Errata: Corrigendum Volume 274, Issue 15, 4008, Article first published online: 19 July 2007
J. Bonaventure, Institut Curie, CNRS UMR 146, Bat. 110, Université Paris Sud, 91400 Orsay, France Fax: +33 1 69 86 53 01 Tel: +33 1 69 86 71 80 E-mail: email@example.com R. Baron, Department of Cell Biology and Orthopaedics, Yale University School of Medicine, PO Box 208044, New Haven, CT 208044, USA Fax: +1 203 785 2744 Tel: +1 203 785 4150 E-mail: firstname.lastname@example.org
Recurrent missense fibroblast growth factor receptor 3 (FGFR3) mutations have been ascribed to skeletal dysplasias of variable severity including the lethal neonatal thanatophoric dysplasia types I (TDI) and II (TDII). To elucidate the role of activating mutations causing TDI on receptor trafficking and endocytosis, a series of four mutants located in different domains of the receptor were generated and transiently expressed. The putatively elongated X807R receptor was identified as three isoforms. The fully glycosylated mature isoform was constitutively but mildly phosphorylated. Similarly, mutations affecting the extracellular domain (R248C and Y373C) induced moderate constitutive receptor phosphorylation. By contrast, the K650M mutation affecting the tyrosine kinase 2 (TK2) domain produced heavy phosphorylation of the nonglycosylated and mannose-rich isoforms that impaired receptor trafficking through the Golgi network. This resulted in defective expression of the mature isoform at the cell surface. Normal processing was rescued by tyrosine kinase inhibitor treatment. Internalization of the R248C and Y373C mutant receptors, which form stable disulfide-bonded dimers at the cell surface was less efficient than the wild-type, whereas ubiquitylation was markedly increased but apparently independent of the E3 ubiquitin-ligase casitas B-lineage lymphoma (c-Cbl). Constitutive phosphorylation of c-Cbl by the K650M mutant appeared to be related to the intracellular retention of the receptor. Therefore, although mutation K650M affecting the TK2 domain induces defective targeting of the overphosphorylated receptor, a different mechanism characterized by receptor retention at the plasma membrane, excessive ubiquitylation and reduced degradation results from mutations that affect the extracellular domain and the stop codon.
Fibroblast growth factor receptor 3 (FGFR3) belongs to a family of four genes (FGFR1–4) encoding receptors with tyrosine kinase activity (RTK). These structurally related proteins exhibit an extracellular domain (ECD) composed of three immunoglobin-like domains, an acid box, a single transmembrane domain and a split tyrosine kinase (TK) domain. Binding of 1 of the 22 fibroblast growth factor (FGF) ligands in the presence of cell-surface heparan sulfate proteoglycans acting as coreceptors, induces receptor dimerization and trans-autophosphorylation of key tyrosine residues in the cytoplasmic domain. Phosphorylated residues serve as docking sites for the adaptor proteins and effectors that propagate FGFR signals via different signalling pathways resulting in the regulation of many cellular processes including proliferation, differentiation, migration and survival [1–4].
Dominant mutations in three members of the FGFR family (FGFR1–3) have been shown to account for two groups of skeletal disorders, namely short-limb dwarfisms and craniosynostoses [5,6]. Mutations in FGFR3 are mostly responsible for long-bone dysplasias including achondroplasia (ACH), the most common form of dwarfism in humans, the milder form hypochondroplasia and the neonatal lethal form thanatophoric dysplasia (TD) types I and II [7,8]. Interestingly, whereas TDII is exclusively accounted for by a single recurrent K650E missense mutation in the TK2 domain, TDI has been ascribed to a series of mutations creating cysteine residues in the ECD (R248C, S249C, G370C, S371C, Y373C) and to base substitutions eliminating the termination codon (X807R/C/G/S/W) . Likewise, substitution of Lys650 by methionine (K650M) can give rise to TDI [10,11] or to a less severe phenotype called severe achondroplasia with developmental delay and acanthosis nigricans (SADDAN) , whereas replacement of lysine by asparagine or glutamine (K650N/Q) is associated with hypochondroplasia . Based on several in vitro and in vivo studies, FGFR3 mutations have been assumed to induce constitutive activation of the receptor either via a ligand-independent process in TD  or by stabilizing ligand-induced dimers resulting in prolonged signalling at the cell surface in ACH [15,16].
In recent years, numerous efforts have been devoted to elucidate how FGFR3 mutations of the highly conserved Lys650 lead to constitutive receptor phosphorylation and can produce three different phenotypes of increasing severity depending on the substituting amino acid [13,17–23]. However, little attention has been paid to mutations creating unpaired cysteine residues in the ECD and the consequences of the stop codon mutation on receptor function remain unknown. In addition, the mechanisms by which FGFR3 mutants are endocytosed and targeted for degradation to attenuate signalling are far from being elucidated. Thorough analyses of other RTKs such as epidermal growth factor receptor (EGFR) or platelet-derived growth factor receptor (PDGFR) have convincingly shown that these receptors become ubiquitylated through recruitment of the E3 ubiquitin ligase casitas B-lineage lymphoma (c-Cbl) [24–26]. This adaptor protein binds to multiple sites in the intracellular domain of the EGF or PDGF receptors ensuring their monoubiquitylation rather than polyubiquitylation after ligand-induced activation [27,28]. This allows receptor endocytosis and subsequent degradation in the lysosome [27,29]. By contrast, no direct interaction between FGFR3 and c-Cbl  or FGFR1 and c-Cbl  has been detected by coimmunoprecipitation, even though constitutive phosphorylation of c-Cbl in COS-7 cells stably expressing the FGFR3 K650E mutant has been described .
In this study, four FGFR3 mutations causing TDI and affecting the extracellular or intracellular domains of the receptor were generated and used for biochemical and immunocytochemical studies in transiently transfected cells. Mutations creating cysteine residues or disrupting the termination codon had mild effects on receptor phosphorylation and glycosylation, whereas conversion of Lys650 into methionine induced strong constitutive phosphorylation of the native nonglycosylated form of the receptor. Such hyperphosphorylation markedly hampered receptor glycosylation at the Golgi level resulting in reduced levels of fully glycosylated receptors at the cell surface of transfected cells. Reversal of this situation following treatment with the FGFR tyrosine kinase inhibitor SU5402 indicated that hyperphosphorylation adversely affected trafficking of the mutant receptor through the Golgi system. Endocytosis and ubiquitylation of the different TDI mutants were also investigated, as was the putative involvement of c-Cbl in this process. Ubiquitylation of the R248C, Y373C and X807R mutant receptors was stronger than the wild-type and apparently independent of c-Cbl. Constitutive phosphorylation of c-Cbl in cells transiently expressing the K650M mutant was shown to affect Tyr731 which lies outside the ubiquitin-conjugating enzyme-binding RING finger domain that is required for E3 ubiquitin ligase activity [25,26,32].
Our results indicate that receptors are constitutively phosphorylated to variable extents and are differentially processed at the intracellular level depending on the domain in which the mutation arises and the level of phosphorylation. Receptors with mutations in the ECD or stop codon are weakly phosphorylated, retained at the cell surface, and strongly ubiquitylated. By contrast, the highly phosphorylated but moderately ubiquitylated K650M mutant is retained intracellularly and unlike other mutants induces constitutive phosphorylation of c-Cbl which, nonetheless, does not seem to directly regulate FGFR3 ubiquitylation.
A series of four mutants (R248C, Y373C, K650M and X807R) reproducing mutations identified in TDI patients and located in different domains of the receptor (Fig. S1) was created by site-directed mutagenesis of the full-length human FGFR3 cDNA and subcloning into the pcDNA3.1 vector. Based on the cDNA sequence of FGFR3 including the 5′-UTR, the X807R mutation that eliminates the regular stop codon was expected to produce an elongated protein of 947 amino acids and containing a highly hydrophobic domain rich in cysteine  (Fig. S1). An extensive search in databases failed to reveal significant homology of the additional 141 amino acid C-terminal tail with other proteins.
We first tested whether the different mutations causing TDI affected receptor biosynthesis and post-translational processing. Twenty-four hours after transient transfection of 293-VnR cells with the wild-type, R248C and Y373C cDNAs, three isoforms with respective molecular masses of 130, 115 and 105 kDa were visible (Fig. 1A,C). When cells were transfected for 48 h, the relative level of the 105 kDa isoform was slightly reduced (Fig. 1A). Transient expression of the X807R mutation gave rise to three isoforms with higher molecular masses than the wild-type and other mutants, ranging from 144 to 119 kDa, in good agreement with the predicted 141 additional residues separating the regular stop codon from the next inframe stop codon. This additional domain apparently decreased the affinity of the anti-FGFR3 serum for the receptor, so that a higher amount of total protein had to be loaded onto the gel in order to obtain a signal equivalent to wild-type and other mutants (Fig. 1C). The 130 kDa isoform of the K650M mutant was only weakly and variably detected in immunoblots. Scanning densitometry of the gel further indicated that the intensity of the 105 kDa band was greatly increased in this mutant at 24 h post transfection (31% of the total signal in K650M versus 10% in wild-type). Similar results were obtained when the same mutants were transiently transfected in chondrogenic ATDC5 cells (Fig. 1F). In order to confirm that the 130 and 115 kDa bands (or 144 and 129 kDa bands in the X807R mutant) corresponded to differently glycosylated forms of the receptor, immunoprecipitated wild-type and mutant receptors were digested with peptidyl N-glycosidase F (PNGase), which completely eliminates glycosyl groups from N-glycosylated proteins, and endopeptidase H (endo H) which cleaves mannose residues from mannose-rich intermediates. Both the 130 and 115 kDa (or 144 and 129 kDa) bands were converted into the nonglycosylated 105 (or 119) kDa isoform by PNGase treatment (Fig. 1B,D). Endo H specifically eliminated the 115 (or 129) kDa band in the wild-type and mutant receptors (Fig. 1D and not shown), indicating that this band represented a partially processed mannose-rich form of the receptor.
To verify that mutations creating cysteine residues in the ECD of the receptor induced formation of disulfide-bonded dimers, lysates from 293-VnR cells transfected with the Y373C mutant were immunoprecipitated with an anti-FGFR3 serum and separated by electrophoresis under nonreducing and reducing conditions. The Y373C mutant, in the absence of ligand, formed dimeric receptors (260 kDa) that disappeared upon dithiothreitol treatment. As expected, no dimer was visible with the wild-type receptor (Fig. 1E). No dimer was detected in cells transfected with the X807R mutant (data not shown).
The degree of constitutive phosphorylation of the mutant receptor is mutation specific
Because several FGFR3 mutations have been reported to variably induce constitutive phosphorylation of the receptor [13,20,33], the extent of receptor phosphorylation and the relationship with glycosylation in 293-VnR cells was assessed by immunoprecipitation of the receptor and immunoblotting with an anti-phosphotyrosine serum. Both the R248C and Y373C mutants showed moderate phosphorylation of the fully glycosylated isoform (130 kDa) in the absence of ligand, whereas FGF was required to induce phosphorylation of the wild-type receptor (Fig. 2A). By contrast, the 105 kDa nonglycosylated isoform of the K650M mutant, and to a lesser extent the 115 kDa mannose-rich intermediate, were heavily phosphorylated 24 h post transfection, whereas the 130 kDa band was not detectably phosphorylated (Fig. 2B). The identity of the phosphorylated bands was confirmed by PNGase treatment of the immunoprecipitated K650M receptor (Fig. 2D). Forty-eight hours after transfection, phosphorylation of the K650M receptor was significantly reduced, but the 105 kDa band remained preferentially phosphorylated (Fig. 2B). Finally, the X807R mutant showed mild constitutive phosphorylation of the 144 kDa mature isoform (Fig. 2C) indicating that this mutant behaved similarly to receptors with mutations in the ECD.
Immunofluorescent staining of 293-VnR and ATDC5 cells expressing the Y373C mutant with anti-FGFR3 and anti-phosphotyrosine sera showed both intracellular and cell-surface phosphotyrosine staining (Figs 2Eb,c and. A similar pattern was observed with the FGF9-activated wild-type (Fig. 2Ed) and the R248C and X807R mutants (not shown), whereas both 293-VnR and ATDC5 cells expressing the K650M mutant had a round morphology and exhibited strong phosphotyrosine signal in the cytoplasm with no detectable cell surface staining (Figs 2Ee,f and. These results were further supported by labelling the plasma membrane with fluoresceine-conjugated cholera toxin and an anti-FGFR3 serum. Marked colocalization of cholera toxin with wild-type FGFR3 was observed, whereas the K650M mutant showed very little overlap (not shown).
Subcellular distribution of wild-type and mutant FGFR3 molecules
To determine more precisely the subcellular localization of the mutant receptors, cells were stained with anti-(peptidyl disulfide isomerase) (PDI) and anti-GM130, markers of the endoplasmic reticulum (ER) and Golgi system, respectively. Costaining with FGFR3 and PDI showed only partial colocalization of the two proteins in cells transfected with the Y373C, R248C and X807R mutants (Fig. 2Eh,j and not shown). The K650M mutant was much more colocalized with PDI than the other mutants (Fig. 2Ei) suggesting that most of the receptor was present in the ER. Costaining with calnexin (another marker of the ER) and Ptyr antibodies gave similar results (not shown). Colocalization of FGFR3 and the cis-Golgi marker GM130 was mostly visible in cells expressing the wild-type and Y373C mutant and to a lesser extent in those expressing the X807R mutant. There was little colocalization of the K650M mutant and GM130, indicating that transfer of this receptor from the ER to the Golgi compartment was less efficient than that of the wild-type receptor and other mutants. Immunostaining of K650M-transfected cells with GM130 and FGFR3 following fragmentation of the Golgi network into mini-stacks by nocadazole treatment showed colocalization of the two proteins in scattered puncta (Fig. 3Ba), confirming that some K650M FGFR3 molecules were present in the cis-Golgi. By contrast, very little overlap was seen between K650M FGFR3 and the trans-Golgi marker p230 (Fig. 3Bb) suggesting that K650M mutant molecules were inefficiently transferred from the cis- to the trans-Golgi compartments.
Effect of brefeldin A treatment on the processing of wild-type and mutant FGFR3 molecules
To further characterize trafficking of the wild-type and mutant FGFR3 molecules through the Golgi apparatus, cells were treated for 1 h with brefeldin A (BFA), a molecule that reversibly disrupts Golgi assembly by inhibiting anterograde transport from the ER to the Golgi . Western blot analysis with an anti-FGFR3 serum of BFA-treated cells expressing the wild-type or Y373C mutant revealed a significant decrease in the 130 kDa fully glycosylated isoform together with an increase in the 115 kDa isoform (Fig. 3A, left), indicating that glycosylation that normally occurs within the Golgi system was prevented by blocking transport from the ER to the Golgi. BFA had no effect on the relative lack of the 130 kDa isoform of the K650M mutant. Endo H digestion of the immunoprecipitated wild-type and Y373C receptors after BFA treatment revealed a partial conversion of the 115 kDa mannose-rich isoform into an endo H-resistant intermediate form (Fig. 3A, left). This was in keeping with previous reports that BFA treatment induces Golgi enzymes (mannosidase II and thiamine pyrophosphatase) to redistribute into the ER, leading to partially processed endo H-resistant glycosylated proteins [34,35]. Unexpectedly, the phosphorylated 115 kDa band of the K650M mutant was partially resistant to endo H digestion in both untreated and BFA-treated cells (Fig. 3A, right). This suggests that some hyperphosphorylated K650M molecules undergo partial processing at the cis/medial-Golgi level to become endo H resistant without being fully glycosylated in the trans-Golgi compartment, and are either retained in the cis/medial-Golgi compartment or sent back to the ER through retrograde transport. Consistent with this possibility, colocalization of FGFR3 K650M with the cis-Golgi marker GM130 was observed in BFA-treated cells (Fig. 3Be), whereas little overlap was detected with the trans-Golgi marker p230 (Fig. 3Bf).
Cell-surface expression and endocytosis of wild-type and mutant receptors
To investigate whether TDI FGFR3 mutations affected cell-membrane localization of the receptor, total 293-VnR cell-surface proteins were first labelled with NHS-biotin, immunoprecipitated with an anti-FGFR3 serum then separated on nonreducing or reducing gels and blotted with avidin D (Fig. 4A). Although the wild-type receptor showed a single 130-kDa band corresponding to the mature monomer, both the R248C and Y373C mutants showed the presence of a 260-kDa dimer in addition to the monomer. The K650M mutant gave only a faint signal with avidin D, consistent with its intracellular retention. We then examined endocytosis of the wild-type and mutant receptors. Cell-surface proteins were labelled by incubating cells with cleavable sulfo-NHS-S-S-biotin for 30 min on ice . Cells were then warmed to 37 °C for increasing times to allow receptor internalization, and the biotin remaining on the cell surface was stripped by washing with glutathione. Biotinylated cells were lysed, the receptors were immunoprecipitated, and the immune complexes were blotted with avidin D to reveal endocytosed molecules. As expected, no biotinylated FGFR3 molecules (wild-type or mutant) were detected when cells were kept at 4 °C (Fig. 4C and not shown). A substantial amount of the biotinylated receptor (130 kDa) was found after 1 h in the absence of ligand, indicating that wild-type FGFR3 is constitutively endocytosed. The signal reached a peak after 2 h then decreased progressively to become undetectable after 5 h (Fig. 4B). The Y373C mutant gave two bands corresponding to the mature 130 kDa monomer and the disulfide-bonded dimer. Internalization was slower than the wild-type, as attested by the delay in reaching the maximum amount of protected biotinylated receptor and the presence of significant amounts of biotinylated receptor after 6 h. Similar results were obtained with the R248C mutant (not shown). Much less biotinylated K650M mutant was detected at any time point because of the reduced amount of mature receptor at the cell surface (Fig. 4C).
Blocking constitutive receptor phosphorylation restores normal maturation and distribution of the K650M mutant
The kinase activity of FGFRs, including FGFR3 [37,38], is inhibited by SU5402, which binds to the kinases' ATP-binding site . We therefore determined whether SU5402 prevented constitutive phosphorylation of FGFR3 mutants, and if so, whether inhibiting receptor phosphorylation altered trafficking of the mutant receptors between different membrane compartments. Cells expressing the Y373C or K650M mutants were treated with different doses of SU5402 for increasing periods. A 25 µm concentration for 16 h was sufficient to totally abolish receptor phosphorylation in cells expressing the Y373C mutant (not shown). Phosphorylation of the K650M mutant, although dramatically reduced, was not completely abrogated (Fig. 5A,B). Increased inhibitor concentrations had no further effect on phosphorylation but affected cell viability (not shown). Immunoblot analysis of the wild-type and K650M mutant receptors following SU5402 treatment and immunoprecipitation with an anti-FGFR3 serum showed the presence of the mature 130 kDa isoform both in the wild-type and mutant (Fig. 5A), indicating that inhibiting the constitutive phosphorylation restored full maturation of the K650M receptor to a significant degree. To firmly establish that SU5402 allowed the K650M receptor to be transported to the plasma membrane and endocytosed, sulfobiotinylation of the mutant receptor with cleavable sulfobiotin was performed after SU5402 treatment. Large amounts of endocytosed receptors were detected after 2–3 h confirming the ability of the mutant receptor to traffic efficiently to the cell surface and be internalized with a kinetic resembling that of the wild-type receptor when hyperphosphorylation was prevented (Fig. 5B).
Excessive ubiquitylation of mutant receptors
Internalized Rtk are usually committed to degradation through ubiquitylation of lysine residues. We therefore studied ubiquitylation of wild-type and mutant receptors by cotransfecting cells with wild-type or mutant FGFR3 and HA-tagged ubiquitin cDNAs. Ubiquitylated receptors identified by blotting with anti-ubiquitin sera appeared as a smear of bands with a lower mobility than the nonubiquitylated receptors. The Y373C mutant gave a stronger signal than the wild-type and the intensity was increased slightly in both cases by treatment with the proteasome inhibitor MG132 (Fig. 6A) indicating that partial degradation of the receptor could occur at the proteasome level. We also analysed ubiquitylation of the Y373C and K650M mutants both in the presence and absence of chloroquine, a lysosomal inhibitor. Unlike Y373C, the K650M mutant was less ubiquitylated than the wild-type receptor and the amounts of ubiquitylated wild-type and mutant FGFR3 were slightly increased by chloroquine treatment (Fig. 6B), suggesting that the lysosomal pathway may also participate to their degradation. The X807R mutant also exhibited an increased ubiquitylation compared with wild-type (not shown) confirming that ubiquitylation levels of the weakly phosphorylated TDI mutant receptors (R248C, Y373C and X807R) were higher than the wild-type. By contrast, the heavily phosphorylated K650M mutant was less ubiquitylated than the wild-type, consistent with its poor expression at the cell surface.
c-Cbl does not mediate the ubiquitylation of FGFR3, but it is constitutively phosphorylated by the K650M mutant
c-Cbl is an adaptor protein and an E3-ubiquitin ligase that is phosphorylated downstream of several growth factor receptors and contributes to their downregulation by mediating their ubiquitylation , suggesting that it may be involved in the ubiquitylation of FGFR3 and/or be phosphorylated by FGFR3 in a basal or ligand-dependent process . We therefore first examined whether c-Cbl might mediate the ubiquitylation of the TDI FGFR3 mutants. Overexpression of c-Cbl with wild-type (stimulated by FGF9) or Y373C mutant FGFR3 did not significantly affect receptor ubiquitylation (Fig. 6C), and the ubiquitinylation of wild-type, Y373C and K650M FGFR3 mutants was not significantly different when either c-Cbl or the oncogenic mutant 70Z-Cbl, which lacks E3-ligase activity and dominant-negatively inhibits ligand-induced EGFR ubiquitylation , were coexpressed with the receptors (Fig. 6D). Consistent with the absence of an effect of c-Cbl or 70Z-Cbl on the ubiquitylation of FGFR3 receptors, myc-tagged c-Cbl failed to coimmunoprecipitate with wild-type FGFR3 (treated or not by FGF9) and FGFR3 mutants (and not shown), indicating that in our cell system, c-Cbl apparently does not directly interact with wild-type FGFR3 or the TDI FGFR3 mutants.
To determine if c-Cbl is phosphorylated downstream of wild-type or mutated FGFR3, we examined lysates from 293-VnR cells coexpressing c-Cbl and wild-type or mutant FGFR3 using immunoblotting or immunofluorescence analysis with an anti-phosphotyrosine serum or an antibody against phospho-Tyr731, a c-Cbl tyrosine residue that is phosphorylated downstream of several receptor and nonreceptor TKs to form a binding site for phosphatidylinositol 3-kinase. No phosphorylation of c-Cbl was seen in cells that expressed the wild-type receptor, the Y373C or the X807R mutant receptors (Figs 7A,B and. Stimulation of the wild-type receptor with FGF9 failed to induce c-Cbl phosphorylation. By contrast, marked c-Cbl tyrosine phosphorylation occurred in cells expressing the K650M mutant (Figs 7A and. Tyrosine 731 was one of the residues phosphorylated in the K650M-expressing cells (Figs 7B and. Cbl phosphorylation in the K650M-expressing cells was not detectably affected by deleting (70Z-Cbl) or mutating (c-CblY371F) Tyr371 (Fig. 7A,C), whose phosphorylation is required for ubiquitylation [32,41]. In fact, phosphorylation of 70Z-Cbl appeared slightly higher than the wild-type c-Cbl. This suggests either that multiple tyrosines in addition to Tyr371 are phosphorylated downstream of FGFR3 K650M or that Tyr371 is not a major site of phosphorylation.
In this study, the effects of TDI-inducing missense mutations on receptor processing, endocytosis and ubiquitylation were investigated by using transiently transfected 293-VnR and ATDC5 cells. Although primary cultured chondrocytes from affected patients would be representative of a more physiological model, the difficulty of efficiently transfecting human chondrocytes and maintaining their differentiated phenotype prompted us to use established cell lines, keeping in mind that overexpression of the receptor in transiently transfected cells may affect their physiological properties. We first demonstrated that replacement of the stop codon by an arginine residue resulted in a stable elongated receptor, which appeared on western blotting as a combination of three bands including the nonglycosylated, mannose-rich and fully glycosylated isoforms, indicating that this elongated receptor underwent the same maturation process as the Y373C and R248C mutants. However, under nonreducing conditions, these two mutants with an additional cysteine in the ECD gave rise to a disulfide-bonded mutant dimer, thus confirming constitutive activation of the receptor . Consistent with previous studies [13,17,20,23], we found that substitution of Lys650 by methionine resulted in a different electrophoretic pattern characterized by a variable but marked reduction in the fully glycosylated isoform and a significant increase in the nonglycosylated and partially glycosylated isoforms. This defective maturation of the receptor resulted in inefficient targeting to the plasma membrane and strong constitutive tyrosine phosphorylation of the nonglycosylated isoform. Similar observations have been reported previously in PC12 cells expressing K650E and K650M chimeric receptors . Inhibition of receptor phosphorylation with SU5402 restored proper receptor maturation and trafficking to the cell surface, suggesting that intracellular retention was a direct consequence of receptor hyperphosphorylation. Support for this hypothesis is provided by the report that eliminating constitutive mouse Fgfr3 phosphorylation by mutating the mechanistically critical Tyr718 in the Fgfr3 activation loop restores normal Fgfr3 receptor maturation . However, we cannot exclude that abnormal constitutive phosphorylation of proteins involved in the trafficking of the receptor, including c-Cbl, could account for its intracellular retention. By contrast, TDI mutations in the ECD or disruption of the termination codon induced a much lower level of phosphorylation of only the fully glycosylated isoform, which did not hamper its maturation, suggesting that factors other than constitutive FGFR3 autophosphorylation are involved in the severity of mutant-associated skeletal disorders.
It is noteworthy that tyrosine phosphorylation of at least four members of the RTK family (e.g. Kit, PDGFRβ, Ros and FLT-3) has been recently reported to lead to defective expression of the mature receptors at the cell surface . Although mechanisms regulating maturation arrest of phosphorylated receptors have not been clearly elucidated, our coimmunolocalization studies pointed to a role for components of the ER–Golgi vesicle transport. Through the use of markers for the ER (PDI) and the Golgi apparatus (GM130, p230), the phosphorylated isoforms of the K650M mutant were identified in both the ER and cis-Golgi compartments but were hardly detectable in the trans-Golgi. These observations differ from those of Lievens et al.  who concluded that mouse mutant K644E/M molecules were trapped in the ER. Disrupting the Golgi apparatus with BFA or nocodazole provided evidence that at least some of the mutant receptors were transported to the Golgi. Nocodazole induces reversible scattering of the juxtanuclear Golgi to peripheral sites via microtubule depolymerization . Colocalization of mutant K650M molecules with the Golgi marker GM130 in mini-stacks dispersed throughout the cytosol indicated that these molecules had reached the cis/medial-Golgi compartment. This conclusion was further supported by the demonstration that mannose-rich K650M receptors showed partial resistance to endo H treatment in the absence as well as in the presence of BFA, whereas other FGFR3 mutants exhibited resistance to endo-H digestion only after BFA treatment. Our observation is consistent with the previous demonstration that cis-Golgi oligosaccharide-modifying enzymes (mannosidase II and thiamine pyrophosphatase) undergo retrograde transport to the ER after BFA treatment . We propose that in the absence of BFA, some heavily phosphorylated K650M molecules were able to reach the cis/medial-Golgi compartment where they were partially processed into endo H-resistant molecules. However, they failed to be efficiently routed to the trans-Golgi network, as documented by the poor colocalization with p230. These molecules were finally recycled to the ER through retrograde transport in a manner similar to Golgi-resident glycosylation enzymes involved in the modification of transiting proteins [43,44]. Direct evidence of defective processing and trafficking of the K650M mutant was provided by labelling wild-type and mutant receptors with membrane-impermeant NHS-biotin. Reduced amounts of the biotinylated K650M mutant receptor were found, whereas the Y373C and R248C mutants were biotinylated at levels similar to the wild-type and formed stable dimers, thus confirming that disulfide-bonded receptors were properly processed and expressed at the cell surface. Whether disulfide bonding between two mutant receptors occurred intracellularly or at the plasma membrane remains to be elucidated.
Analysis of receptor endocytosis through the use of cleavable biotin indicated that internalization of disulfide-bonded mutant receptors was slower than the wild-type. A small amount of the biotinylated K650M mutant was detected, in keeping with its defective expression at the cell surface. Treatment with SU5402 was able to at least partially restore trafficking of the K650M mutant receptor to the cell surface and its subsequent endocytosis. Retention of the disulfide-bonded dimers at the cell surface was indicative of defective receptor internalization, allowing ligand-independent prolonged signalling to target molecules.
Mechanisms that control receptor endocytosis are multiple and complex . Ubiquitylation is considered one of the critical signals for endocytosis and degradation in the lysosome or the proteasome [27,46]. Consistent with data from Monsonego-Ornan et al.  on the G380R ACH mutant, ubiquitylation of the TDI mutants (R248C, Y373C and X807R) was found to be higher than wild-type, but these results differed from those of Cho et al.  who reported reduced ubiquitylation of the ACH mutant in stably transfected cells. Discrepancies between these studies may be due to the two different cell types (HEK293 versus COS-7 cells) and the use of retroviruses for stable transfection of cDNA constructs versus transient transfection of plasmids. Polyubiquitylation is responsible for the internalization and proteasomal degradation of several plasma membrane proteins , but monoubiquitylation has been recently identified as the main mechanism regulating RTK endocytosis and degradation [27,28] and is associated in yeast with proteasome-independent functions including protein trafficking . Although it is not known whether FGFR3 mutants are monoubiquitylated, polyubiquitylated or both, it is tempting to speculate that highly ubiquitylated R248C, Y373C and X807R receptors could be preferentially monoubiquitylated on their 26 lysine residues lying in the intracellular domain  and transferred to early endosomes. From this compartment, part of the ubiquitylated mutant molecules could be sorted for degradation with a lesser efficiency than moderately ubiquitylated wild-type receptors, whereas a higher number of mutant molecules than wild-type would be recycled back to the plasma membrane.
The E3-ubiquitin ligase c-Cbl is directly involved in the ubiquitylation of several RTKs [24–26,32] and may participate in the downregulation of FGFR1 via an indirect interaction with the phosphorylated docking protein FRS2α[3,31]; but definitive evidence that c-Cbl is responsible for ubiquitylation of FGFR3 is still missing. We conclude that c-Cbl does not play a key role in the ubiquitylation process of TDI FGFR3 mutants in our cell system because: (a) the extent of ubiquitylation of wild-type and TDI FGFR3 mutants was similarly unaffected by cotransfecting c-Cbl or the dominant-negative ubiquitylation-deficient 70Z-Cbl (Fig. 6C,D); and (b) c-Cbl failed to coimmunoprecipitate with wild-type and TDI FGFR3 mutants, consistent with previous observations on ACH and TDII mutants . However, the possible involvement of the adaptor proteins FRS2 and Grb2 in the ubiquitinylation process cannot be excluded [31,40]. Alternatively, other E3 ubiquitin ligases such as the von Hippel–Lindau protein, which regulates surface localization of FGFR1 , might be involved in FGFR3 ubiquitylation.
Phosphorylation of Tyr731, one of several phosphorylated tyrosine residues located in the C-terminal half of c-Cbl, most likely resulted from intracellular retention of the K650M FGFR3 mutant, even though it did not involve a direct interaction between the two proteins. Because several Src-like kinases including Src, Fyn and Yes have been shown to phosphorylate c-Cbl on Tyr731 [50–52], we hypothesized that c-Cbl phosphorylation would be mediated via a tripartite complex involving K650M FGFR3 and a Src-like kinase. The observation that c-Cbl was able to interact with FGFR2 and Fyn or Lyn in osteoblastic cells  and the demonstration, using a phosphoproteomic approach, that FGFR1, when phosphorylated, induced phosphorylation of both Cbl-b and Fyn , are consistent with this hypothesis. Hence, unlike other TDI mutants, the K650M mutant could elicit signalling via an alternative internal pathway involving c-Cbl and a Src-like kinase.
Taken together, the data reported here provide evidence that TDI is caused by mutations affecting the receptor in at least two different ways. Conversion of Lys650 into methionine in the TK2 domain induces hyperphosphorylation and marked intracellular retention of the mutant receptor leading to phosphorylation of target signalling molecules including c-Cbl. By contrast, mutations creating cysteine residues in the ECD or elongating the receptor result in delayed endocytosis, excessive ubiquitylation and reduced degradation of the mutant proteins, but have lower impact on FGFR3 phosphorylation.
DNA constructs and plasmids
Full-length wild-type human FGFR3 cDNA cloned into pLNCX was kindly provided by M. Hayman (State University, New York, NY) and subcloned into pBSII. Two different strategies were used to obtain point mutations in different subdomains of the receptor. Total RNA extracted from cultured cells of TDI patients carrying the R248C or Y373C mutations were reverse transcribed with two different set of primers. RT-PCR products of the mutant allele were cloned into TOPO TA cloning vector (Invitrogen, Carlsbad, CA) then digested with RsrII and PmlI (for R248C) or with PmlI and MluI (for Y373C). DNA fragments were subcloned into the FGFR3 pBSII vector at the RsrII/PmlI sites or PmlI/MluI sites. Wild-type and mutant FGFR3 cDNAs were then transferred from pBSII to pcDNA3.1 at the HindIII/EcoRI restriction sites.
Single-point mutations in the intracellular domain, namely K650M and X807R were generated by site-directed mutagenesis (Quick Change® site-directed mutagenesis, Stratagene, La Jolla, CA) according to the manufacturer's instructions. Sequences of the primers used for mutagenesis are shown in. Mutagenesis for the K650M mutant was performed on the BsaBI/SphI fragment of FGFR3 in pBSII. For X807R mutagenesis, the SpeI/SphI fragment of FGFR3 in pBSII was used. The mutant FGFR3 cDNA was then transferred to pCDNA3.1. The presence of mutations was confirmed by sequencing on an ABI prism 3100 (Applied Biosystems, Foster City, CA).
Generation of plasmids containing full-length myc-tagged c-Cbl and c-Cbl mutants (c-70Z-Cbl and c-CblY371F) has been described previously [41,55].
Cell lines and transfection
Human embryonic kidney cells stably expressing the vitronectin receptor (293-VnR) were cultured in DMEM supplemented with 10% fetal bovine serum and antibiotics. These cells rather than HEK293 cells were used as they attach more tightly to plastic surfaces. The patterns of expression and post-translational processing of wild-type and mutant FGFR3, determined by western blot, were comparable in the two cell lines, indicating that the presence of elevated levels of the VnR did not affect the pathways studied in these experiments. ATDC5 cells were cultured in a 1 : 1 mixture of DMEM and Ham's F12 medium containing 5% fetal bovine serum, insulin (10 µg·mL−1), ferritin (10 µg·mL−1), selenium (1 ng·mL−1) and antibiotics. Cells at 60% confluency were transiently transfected with wild-type or mutant FGFR3 cDNAs in the presence of Fugene 6 (Roche, Indianapolis, IN) according to the manufacturer's instructions. Cells were collected after 24 or 48 h. In some experiments, BFA (Epicentre Technologies, Madison, WI) was added to transfected cells for 1 h at a final concentration of 5 µg·mL−1.
The tyrosine kinase inhibitor SU5402 (a gift from G. McMahon, SUGEN, San Francisco, CA) was dissolved in dimethylsulfoxide and added to transfected cells for 16 h at a final concentration of 25 µm. Control cells were incubated with dimethylsulfoxide alone at a final concentration of 1%. Nocodazole treatment (10 µg·mL−1) was performed 24 h post transfection, for 2 h at 37 °C.
Immunoblotting and immunoprecipitation
Transfected cells were washed in NaCl/Pi and lysed in radioimmune precipitation assay buffer (50 mm Tris HCl pH 7.6, 150 mm NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, 1 mg·mL−1 pepstatin A, 1 mg·mL−1 leupeptin, 1 mg·mL−1 aprotinin, 2 mm phenylmethanesulfonyl fluoride, 1 mg·mL−1 sodium orthovanadate), then clarified by centrifugation for 30 min at 12 000 g. Aliquots of lysates were reserved for immunoblotting and the rest of the lysates were immunoprecipitated for 4 h at 4 °C with an anti-FGFR3 serum raised against the cytoplasmic domain (Sigma, St Louis, MO). Immune complexes were bound to Protein G agarose beads and washed three times with radioimmune precipitation assay buffer, then heated at 95 °C for 10 min in 4× loading buffer (Invitrogen). Total cell lysates or immunoprecipitates were resolved by electrophoresis on 4–12% gradient NU-PAGE gels (Invitrogen). Proteins were transferred to poly(vinylidene) difluoride membranes (Immobilon, Millipore, Bedford, MA), incubated with primary antibodies followed by horseradish peroxidase (HRP)-conjugated secondary antibodies and the bands detected by enhanced chemiluminescence (Amersham Pharmacia Biotech, Piscataway, NJ).
The following primary antibodies were used for immunoprecipitation and immunoblotting: rabbit anti-FGFR3 (Sigma), mouse anti-(phosphotyrosine P-Tyr-102) (Cell Signaling Technology, Beverly, MA); mouse anti-myc from 9E10 hybridoma (Roche Molecular Biochemicals); mouse anti-Cbl, mouse anti-GM130 and mouse anti-p230 (BD Biosciences, Franklin Lakes, NJ), rabbit anti-(phospho-Cbl tyrosine 731) (Cell Signaling), mouse anti-(peptidyl disulfide isomerase) (Affinity Bioreagents, Golden, CO) and mouse anti-ubiquitin (Chemicon, Temecula, CA).
Deglycosylation of FGFR3 isoforms
FGFR3 was immunoprecipitated from cell lysates using anti-FGFR3 serum. Immune complexes bound to protein G–agarose beads were resuspended in 50 mm sodium citrate, pH 5.5, supplemented with 1% SDS and 1%β-mercaptoethanol and heated for 10 min at 95 °C. Endo H (Roche) was added at a final concentration of 50 mU and the mixture was incubated at 37 °C for 2 h. Peptidyl N-glycosidase F (PNGase F) treatment was achieved by diluting 1 vol. of the sodium citrate/SDS/β-mercaptoethanol solution with 1 vol. of sodium citrate 50 mm, pH 5.5, containing 1% NP-40. Then 5 U of PNGase F solution (Roche) were added followed by incubation for 2 h at 37 °C. Enzymatic activities were blocked by adding 4× loading buffer.
293-VnR cells were seeded in Labtek chambers (BD Bio-sciences) at a density of 15 000 cells·well−1. Cells were allowed to reach 60% confluency, then transfected with wild-type or mutant FGFR3 cDNAs using Fugene 6 (0.5 µL·well−1). After 24 h, cells were fixed with 4% paraformaldehyde, permeabilized for 15 min with 0.1% Triton X-100 in NaCl/Pi and incubated for 30 min with 10% sheep serum in NaCl/Pi. The following sera were used for immunostaining: rabbit anti-FGFR3 (1 : 400), mouse anti-(phosphotyrosine P-Tyr102) (1 : 200), mouse anti-GM130 (1 : 100), mouse anti-p230 (1 : 100), mouse anti-(peptidyl disulfide isomerase) (1 : 100). Appropriate second sera: anti-(rabbit Alexa fluor green 458), anti-(mouse Alexa fluor red 561) (Molecular Probes, Eugene, OR) were added at a 1/400 dilution and incubated at room temperature for 2 h. 4′,6-Diamidino-2-phenylindole was used for nuclear counterstaining. Glass slides were mounted and photographed using an inverted Olympus microscope.
293-VnR cells transiently transfected with wild-type or mutant FGFR3 cDNAs were washed twice with cold NaCl/Pi then incubated at 4 °C for 30 min with either NHS-biotin or cleavable sulfo-NHS-S-S-biotin (Uptima, Montluçon, France) at a 0.5 mg·mL−1 concentration in NaCl/Pi. Coupling of NHS-biotin was blocked by washing with 15 mm glycine in NaCl/Pi. When cells were treated with sulfo-NHS-S-S-biotin, a previously described procedure for analysis of endocytosis was used . Briefly, after biotinylation of cell surface proteins, excess biotin was quenched by incubating cells for 10 min with 50 mm Tris/HCl, pH 7.5 at 4 °C. Cells were re-incubated at 37 °C in fresh DMEM for various times (0–6 h) to allow receptor endocytosis. Biotin was then cleaved from proteins on the cell surface by washing with 50 mm glutathione, 75 mm NaCl, 75 mm NaOH, 10% fetal bovine serum. Cells were then washed with 50 mm iodoacetamide in 1% BSA to quench residual glutathione. Cells were lysed with RIPA buffer and lysates were immunoprecipitated with anti-FGFR3 sera. Precipitated proteins were separated on NuPAGE gels under nonreducing conditions, transferred to poly(vinylidene difluoride) membranes and probed with HRP-conjugated avidin D (Vector Laboratories, Burlingame, CA).
293-VnR cells were cotransfected with wild-type or mutant FGFR3 cDNAs and HA-tagged-ubiquitin cDNA (a gift of D. Bohmann, Rochester, NY). In some experiments cells were also cotransfected with c-Cbl or the mutant 70Z-Cbl. At 24 h post transfection, cells were treated for 1 h with the proteasome inhibitor MG132 (Biomol Research Laboratories, Plymouth Meeting, PA) at a final concentration of 50 µm in 0.1% dimethylsulfoxide or with the lysosome inhibitor chloroquine (Sigma) at a final concentration of 500 µm. Cell lysates were immunoprecipitated with anti-FGFR3 or anti-HA sera (Sigma) and analysed by immunoblotting with anti-HA, anti-ubiquitin or anti-FGFR3 sera. Treatment with 10 µg·mL−1 cycloheximide for 1 h followed by incubation in fresh cyclohexamide-free medium was performed when required to block protein synthesis.
We are grateful to Dr M. Hayman (State University, New York, NY) and Dr D. Bohmann (University of Rochester, NY) for providing plasmids and to Dr G. McMahon (SUGEN, San Francisco, CA) for providing the SU5402 TK inhibitor. We thank Dr Archana Sanjay for helpful suggestions. Part of this work was supported by the European Skeletal Dysplasia Network (grant QLG1-CT-2001-02188) and by the Philip Foundation.