Different intracellular trafficking of FGF1 endocytosed by the four homologous FGF receptors

The large family of fibroblast growth factors (FGFs) comprises in humans 22 structurally related heparin binding polypeptides which are involved in the regulation of many key cellular processes (Powers et al., 2000). The FGFs mediate their biological effects through binding to high-affinity cell-surface receptors, FGFRs. The FGFR family constitutes a variety of polypeptides encoded by four related genes (Johnson and Williams, 1993; Ornitz and Itoh, 2001). The receptors share common structural features and consist of an extracellular ligand binding domain, a transmembrane domain and a cytoplasmic region. The extracellular domain contains a unique acidic region and two or three immunoglobulin-like domains (D1-D3), dependent on alternative splicing. The cytoplasmic region contains a split tyrosine kinase domain (Johnson and Williams, 1993).

Binding of FGFs to FGF receptors (FGFRs) is stabilized by heparan sulfate proteoglycans and results in a dimer receptor-ligand complex that activates the intracellular tyrosine kinase domain by autophosphorylation. The autophosphorylation triggers the transient assembly of a large intracellular complex, which activates downstream signalling pathways such as phospholipase C-γ/protein kinase C, phosphoinositide 3-kinase/Akt and Ras/MAPK (mitogen-activated protein kinase) (Klint and Claesson-Welsh, 1999; Schlessinger, 2004). Depending on the target cell type, FGF signalling can induce cell proliferation, differentiation, survival and motility (Basilico and Moscatelli, 1992).

Signalling from activated transmembrane receptors is attenuated by degradation in lysosomes. Lysosomal targeting of tyrosine kinase receptors is best illustrated for the epidermal growth factor receptor (EGFR) and involves the attachment of ubiquitin to lysine residues in the cytoplasmic tail of the activated receptor (Levkowitz et al., 1998). Upon internalization, the receptors appear in early/sorting endosomes where the receptors destined for degradation in the lysosomes become ubiquitylated. As a result, they are recognized by Hrs and the ESCRT (endosomal sorting complex required for transport) complexes and internalized into the endosomes by membrane invagination (Raiborg et al., 2003). Multivesicular bodies fuse with late endosomes and the endocytosed material is then sorted to lysosomes for degradation.

Receptors that are not retained in the sorting endosomes recycle either directly or via the endocytic recycling compartment, ERC, back to the cell surface. Most receptors known to recycle possess no signalling activity and are often associated with uptake of nutrients. The transferrin receptor, TfR, is known to recycle via the ERC and is often used as a marker for the recycling endocytic pathway (Yamashiro and Maxfield, 1984). The importance of ubiquitin as a signal for lysosomal sorting is illustrated by experiments where transferrin receptors were fused to ubiquitin and found to be sorted into the degradative pathway (Raiborg et al., 2002).

From what is known about the endocytosis of the different FGFRs, it appears that they may utilize different mechanisms for internalization and that this also may vary between different cell types (Gleizes et al., 1996; Marchese et al., 1998; Citores et al., 1999; Citores et al., 2001; Reilly et al., 2004). However, irrespective of the mechanism of endocytosis, FGF/FGFR complexes have been observed in early endosomes/sorting endosomes approximately 10 minutes after internalization (Gleizes et al., 1996; Citores et al., 1999; Belleudi et al., 2002). Subsequent to their presence in sorting endosomes, KGF (FGF7) and KGFR (a splicing variant of FGFR2), were found to be sorted to late endosomes in HeLa cells (Belleudi et al., 2002). FGF2 has also been observed in late endosomes and lysosomes in BHK cells (Gleizes et al., 1996). However, FGF1/FGFR4 in COS cells were found to accumulate in intracellular structures identified as the recycling compartment (Citores et al., 1999). It was found that binding of FGF to FGFR1 and FGFR3 induces ubiquitylation of the receptors and that this contributes to their downregulation (Mori et al., 1995; Monsonego-Ornan et al., 2002; Wong et al., 2002; Cho et al., 2004). Activated FGFR3 has recently been reported to be targeted for lysosomal degradation through c-Cbl-mediated ubiquitylation, whereas FGFR3 harbouring mutations associated with skeletal disorders were found to be less ubiquitylated and escape lysosomal targeting (Cho et al., 2004).

To compare the intracellular fate of FGF internalized by the four related tyrosine kinase FGFRs, HeLa cells transfected with any one of the four FGFRs were chosen as a model system, and FGF1, which binds equally well to the four FGFRs (Ornitz et al., 1996), was used as a ligand. The present work shows that FGF1 endocytosed by the four receptors is indeed sorted differently and that different levels of ubiquitylation appear to be the molecular mechanism responsible for the different sorting.

Rabbit anti-FGFR1, anti-FGFR2, anti-FGFR3 and anti-FGFR4 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit antibodies against tyrosine 653/654 phosphorylated FGFRs were from Cell Signalling (Beverly, MA). Mouse anti-EEA1 antibodies were obtained from Transduction laboratories (Lexington, KY) and mouse anti-LAMP-1 antibodies were from Developmental Studies Hybridoma Bank (Iowa City, IA). Mouse anti-myc antibodies were from 9E10 hybridoma (Evan et al., 1985). Mouse anti-TfR antibodies were from Boehringer Mannheim (Mannheim, Germany). The secondary antibodies Cy2-conjugated anti-mouse IgG, Cy2-conjugated anti-rabbit IgG, HRP-conjugated anti-mouse IgG and HRP-conjugated anti-rabbit IgG were from Jackson Immuno-Research Laboratories (West Grove, PA).

Cy3-maleimide, heparin-Sepharose, streptavidin-Sepharose, ECL plus western blotting system were from Amersham Biosciences (Buckinghamshire, UK). APS, TEMED, 40% acrylamide/bisacrylamide and Restore Western Blot Stripping Buffer were from BIO-RAD (Hercules, CA). Fugene 6 was from Boehringer Mannheim (Indianapolis, IN). Dulbecco's Modified Eagle Medium (DMEM), streptomycin and penicillin were from GIBCO, Invitrogen (Carlsbad, CA). Alexa 488 EGF and Alexa 647 Transferrin were from Molecular Probes, Invitrogen. Restriction enzymes were from New England Biolabs, (Beverly, MA). Mowiol was from Novabiochem Corporation (La Jolla, CA). Fetal calf serum was from PAA Laboratories GmbH (Linz, Austria). Easytag Methionine L-[³⁵S] was from Perkin Elmer (Boston, MA). Leupeptin to use on live cells was from Peptide Institute (Osaka, Japan). Ez-link sulfo-NHS-LC-Biotin was from Pierce (Rockford, IL). Rabbit Reticulocyte Lysate System was obtained from Promega Corporation (Madison, WI). Complete EDTA free protease inhibitor cocktail tablets were from Roche Diagnostics (Penzberg, Germany). T3 RNA polymerase and T7 RNA polymerase were from Stratagene (La Jolla, CA). Other chemicals were from Sigma-Aldrich (St Louis, MO). Alexa 488-labelled FGF2 was a generous gift from Jedrzej Malecki, this Institute. [¹²⁵I]FGF1 was a generous gift from Malgorzata Zakrzewska, Institute of Biochemistry and Molecular Biology, University of Wroclaw.

pcDNA3-hFGFR1: cDNA encoding hFGFR1 IIIc was cut out from pSV7d (Wennstrom et al., 1991) with EcoRI and XbaI in two fragments, and ligated into pcDNA3 (Invitrogen, Carlsbad, CA) cut with the same enzymes. pcDNA3-hFGFR2: cDNA encoding hFGFR2 IIIc lacking D1 was cut out from pBluescript (RZPD, Berlin, Germany, Clone ID: IMAGp998N0911701Q3) with NotI and SpeI, and ligated into pcDNA3 cut with NotI and XbaI. The pcDNA3-hFGFR3 IIIc construct was a generous gift from Avner Yayon, ProChon Biotech Ltd, Israel (Adar et al., 2002). The pcDNA3-hFGFR4 and the pB-FGF1 construct has been described previously (Wesche et al., 2000; Klingenberg et al., 2000). The pTriEX-2-FGF1 construct was a generous gift from Camilla Skiple Skjerpen, this Institute, and the pcDNA3-myc-tagged-ubiquitin was a generous gift from Harald Stenmark, this Institute.

HeLa cells were propagated in DMEM, supplemented with 10% (vol/vol) fetal calf serum and 100 U/ml penicillin and 100 μg/ml streptomycin in a 5% CO₂ atmosphere at 37°C.

Transient expression of the different receptors was performed by transfecting HeLa cells with the plasmid DNA (pcDNA3 with appropriate inserts) by using Fugene 6 transfection reagent according to the manufacturer's protocol. Cells were seeded into plates (BD Biosciences, San Jose, CA and Nalge Nunc International, Rochester, NY) the day preceding the transfection and experiments were performed 15-24 hours after transfection.

FGF1 was labelled with Cy3-maleimide according to the manufacturer's protocol. Not transfected or transiently transfected HeLa cells grown on coverslips at 37°C were incubated with Cy3-FGF1 for 2 hours in HEPES medium at 4°C in the presence of 50 U/ml heparin. The cells were then washed three times in PBS and incubated for different periods of time in DMEM with 0.3 mM leupeptin at 37°C. The cells were fixed in 3% paraformaldehyde in PBS for 15 minutes, washed three times in PBS and mounted in Mowiol. In some cases the cells were in addition to Cy3-FGF1 incubated with Alexa 488 EGF and Alexa 647 transferrin in the presence of 50 U/ml heparin and 0.3 mM leupeptin. When antibodies were used to visualize structures within the cell, the fixation was quenched with 50 mM NH₄Cl in PBS for 15 minutes and the cells were permeabilized with 0. 05% saponin in PBS for 5 minutes. The cells were then incubated with primary antibody in 0. 05% saponin in PBS for 20 minutes, washed three times in 0. 05% saponin in PBS and incubated for additional 20 minutes with the secondary antibody coupled to a fluorophore. After washing once in 0. 05% saponin and twice in PBS, the cells were mounted in Mowiol and examined with a Zeiss LSM 510 META confocal microscope (Zeiss, Jena, Germany). Images were prepared with Adobe Photoshop 7. 0 (Adobe, San Jose, CA) and Zeiss LSM Image Browser (Version 3).

Images of transfected, randomly chosen cells were divided into squares, and every fifth square within the chosen cell was examined. Red structures indicating internalized Cy3-labelled FGF1 were compared with structures of the different markers and the proportion of red structures that colocalized with structures of the specific marker was calculated. The mean and s.d. were calculated from 15 cells in each case.

HeLa cells not transfected or transiently transfected with FGFR1, FGFR2, FGFR3 or FGFR4 were washed three times in PBS and cell-surface proteins were biotinylated with 0.5 mg/ml Ez-link sulfo-NHS-LC-Biotin in PBS for 15 minutes at 4°C. The biotinylation reaction was quenched with 50 mM Tris-HCl pH 8.0. The cells were washed twice with PBS and then incubated for the indicated periods of time in DMEM containing 100 ng/ml FGF1 and 20 U/ml heparin. The cells were washed with PBS and lysed on ice in lysis buffer (0.1 M NaCl, 10 mM Na₂HPO₄, 1% Triton X-100, 1 mM EDTA, supplemented with complete protease inhibitors, pH 7.4) for 20 minutes. The lysate was centrifuged to remove nuclei and then biotinylated proteins were pulled down from the supernatant with streptavidin-Sepharose beads at 4°C over night. The beads were then washed three times in PBS containing 0.1% Tween 20 and finally resuspended in 15 μl of reducing SDS-PAGE sample buffer. Proteins were separated by 7% SDS-PAGE as described by Laemmli (Laemmli, 1970) and transferred to Immobilon-P PVDF membrane (Millipore Corporation, Bedford, MA), which was probed with anti-FGFR1, anti-FGFR2, anti-FGFR3 or anti-FGFR4 primary antibodies and anti-rabbit HRP-conjugated secondary antibody. Immunoactivity was detected by using ECL plus western blotting system and Chemi Genius Image Acquisition System (Syngene, Cambridge, UK). To compare the intensity of bands of interest on the membrane, ImageQuant software (Amersham Biosciences, Buckinghamshire, UK) was used. Background correction was performed by subtracting values obtained by scanning adjacent areas of the membrane with the same size but containing no visible bands from those obtained with the bands of interest.

The [³⁵S]methionine-labelled 18 kDa, long form of FGF1 and the 16 kDa form of FGF1 was produced by transcription using T7 RNA polymerase or T3 RNA polymerase, respectively, and translation in a rabbit reticulocyte lysate supplemented with Easytag Methionine L-[³⁵S] according to the manufacturer's protocol.

HeLa cells, not transfected or transiently transfected with the different FGFRs, were incubated at 37°C with the [³⁵S]methionine-labelled 16 kDa form of FGF1 and 50 U/ml heparin for 1 hour to allow binding and endocytosis of FGF1. The cells were washed with PBS and lysed. FGF1 was extracted from the lysate by binding to heparin-Sepharose and analysed by 13% SDS-PAGE.

HeLa cells, transiently transfected with the different FGFRs, were incubated with the [³⁵S]methionine-labelled 18 kDa form of FGF1 and 20 U/ml heparin at 37°C for 1 hour to allow endocytosis via high-affinity receptors. Then the cells were washed twice with a high salt, low pH buffer (2 M NaCl, 20 mM NaAc, pH 4. 0) and once with PBS on ice to remove excess and cell-surface bound FGF1 (Klingenberg et al., 2000). The cells were then either lysed immediately in lysis buffer or incubated further in growth medium with or without 100 μM chloroquine at 37°C for 3 or 6 hours before lysis. [³⁵S]Methionine-labelled FGF1 was extracted from the lysate by adsorption to heparin-Sepharose and analysed by 15% SDS-PAGE. The proteins on the gels were fixed in fixative (25% methanol, 7.5% acetic acid) and then the gels were dried. STORM Phosphorimager scanning and Image Quant, Version 5.0 (Molecular Dynamics, Amersham Biosciences, Buckinghamshire, UK) software were used to estimate the relative amount of radioactive FGF.

To study degradation and recycling of FGF1, HeLa cells were transfected with FGFR1 or FGFR4 and incubated at 37°C for 20 minutes in growth medium containing 100 ng/ml [¹²⁵I]FGF1 and 40 U/ml heparin. To remove surface bound FGF1, the cells were washed twice with a high salt/low pH buffer and twice in PBS. The cells were further incubated in growth medium at 37°C for indicated periods of time. The medium was then removed and collected. The cells were incubated for 5 minutes with high salt/low pH buffer and the buffer was collected. Finally, the cells were collected after solubilisation in 0.1 M KOH at 37°C for 30 minutes. The collected medium, the collected high salt/low pH buffer wash and the collected dissolved cells were adjusted to 5% trichloroacetic acid (TCA) and kept on ice for 20 minutes and then centrifuged at 4°C for 5 minutes. The radioactivity present in the supernatants (TCA-soluble fraction) and pellets (TCA-insoluble fraction) was determined in a gamma counter. Degradation was measured for each time point as the amount of radioactivity in the TCA-soluble fractions and expressed as a percentage of the total radioactivity in the culture. Recycling was measured for each time point as the amount of radioactivity in the TCA-insoluble fractions in the medium and in the high salt/low pH buffer and expressed as a percentage of the total radioactivity in the culture.

HeLa cells cotransfected with myc-ubiquitin and FGFR1, FGFR4 or empty vector were starved for 16 hours and then washed three times in PBS. Cell-surface proteins were biotinylated with 0.5 mg/ml Ez-link sulfo-NHS-LC-Biotin in PBS for 15 minutes at 4°C. The biotinylation reaction was quenched with 50 mM Tris-HCl pH 8. 0 and the cells were washed twice with PBS. The cells were incubated for 2 hours in 200 ng/ml FGF1, 20 U/ml heparin and 0.3 mM leupeptin at 37°C in DMEM without serum. The cells were then washed once in DMEM without serum and lysed at 95°C for 5 minutes in 1% SDS in PBS. The lysate was decanted into QIAshredder columns and centrifuged for 2 minutes at 4°C. Equal amounts of lysate and 2× pull down-buffer (2% Triton X-100, 0.5% sodium deoxycholate, 2 mM EDTA, 40 mM NaF, 1% bovine serum albumine, 2 mM N-ethylmaleimide supplemented with protease and phosphatase inhibitors) were added to streptavidin-Sepharose beads to pull down biotinylated proteins. After tumbling 1 hour at 4°C, the beads were washed twice in 1× pull-down buffer (0.5% SDS and 50% 2× pull-down buffer in PBS) and once in 1:10 diluted PBS. The proteins that remained bound to the streptavidin-Sepharose beads were submitted to 7% SDS-PAGE and then transferred to a PVDF membrane which was probed with anti-myc primary antibody and anti-mouse HRP-conjugated secondary antibody to detect the level of ubiquitylation of internalized FGFRs. Immunoreactivity was detected using ECL plus western blotting system and Chemi Genius Image Acquisition System. The membrane was stripped twice and reprobed with anti-phospho FGFR primary antibody and anti-rabbit HRP-conjugated secondary antibody to detect the level of internalized receptors, and with anti-transferrin receptor primary antibody and anti-mouse HRP-conjugated secondary antibody to verify equal loading of the gel. To ensure equal expression of ubiquitin, the cells were analysed by immunofluorescence microscopy.

Upon ligand binding to the FGFRs, the ligand-receptor complexes are internalized (Sorokin et al., 1994; Munoz et al., 1997) and transported to various intracellular compartments. Because FGF1 binds equally well to the four high-affinity FGFRs (Ornitz et al., 1996), this ligand was labelled with the fluorescent dye Cy3 and the fluorescent growth factor was then used as a marker to explore the intracellular trafficking of FGF1 and the four FGFRs. Cy3-labelled FGF1 has previously been shown to retain its binding capacity towards the FGFRs and heparin sulfate proteoglycans (Citores et al., 1999).

The distribution of fluorescent growth factor-receptor complexes was studied in HeLa cells transiently transfected with the different FGFRs and incubated with Cy3-FGF1 for different periods of time. HeLa cells do not express detectable amounts of endogenous FGFRs. To avoid FGF1 binding to cell-surface heparan proteoglycans and to facilitate binding to the high-affinity FGFRs, heparin was added to the extracellular medium. The data in Fig. 1A show that fluorescent FGF1 binds to the surface of transfected cells when treated with the growth factor at 4°C in the presence of heparin. There was no detectable binding to untransfected cells (Fig. 1A, first panel). Uptake of similar amounts of radiolabelled FGF1 in HeLa cells transfected with any of the four receptors indicates that the transfected cells express comparable levels of the four FGF receptors (Fig. 1B). In addition, very little FGF1 was associated with untransfected cells. When the cells were incubated at 37°C, the fluorescent growth factor appeared as intracellular dots, indicating uptake into vesicles (Fig. 2).

To determine whether FGF1 and the different FGFRs remain in the same compartments after internalization, we carried out double-labelling experiments where cells were allowed to take up Cy3-labelled growth factor for 2 hours at 37°C in the presence of the inhibitor of lysosomal degradation, leupeptin, and then stained with antibodies against the different FGFRs. There was considerable overlap between internalized FGF1 and FGFRs, as visualised in double-staining experiments when spots labelled with both fluorophores appeared yellow (Fig. 2). This indicates that the ligand and the receptor are in the same intracellular compartments as previously reported for KGF/KGFR and FGF1/FGFR4 (Marchese et al., 1998; Citores et al., 1999; Belleudi et al., 2002).

To follow the endocytic pathway and to identify the intracellular compartments where the different FGF1/FGFR complexes are localized upon internalization, the transiently transfected HeLa cells were allowed to bind Cy3-labelled FGF1 at 4°C and then they were incubated for different periods of time at 37°C. The cells were subsequently fixed and stained with antibodies against markers for different intracellular compartments. As shown in Fig. 3, incubation for 15 minutes at 37°C resulted in good overlap of EEA1, a protein associated with early/sorting endosomes (Mu et al., 1995), and endocytosed Cy3-FGF1. Quantitation of the colocalization showed that about 70-80% of the FGF1-positive structures in cells transfected with either of the four FGFRs were positive for EEA1 after 15 minutes of endocytosis. Thus, all the four FGF1/FGFR complexes reach the sorting endosomal compartment.

After a 2 hour chase in the presence of leupeptin to inhibit degradation in the lysosomes, the major part of the internalized FGF1 in cells transfected with FGFR1-3 was found to colocalize with antibodies against LAMP-1. LAMP-1 is a protein associated with late endosomes/lysosomes (Geuze et al., 1988). In the case of FGFR4-transfected cells there was less colocalization (Fig. 4A). Quantitation showed that about 90% of the FGF1-positive structures in FGFR1-transfected cells were LAMP-1 positive, whereas in the case of FGFR4 only about 45% were LAMP-1 positive (Fig. 4B). In the case of FGFR2 and FGFR3 about 70% of the FGF1-positive structures were also positive for LAMP-1. These findings indicate that subsequent to their uptake in early/sorting endosomes the four FGFRs are sorted differently. The major part of internalized FGFR1-3 is sorted to lysosomes, whereas the major part of internalized FGFR4 is not.

To further study the different sorting of the receptors and to determine the localization of the FGF1/FGFR4 complex subsequent to its appearance in early/sorting endosomes, the endocytic pathway followed by the fluorescent FGF1 was compared with those taken by epidermal growth factor (EGF) and transferrin (Tf). EGF receptors and its ligand, EGF, progress to lysosomes upon internalization (Futter et al., 1996), whereas the transferrin receptor and its ligand, Tf, are known to be recycled from early/sorting endosomes via the endosomal recycling compartment back to the cell surface (Hopkins, 1983).

HeLa cells transfected with the different FGFRs were incubated for 2 hours at 37°C with Alexa 488-labelled EGF and Cy3-labelled FGF1 in the presence of leupeptin. Alexa 647-labelled transferrin was added after 90 minutes. Colocalization was shown in overlay experiments when dots labelled with Cy3-FGF1 and Alexa 488 EGF appeared yellow and dots labelled with Cy3-FGF1 and Alexa 647-labelled transferrin appeared purple. Fluorescent FGF1 endocytosed by FGFR1-3 showed considerable overlap with fluorescent EGF, indicating that the major part of internalized FGF1/FGFR1-3 complexes accumulates in lysosomes However, fluorescent FGF1 endocytosed by FGFR4 showed a notable overlap with transferrin, indicating that a great part of the internalized FGF1/FGFR4 complexes accumulate in the endocytic recycling compartment (Fig. 5A).

Quantitation of colocalization was performed as described in Materials and Methods. As shown in Fig. 5B approximately 65% of the FGF1-positive structures in cells transfected with FGFR1 colocalized with intracellular structures containing EGF, whereas only about 8% of the FGF1-positive structures colocalized with intracellular structures containing transferrin. In cells transfected with receptor 2 or 3 between 40% and 50% of the FGF1-positive structures also contained EGF, whereas about 20% of the FGF1-positive structures contained transferrin. In the case of cells transfected with FGFR4 only about 20% of the FGF1-positive structures contained EGF, whereas about 50% of the FGF1-positive structures contained transferrin. Between 20% and 30% of the FGF1-positive structures in the transfected cells contained neither EGF nor transferrin and a small fraction of about 5% of the FGF1-positive structures contained both EGF and transferrin.

To decide whether the observed difference in the sorting of FGF1 internalized by the four FGFRs is specific for FGF1, cells transfected with FGFR1 or FGFR4 were allowed to endocytose Cy3-labelled FGF1 and Alexa 488-labelled FGF2 for 2 hours in the presence of leupeptin. The two fluorophore-labelled ligands showed almost complete colocalization in both cases (data not shown). This finding suggests that the different sorting does not depend on the ligand.

The different sorting of the four related FGFRs could result in different kinetics of degradation of FGF1 as well as the receptors. The degradation of the four FGFRs and FGF1 was analysed in FGFR-transfected HeLa cells.

To analyse the degradation of internalized receptors, cells transfected with the different FGFRs were submitted to biotinylation of proteins exposed at the cell surface. The cells were then treated with FGF1 for increasing periods of time and then lysed. The biotinylated proteins in the lysates were collected by adsorption to streptavidin-Sepharose followed by SDS-PAGE and immunoblotting with the appropriate anti-receptor antibodies. We observed for each FGFR a gradual reduction in the intensity of the bands corresponding to the receptors, indicating degradation of the internalized receptors (Fig. 6A). There was, however, a large difference in the rate of degradation. Thus, the intensity of the bands corresponding to FGFR1 was strongly reduced after 2 hours, whereas that corresponding to FGFR4 was only slightly reduced after 6 hours. The degradation of FGFR2 and FGFR3 was slower than the degradation of FGFR1 but considerably faster than that of FGFR4.

The bands representing FGFRs in Fig. 6A were quantitated from the gels and plotted as a percentage of the amount of receptors at time zero. The mean values from three independent experiments (Fig. 6B) show that the amount of FGFR1 decreased from 100% to ∼15% in 2 hours, whereas the amount of FGFR4 was only reduced to ∼80% in 6 hours. In the case of FGFR2 and FGFR3, 30-40% of the initial amount remained after 6 hours.

To investigate the degradation of FGF1, transfected cells were incubated with the radiolabelled 18 kDa form of FGF1 for 1 hour to allow endocytosis of the growth factor-receptor complex to occur. The cells were then washed to remove surface-bound FGF1 and further incubated for increasing periods of time. Finally, the cells were lysed and solubilised proteins were adsorbed to heparin-Sepharose and analysed by SDS-PAGE. The degradation of FGF1 can be seen in Fig. 7A as a stepwise conversion of the 18 kDa form of FGF1 into the shorter 16 kDa form followed by further degradation.

The further degradation of FGF1 seems to occur more slowly, indicating that the 16 kDa form of FGF1 is more resistant to degradation than the 18 kDa form. The degradation of internalized FGF1 was inhibited by the weak base chloroquine (shown only for FGFR1), suggesting that the digestion occurred in a lysosomal compartment.

After 6 hours only a small fraction of FGF1 remained as the 18 kDa form in cells transfected with FGFR1, whereas a significant amount of the 18 kDa form was present in cells transfected with FGFR4. The amount of the 18 kDa form of FGF1 that remained in the cells was calculated for each receptor type and each time point, and expressed as a percentage of the amount of the 18 kDa form at time zero. The values plotted in the graph in Fig. 7B are average values from five (FGFR1 and FGFR4) and three (FGFR2 and FGFR3) independent experiments. Approximately 50% of FGF1 was in the 18 kDa form in FGFR4 transfected cells after 3 hours, whereas only 15% of FGF1 remained as the 18 kDa form after 3 hours in cells transfected with FGFR1. The fraction of the 18 kDa form in cells transfected with FGFR2 or FGFR3 was reduced to about 30% after 3 hours. The results indicate that FGF1 endocytosed by FGFR4 is more slowly degraded than FGF1 endocytosed by FGFR1. FGF1 endocytosed by FGFR2 or FGFR3 seems to be more slowly degraded than FGF1 endocytosed by FGFR1, but faster than FGF1 endocytosed by FGFR4.

The degradation and the recycling of FGF1 was further analysed in HeLa cells transfected with FGFR1 or FGFR4. The transfected cells were allowed to internalize [¹²⁵I]FGF1 for 20 minutes, washed to remove excess and cell-surface-associated [¹²⁵I]FGF1 and incubated further for the periods of time indicated. The degradation of FGF1 was measured as the percentage of radioactivity in the culture that was TCA soluble. Recycled FGF1 was measured as the percentage of the total amount of TCA-insoluble radioactivity in the culture that was present in the medium or released from the cell surface by a high salt/low pH wash. About 60% of FGF1 internalized by FGFR1 was found to be degraded after 5 hours, whereas only about 30% of FGF1 internalized by FGFR4 was degraded after 5 hours (Fig. 8A). The rate of recycling for FGF1 internalized by FGFR4 was higher than for FGF1 internalized by FGFR1 (Fig. 8B). Altogether, the data indicate that cells transfected with FGFR1 degrade FGF1 more efficiently than cells transfected with FGFR4 and that FGF1 endocytosed by FGFR4 shows a higher extent of recycling than FGF1 internalized by FGFR1.

The attachment of ubiquitin to lysines in the intracellular part of a membrane protein is thought to function as a signal for lysosomal degradation (Raiborg and Stenmark, 2002; Haglund et al., 2003).

An amino acid sequence alignment of the intracellular part of the four receptors revealed several conserved lysines in FGFR1-3 that were absent in FGFR4 (Fig. 9). The intracellular domain of FGFR1 and of FGFR2 contains 29 lysines, the intracellular part of FGFR3 contains 25 lysines, whereas only 16 lysines are present in the intracellular domain of FGFR4. Because lysines are potential ubiquitylation sites, it is possible that FGFR4 is less ubiquitylated than FGFR1.

To test this, HeLa cells co-expressing myc-tagged ubiquitin and either FGFR1, FGFR4 or empty vector were incubated without serum overnight. This was included to avoid stimulation and possibly ubiquitylation of surface proteins by factors in the serum.

Cell-surface proteins were then biotinylated and the cells were treated with FGF1 for 2 hours. Leupeptin was added to prevent lysosomal degradation of the receptors. The cells were then lysed and biotinylated proteins were collected and analysed by western blotting using anti-myc antibody (Fig. 10). Ubiquitylation of the receptors was detected as a smear of bands migrating more slowly than the nonubiquitylated receptors. The signal was much stronger for FGFR1 than for FGFR4, indicating that FGFR1 is ubiquitylated to a higher extent than FGFR4. Little ubiquitylation was detected in cells transfected with the empty vector.

To test whether equal amounts of FGFR were present at the membrane in the two cases, the membrane was stripped and reprobed with an anti-phospho FGFR antibody that was raised by immunizing rabbits with a synthetic peptide corresponding to residues surrounding tyrosine 653/654 of human FGFR1, which are conserved in FGFR1 and FGFR4. The results indicate the presence of similar amounts of FGFR1 and FGFR4 (Fig. 10). The membrane was stripped and reprobed with anti-transferrin receptor antibodies to verify approximately equal loading on the gel (Fig. 10). Similar expression of the mycubiquitin construct was confirmed by immunofluorescence microscopy (data not shown).

Altogether, the observations show that FGFR4 is less ubiquitylated than FGFR1 and they suggest that different levels of ubiquitylation is the molecular mechanism determining the different sorting of the receptors.

The present work shows that FGF1 internalized by FGFR1 is targeted for lysosomal degradation whereas the majority of FGF1 internalized by FGFR4 escapes into a recycling pathway. In cells expressing FGFR2 or FGFR3, somewhat less FGF1 is sorted to degradation than in cells expressing FGFR1. Furthermore, FGF1 endocytosed by FGFR4 was more slowly degraded than FGF1 endocytosed by FGFR1, FGFR2 or FGFR3. Also, FGFR4 as such was more slowly degraded than the other receptors.

Targeting of receptors for lysosomal degradation has been associated with the attachment of mono-ubiquitin to multiple lysines in the intracellular part of the receptors (Haglund et al., 2003). Consistent with the observed different sorting of the FGFRs, FGFR4 has fewer potential ubiquitylation sites (lysines) in its intracellular part than the other FGFRs. The present work shows that FGFR4 is indeed less ubiquitylated than FGFR1. This indicates that different levels of ubiquitylation of the FGFRs determine their intracellular sorting.

The present findings are in accordance with previous data concerning the trafficking of endocytosed FGFRs. Endocytosed KGF and KGFR in HeLa cells and FGF2 and FGFR3 in RCJ cells have previously been reported to enter the lysosomes (Belleudi et al., 2002; Cho et al., 2004), whereas internalized FGF1 and FGFR4 were found to accumulate in the recycling compartment in COS cells (Citores et al., 1999). It has also been reported that binding of FGF to FGFR1 in HeLa cells (Wong et al., 2002) and PAE cells (Mori et al., 1995) and binding of FGF to FGFR3 in COS-7 cells (Cho et al., 2004) and 293T cells (Monsonego-Ornan et al., 2002) induces ubiquitylation of the receptors and that this contributes to their downregulation. Taken together, these findings indicate that the distinct sorting of the FGFRs depends on receptor type rather than cell type or ligand.

Late endosomes and lysosomes contain large amounts of glycoproteins such as the lysosome-associated membrane protein-1 (LAMP-1) (Geuze et al., 1988). The staining of these compartments for laser scanning confocal microscopy analysis with primary antibodies against LAMP-1 and fluorophore-conjugated secondary antibodies gave a dense pattern of LAMP-1-positive structures. The dense pattern may have caused an overestimation of the degree of colocalization between FGF1-positive structures and LAMP-1-positive structures. In FGFR4-transfected cells approximately 45% of the FGF1-positive structures were positive for LAMP-1. However, when the distribution of fluorophore-labelled FGF1 internalized via FGFR4 was compared with the distribution of internalized fluorophore-labelled EGF, a marker for lysosomal trafficking, only 20% of the FGF1-positive structures contained EGF.

When continuous uptake of fluorophore-labelled FGF1 was allowed in cells transfected with the different receptors, between 20% and 30% of the FGF1-positive structures inside the cells colocalized neither with the marker for lysosomal trafficking (EGF) nor with transferrin, the marker for the recycling pathway. It is likely that the amount of overexpressed FGFRs exceeds the amount of the other receptors at the cell surface. Therefore, free FGFRs could still be present at the cell surface, ready to bind and internalize ligand when most of the EGF and transferrin receptors are already located in intracellular vesicles. Some of the FGF1-positive structures inside the cells also contained both fluorophore-labelled EGF and transferrin. This is probably material in early endosomes.

The kinetics of degradation of FGF1 as well as that of the receptors were analysed in FGFR-transfected HeLa cells and the results correlate with the different sorting observed for FGF1 endocytosed by the four FGFRs -that is, degradation was faster for receptor/ligand sorted to lysosomes (particularly FGFR1) than for recycling receptor/ligand. The degradation of internalized FGF1 appeared first as a conversion of the 18 kDa form of FGF1 to the shorter 16 kDa form. Further degradation of the short form of FGF1 occurred more slowly, indicating that the 16 kDa form of FGF1 is more resistant to degradation than the 18 kDa form. This seems to be the case for FGF2 also. Internalized FGF2 in BCE cells has earlier been reported to be rapidly cleaved from an 18 kDa form to a 16 kDa form, and the 16 kDa form was then found to be more slowly degraded, with a half-life of approximately 8 hours (Moscatelli, 1988). It is not clear which FGFR were expressed on the cells in that case.

Furthermore, if the recycled ligand is allowed to detach from the receptor at the cell surface, the observed degradation of FGF1 might be slightly overestimated. It is therefore possible, particularly in the case of FGFR4, that FGF1 is actually more slowly degraded than apparent from Fig. 7.

All the four FGFRs have been found to have distinct patterns of distribution in many human tissues. The most widespread expression has been observed for FGFR1, whereas FGFR4 was found to have a more limited distribution (Hughes, 1997). Gene deletion and mutation studies in mice have implicated FGFR1-3 in numerous developmental events, whereas FGFR4 seems to play a more modest role in development (Goldfarb, 1996; Xu et al., 1999; Ornitz and Marie, 2002). It is possible that this is due to the different sorting of the FGFRs. The recycling of FGFR4 could prolong the signalling. It is therefore possible that FGFR4 is less suited for processes where rapid downregulation is necessary. On the other hand, the recycling of FGFR4 could provide a mechanism for gradient formation during developmental processes. A simple model of gradient formation implies that morphogens dilute as they diffuse between cells. Recent data, however, suggest that movement of morphogens could also occur by transcytosis where endocytosed morphogens can be re-secreted and move forward into the target tissue (Entchev et al., 2000). FGFs could therefore, after binding and activation of FGFR4 in one cell, be recycled and activate neighbouring cells and then spread through the tissue.

FGFRs have been found overexpressed or mutated to constitutively active forms in numerous human cancers (Powers et al., 2000). FGFR1 is the most studied FGFR in human cancer and several reports closely link abnormal expression of FGFR1 to cancer progression (Becker et al., 1992; Yamaguchi et al., 1994; Giri et al., 1999). But also an increasing number of reports links overexpression of FGFR4 with poor prognosis in several human tumour types (Jaakkola et al., 1993; Facco et al., 1998; Yamada et al., 2002; Gowardhan et al., 2005). Several mechanisms are involved in the attenuation of signalling from activated receptors. One of them is the degradation of the receptors in lysosomes. Accurate downregulation of activated signalling receptors is important to prevent increased signalling and enhanced cancer progression (Bache et al., 2004). It is therefore possible that elevated levels or constitutively active forms of recycling receptors could predispose cancer patients for accelerated disease progression because the receptors are not efficiently downregulated. A recent study of the four FGFRs in human thyroid cancer indicates that FGFR4, in contrast to the other FGFRs, may promote the progression of these tumours. Thus, FGFR4 was found to be strongly expressed in the more aggressive tumour types and in the most rapidly proliferating cells (St Bernard et al., 2004). A mechanism for FGFR4 to induce cell proliferation in breast tumour cells through regulation of cyclin D1 translation in cooperation with ErbB2 has recently been proposed (Koziczak and Hynes, 2004; Koziczak et al., 2004).

The expression of ErbB2/HER2, a member of the EGF receptor family, is frequently elevated in human cancers (Yarden and Sliwkowski, 2001). ErbB2 does not bind ligand, but instead acts as a heterodimerization partner for ligand-activated members of the EGF receptor family, thus amplifying the signalling. As FGFR4, ErbB2 is rapidly recycled upon endocytosis. Overexpression of ErbB2 potentiates EGFR signalling by diverting EGFR from EGF-induced downregulation (Worthylake et al., 1999). The ability of ErbB2 to shunt ligand-activated receptors to recycling may explain its oncogenic potential. It is possible that overexpression of FGFR4 followed by possible heterodimerization could in a similar way lead to impaired downregulation of the other FGFRs, thereby increasing their oncogenic potential.

Several monoclonal antibodies that inhibit ErbB2-transformation in vitro and have antitumour properties in vivo are proposed to trigger surface downregulation and endocytic degradation of ErbB2 (Klapper et al., 2000; Austin et al., 2004). Similar therapeutic strategies might be introduced to inhibit FGFR and particularly FGFR4 signalling in tumours where elevated levels of FGFR signalling are associated with poor prognosis.

However, examination of constitutively activated derivatives of FGFR1, FGFR3 and FGFR4 in which a myristylation signal was substituted for the extracellular and transmembrane domains, thereby targeting the kinase domain to the plasma membrane, revealed that FGFR4 was much less transforming than activated FGFR1 and FGFR3 (Hart et al., 2000). Because FGFR1 also exhibits higher signalling activity than FGFR4 it has been suggested that FGFR1 is the most potent mutagenic member of the FGFR family (Vainikka et al., 1994; Wang et al., 1994). Other mechanisms for attenuating signals may therefore play a role to limit the signalling from the FGFR4.

The present work shows that endocyotosed FGF1 is sorted to recycling or degradation depending on the receptor type. Many growth factors can bind to more than one receptor, but in many cases the different roles of the separate receptors are unclear. The sorting of ligands and receptors to recycling or degradation affects their half-lives and may thus regulate their signalling. Although the exact biological role of the different trafficking of the FGFRs remains to be elucidated, different intracellular sorting of the ligand-receptor complexes is likely to be important for their signalling properties.

E.M.H. is a pre-doctoral fellow and V.S. is a post-doctoral fellow of the Norwegian Cancer Society. This work has been supported by the Research Council of Norway, the Novo Nordisk Foundation, the Blix Fund for the Promotion of Medical Research, the Rachel and Otto Kr. Bruuns Fund, Torsteds Fund and the Jahre Foundation.