anti-phospho-ERK (Santa Cruz Biotechnology, sc-7383) antibodies. Secondary antibodies conjugated with 680 or 800 infrared fluorescent dyes (IRDye, LI-COR Biosciences, Lincoln, NE, USA) were used for simultaneous analysis of ERK and pERK in separate fluorescent channels using the Odyssey Infrared Imaging System (LI-COR). The quantitative analysis was carried out following the Odyssey LI-COR instruction manual. The values ± SEM of the relative amount of pERK with respect to the total amount of ERK, normalized against the value of the relative pERK/ERK amount of the cells exposed to the c.m. of CHO CT cells considered as 100%, are represented.
Quantification and statistics
Chemotaxis assay data are expressed as percentage of migrating NPs relative to control conditions ± SEM, considered as 100%. All data were analysed using the Sigmastat software package (SPSS Inc., Chicago, IL, USA). anova test (Bonferroni correction) or the corresponding test on ranks were used for the chemotaxis assays versus CT conditions (minimal statistical significance was fixed at p <0.05).
Role of the N-terminal domains of anosmin-1, WAP and FnIII.1 in the binding to FGFR1
The implication of the WAP and FnIII.1 domains of anosmin-1 in FGFR1 binding has previously been described (Hu et al. 2009; Murcia-Belmonte et al. 2010). Therefore, we tested the effect of two missense mutations within these domains described in KS patients (WAP C172R and FnIII.1 N267K) on the binding of the N-terminal region of anosmin-1 to FGFR1 (Fig. 1a). In a GST pull-down assay using the combination WAP-FnIII.1 (K130–K285) as a positive control, the C172R substitution had no effect on the binding of this region to FGFR1 (Fig. 1b). On the contrary, the N267K substitution seemed to have a drastic effect, eliminating the binding to FGFR1 almost completely (Fig. 1b). To study in more detail the implication of the different domains in the binding of the N-terminal region of anosmin-1 to FGFR1, a recombinant GST–anosmin-1 fusion protein was constructed spanning the WAP domain (K130–K180) fused in frame to the FnIII.3 domain (G413–K527, GST-WAP-FnIII.3, Fig. 1a). In the subsequent GSTpd using the WAP-FnIII.1 plasmid as a positive control, the recombinant protein GST–WAP-FnIII.3, was able to interact with FGFR1 (Fig. 1c). The amount of FGFR1 pulled down by the different proteins relative to the WAP-FnIII.1 protein was semiquantified, confirming that the C172R mutation had no effect in the binding and that the N267K reduced it greatly. The recombinant protein WAP-FnIII.3 still retained about 50% of the binding capacity to the receptor (Fig. 1d).
Effect of the C172R/N267K mutations in the function of anosmin-1 and the role of the FnIII.1, FnIII.3 and CR domains
We next analysed the effect of the two mutations described above (C172R and N267K) (Fig. 2a), on the chemotropic effect on rat SVZ NPs mediated by anosmin-1 via FGFR1 and conserved by the N-terminal region of the protein, A1Nt (Murcia-Belmonte et al. 2010). The presence of all the proteins used was confirmed by western blot in CHO cell lysates and c.m. (Fig. 2b). Full-length proteins harbouring any one of these mutations lost the chemoattractive effect on rat SVZ NPs (Fig. 2c and d), but surprisingly, the mutant N-terminal proteins, A1NtC172R and A1NtN267K, retained the effect (Fig. 2c and d). As in the GSTpd the recombinant protein WAP-FnIII.3 was still able to bind to FGFR1 (Fig. 1c and d), a recombinant N-terminal anosmin-1 protein with a C-terminal HA tag in which the FnIII.1 domain was replaced with the FnIII.3 (A1NtFnIII.3; Fig. 3a) was tested for chemotropism. The presence of the different proteins in whole-cell lysates and c.m. was confirmed by western blot (Fig. 3b). In the chemotaxis assay performed, this protein lacked the chemotropic effect observed with the N-terminal region of anosmin-1 (Fig. 3c and d).
The likely regulatory role of the cysteine-rich region in the activity of anosmin-1 has been proposed before (Andrenacci et al. 2006; Murcia-Belmonte et al. 2010). To further investigate the role of the CR region in the effects of this protein mediated via FGFR1, we generated a mutant form of anosmin-1 lacking this domain (A1ΔCR, deletion spanning residues R46–L128, see 'Materials and methods' and Fig. 3a). This deletion suppressed the chemotropic effect of anosmin-1 on SVZ NPs (Fig. 3c and d). By western blot, we confirmed the presence of the protein in CHO cell lysates and the c.m. used in the experiments (Fig. 3b).
Nonsense mutations found in patients suffering from Kallmann syndrome render non-functional proteins
Several nonsense mutations have been described in KS patients both in FnIII.2 and FnIII.3 (Hu and Bouloux 2011). We have generated expression plasmids carrying C-terminal HA-tagged versions of anosmin-1 recapitulating some of these mutations: Y328X in FnIII.2 (Georgopoulos et al. 2007) and Q421X (Hardelin et al. 1993), R478X (Albuisson et al. 2005), S548X (Versiani et al. 2007) and E514X (Trarbach et al. 2005) in FnIII.3 (Fig. 4a). The expression and secretion of the different proteins used in the study was assessed by western blot in whole-cell lysates and c.m. of CHO cells transfected with the different plasmids (Fig. 4b). When used in chemotaxis assays, all of these proteins failed to exert the chemotropic effect on rat SVZ NPs as compared with the N-terminal truncated form of anosmin-1, A1Nt (M1–A289), which is even smaller than all the mutant proteins studied (Fig. 4c and d).
FGF2 and anosmin-1 exert their chemotropic effect through the activation of the ERK MAPK pathway
It has been shown before that the activation of FGFR1 by anosmin-1 triggers the activation of ERK and p38 MAPK pathways and that both are necessary for the neurite outgrowth phenotype observed in FNC-4B cells stimulated with anosmin-1 (González-Martínez et al. 2004). To investigate the FGFR1 downstream signalling involved in the chemotropic effect elicited by FGF2 and anosmin-1 through FGFR1 on rat SVZ NPs, we performed chemotaxis assays using specific inhibitors of the ERK MAPK pathway. The chemoattraction on SVZ NPs exerted by both FGF2 and anosmin-1 is blocked by the MEK1/2 inhibitor U0126 (Fig. 5a and b). To confirm that FGF2 and anosmin-1 in fact activated the ERK MAPK pathway, dissociated rat SVZ cells plated on poly-l-lysine-coated P12-well plates were treated with FGF2 and anosmin-1 in the presence or not of the FGFR blocker SU5402. Both FGF2 and anosmin-1 were able to activate the ERK pathway, and the treatment of the cells with the FGFR blocker SU5402 prevented this activation from happening (Fig. 5c and d).
Contribution of the different domains of anosmin-1 to the binding to FGFR1 and the function of the protein
So far, most of the biological effects in which anosmin-1 participates seem to be mediated through the activation or modulation of FGFR1 signalling (Dodé et al. 2003; González-Martínez et al. 2004; Bribián et al. 2006; Ayari and Soussi-Yanicostas 2007; Hu et al. 2009; García-González et al. 2010; Murcia-Belmonte et al. 2010). Not surprisingly, interaction between these two proteins has been described by CoIP (González-Martínez et al. 2004; Bribián et al. 2006; Ayari and Soussi-Yanicostas 2007). Some efforts have been made to elucidate the role of the different domains of anosmin-1 in the interaction with FGFR1 and, in this respect, it has been shown by SPR that the N-terminal region of anosmin-1 (comprising the CR, WAP and FnIII.1 domains) interacts with some of the extracellular domains (D2–D3) of FGFR1 (Hu et al. 2009). More recently, we have demonstrated that the interaction of the N-terminal region is circumscribed to the WAP-FnIII.1 domains, as both of them are necessary for the interaction, but none of them alone is able to interact with FGFR1 (Murcia-Belmonte et al. 2010). A regulatory role for the CR domain has also been suggested that does not seem to be required for the interaction, at least in vitro (Andrenacci et al. 2006; Murcia-Belmonte et al. 2010).
The GST pull-down results obtained here indicated that the C172R mutation within the WAP domain (Oliveira et al. 2001) did not have a profound effect on the interaction with FGFR1, but the N267K substitution within FnIII.1 (Hardelin et al. 1993) reduced it drastically. Previous studies have established that the C172R substitution does not affect the binding of full-length anosmin-1 to heparin (Hu et al. 2004). In addition, it has been shown that the N267K mutation prevents full-length anosmin-1 from binding to FGFR1 (Hu et al. 2009), which is supported by our results. The C172R mutation has been suggested to be physiologically relevant as it would disrupt a highly conserved Cys157–Cys172 intramolecular disulphide bond (Oliveira et al. 2001; Hu et al. 2004). The N267K substitution in FnIII.1 is adjacent to a predicted HS-binding site and could either disrupt the protein folding, preventing secretion or alter the function of anosmin-1 (Robertson et al. 2001). Our present results indicated that the expression and secretion of the mutant proteins studied, A1C172R and A1N267K, are not affected. These results are in agreement with previous reports showing the secretion of the full-length proteins with the N267K, E514K and F517L substitutions (Robertson et al. 2001; Cariboni et al. 2004; Murcia-Belmonte et al. 2010), ruling out a synthesis or secretion defect.
Surprisingly, while in the full-length protein both mutations (A1C172R and A1N267K) blocked the chemotropic effect of anosmin-1 on SVZ NPs, none of them seemed to affect the chemoattraction exerted by the N-terminal region of anosmin-1 (A1NtC172R and A1NtN267K). Although the lack of chemoattraction displayed by the A1N267K protein confirmed previous reports (Cariboni et al. 2004), contradictory results regarding the effect of mutations in the WAP and FnIII.1 domains have been reported before. Thus, in Drosophila, the over-expression of mutant forms of anosmin-1 with mutations in cysteine residues within the WAP domain (C127/128S) or the S263K substitution in the FnIII.1 domain, equivalent to the human N267K mutation, produce similar phenotypic defects to the over-expression of the WT protein (Andrenacci et al. 2006). In C. elegans, the over-expression in AIY interneurons of the WT and a WAP mutant protein, produces an axon branching defect that is not retained by the full-length protein harbouring the S241K mutation in FnIII.1. When these mutant proteins are expressed panneuronally, the WT and the S241K mutant have the same misrouting defect in sensory neurons, but the WAP mutant presents an impaired potential to misroute the same axons (Bülow et al. 2002). Both the full-length and the truncated N-terminal region of the protein harbouring the C172R mutation fail to exert neurite outgrowth on FNC-B4 cells promoted by anosmin-1 (González-Martínez et al. 2004). They also cancel the proliferative effect of anosmin-1 on PC-3 cells, suggesting the requirement of an intact WAP domain. Nevertheless, they are able to increase the amidolytic activity of the serine protease uPA in vitro (Hu et al. 2004). It has been suggested that the function of the different anosmin-1 domains could be determined or conditioned by the extracellular environment, and particularly by the HS composition. Thus, anosmin-1 could be involved in different phenotypic effects depending on the cell type and the fine structure of the HS glucidic moiety, which would modulate the interaction with different receptors or binding proteins (Bülow et al. 2002; Andrenacci et al. 2006; Tornberg et al. 2011). Therefore, it could be speculated that the biological response exerted by the truncated N-terminal anosmin-1 with the mutations studied could be different from that elicited by the full-length protein, as it lacks three of the FnIII domains. This, together with the mutations in the N-terminal domains could make these proteins present a different binding capacity to FGFR1 and to other receptors or molecules of the ECM. Altogether, these data indicated that C172R, while not affecting the binding to FGFR1, prevents anosmin-1 from working properly. The N267K mutation, at least in full-length anosmin-1, renders a non-functional protein that is not able to bind to FGFR1 (Hu et al. 2009).
We have previously shown that the third FnIII repeat of anosmin-1 interacts with FGFR1 (Murcia-Belmonte et al. 2010). Mutations within this domain present in KS patients, E514K and F517L (Georgopoulos et al. 1997; Maya-Nuñez et al. 1998), greatly reduce the interaction of the full-length protein with the receptor (Hu et al. 2009), or completely abolish the binding of the FnIII.3 domain to FGFR1, suggesting a high specificity in the binding to FGFR1 (Murcia-Belmonte et al. 2010). Not surprisingly, these mutations provoke the total loss of the chemoattractive effect displayed by anosmin-1 on GN11 cells (Cariboni et al. 2004) and SVZ NPs, probably because of an inefficient FGFR1 binding (Murcia-Belmonte et al. 2010). The recombinant GST fusion protein WAP-FnIII.3 showed a reduced binding to FGFR1, but was still able to bind to the receptor. In the chemotaxis assay, this recombinant N-terminal truncated anosmin-1 (A1NtFnIII.3) did not show any effect. All this would suggest that although still capable of binding to FGFR1 through FnIII.3, the recombinant protein is not able to activate the receptor. The A1NtFnIII.3 protein would be directed to the specific region of FGFR1 where FnIII.3 binds and would anchor the CR and WAP domains to a place within the protein different from the binding place where FnIII.1 would hypothetically direct them. This would provoke a lack of functionality because of a mislocalized binding to the receptor. As mentioned above, the N-terminal cysteine-rich region has been proposed to play a regulatory role facilitating or preventing the interaction of the FnIII domains with interacting molecules. Together with the WAP domain, the CR domain could assist in the interaction by helping FnIII.1 adopt an optimal conformation (Robertson et al. 2001; Andrenacci et al. 2006; Hu et al. 2009; Murcia-Belmonte et al. 2010). Our own data suggest that the CR is not necessary for the interaction of the N-terminal region with FGFR1 (Murcia-Belmonte et al. 2010). In fact, a CR mutated form of Drosophila anosmin-1 with the C85S substitution still retains some activity in respect to the phenotype analysed (Andrenacci et al. 2006). The failure of the mutant anosmin-1 without the CR domain (A1ΔCR) to induce chemotropism via FGFR1 on rat NPs would indicate a more distinct role for this region in the function of anosmin-1. The CR domain, although not required for the physical interaction with the receptor, is essential for the effect analysed here (Fig. 6).
Nearly 60 mutations spread throughout the whole KAL1 gene have been identified associated to KS patients. Two-thirds of them are deletions, frameshift or nonsense mutations that are expected to affect the overall length or composition of the protein produced, sufficient to account for the disease (Hu and Bouloux 2011). We have generated truncated proteins recapitulating some of the nonsense mutations found in the FnIII.2 and FnIII.3 domains and tested them in chemotaxis assays. None of the truncated proteins was able to induce the chemoattraction exerted by the N-terminal region of anosmin-1 (A1Nt) over SVZ NPs. It has been proposed that at the cell-surface anosmin-1 would undergo proteolytic cleavage at positions 345–350 (KKKRRK) within the FnIII.2 domain, resulting in a C-terminal diffusible peptide of 45 kDa and a truncated N-terminal protein (Rugarli et al. 1996). Our data indicated that a slightly smaller (A1Y328X) and a larger truncated protein (A1Q421X), which includes the second FnIII repeat, have no activity. Therefore, the proposed truncated N-terminal peptide resulting from the proteolytic cleavage would also be a non-functional protein, and probably the result of non-specific proteolysis or degradation leading to the regulation of the protein activity. The fact that the truncated A1Nt protein conserves its biological activity in some scenarios (Bülow et al. 2002; González-Martínez et al. 2004; Hu et al. 2004; Murcia-Belmonte et al. 2010) could be because of the possibility that A1Nt is more stable in this form than larger truncated proteins produced by the nonsense mutations studied, likelihood already suggested in the case of similar mutations analysed in Drosophila (Andrenacci et al. 2006). The chemoattraction on SVZ NPs exerted by anosmin-1 is abolished by the E514K or F517L substitution. This suggests that in addition to impede the binding of FnIII.3 to FGFR1, these substitutions could produce a conformational change in the protein that would destabilize the union of the N-terminal domains of anosmin-1 to FGFR1, which would produce a non-functional protein (Murcia-Belmonte et al. 2010). All these data suggest that the presence of any additional sequence behind the FnIII.1 domain could destabilize the protein and/or impede the correct binding to FGFR1 or to other potential partners (Fig. 6).
An unresolved question is the physiological relevance of the truncated A1Nt protein. This fact raises the possibility of the existence of mutations within the KAL1 gene that would produce a truncated N-terminal protein lacking some of the biological activities promoted by the missing FnIII domains, but still functional regarding the effects dependant on the N-terminal domains and that could evade detection in the genetic analysis of KS patients.
The degree of anosmia and hypogonadotropic hypogonadism (HH) in KS patients can vary significantly, even between monozygotic twins or siblings sharing identical mutations (Kim et al. 2008). Thus, so far, it has been difficult to correlate genotype with phenotype, which suggest that genetic and epigenetic variability may play a significant role in the development of the disease (Versiani et al. 2007). In general, mutations in the KAL1 gene are associated with HH and some degree of anosmia, whereas cleft palate and dental agenesis could be characteristics of FGFR1 mutations, which show more variable degrees of HH with or without anosmia (Quinton et al. 2001; Dodé and Hardelin 2004; Sato et al. 2004; Pitteloud et al. 2006; Zenaty et al. 2006; Kim et al. 2008).
Roughly 30% of KS patients present renal agenesis, a satellite symptom specially common in X-linked KS patients, although it has been also described in sporadic and familial cases of KS with an apparent normal KAL1 gene (Maya-Nuñez et al. 1998; Söderlund et al. 2002; Sato et al. 2004; Versiani et al. 2007). Bimanual synkinesis (up to 75% of X-linked cases) is also more likely to be associated with KAL1 mutations (Kim et al. 2008), but again there are patients with this symptom and no described mutation in the coding region of the KAL1 gene (Maya-Nuñez et al. 1998; Söderlund et al. 2002). Therefore, defects present in the promoter or untranslated regions of the KAL1 gene, as well as the participation of different genes in the X-linked form of KS, cannot be ruled out to explain these observations (Maya-Nuñez et al. 1998). It is also difficult to establish a relationship between phenotype and the presence of a particular symptom with the occurrence of a specific mutation within the KAL1 gene. For example, renal agenesis has been reported in patients with nonsense mutations at different levels of the protein: R191X and L250X in FnIII.1, R613X in FnIII.3 (Trarbach et al. 2005); E320X in FnIII.2 and R457Xin FnIII.3 (Albuisson et al. 2005); deletion of exons 5–10 spanning FnIII.1-3 (Trarbach et al. 2005) or missense mutations, E514K in FnIII.3 (Maya-Nuñez et al. 1998). The same is true for bimanual synkinesis described in patients with mutations in different regions of the protein: C163Y in WAP (Sato et al. 2004); C172R in WAP (Oliveira et al. 2001); R191X in FnIII.1 and R423X in FnIII.3 (Albuisson et al. 2005); whole-gene deletion (Sato et al. 2004).
Active role of the ERK MAPK pathway in FGF2/anosmin-1 promoted chemotropism
Anosmin-1 is able to induce neurite outgrowth via FGFR1 in FNC-4B NPs derived from human foetal olfactory neuroepithelium, an effect that is mediated through the downstream activation of the ERK and p38 MAPK pathways (González-Martínez et al. 2004). Neurite outgrowth and neuronal migration require the reorganization of the F-actin cytoskeleton and, in fact, anosmin-1 can induce filopodia formation and actin cytoskeleton reorganization in these cells, possible after the activation of the Cdc42–Rac1 pathway (González-Martínez et al. 2004). Here, we showed that the chemotropic effect on rat SVZ NPs, mediated by both FGF2 and anosmin-1 via FGFR1, is blocked by the specific inhibitor of the ERK MAPK pathway U0126, indicating that the activation of this pathway is essential for the migration of these cells after FGFR1 stimulation. The role of MAPK activation in neuronal migration has been reported before. Specifically, in mouse SVZ NPs, the neurotrophin brain-derived neurotrophic factor via TrkB promotes migration through the activation of cyclic adenosine monophosphate response element-binding protein, which requires both phosphatidylinositol 3-kinase (PI3K) and ERK signalling pathways (Chiaramello et al. 2007). Mouse cortical neuron migration induced by hepatocyte growth factor via Met requires the activation of different signalling pathways including ERK, PI3K/Akt and Rac1/p38 (Segarra et al. 2006). Human neural progenitor cell migration controlled by the EGFR/PKC pathway is also dependent on ERK1/2 activation (Moors et al. 2007). The GTPases of the RHO family are key regulators of the actin/microtubule cytoskeletons, cell polarity and adhesion, playing a critical role in CNS neuronal migration (Govek et al. 2011). As mentioned above, the activation of Cdc42–Rac1 by anosmin-1 has previously been described. Thus, the implication of these pathways in the chemotropism exerted by anosmin-1 via FGFR1 should be studied in more detail in the future.
We are grateful to Dr. José Ángel Rodríguez Alfaro, Dr. Javier Mazarío, Miss Isabel Machín, Mr. Rafael Lebrón and Mr. Jacinto Sarmentero for their technical assistance. This research was supported by grants from the Spanish Ministerio de Economía, Innovación y Competitividad MINECO (SAF2009-07842, ADE10-0010, RD07-0060-2007), Spain, from the Gobierno de Castilla-La Mancha (PI2009/29) and the Fundación Eugenio Rodíguez Pascual. VMB and DGG are PhD students hired by Gobierno de Castilla-La Mancha (MOV2007-JI/19 and MOV2010-JI/11 respectively) and currently hired under RD07-0060-2007 and by SESCAM respectively. FdCS is a CSIC staff scientist in special permission and hired by SESCAM, Gobierno the Castilla-La Mancha, Spain. PFE was a researcher hired by SESCAM, Gobierno de Castilla-La Mancha, currently hired under ADE10-0010.