FMRP is an RNA binding protein whose absence produces pathological manifestations of the fragile-X syndrome. FMRP is a component of mRNP complexes found in association with actively translating polyribosomes, RNA complexes trafficking in neurites, RNA granules in cytoplasm and, in Drosophila, with the RNAi machinery. We report here the identification and characterization of a novel FMRP-interacting protein associated to polyribosomes as a component of mRNP complexes containing FMRP. We named this protein 82-FIP (82-kD FMRP Interacting Protein). FMRP interacts with 82-FIP through a novel interaction motif located in its N-terminal region. The distribution of 82-FIP in different areas of the brain is very similar to that of FMRP. However, unlike FMRP, 82-FIP is found in both nucleus and cytoplasm in some neurons, while it appears only cytoplasmic in others. Subcellular distribution of 82-FIP is cell cycle-dependent in cultured cells, suggesting that the composition of some FMRP-containing RNP complexes may be cell cycle-modulated.
Absence of FMRP (Fragile-X Mental Retardation Protein) causes the fragile-X syndrome, the most common monogenic form of mental retardation. This X-linked syndrome is characterized by moderate to severe mental retardation in affected males, while 60% of carrier females present mild to moderate mental retardation. Affected boys often manifest hyperactive and/or autistic behaviour (1,2). In addition to cognitive and behavioural defects, fragile-X male patients are characterized by facial dysmorphism, post pubertal macroorchidism and mild connective tissue dysplasia (1). Silencing of FMRP expression in fragile-X patients is due to a CGG repeat hypermethylated expansion localized in the 5′ UTR of the FMR1 gene, blocking its transcription (2–4). The brain cortex of fragile-X patients displays dendritic spines, which are longer, thinner and denser than normal. These abnormalities have been postulated to be at the basis of the mental retardation of fragile-X patients (5–7). Fmr1-null mice exhibit similar neuroanatomical features, although these animals present only a subtle behavioural defect (8,9). Moreover, it has been shown that dFXR, (the FMR1 homologue in Drosophila melanogaster) (10–12) is involved in the process of neurite extension, guidance and branching (11,13).
FMRP is highly expressed in neurons and in spermatogonia (14,15) and is predominantly localized in the cytoplasm (14), although a small proportion of the protein (or some specific isoforms) may be present in the nucleus (16–18). FMRP is an RNA binding protein that has high affinity for RNA G-quartet structures (19,20). FMRP is endowed with a nuclear export signal (NES) and a nuclear localization signal (NLS) (21,22). Treatment with leptomycin B confirmed that the protein has the ability to shuttle between nucleus and the cytoplasm, suggesting that it may be involved in the process of RNA export from the nucleus (23,24). There is increasing evidence that FMRP is involved in translational control. (i) Preincubation of mRNAs with FMRP leads to translational inhibition both in a cell-free system and in Xenopus oocytes (25,26). This effect is not mRNA specific, but is impaired by the I304N mutation, present in a patient with a severe form of the fragile X syndrome (26). In addition, in transfection experiments, expression of high levels of FMRP results in repression of reporter genes (27). (ii) A subset of mRNAs has been found with altered polysomal distribution in lymphoblastoid cells lacking FMRP, including some mRNAs that contain a G-quartet stretch (20,28). (iii) Translation of some mRNAs, including MAP1B (Microtubules Associated Protein 1B), appears moderately down-regulated by Fmrp in mouse brain via the non-coding RNA BC1 (29). (iv) In D. melanogaster, the dFMR1 protein, the ortholog of the common ancestor of FMRP, FXR1P and FXR2P (the two proteins homologues of FMRP), was observed to down-regulate expression of the futsch protein (homologue to mammalian MAP1B) (11). In addition, two different subsets of FMRP-associated mRNAs have been described recently (30,31). In brain of the Fmrp-null mice, some of these mRNAs or their encoded proteins show discrete changes in abundance and/or differential subcellular distribution (30,31).
FMRP was observed to be a component of granule-like structures that contain repressed mRNAs in mammal cells after exposure to stressful conditions (27) and to be associated with the RNAi machinery in Drosophila (32,33). In neurons, a minor fraction of FMRP is present in dendrites and this led to the hypothesis that FMRP is involved in translational control at the post-synaptic level (34). Recent results have shown that in PC12 cells FMRP is involved in the transport of mRNAs in neurites probably being involved in the trafficking of mRNA from the neuronal cell body to synaptic sites (27,35).
Several FMRP interacting proteins have been characterized in recent years: FXR1P and FXR2P are closely similar to FMRP and their functions are presumed to be related (36); NUFIP1, a nuclear RNA binding protein (37,38); CYFIP1 and 2, two homologous proteins that link FMRP to the RhoGTPase pathway in mammals (39,40) and in Drosophila (41). Furthermore, other proteins have shown to be components of FMRP containing complexes, although evidence for direct interaction is generally lacking (27,42–44). In Drosophila the dFMR protein was found to be associated with the RNA-induced silencing complex (RISC) together with Argonaute 2 (AGO2) and Dicer (32,33). In the present study we report the identification of a novel interactor of FMRP that we named 82-FIP (82 kDa FMRP Interacting Protein) and which, similarly to NUFIP1 (37) and CYFIP1 (39), binds specifically to FMRP but not to FXR1P and FXR2P (36). 82-FIP is associated with polyribosomes, is a component of FMRP-containing mRNP and is localized both in the cytoplasm and the nucleus. We show here that 82-FIP distribution is cell cycle-modulated, being cytoplasmic in the G2/M phase and accumulating in nucleus during the G1 phase. These findings also suggest that the composition of the FMRP-containing complexes is cell cycle-dependent. A differential subcellular distribution of 82-FIP is also observed in neurons for some different areas of the brain.
Identification of a novel FMRP interacting protein
In order to identify FMRP-interacting proteins we previously described a two-hybrid screening in yeast using a mouse embryonic library (37). We described the cloning and characterization of NUFIP1 (37), CYFIP1 and CYFIP2 (39). A small cDNA fragment (clone 80) was also identified in this screening (Fig. 1A). Initial analyses showed an interaction between the sequence encoded by clone 80 and FMRP in a GST pull-down assay (not shown). In the yeast two-hybrid assay we observed that, similarly to NUFIP1 (Fig. 1B) (37) and CYFIP1 (39), 82-FIP interacts with FMRP but not with FXR1P and FXR2P (Fig. 1B), despite the high level of homology of the three proteins in their N-terminal region (Fig. 1B). In contrast, CYFIP2 interacts with the three FXR proteins, as expected (Fig. 1B) (39). The strong interaction between clone 80 product and the FMRP full-length protein or its N-terminal portion (the first 218 amino acids) (Fig. 1B), induced us to pursue the characterization of the corresponding full-length protein.
We found that the FMRP-interacting sequence lies inside a novel protein that we named 82-FIP (accession no. AJ493465; Fig. 1C). Its cDNA which was already described as KIAA1321 (accession no. AB037742), is homologous to several EST clones and to the NT_010799 genomic clone from human chromosome 17. The human KIAA1321 clone encodes a protein of 695 amino acids, which is highly similar to the clone (accession no. AK122494) defining the orthologous mouse protein of 692 amino acids (94.6% identity). The amino acid sequences of the 82-FIP proteins show no homology to proteins of known function or to any known functional domain. In order to define the conserved domains we compared the human and mouse 82-FIP sequences with the Fugu homologue (accession no. AJ51590, f82-fip). f82-fip encodes a putative protein of 766 amino acids and shows an overall identity with the human homologue of 40% (not shown). Interestingly, both human and mouse 82-FIP proteins show a very high level of conservation with f82-fip in the region of interaction with FMRP (Fig. 1A).
Definition of a novel protein–protein interacting domain in FMRP
In order to evaluate whether full-length 82-FIP interacts with FMRP, we performed a series of pull-down assays. The FMRP full-length protein (ISO1) was produced in a baculovirus system as a glutathione S-transferase (GST) fusion protein (37), while 82-FIP was produced and labelled with [35S]methionine by in vitro transcription–translation using a rabbit reticulocyte lysate system. The GST–FMRP fusion protein immobilized on glutathione–Sepharose was incubated with labelled 82-FIP or, as a positive control, with FXR1P a known interactor of FMRP (36), or with luciferase, as a negative control. 82-FIP was retained specifically by the immobilized GST-FMRP with an efficiency comparable to that shown by FXR1P (Fig. 2A), while no binding was observed to the GST column alone.
We confirmed the FMRP/82-FIP interaction by immunoprecipitating protein extracts from COS cells transfected with a vector expressing ISO7 FMRP (16), using two anti 82-FIP antibodies raised against synthetic peptides corresponding to the N- and to the C-terminal regions of the protein (see Materials and Methods). Cell extracts from FMRP-ISO7 transfected COS cells were immunoprecipitated using the no.1666 anti 82-FIP serum, followed by immunoblotting with the anti-FMRP mAb1C3 (14) and with the polyclonal no. 1667 anti-82-FIP. ISO7 was co-immunoprecipitated by the anti-82-FIP antibody, but not by non-immune IgG (Fig. 2B). This analysis indicates that interaction between 82-FIP and FMRP also takes place in vivo.
To define the 82-FIP-interaction domain in FMRP we designed chimeric clones, in which different portions of the cDNA FMR1 coding for the N-terminal region of the protein have been substituted by the corresponding regions of FXR1 or FXR2. The same approach was previously used to define the interaction domain of FMRP with CYFIP1 (39). FMRP still interacts with 82-FIP when the sequence encoded by exon 7 (residues 173–218) in FMRP was substituted with the corresponding sequence from FXR1P (not shown), but this interaction was disrupted when sequences corresponding to FMRP exons 4–7 (residues 66–218) were substituted by the corresponding sequences from FXR1P (not shown). In the same series of experiments we have tested also the interaction of CYFIP2 and NUFIP1 with these chimeric proteins. As expected, CYFIP2 interacts with the FMRP-exon 7 sequences (39), while NUFIP1 interacts with FMRP exon 4–7 sequences, as does 82-FIP. To confirm these results, a recombinant protein corresponding to the first 134 amino acids of FMRP fused at its N-terminal with a His-tag/FMRP(1–134) was tested in a pull-down assay to determine interactions with 82-FIP and NUFIP1. We found that 82-FIP and NUFIP1 were retained specifically by the FMRP(1–134) protein immobilized on Ni-NTA agarose beads (Fig. 2C). These results allow us to restrict a novel interacting site in FMRP between residues 66 and 134, the region corresponding to exons 4 and 5 (45).
Expression of 82-FIP
82-FIP mRNA is expressed in a very wide range of tissues as revealed by RT-PCR analysis (www.kazusa.or.jp/huge/gfpage/KIAA1321/) or by northern blot (not shown). To determine the distribution of 82-FIP in brain, immunohistochemistry was performed on adult mouse brain sections using the no. 1666 antibody, which recognizes murine 82-FIP. The protein appears localized in both nucleus and cytoplasm in most neurons, with a distribution matching that of FMRP. In the cortex, it is distributed in a diffuse way in the nucleus and in the cytoplasm (Fig. 3A–C). Interestingly, 82-FIP is only localized in the cytoplasm in neurons of the dentate gyrus (not shown) in the olfactive bulb, in the ependymal epithelium and in the granular layer of the cerebellum (Fig. 3D–F). In Purkinje cells, 82-FIP is distributed in both cell compartments and, in addition, is localized in nuclear dots adjacent to the nucleolus (Fig. 3D–F). We did not observe colocalization between these 82-FIP containing structures and Cajal bodies (or SC35-containing speckles; data not shown).
Sub-cellular localization of 82-FIP is cell-cycle dependent in growing cells
The localization of the endogenous 82-FIP protein was studied by immunostaining using the no. 1666 antibody. The protein was found localized in the nucleus and in the cytoplasm, particularly in the perinuclear region (Fig. 4A). When analysed by confocal microscopy, the colocalization of 82-FIP with endogenous FMRP is evident in the cytoplasm and in the perinuclear region (Fig. 4C).
In exponentially growing COS and 3T3 cells we observed three different patterns of nucleo-cytoplasmic distribution of 82-FIP: (i) most abundant in the nucleus, (ii) diffuse in both nucleus and cytoplasm, (iii) most abundant in the cytoplasm (Fig. 5L). In order to analyse whether these different intracellular localizations are cell cycle-dependent, COS cells were synchronized in the G1, S or G2/M phases of the cell cycle by mimosine, thymidine or nocodazole treatments, respectively (46). Cell-cycle synchronization was verified by flow-cytometry. Synchronization in G1 produced 82-FIP accumulation in the nucleus (Fig. 5A), whereas the protein was localized in the cell cytoplasm in cells blocked in G2/M (Fig. 5D) or was diffusely localized in both the nucleus and the cytoplasm (Fig. 5G) in cells blocked in S phase. Statistical analysis of observations illustrated in Figure 5 is reported in Table 1. In the same experiments we also tested the sublocalization of FMRP using the mAb 1C3 antibody and we observed that its cytoplasmic localization remained unchanged in the different cell cycle phases (data not shown). Synchronization of mouse 3T3 fibroblasts in G0/G1 by serum deprivation (0.1% FCS for 48 h) also produced 82-FIP accumulation in the nucleus (not shown).
82-FIP is present in FMRP-containing complexes associated to polyribosomes
Since cytoplasmic FMRP is associated to actively translating polyribosomes (47), we tested whether 82-FIP cofractionated with FMRP after sedimentation onto sucrose density gradient. Indeed, in the presence of Mg2+ 82-FIP was detected in fractions corresponding to polyribosomes, similarly to FMRP (Fig. 6A). In the presence of EDTA, that dissociates polyribosomes, the two proteins were recovered together in fractions with lower sedimentation values (Fig. 6B). The presence of 82-FIP in polyribosomes fractions was not altered when cytoplasmic extracts derived from fragile-X lymphoblastoid cells were analysed (Fig. 6A and B), indicating that FMRP is not required for the stowing of 82-FIP with polyribosomes.
Since it was previously shown that FMRP association with polyribosomes is mRNA dependent (47), we therefore performed oligo(dT) chromatography using enriched fractions of polyribosomes prepared from normal and fragile-X lymphoblastoid cells to assess whether this is also the case for 82-FIP. Indeed, we observed that 82-FIP is retained onto the oligo (dT) column and is eluted together with FMRP at the same salt concentration. In addition, in the absence of FMRP the behavior of 82-FIP is not altered (Fig. 7A). These results suggest that, similarly to FMRP, 82-FIP is associated with mRNA as well as with other interacting proteins present in RNP complexes. To investigate the capacity of 82-FIP to bind to RNA we incubated an in vitro synthesized 82-FIP with RNA homopolymers immobilized on agarose beads. After extensive washing of the different homopolymers, we observed that 82-FIP was specifically retained by poly(G) homopolymers (Fig. 7B).
Recent studies on FMRP in mammals and fly have identified some novel aspects of its function (40). FMRP directly interacts with other mRNA binding proteins in ribonucleo protein complexes associated with polyribosomes (47,48), in RNA complexes trafficking in neurites (35), in cytoplasmic RNA granules (27) and in the RISC complex in Drosophila (32,33). These findings suggest that several FMRP containing mRNPs complexes may exist and have distinct functions. The systematic identification of FMRP interacting proteins is thus important for the characterization of the composition and function of such mRNA complexes. In the present study we have characterized a novel FMRP-interacting protein, 82-FIP, that has been identified by a two-hybrid screening. We show that 82-FIP and FMRP interact in vitro and in vivo and are both localized in FMRP-containing mRNP complexes associated with polyribosomes. 82-FIP present in poly(A)-mRNP is retained by oligo(dT) chromatography and binds to poly(G) RNA homopolymers, suggesting that it has affinity for mRNA. Since no known motifs for RNA binding have been identified in this protein, we suggest that it contains a novel type of RNA binding domain. Both 82-FIP and NUFIP1 interact with a region in the N-terminal portion of FMRP (between residues 66–134), which defines a novel protein–protein interaction motif in FMRP. This finding suggests that a competition between 82-FIP and NUFIP1 for binding to FMRP might exist, as already proposed for proteins belonging to the FXR and CYFIP families (39). The interaction domain in FMRP bound by 82-FIP and NUFIP1 partially overlaps the non-classical FMRP NLS (22). The possibility then exists that FMRP can enter the nucleus in association to one of these two proteins (or both) and not by binding to importin. Conversely, this interaction can mask the NLS, retaining FMRP in the cytoplasm.
Interestingly, 82-FIP is localized both in the nucleus and in the cytoplasm having a cell cycle-dependent distribution. Cell cycle control depends on a protein-kinase-based machinery comprising two classes of proteins: the cyclin-dependent protein kinases (CdK) and the cyclins. It will be interesting to test whether also 82-FIP is phosphorylated in a cell cycle-dependent manner. The activity of dFMRP, the Drosophila homologue of FMRP, seems to be modified by phosphorylation (49). Other examples exist of proteins whose subcellular localization is regulated by phosphorylation, one being the winged helix transcription factor FKHR1, which is localized in the cytoplasm only after phosphorylation (50).
We have also observed that 82-FIP subcellular distribution is peculiar in brain, being either only cytoplasmic or present in both the nucleus and the cytoplasm, depending on the different areas of the brain. Since neurons are non-cycling cells, 82-FIP subcellular localization in brain is likely to be promoted by a mechanism other than cell cycle regulation.
The fact that FMRP intracellular distribution appears invariant in conditions that affect 82-FIP intracellular localization suggests that 82-FIP/FMRP interaction is specific of a subset of FMRP-containing mRNPs. Among the proteins interacting with FMRP, NUFIP1 has been, up to now, the only one shown to be associated to mRNPs directly in the nuclear compartment (38). The subcellular distribution of 82-FIP and its ability to bind RNA in vitro and to be associated to mRNA complexes suggest that it may represent another link between FMRP and nuclear RNA. The nuclear role of 82-FIP might be linked to RNA metabolism specific to this cellular compartment (maturation, storage, mRNP assembly).
Like the other FMRP interacting proteins, 82-FIP might have a role in the development of the nervous system and in cognitive function. Indeed, FXR2P-null mice show some behavioural phenotypes similar to those observed in Fmr1 knockout mice (51). The dCYFIP Drosophila-null flies show abnormalities in neuronal morphogenesis (41), as also reported in dFMR mutant flies (11,13). Based on these observations and by analogy, we speculate that mutations and defects in FMRP-interacting proteins might also be implicated in forms of mental retardation or cognitive impairment.
MATERIALS AND METHODS
Two-hybrid screening in yeast
The screening and the following liquid β-galactosidase assay were performed in 10 independent measurements, as previously described (37).
Vectors construction and proteins purification
The full-length cDNA was obtained from the Kazusa Institute as KIAA1321 plasmid. This plasmid was digested with XhoI and BglII and the insert subcloned in pTL1 (52) digested with the same restriction enzymes, producing the pTL-82-FIP plasmid.
The chimeric clones FMR1/FXR1 were constructed as previously described (39). FMRP(1–134) fragment was amplified by PCR with engineered NcoI on 5′ ends and NotI on 3′ ends from the complete cDNA. The constructs were cloned in a pET 9-derived plasmid as fusion proteins either with a six histidine tag and/or GST. All clones were expressed in E. coli BL21(DE3). The cells were grown in LB medium with either ampicillin (100 mg/l) or kanamycin (30 mg/l), induced for 3–4 h by addition of 0.5 mM of isopropyl-β-d-thiogalacto-pyranoside (IPTG) when the cultures reached an optical density of 0.8 at 600 nm. The cells were harvested (Beckman centrifuge) and resuspended in lysis buffer (20 mM Tris–HCl at pH 8, 200 mM NaCl, 0.2% IGEPAL CA-630, 2 mM β-mercaptoethanol, lysozyme from Sigma, DnaseI and antiproteases from Boehringer Mannheim), sonicated (Branson sonifier, model 250/450) for 5 min at amplitude 5 and centrifuged at 18 000 rpm for 40 min. The proteins were purified by affinity chromatography (using Ni-NTA gel or glutathione–Sepharose). The purity of the recombinant proteins was checked by SDS–PAGE after each step of purification and by mass spectroscopy of the final products. GST–FMRP was purified as previously described (37).
Antibodies against human 82-FIP
Two synthetic peptides—KHEQKHTLQQHQETPKKKT and LEPSHIGDLQKADTSSQ—corresponding to amino acids 73–92 and 592–609 of human 82-FIP, respectively, were coupled to ovalbumin (Ovalbumin-MBS; Aldrich) and used for immunization of rabbits using standard protocols. The antisera were purified on an affinity column coupled to the same peptide used for immunization, according to the manufacturer's instructions (Sulfolink Coupling Gel-PIERCE). The specificity of purified rabbit antisera was determined by immunoblot analysing total cell extracts from COS cells transfected with the vector encoding 82-FIP and cell extracts from mock-transfected COS, HeLa and 3T3 cells. The two antisera recognize a polypeptide of the same molecular weight in all analysed extracts. 82-FIP expression also appears high in non-transfected COS cells.
Pull-down assays and immunoprecipitation
Full-length and deletion constructs were produced by in vitro transcription–translation in rabbit reticulocyte lysate in the presence of [35S]methionine (ICN), according to the manufacturer's instructions (Promega).
In vitro translated proteins were mixed with 1 µg of GST–FMRP produced in a baculovirus system, 1 µg of GST, 1 µg of His-FMRP(1–134). Pull-down assays were carried out in the following buffer as described (37): [20 mM Tris–HCl, pH 7.5, 50/100/150/200/400 mM NaCl, 5 mM EGTA, 1% TritonX-100, 1mM phenylmethylsulfonyl fluoride (PMSF)].
COS cells transfected with the eukaryotic vector expressing FMRP-ISO7 (16) were lysed in the following buffer: 350 mM NaCl, 20 mM Tris–HCl pH 7.5, 2% TritonX-100, 1 mM PMSF. The immunoprecipitation was carried out in lysis buffer as essentially described (37) and as indicated in the legends to the figures. The proteins bound to the beads were separated by electrophoresis on 10% SDS–polyacrylamide gels. FMRP was visualized by immunoblot using the 1C3 antibody (14), 82-FIP was revealed by the no. 1667 antibody.
Immunofluorescence and immunostaining
Transfection of COS cells, cell fixation and immunodetection with the 1C3 monoclonal antibody were carried out as previously described (16). Polyclonal no. 1666 antibody was diluted 1 : 1000, anti L7a anti-serum (53) 1 : 10 000. Double immunofluorescence experiments were performed by separate and sequential incubations of each primary antibody diluted in phosphate-buffered saline (PBS) at 4°C overnight, followed by the specific secondary antibody coupled to Texas Red or Alexa 488 (Molecular Probes) incubated at room temperature for 1 h.
Fixation of histological sections was performed for 10 min in 50% acetone. After two washes in 100% acetone, the samples were washed in PBS and then post-fixed 10 min in extract 4% paraformaldehyde (PFA). After permeabilization in PBS–0.5% Tween 20 for 20 min, samples were incubated with 10% normal goat serum (NGS) for 1 h. The primary antibody (no. 1666 diluted 1/200) was incubated overnight at room temperature in the following mix: PBS, 1% NGS, 0.5% Tween 20. Samples were washed four times in PBS and then incubated for 2 h with the secondary antibody coupled to a fluorochrome. After washing in PBS, samples were stained with Hoechst 33342 dye.
Samples were analysed with a Leica (Wetzlar, Germany) fluorescence microscope. Confocal images were obtained with a Leica TCS4D microscope. Images were colorized and merged using the Adobe Photoshop software program.
Polyribosomes analyses and oligo(dT) selection of mRNPs
Polyribosomes were concentrated after preparative ultracentrifugation of cytoplasmic extracts through a 45% (w/w) sucrose cushion. The enriched polysomal pellets were resuspended in buffer containing 5 mM MgCl2 or 25 mM EDTA and analysed by sedimentation velocity through linear sucrose gradients as described (47). Polyribosomes from 150 to 500S obtained from sucrose gradients loaded with normal or fragile-X lymphoblastoid cell extracts were concentrated by ultracentrifugation and then resuspended in a 100 mM KCl, 25 mM Tris–HCl, pH 7.4, 25 mM EDTA solution, and aliquots of 1 ml (containing 5 A260 units) were analysed by oligo(dT)-cellulose retention as described (47).
COS cells were synchronized in different phases of cell cycle as previously described (46).
We thank Enzo Lalli for discussion, Andrew Ziemiecki for anti L7a antibody, Takahiro Nagase and Kazusa DNA Research Institute for the KIAA1321 gene. We are grateful to Jochen Barths, Laurent Bianchetti, Marcel Boeglin, Didier Hentsch and Isabelle Kolb-Cheynel for help. This study was supported by funds from the Human Frontier Science Program (RGP0052/2001), NIH (R01 HD40612-01), Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Canadian Institutes of Health Research, by the FRAXA Foundation, by Université Louis Pasteur and by the Hôpital Universitaire de Strasbourg (HUS). M.C. is the recipient of an ‘Allocation de Recherche de l'Ecole Normale Superieure (Paris)’, A.S. was supported by the Ernst Schering Research Foundation and by the Fondation pour la Recherche Médicale and M.-E.H. holds a scholarship from the Canadian Institutes of Health Research.
To whom correspondence should be addressed. Tel: +33 388653414; Fax: +33 388653246; Email: firstname.lastname@example.org