Modulation of intracellular transport of acidic fibroblast growth factor by mutations in the cytoplasmic receptor domain

Acidic fibroblast growth factor (aFGF) belongs to a large group of growth factors that bind to and activate transmembrane receptors with a cytoplasmic tyrosine kinase domain (Burgess and Maciag, 1989). In mammals there are four closely related FGF-receptor (FGFR) genes and a number of splice variants of the receptors (Johnson and Williams, 1993; Partanen et al., 1991; Mason, 1994). In addition to binding to the FGF-receptors, aFGF binds to surface heparans abundantly present in most mammalian cells (Burgess and Maciag, 1989). This binding can be competed out by adding heparin to the medium. We have earlier presented evidence that aFGF bound to FGF-receptors is to some extent translocated to the cytosol and to the nucleus (Wiedlocha et al., 1995; Wiedlocha et al., 1996; Klingenberg et al., 1998; Klingenberg et al., 2000a; Klingenberg et al., 2000b). This is not the case with aFGF bound to surface heparans (Wiedlocha et al., 1995).

In attempts to elucidate the translocation mechanism of aFGF, we are studying endocytic uptake and intracellular routing of the growth factor. In an earlier paper (Citores et al., 1999), we presented evidence that, in FGFR4-transfected COS cells, aFGF is internalized by a mechanism not involving coated pits. This is in accordance with earlier studies of the uptake of basic FGF which was found to be mainly present in caveolae and not in clathrin-coated pits in BHK cells (Gleizes et al., 1996). By contrast, in NIH3T3 cells, the uptake of KGF, a member of the FGF family, was reported to occur from coated pits. In the cold, KGF was found to bind to receptors that were mainly localized outside coated pits, but upon incubation with KGF at 37°C the growth factor was rapidly transferred to coated pits and endocytosed (Marchese et al., 1998). The most likely conclusion from these experiments is that FGF receptors can be endocytosed by two mechanisms, one that involves coated pits and another that does not.

Previous studies found that a tyrosine residue (Y766) close to the C-terminus of FGFR1 is essential for endocytosis in rat L6 myoblasts and hematopoietic Ba/F3 cells (Sorokin et al., 1994). In a deletion mutant where this residue was absent, and in one where it was mutated to alanine, the rate of endocytosis was strongly reduced. This was also the case with a receptor that was kinase negative due to a point mutation in the kinase domain.

We previously found that FGFR4 partly copurifies with structures enriched in caveolin (Citores et al., 1999). All four FGFR contain a putative caveolin-binding sequence in the cytoplasmic domain of FGFR4 (Couet et al., 1997). To study if this sequence is required for cofractionation with the caveolin-rich material, we mutated key amino acids in this region.

We also found earlier that aFGF endocytosed in cells expressing the wild-type receptors is transported to a juxtanuclear organelle that was identified as the recycling endosome compartment (Citores et al., 1999). We have here studied the intracellular trafficking of aFGF in cells transfected with receptors that have a mutated or deleted cytoplasmic domain.

PMSF, pronase, trypsin, tetracycline, puromycin, geneticin and human transferrin were obtained from Sigma, St Louis, MO; disuccinimidyl suberate was from Pierce, Rockford, IL; protein A-Sepharose CL-4B and heparin-Sepharose were from Pharmacia, Uppsala, Sweden; restriction endonucleases were from New England Biolabs, Beverly, MA. Na¹²⁵I was obtained from the Radiochemical Centre, Amersham International, Buckinghamshire, UK. Anti-FGFR4 and anti-phosphotyrosine antibodies were from Santa Cruz Biotechnology, Santa Cruz, CA; anti-caveolin 1 antibodies were from Transduction Laboratories, Lexington, KY; human anti-EEA1 was obtained from Judy Callaghan, Institute for Cancer Research, Oslo, Norway. Monoclonal anti-LAMP-1 antibodies developed by J.T. August and J.E.K. Hildreth were obtained from The Developmental Studies Hybridoma Bank, Dept of Biological Sciences, Iowa City, IA. The secondary antibodies used, FITC-labelled anti-human IgG and FITC-labelled anti-mouse IgG were obtained from Jackson Immuno-Research Laboratories Inc., West Grove, PA. Fugene-6 reagent was from Boehringer Mannheim; unlabelled aFGF was produced in bacteria, purified on a heparin-Sepharose column as described (Wiedlocha et al., 1996) and labelled chemically with ¹²⁵I (Fraker and Speck, Jr, 1978). aFGF metabolically labelled with [³⁵S]methionine was synthesized in a cell-free system as earlier described (Wiedlocha et al., 1994). aFGF labelled with CY3 (Amersham) and human transferrin labelled with alexa (Molecular Probes, Eugene, OR) were prepared according to the procedure giving by the company. Oregon Green 514 EGF was from Molecular Probes.

COS-1 cells were propagated in Dulbecco’s modified essential medium (DMEM) with 10% (v/v) fetal calf serum in a 5% CO₂ atmosphere at 37°C. The HeLa cell line stably transformed with cDNA for dyn^K44A, where the promoter is negatively controlled by tetracycline (Damke et al., 1994) was kindly provided by Sandra Schmid, Scripps Research Institute, La Jolla, CA. The cells were cultured in DMEM supplemented with 10% fetal calf serum, 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM L-glutamine, 400 μg/ml geneticin, 200 ng/ml puromycin, and 1 μg/ml tetracycline. For experiments, HeLa-dyn^K44A cells were seeded with and without tetracycline and used 48 hours later.

Transient expression of the different FGF receptors and human transferrin receptor was performed by transfecting COS-1 and HeLa dyn^K44A cells with plasmid DNA (pcDNA3 with appropriate inserts) by using Fugene-6 transfection reagent according to the procedure given by the company. The cells were used for experiments 48 hours after transfection.

cDNA for human transferrin receptor in pcDNA3 was a kind gift from Toril Bremnes, Institute for Cell Biology, University of Oslo. pEGFR-wt encoding EGF receptor, was obtained from A. Sorkin (Carter and Sorkin, 1998). pcDNA3-FGFR4-K503R has been described earlier (Muñoz et al., 1997). pcDNA3-FGFRalan4 was obtained by cloning a mutated cDNA obtained by PCR into pFGFR4-K503R that had been cut with BglII. The mutations were introduced by using pcDNA3-FGFR4 as template: as forward primer, the oligonucleotide 5′-CTGGCCGGCCTCGTGAGTCTAGATC-3′ and as reverse primer, 5′-GTACACACACCAGAGTGACGTCGCGTCTGCTGGCATCCT-GCTAGCGGAGATCTTCACCCTC-3′. pcDNA3R4ΔBgl and pcDNA3R4ΔB1 have been described earlier (Klingenberg et al., 2000a).

Caveolin-enriched membrane fractions were prepared using a detergent-free method, as described previously (sSong et al., 1996). The cells (two 150 mm dishes) were washed twice with HEPES medium containing 10 U/ml heparin and the cells were kept at 4°C for 3 hours in the same buffer containing 50 ng/ml of ¹²⁵ I-aFGF. After washing once with HEPES medium and once with PBS, the cells were treated for 20 minutes at 4°C with 0.3 mM disuccinimidyl suberate in PBS. After crosslinking, the cells were washed with cold 25 mM Tris buffer, pH 7.4, and twice with PBS. The cells were scraped into 2.2 ml of 0.5 M sodium carbonate, pH 11, and homogenized using first a syringe, then a Dounce homogenizer (20 strokes) and finally a sonicator (3×20-second bursts). The homogenate was then adjusted to 45% sucrose by the addition of 2.2 ml of 90% sucrose prepared in MBS (25 mM MES, pH 6.5, 0.15 M NaCl) and placed in the bottom of an ultracentrifuge tube. A discontinuous sucrose gradient was formed by overlaying this solution with 4 ml of 35% sucrose and 4 ml of 5% sucrose (both in MBS). The tubes were centrifuged at 150,000 g in an SW40 rotor for 19 hours at 4°C, and then 12-13 1-ml fractions were collected manually from the top of the gradient. Aliquots (50 μl) of the fractions were subjected to SDS-PAGE on 7.5% or 10% gels. The ligand-receptor complexes were detected by autoradiography, and caveolin was visualized by western blot after transfer of the protein from the gel to a nitrocellulose membrane.

COS-1 cells transiently transfected with FGFR4 and FGFRalan4, growing in 6-well plates were kept for 2 hours on ice with [³⁵S]methionine-labelled aFGF and 10 U/ml heparin, in the presence or absence of unlabelled aFGF, and then washed five times with PBS. To measure binding, cells were lysed and aFGF was precipitated with heparin-Sepharose and analyzed by SDS-PAGE and autoradiography.

COS cells were serum-starved for 3 hours and then treated for 10 minutes with aFGF at 37°C, or left untreated. The cells were then washed, lysed in the presence of phosphatase- and protease-inhibitors (Wiedlocha et al., 1994) and centrifuged for 5 minutes at 15,800 g at 4°C. The supernatant was centrifuged again and the last supernatant was incubated with anti-FGFR4 antibody. Immunocomplexes were collected with protein A-Sepharose CL-4B, subjected to SDS-PAGE and transferred to a nitrocellulose membrane. The membrane was probed with mouse anti-phosphotyrosine antibody (PY99) and visualized with a horseradish peroxidase-conjugated second antibody and enhanced chemoluminescence.

Cells were starved for 3 hours in serum-free medium and then stimulated for 10 minutes at 37°C with 50 ng/ml aFGF in the presence of 10 U/ml heparin. After stimulation, the cells were lysed on ice for 15 minutes in Tris-lysis buffer (25 mM Tris, pH 7.5, 20 mM NaCl, 2 mM dithiothreitol (DTT), 1 mM EGTA, 0.5 mM Na₃VO₄, 0.5% Triton X-100, 1 μg/ml aprotinin, 1 mM PMSF). The lysates were centrifuged, and the supernatants were incubated for 2 hours at 4°C with anti-FGFR4 antibody. The immunocomplexes were then adsorbed to Protein A-Sepharose CL-4B and washed three times in the same buffer and once in kinase buffer (20 mM Tris, pH 7.6, 10 mM MgCl₂, 2 mM MnCl₂, 1 mM DTT, 1 mM EGTA, 0.1 mM Na₃VO₄, 1 μg/ml aprotinin). For the kinase reaction, immunoprecipitates were incubated at 30°C for 10 minutes in 50 μl kinase buffer containing 1 μCi of [γ-³²P]ATP (Vainikka et al., 1996).

The reactions were stopped by boiling with sample buffer, and the samples were then analyzed by SDS-PAGE and autoradiography.

Confluent cultures of HeLa dyn^K44A cells (Damke et al., 1994; Damke et al., 1995) transfected with wild-type and mutant FGFR and grown in 35 mm plates were incubated with HEPES medium and 10 U/ml heparin for 5 minutes at 37°C. After this, the cells were incubated at 37°C with 100 ng/ml ¹²⁵I-aFGF for 20 minutes. The cells were then washed three times with cold PBS, and incubated for 6 minutes in a solution containing 2 M NaCl, 20 mM Na-acetate, pH 4, to release surface bound aFGF (Klingenberg et al., 2000a). The cells were then washed once in the same buffer and solubilized in 0.1 M NaOH, and the radioactivity both in the salt/acid wash and in the cells was measured. The rate of internalization is expressed as the ratio between the internalized and total cell-associated ligand (percent of total cell-associated radioactivity). Endocytosis of transferrin was measured after incubation for 5 minutes with transferrin (100 ng/ml, labelled with ¹²⁵I or with rutenium; Kobrynski et al., 1996) in HEPES medium, as previously described (Llorente et al., 1998).

To study degradation and recycling of aFGF, COS or HeLa dyn^K44A cells expressing wild-type or mutant FGFR were incubated with ¹²⁵I-aFGF for 20 minutes as above. Then the medium was removed and the monolayers were washed with 2 M NaCl, 20 mM Na-acetate, pH 4, to remove cell surface-bound ¹²⁵I-aFGF. Then 1 ml of HEPES medium was added and the cells were incubated further at 37°C. After the times indicated, the medium was removed and assayed for degraded ¹²⁵I-aFGF as trichloroacetic acid-soluble radioactive material. The cells were then washed with low pH/high salt buffer to remove surface bound growth factor. The recycled ¹²⁵I-aFGF was measured as trichloroacetic acid-insoluble radioactive material in the medium and in the buffer.

COS cells transfected with FGFR or FGFRalan4 (one 150 mm dish for each experiment) were incubated with HEPES medium containing 50 ng/ml of ¹²⁵I-aFGF and 10 U/ml heparin at 37°C or with ¹²⁵I-transferrin for 1 hour at 4°C and then incubated at 37°C for 20 minutes. After washing three times with PBS, the cells were resuspended in 4 ml homogenization buffer (0.25 M sucrose, 0.01 M HEPES and 1 mM EDTA) and homogenized in a 5 ml syringe with a 25G needle. The homogenate was centrifuged for 10 minutes at 2,500 rpm in an Eppendorf centrifuge. Postnuclear fractions were layered on top of linear sucrose gradients (21-54% sucrose) and were then centrifuged at 20,000 rpm for 6 hours at 4°C in a Beckmann SW28 rotor. After centrifugation, 18×2-ml fractions were collected from the top of each gradient and the radioactivity of the fractions was measured. The amount of β-N-acetylglucosaminidase in each fraction was determined as described previously (Beaufay et al., 1974). Each fraction was then treated with 20 μl heparin-Sepharose to bind the growth factor, and the absorbed material was analyzed by SDS-PAGE and autoradiography.

Cells were grown on coverslips and incubated with Cy3-aFGF, alexa-transferrin or Oregon green labelled EGF. Incubation with Cy3-aFGF was in the presence of 10 U/ml heparin, unless otherwise stated. The cells were washed with PBS and fixed with 3% paraformaldehyde in PBS for 15 minutes at room temperature. For double staining experiments, paraformaldehyde fixed cells were permeabilized with 0.05% Saponin in PBS for 5 minutes. Coverslips were then incubated in PBS/0.05% Saponin at room temperature for 20 minutes with the primary antibody, washed and then incubated with the secondary antibody. After staining, the coverslips were mounted in Mowiol and examined with a Leica (Wetzlar, Germany) confocal microscope. Images were taken at ×100 magnification and captured as images at

1024×1024 pixels. Montages of images were prepared with the use of Photoshop 4.0 (Adobe, Mountain View, CA).

To investigate whether endocytosis of aFGF takes place through clathrin-coated pits and caveolae or by another mechanism, we transiently transfected HeLa dyn^K44A cells with wild-type and different mutants of FGFR4 (Fig. 1A). Crosslinking experiments after preincubation of the cells with ¹²⁵I-labelled aFGF demonstrated that all receptor mutants were expressed at the surface of the cells (Fig. 1B). HeLa dyn^K44A cells express the dominant negative dynamin K44A mutant under tetracyclin regulation. When the mutant dynamin is induced by removal of tetracyclin, the cells are defective in clathrin-mediated endocytosis as well as in endocytosis from caveolae (Damke et al., 1994; Henley et al., 1998; Oh et al., 1998).

Endocytosis of ¹²⁵I-transferrin, which occurs from coated pits (Hopkins, 1983), was inhibited by more than 90% by overexpression of the mutant dynamin. By contrast, ¹²⁵I-aFGF uptake was reduced to about half in cells expressing the wild-type FGFR4, whether or not heparin was present (Fig. 2A). These data indicate that the endocytic uptake of aFGF in these cells occurs only partly by endocytosis from clathrin-coated pits or caveolae.

In cells transfected with kinase-negative receptor mutants (FGFRalan4, FGFR4-K503R, FGFR4ΔBgl and FGFR4ΔEag/BamH, see below) and under conditions where the mutant dynamin was not induced, the rate of endocytosis was considerably reduced compared with that in cells transfected with wild-type receptor (Fig. 2A). Furthermore, expression of the mutant dynamin reduced almost by half the rate of endocytotic uptake of growth factor in cells transfected with wild-type receptor, whereas in cells transfected with FGFR4alan4, FGFR4-K503R and FGFR4ΔBgl the internalization of aFGF was only slightly inhibited. This indicates that endocytosis by these receptor mutants occurs mainly by an endocytosis mechanism not involving clathrin-coated pits or caveolae.

When internalization was measured in cells expressing FGFR4ΔB1, where only the 11 C-terminal amino acids are deleted and where the kinase activity is intact (Klingenberg et al., 2000a), the pattern was similar to that obtained with the wild-type receptor (Fig. 2A). When essentially the whole cytoplasmic domain was removed (FGFR4ΔEag/BamH) the endocytosis rate was strongly reduced compared with the control, whether or not the clathrin-dependent endocytosis was blocked by expression of mutant dynamin.

In an earlier paper (Citores et al., 1999), we found that in transfected COS cells there was essentially no reduction of the endocytic uptake of aFGF when we inhibited endocytosis from coated pits by acidification of the cytosol. To test if this is the case in HeLa cells transfected with FGFR4, we measured uptake of labelled aFGF in cells that had been acidified by preloading with increasing concentrations of NH₄Cl followed by its removal (Sandvig et al., 1987). The data in Fig. 2B show that, in these cells, the uptake of aFGF was reduced by about 40% after acidification of the cytosol. The uptake of transferrin was reduced by more than 95% under the same conditions. The results therefore indicate that there is a difference between COS cells and HeLa cells in the way they endocytose aFGF. COS cells appear to endocytose the growth factor mainly from non-clathrin-coated areas of the membrane, whereas HeLa cells appear to endocytose the growth factor by both mechanisms.

We next measured the ability of HeLa dyn^K44A cells transfected with the different receptors to recycle endocytosed aFGF back to the surface and to the culture medium. Recycling of aFGF, measured as TCA-insoluble radioactivity in the medium, as well as released from the cell surface by a high salt/low pH wash, occurred as a rapid phase that lasted for less than 20 minutes and a slow phase that lasted from one to several hours. In cells transfected with R4ΔEag/BamH the recycling was more extensive than in cells transfected with the other constructs (Fig. 3A). In COS cells, there was not much difference between cells transfected with the different mutants and cells transfected with wild-type receptor (Fig. 4A).

We also tested the degradation of aFGF internalized by cells transfected with the different receptors. The data in Fig. 3B and Fig. 4B demonstrate that cells transfected with FGFRalan4 were more efficient in degrading the internalized growth factor than cells transfected with the wild-type and with the other receptor mutants. Bafilomycin A₁, an inhibitor of the vacuolar H⁺-ATPase (Yoshimori et al., 1991), inhibited degradation of aFGF in cells transfected with the wild-type and mutant receptors (data not shown) indicating that degradation takes place in an acidic compartment. In HeLa dyn^K44A cells, essentially the same degradation rates were obtained whether or not the mutant dynamin was expressed by removal of tetracyclin (data not shown).

Because the endocytic uptake of aFGF from uncoated areas of the membrane is the predominant process in COS cells, whereas, in HeLa cells, both mechanisms are operating, in the following studies we have used mainly transfected COS cells. Control experiments demonstrated that COS cells transfected with the different receptor mutants endocytosed ¹²⁵I-aFGF to a similar extent, except in the case of FGFR4ΔEag/BamH, where the rate of uptake was reduced to about half.

It has recently been shown that a potential caveolin-binding sequence is present within the kinase domain of many protein kinases, including tyrosine kinase receptors (Couet et al., 1997). To study if this sequence is important for the trafficking of aFGF, we prepared a receptor mutant (FGFRalan4) where three of the aromatic residues in the putative caveolin-binding sequence were changed to alanines (Fig. 5A). The expressed mutant receptor bound [³⁵S]methionine-labelled aFGF to the same extent as wild-type FGFR4 (Fig. 5C). We then studied if the mutant receptor cofractionated with caveolin-containing membrane structures, as measured by floatation in a discontinuous sucrose gradient. The data in Fig. 5B show that a considerable part of ¹²⁵I-aFGF crosslinked to FGFRalan4 and was found in the same fraction as caveolin, similar to the results obtained earlier with cells transfected with the wild-type receptor (Citores et al., 1999). The data indicate that the putative caveolin-binding amino acid stretch does not account for the presence of the receptor in the caveolin-containing fractions.

Subdomain IX in the kinase domain of the FGF-receptor contains a conserved short stretch that includes the nearly invariant residues corresponding to Asp671 and Gly676 (Couet et al., 1997; Hanks et al., 1988). Because the amino acid stretch mutated in FGFRalan4 is present within this highly conserved region, we studied if the mutation influences the kinase activity of the receptor. We took advantage of the observation that when aFGF-treated cells expressing FGFR4 are lysed and immunoprecipitated with an antibody against the C-terminus of the receptor, an 85 kDa serine kinase is associated with the immunoprecipitate, owing to its interaction with the activated and autophosphorylated FGFR4 (Vainikka et al., 1996). Upon subsequent incubation of the immunoprecipitate with [γ-³²P]ATP, the receptor is specifically labelled at serine residues. Therefore, phosphorylation of the receptor in vitro indicates that the receptor had become tyrosine phosphorylated in vivo. A phosphorylated band with a migration rate corresponding to FGFR4 was obtained in cells transfected with FGFR4 (Fig. 5D, lane 2), but not in cells transfected with FGFRalan4 (lane 3). Anti-FGFR4 antibodies immunoprecipitated similar amounts of the two receptors, as measured by western blot (not shown). Because the lack of phosphorylation of the mutant receptor could be caused by a problem in the association of FGFRalan4 with the serine kinase, we used a more direct assay. In this case, the receptor was immunoprecipitated from the cell lysate with anti-FGFR4 and analyzed by western blot probed with anti-phosphotyrosine. In FGFR4-transfected cells, a phosphorylated band corresponding to FGFR4 was observed, whether or not aFGF was added, probably owing to high receptor density, but this was not the case in FGFRalan4-transfected cells (Fig. 5E). Western blot with anti-FGFR4 demonstrated that the wild-type receptor and the FGFRalan4 mutant were expressed to about the same extent (Fig. 5F). These data indicate that the tyrosine kinase domain of FGFRalan4 had been inactivated by the mutations, probably owing to a conformational change.

To further study the role of the kinase and of the putative caveolin-binding domain, COS-1 cells were transfected with a kinase-negative point mutant of the receptor, FGFR4-K503R (Fig. 6A), and with a deletion mutant (FGFR4ΔBgl) where most of the cytoplasmic part, including the whole kinase domain, had been removed (Fig. 6B). In both cases, part of the mutant FGF receptors co-fractionated with caveolin at the interphase between the low and the high sucrose density of the gradient (fractions 4 and 5). Clearly therefore, the putative caveolin-binding domain is not the reason that FGFR4 is partially found in the caveolin-containing fractions.

In an earlier paper (Citores et al., 1999), we demonstrated that upon endocytosis the wild-type receptor FGFR4 accumulates in a juxtanuclear region that was identified as the recycling endosome compartment. To investigate what properties of the growth factor receptor determine the intracellular transport route, we studied the trafficking of Cy3-labelled aFGF in cells transfected with receptors that had different mutations in the cytoplasmic domain. The FGFRalan4 mutations completely changed the routing of internalized growth factor. In this case, the growth factor remained in vesicular structures dispersed over the whole cytoplasm (Fig. 7A, right). Bafilomycin A₁, which inhibits acidification of endosomes and could therefore alter the routing, did not affect this distribution even after incubation for 15 hours (Fig. 7B, right). In the absence of bafilomycin A₁, the growth factor was degraded after this long incubation period (not demonstrated). The kinase negative mutant (FGFR4-K503R) displayed altered trafficking of the growth factor resembling that observed with FGFRalan4 (Fig. 7C, left), although accumulation in the juxtanuclear compartment also occurred (see below).

Kinase-negative mutants can also be obtained by deleting most of the intracellular part of the receptor, including the kinase domain (FGFR4ΔBgl) (Fig. 1A). Surprisingly, in this case the routing of the growth factor to the juxtanuclear area was not different from that of the wild-type FGFR4 (Fig. 7C, right). Similar distribution of aFGF was found in cells transfected with the almost full length receptor, FGFR4ΔB1, which has intact kinase (Fig. 7D, left). When essentially the whole cytoplasmic domain was removed (FGFR4ΔEag/BamH), the receptor remained mostly at the cell surface in cells with high expression level (not demonstrated). By contrast, in cells expressing fewer receptors, transport to the juxtanuclear area was observed (Fig. 7D, right).

In attempts to characterize the vesicles containing aFGF in cells transfected with kinase-negative receptors, transfected cells were incubated with Cy3-aFGF and then labelled with antibodies to EEA-1, a marker for early endosomes (Mu et al., 1995), or antibodies to LAMP-1, a marker for late endosomes and lysosomes (Geuze et al., 1988). As shown in Fig. 8, in cells transfected with FGFRalan4 there was little colocalization of Cy3-aFGF with EEA-1, but good colocalization with LAMP-1. This indicates that much of the growth factor was in late endosomes and lysosomes in cells transfected with FGFRalan4. In addition, in cells transfected with FGFR4-K503R there was considerable colocalization with LAMP-1, whereas, in cells transfected with wild-type receptor, this was not the case.

In cells transfected with FGFRalan4 there was little overlap of labelled aFGF and transferrin compared with cells transfected with wild-type receptor (Fig. 9), indicating that the growth factor did not reach the recycling endosome compartment in cells transfected with FGFRalan4. An intermediate picture was obtained in cells transfected with FGFR4-K503R, whereas, in cells transfected with FGFR4ΔBgl, Cy3-aFGF was colocalized with transferrin in the recycling endosome compartment as in cells transfected with the wild-type receptor.

To test the findings biochemically, subcellular fractionation was performed to determine the fate of the internalized ¹²⁵ I-aFGF in COS cells transfected with FGFR4 and FGFRalan4. After 2 hours of incubation with aFGF, the postnuclear supernatants were fractionated on linear sucrose gradients and the radioactivity in each fraction of the gradients was measured (Fig. 10A,B). To determine the position of endosomes and lysosomes in the gradient, the cells were incubated with ¹²⁵I-transferrin for 1 hour at 4°C and then transferred at 37°C for 20 minutes. In this case, a large amount of the radioactivity in the gradient was found in fractions 7-11, which contain early and recycling endosomes. In addition, the distribution of the lysosomal marker β-N-acetylglucosaminidase is shown (Fig. 10B). The activity of this enzyme appeared in fractions 11-15, whereas the activity in fractions 7-11 was very low, indicating that these fractions were not contaminated much by lysosomal content.

In FGFR4-transfected cells incubated with aFGF, the radioactivity appeared at a density corresponding to that of endosomes and lysosomes (Fig. 10A), whereas, in FGFRalan4 transfected cells, most ligand sedimented at a density corresponding to lysosomes (Fig. 10A).

In agreement with this, analysis of the content of the fractions after adsorption to heparin-Sepharose followed by SDS-PAGE and autoradiography indicated a clearly different distribution of aFGF in the two transfected cell cultures (Fig. 10C,D). Whereas a considerable part of the aFGF was found in the fractions containing transferrin in FGFR4-transfected cells, there was little aFGF in these fractions in the FGFRalan4-transfected cells. In FGFRalan4-transfected cells the growth factor appeared concentrated in the lysosomal fractions. It should be noted, however, that even in the cells transfected with the wild-type receptor a large part of the aFGF was present in the lysosomal fractions. This was more evident when the total radioactivity in the fractions was measured (Fig. 10A) than when the amount of intact growth factor was measured (Fig. 10C,D), indicating that the material in the lysosomes was partially degraded.

The data are consistent with the possibility that, like transferrin, a considerable part of aFGF accumulates in the recycling endosome compartment in cells transfected with the wild-type receptor, while the growth factor taken up by FGFRalan4-transfected cells accumulates mainly in lysosomes.

We studied the endocytic uptake and transport of aFGF in HeLa dyn^K44A cells transfected with wild-type and mutant FGFR4. For this purpose, Cy3-labelled growth factor was incubated with HeLa dyn^K44A cells under conditions where the uptake from coated pits was blocked by the expression of mutant dynamin. In cells transfected with the wild-type receptor (FGFR4), the growth factor was internalized and transported to the juxtanuclear organelle previously identified in COS cells as the recycling endosome compartment (Fig. 11).

Similar results were obtained when the expression of the mutant dynamin was suppressed by the presence of tetracyclin (data not shown). In the case of the mutant receptor lacking the cytoplasmic part, FGFR4ΔEag/BamH, some aFGF was transported to the juxtanuclear location, but the majority of the growth factor remained at the cell surface (Fig. 11), in accordance with the data in Fig. 2 showing that this mutant is inefficiently endocytosed. By contrast, in cells transfected with the two full length but kinase-negative receptors, FGFRalan4 and FGFR4-K503R, the growth factor was not detectably transported to the juxtanuclear organelle, but remained in vesicular structures spread out over the cytoplasm (Fig. 11).

The experiments presented so far were carried out in the presence of heparin to prevent binding of aFGF to surface heparans that would make the interpretation of the receptor-dependent uptake difficult. In the absence of heparin, the growth factor binds extensively even to cells that lack the specific FGF receptors and we decided to study uptake and intracellular transport of aFGF taken up by this pathway in COS cells lacking specific FGF receptors. The data in Fig. 12A demonstrate that rather than accumulating in the recycling compartment, like transferrin, the growth factor taken up by the heparan pathway accumulated in vesicles that contained the marker for late endosomes and lysosomes, LAMP-1. Little growth factor was present in vesicles containing the early endosome marker EEA-1, indicating that the growth factor was rapidly transported to the lysosomes. In addition, there was no co-localization with transferrin. It is therefore clear that aFGF taken up after binding to cell surface heparans is transported very differently from growth factor bound to wild-type FGFR4, but similarly to growth factor bound to FGFRalan4.

Upon internalization of the receptor for EGF induced by ligand binding, the receptor is rapidly downregulated by targeting to lysosomes (Levkowitz et al., 1998). To compare the intracellular distribution of EGF with that of heparan-bound aFGF, we allowed COS cells transiently transfected with EGFR but lacking FGF receptors to take up the two growth factors that were labelled with different fluorescent dyes. Figure 12B demonstrates that EGF is sorted like heparan-bound aFGF.

We have studied endocytosis and intracellular routing of aFGF in cells transfected with wild-type and mutant FGFR4. To measure endocytosis by clathrin-dependent and -independent endocytosis, we took advantage of the finding that cells expressing a dominant negative mutant of dynamin cannot endocytose by the clathrin-dependent pathway (Damke et al., 1994) and from caveolae (Henley et al., 1998; Oh et al., 1998).

We found that in HeLa dyn^K44A cells transfected with the wild-type FGFR4, endocytosis was reduced almost by half when the mutant dynamin was expressed. This was also the case with a mutant receptor that lacks the C-terminal 11 amino acids, but has intact kinase. Our findings are in accordance with previous studies showing that phosphorylation of a tyrosine residue close to the C-terminus of the receptor is necessary for efficient uptake of the growth factor (Sorokin et al., 1994). This residue is retained in FGFR4ΔB1. By contrast, in HeLa dyn^K44A cells transfected with receptors that had an inactivated or deleted kinase domain, there was little difference in endocytic uptake of aFGF, whether or not the cells expressed the mutant dynamin. Clearly, the growth factor is to a large extent endocytosed by a route not involving clathrin-coated pits or caveolae, and this pathway is not dependent on the kinase activity of the receptor.

We have recognized a difference between COS and HeLa cells in that, in the case of the COS cells, the uptake of aFGF occurs only by an endocytosis mechanism not involving clathrin-coated pits, whereas both mechanisms are active in HeLa cells. Endocytosis by an alternative mechanism not involving clathrin-coated pits is involved in the internalization of several bacterial and plant toxins such as cholera toxin, tetanus toxin and ricin (Olsnes et al., 1999). Ricin is endocytosed by both mechanisms in a similar way to aFGF in HeLa cells (Llorente et al., 1998). It is possible that both mechanisms are operative in COS cells but that only a small fraction of the growth factor is taken up by the clathrin-coated pit pathway and that this was not detectable in our assay. It is unlikely that the alternative endocytosis of aFGF occurs from caveolae because this uptake route is blocked by expression of the mutant dynamin K44A (Henley et al., 1998).

All four FGFRs contain a putative caveolin-binding domain. We mutated this domain, expressed the construct in cells and studied the distribution of the receptor in a discontinuous sucrose gradient. The mutation did not affect the cofractionation of the receptor with caveolin. In mutants where the putative caveolin-binding domain was deleted, the receptor exhibited the same distribution in the gradient. This indicates that other parts of the receptor, perhaps the transmembrane domain, must bind to structures of low buoyant density.

In the case of FGFRalan4, we did not observe transport of the growth factor to the recycling compartment as we did with the wild-type receptor. Instead, the growth factor was transported to lysosomes. In cells transfected with FGFR4-K503R, the growth factor was partly found in lysosomes. Surprisingly, when the major part of the cytoplasmic domain of the receptor, including the whole kinase domain, was removed, the receptor migrated to the recycling compartment, apparently in the same manner as the wild-type receptor. The possibility therefore exists that the full-length receptor interacts with some component in endosomes that impedes its proceeding further to the recycling compartment and that this interaction is eliminated by the kinase activity of the receptor. The release could depend on phosphorylation of a tyrosine residue in the receptor or on a putative binding protein in the endosomes. The deletion mutants, being devoid of the receptor part that interacts with the putative endosomal binding protein, could proceed to the recycling compartment without the requirement for phosphorylation. Further experiments are required to elucidate the nature of the interaction.

The endocytosed aFGF was partially recycled back to the cell surface and released into the medium. This process was inefficient compared with the rate obtained with transferrin (Ciechanover et al., 1983). Interestingly, the full-length kinase-negative mutants FGFRalan4 and FGFR4-K503R that were unable or inefficient in reaching the recycling compartment, recycled aFGF back to the medium at approximately the same rate as the wild-type receptor. The reason for this is probably that, in these mutants, the recycling occurs directly from the sorting endosomes. In cells transfected with the wild-type receptor, most of the recycling could occur at the level of the sorting endosomes (Ren et al., 1998; Hao and Maxfield, 2000). Degradation of aFGF occurred more rapidly in cells transfected with FGFRalan4 than in cells transfected with the wild-type receptor, in accordance with its transport to late endosomes and lysosomes.

Cells transfected with the mutant receptor, FGFR4ΔEag/BamH, endocytosed aFGF at a strongly reduced rate. Because the mutant receptor was well expressed, this resulted in accumulation of Cy3-labelled aFGF at the cell surface. Compared with wild-type receptor, a larger part of the endocytosed growth factor was recycled back to the cell surface and released into the medium in cells transfected with this mutant. It therefore appears that this mutant lacks a signal for endocytosis and retention in the vesicular compartments that the other mutants and the wild-type receptor possess. In particular, FGFR4ΔBgl contains only 56 amino acids more of the juxtamembrane cytoplasmic tail than FGFR4ΔEag/BamH (the latter construct contains in addition the 13 most C-terminal amino acids of the receptor). The stretch missing in FGFR4ΔEag/BamH contains a dileucine and a leucine-valine motif. Preliminary experiments deleting these signals gave a gradual reduction in endocytosis rate, indicating that both are necessary for maximal activity.

In the absence of heparin, the growth factor binds abundantly to surface heparans. We have earlier demonstrated that, in contrast to binding to the specific receptors, binding to heparans does not result in translocation to the cytosol and nucleus. Instead, the growth factor is degraded (Wiedlocha et al., 1996). Here, we demonstrate that heparan-bound aFGF is transported to lysosomes upon internalization. In contrast to receptor-bound growth factor, which is transported to the recycling compartment, the heparan-bound growth factor is not. Thus, in addition to mediating signal transduction and translocation of aFGF, FGFR4 clearly plays a role in specific intracellular transport of the aFGF-FGFR4 complex, which may be important for signalling.

In the case of EGF that is normally routed together with its receptor to the lysosomes for degradation (Levkowitz et al., 1998), it was recently found that threonine phosphorylation of the receptor by protein kinase C resulted in transport of the receptor to the recycling compartment (Bao et al., 2000). Whether or not phosphorylation of a residue in the FGF receptor is involved in its transport to the recycling compartment is not known.

It is not clear if endocytosis and transport to the recycling compartment is involved in the translocation of aFGF to the cytosol and nucleus. Conceivably, the translocation could occur directly across the surface membrane (e.g. from the caveolin-containing lipid rafts). Bacterial and plant toxins that can cross biological membranes and reach the cytosol, where they act enzymatically, are translocated either from the acidified endosomes or from the endoplasmic reticulum, in those cases where the mechanism has been sufficiently elucidated (Olsnes et al., 1999). There are a large number of protein toxins that enter the cytosol and, in most cases, the translocation mechanism is not known. It is possible that some of the toxins employ the same entry mechanism as aFGF.

L. C. and O. K. are Post-doctoral Fellows, and J. W. and D. K. are Pre-doctoral Fellows of The Norwegian Cancer Society. This work was supported by Novo Nordisk Foundation, The Norwegian Research Council, Blix Fund for the Promotion of Medical Research, Rachel and Otto Kr. Bruun’s legat and by The Jahre Foundation. The skilful technical assistance of Mette Sværen and the kind help of Alicia Llorente is gratefully acknowledged. We thank Kirsten Sandvig, Pål Falnes and Harald Stenmark for critical reading of the manuscript.