Translocation of cytosol of exogenous, CAAX-tagged acidic fibroblast growth factor.

Acidic fibroblast growth factor (aFGF) added externally to cells has been proposed to enter the nucleus and stimulate DNA synthesis, but it has remained controversial whether or not exogenous aFGF has the capability of crossing cellular membranes. To test this, a novel principle to study translocation of proteins to the cytosol was developed by fusing a C-terminal farnesylation signal, a CAAX tag (C = Cys, A = an aliphatic amino acid, and X = any amino acid), onto aFGF. Farnesylation is only known to occur in the cytosol and possibly in the nucleus. When incubated with NIH3T3 cells overnight, about one-third of the cell-associated, CAAX-tagged growth factor was farnesylated, indicating that efficient translocation had taken place. Binding to specific FGF receptors was required for translocation to occur. Part of the farnesylated growth factor was found in the nuclear fraction. The data indicate that CAAX-tagged aFGF added externally to cells is able to cross cellular membranes and enter the cytosol and the nucleus.

Acidic fibroblast growth factor (aFGF), 1 which promotes cell differentiation and proliferation (Burgess and Maciag, 1989;Basilico and Moscatelli, 1992), binds to transmembrane receptors containing a cytoplasmic tyrosine kinase domain that is activated upon binding (Johnson and Williams, 1993;Ruta et al., 1989;Mason, 1994). It has remained controversial if aFGF transmits signal only through the receptor or if the growth factor is also translocated into the cells and to the nucleus to stimulate DNA synthesis (Imamura et al., 1990;. We recently demonstrated that aFGF fused to diphtheria toxin A fragment could be translocated ("microinjected") by the diphtheria toxin pathway to the cytosol and to the nucleus and stimulate DNA synthesis in cells lacking functional aFGF receptors (Wiȩdłocha et al., 1994). Clearly, therefore, intracellular growth factor can convey a biological signal.
Morphological and biochemical evidence for nuclear target-ing of externally added aFGF and certain other growth factors has been provided by several groups (Imamura et al., 1990;Baldin et al., 1990;Kimura, 1994;Wiȩdłocha et al., 1994;Moroianu and Riordan, 1994;Imamura et al., 1994). A major problem in such studies is to demonstrate that the protein has really crossed cellular membranes to gain access to the cytosol or to the nucleoplasm. Alternatively, it could be present in intracellular vesicles or cisternal compartments that may have a juxtanuclear location (Prudovsky et al., 1994), or it could be trapped in cytoskeletal material adhering to the nucleus upon disrupting or dissolving the cells.
To test for translocation in a more rigorous manner, we have developed a novel principle to study whether or not a protein is exposed to the cytosol. We added to aFGF a C-terminal CAAX (C ϭ Cys, A ϭ an aliphatic amino acid, and X ϭ any amino acid) motif, which, upon exposure to the cytosol, will be modified by attachment of a prenyl group followed by proteolytic removal of the last three amino acids and carboxyl methylation of the appearing C-terminal cysteine (Clarke, 1992;Cox and Der, 1992). Prenylating enzymes have been observed only in the cytosol (Reiss et al., 1990;Schaber et al., 1990;Casey et al., 1991) and possibly in the nucleus (Lutz et al., 1992;Sinensky et al., 1994), and there is no indication that prenylation occurs at the cell surface or inside compartments of the endocytic pathway. In the present study, we have found that externally added CAAX-tagged aFGF is efficiently farnesylated by cells, indicating that translocation of the growth factor to the cytosol and to the nucleus does indeed take place.
Cell Cultures-NIH3T3 cells and the human osteosarcoma cell line U2OS Dr1 were propagated as earlier described (Wiȩdłocha et al., 1994). Cells were seeded into Costar (Cambridge, MA) microtiter plates or 25-cm 2 Falcon flasks the day preceding the experiments.
Transfection of Cells-The expression plasmid LTRFGFR4 encoding FGF receptor 4 (Partanen et al., 1991) was a kind gift from Dr. Alitalo. pMamNeo encoding the neomycin phosphotransferase gene was obtained from Clontech. U2OS Dr1 cells were seeded out in 10-cm Petri dishes (10 6 cells per dish) in Dulbecco's modified minimal essential medium(DMEM) containing 5% fetal calf serum (FCS). The next day the cells were washed in serum-free medium, and then 2.5 ml of serumfree DMEM, containing 5 g of LTRFGFR4 and 5 g of pMamNeo DNA, and 20 l of a 1 mg/ml aqueous solution of DOTAP (Boehringer Mannheim), which had previously been vortexed and left for 5 min at room temperature, was added. The cells were incubated for 5 h at 37°C with occasional careful shaking. Then 250 l of FCS and 2.5 ml of medium containing 5% FCS was added, and the cells were incubated overnight. The next day, the medium was changed, the cells were trypsinized, diluted 1:3, and seeded out into new Petri dishes. After 4 days the medium was removed, and DMEM containing 5% FCS and 1 mg/ml gentamycin was added. Small colonies developed after 2 weeks. The cells were then incubated further in DMEM containing 0.5% FCS and 10 ng/ml aFGF. Colonies that grew under these conditions were tested for aFGF-stimulated incorporation of [ 3 H]thymidine as described (Wiȩdłocha et al., 1994). One transfectant (U2OS Dr1 R4) was selected.
pTrc-dtA-cax in E. coli DH5␣ was induced as above, and the bacteria were harvested, washed once with PBS (10 mM sodium phosphate, pH 7.4, 0.14 M NaCl), resuspended in PBS, and sonicated. After centrifugation, the clear supernatant was chromatographed on a Sephacryl S-100HR column equilibrated with 20 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 1 mM dithiothreitol. Fractions containing dtA-CAAX were collected, dialyzed against 20 mM phosphate buffer, pH 6.0, 1 mM dithiothreitol, and chromatographed on a Q-cartridge (Bio-Rad) with a linear NaCl gradient (0 -1 M) in the same buffer.
Measurements of DNA Synthesis-Cells growing in 24-well microtiter plates (5 ϫ 10 4 cells/well) were preincubated for 48 h in serum-free medium at 37°C. Then, the cells were treated with increasing amounts of aFGF or aFGF-CAAX in the absence and presence of 10 units/ml heparin, and the incubation was continued for 24 h at 37°C. During the last 6 h, the cells were incubated with 1 Ci/ml [ 3 H]thymidine (Amersham) as described (Imamura et al., 1990), and the incorporated radioactivity was measured.
In Vitro Transcription and Translation-Plasmid DNA was linearized downstream of the encoding gene and transcribed in a 20-l reaction mixture with T3 RNA polymerase as described (McGill et al., 1989). The mRNA was precipitated with ethanol and dissolved in 10 l of H 2 O containing 10 mM dithiothreitol and 0.1 unit/l RNasin. The translation was performed for 1 h at 30°C in micrococcal nuclease-treated rabbit reticulocyte lysate (Promega) using 5 l of the dissolved mRNA per 100 l of lysate and 1 M [ 35 S]methionine (Amersham).
In Vitro Farnesylation-For in vitro labeling, 0.2 Ci of [ 3 H]farnesyl pyrophosphate (DuPont NEN) and 2 ng of aFGF-CAAX or aFGF were added to 20 l of reticulocyte lysate (Promega). The mixture was adjusted to contain 5 mM MgCl 2 . In some cases, 1 l of dog pancreatic microsomes (Promega), 50 M B581, or both were also added. The mixtures were then incubated for 30 min at 37°C. Labeled proteins were recovered from the reaction mixture by immunoprecipitation with anti-aFGF antibody (Sigma) adsorbed to protein A-Sepharose and analyzed by SDS-PAGE and fluorography (Wiȩdłocha et al., 1992).
Analysis of in Vivo Farnesylation-Cells growing in 25-cm 2 flasks were serum-starved for 36 h and then preincubated overnight in serumfree medium containing 1 Ci/ml [ 3 H]mevalonic acid (DuPont NEN), 1 Ci/ml [ 14 C]mevalonic acid, and 4 g/ml lovastatin (to prevent formation of endogenous mevalonic acid). aFGF or aFGF-CAAX was added, and the cells were incubated overnight. In some cases the growth factor was added simultaneously with the labeled mevalonic acid. The cells were then washed twice with HEPES medium containing 5 units/ml heparin and three times with HEPES medium without heparin, then lysed (Wiȩdłocha et al., 1994) and centrifuged for 5 min at 14,000 rpm at 4°C. The supernatant was centrifuged once more, and the second supernatant (the cytoplasmic fraction) was rotated for 2 h at 4°C with 30 l of heparin-Sepharose (Pharmacia) or protein A-Sepharose C-4B that had previously been treated with 2 l of rabbit anti-bovine aFGF (Sigma). In the case of heparin-Sepharose, the matrix was subsequently washed with 0.7 M NaCl to remove less strongly bound material. The nuclear pellets were washed twice in lysis buffer containing 0.4 M sucrose and 1 mM MgCl 2 , layered over 0.8 ml of 0.7 M sucrose and centrifuged for 15 min at 3,000 rpm at 4°C. The nuclei were then sonicated and extracted with 0.5 M NaCl. After clarification by centrifugation for 5 min at 14,000 rpm at 4°C, the extract was diluted in PBS containing 0.1% Triton X-100 and subjected to heparin-Sepharose ad-sorption or immunoprecipitation as above. In each case, the adsorbed material was analyzed by SDS-PAGE and fluorography (Wiȩdłocha et al., 1992).
Triton X-114 Partitioning of Farnesylated Proteins-Triton X-114 equilibrated with PBS was diluted to a 20% stock solution (Madshus et al., 1984). PBS (300 l) and 100 l of the Triton X-114 stock solution were mixed and kept on ice. Protein was added, and the mixture was kept at 0°C for further 15 min. To separate the phases, the solution was incubated for 15 min at 37°C and then centrifuged for 2 min at room temperature in an Eppendorf centrifuge. The upper (water) phase was transferred to a new tube, and the lower (Triton) phase was washed at 37°C with 500 l of PBS. Both phases were then diluted to 1 ml with PBS (0°C), followed by immunoprecipitation with anti-aFGF antibodies adsorbed to protein A-Sepharose or with heparin-Sepharose. The precipitated material was analyzed with SDS-PAGE and fluorography.
To study partitioning of proteins from cells, the cells were lysed directly in the Triton/PBS mixture in the presence of 1 mM phenylmethylsulfonyl fluoride, and, after removal of the nuclei by centrifugation, the procedure for phase separation described above was followed.
SDS-PAGE-Polyacrylamide gel electrophoresis in the presence of SDS was carried out in 12% gels as described by Laemmli (1970). After electrophoresis, the gels were fixed for 30 min in 27% methanol, 4% acetic acid and then incubated for 30 min in 1 M sodium salicylate, 2% glycerol, pH 5.8. Kodak XAR-5 film was exposed to the dried gel in the absence of intensifying screen at Ϫ80°C for 3-4 weeks.

Acidic Fibroblast Growth Factor with a CAAX Tag Is
Biologically Active-The farnesylation signal (Cys-Val-Ile-Met) of the K-Ras-4B protein has been shown to be sufficient for farnesylation both in vitro (Goldstein et al., 1991) and in vivo (Hancock et al., 1991a), and we added this CAAX motif to the C terminus of aFGF. When recombinant aFGF-CAAX and [ 3 H]farnesyl pyrophosphate were incubated in a reticulocyte lysate (as a source of farnesyl transferase), a labeled 15-kDa band was obtained (Fig. 1, lane 2), migrating somewhat more rapidly than non-farnesylated [ 35 S]methionine-labeled aFGF-CAAX (lane 1). When microsomes were added as a source of enzymes required for cleavage and carboxyl methylation (Hancock et al., 1991b), the migration rate of the farnesylated protein was slightly increased (lane 3), indicating that complete processing had occurred. In the presence of the farnesylation inhibitor B581, a peptidomimetic agent resembling the farnesylation signal (Garcia et al., 1993), the growth factor was not labeled (lanes 4 and 5). Also, aFGF without a CAAX sequence was not labeled, whether or not microsomes were present (lanes 6 and 7).
To test the biological activity of aFGF-CAAX, we measured its ability to bind to and stimulate DNA synthesis in NIH3T3 cells, which express functional FGF receptors (Imamura et al., 1990). Scatchard analysis of binding data obtained in the presence of 10 units/ml heparin showed that NIH3T3 cells contain 74,000 binding sites for aFGF with a K d of 150 pM (data not shown). As shown in Fig. 2A, aFGF and aFGF-CAAX were equally efficient in competing out the binding of 125 I-aFGF to the cells. Furthermore, aFGF-CAAX, like aFGF, stimulated DNA synthesis in serum-starved NIH3T3 cells in a dose-dependent manner, both in the absence and presence of heparin ( Fig. 2B). Although aFGF-CAAX was less active than aFGF at low concentrations, it appears that the addition of a C-terminal CAAX motif does not seriously impair the biological functions of the growth factor.
CAAX-tagged aFGF Added Externally to Cells Is Farnesylated in Vivo-To investigate whether aFGF-CAAX added externally to cells becomes farnesylated, NIH3T3 cells were incubated overnight with aFGF or aFGF-CAAX in the presence of labeled mevalonate, a precursor of the farnesyl group. As shown in Fig. 3A, a labeled band with migration rate corresponding to 15.5 kDa was immunoprecipitated with anti-aFGF from lysed cells that had been treated with aFGF-CAAX (lane 2) but not in cells treated with aFGF (lane 3). The labeled protein was, like aFGF (Imamura et al., 1990), adsorbed to heparin-Sepharose at high ionic strength (lane 4). In the case of aFGF, no labeled material was adsorbed to heparin-Sepharose (lane 5).
To compare the labeling of aFGF-CAAX with that of endogenous prenylated proteins, we analyzed the total trichloroacetic acid-precipitable material in the cytoplasmic fraction. The results (lane 6) showed a labeled band migrating as expected for prenylated aFGF-CAAX (position indicated with an asterisk). There was no band at the same position in cells treated with aFGF (lane 7). The labeling extent of externally added aFGF-CAAX was comparable to that of prenylated endogenous proteins in the cells, such as the small G proteins (lane 6). There was no labeling when aFGF-CAAX was added to lysed cells on ice, excluding the possibility that the labeling seen in Fig. 3A could have occurred after lysis of the cells (data not shown).
We and others (Zhan et al., 1992;Wiȩdłocha et al., 1994) have earlier found that when NIH3T3 cells are incubated with labeled aFGF, part of the growth factor is recovered in the nuclear fraction. In the experiments in Fig. 3B, we therefore analyzed both the cytoplasmic and the nuclear fractions for material adsorbing to heparin-Sepharose and to immobilized anti-aFGF. When cells were incubated with aFGF in the presence of labeled mevalonate, as expected no labeled growth FIG. 2. Ability of aFGF-CAAX to bind to and stimulate DNA synthesis in NIH3T3 cells. A, to measure binding, 5 ng/ml 125 I-aFGF and increasing concentrations of unlabeled aFGF or aFGF-CAAX in the presence of 10 units/ml heparin were incubated for 2 h at 4°C with cells growing as monolayers in 24-well microtiter plates at a density of 5 ϫ 10 4 cells/well. The cells were washed, and the bound radioactivity was measured. B, to measure [ 3 H]thymidine incorporation, cells were preincubated for 48 h in serum-free medium at 37°C; then, increasing amounts of aFGF or aFGF-CAAX were added with and without 10 units/ml heparin, and the incubation was continued for 24 h at 37°C. During the last 6 h, the cells were incubated with 1 Ci/ml [ 3 H]thymidine, and the incorporated radioactivity was measured.

FIG. 3. Farnesylation of aFGF-CAAX in NIH3T3 cells.
A, serum-starved cells were incubated with 10 units/ml heparin and 100 ng/ml aFGF or aFGF-CAAX overnight in the presence of lovastatin and labeled mevalonic acid. The cells were lysed, and the postnuclear supernatant was divided into three parts. One part was treated with anti-aFGF adsorbed to protein A-Sepharose, and one part was treated with heparin-Sepharose, which was subsequently washed with 0.7 M NaCl. The third part was precipitated with 5% trichloroacetic acid. The precipitated material was analyzed by SDS-PAGE and fluorography. The position of small GTP-binding proteins (Small G) and of farnesylated aFGF-CAAX (asterisk) are indicated. B, increasing amounts of aFGF or aFGF-CAAX were added to 2 ϫ 10 6 serum-starved cells in the absence and presence of 50 M B581. The cells were incubated overnight and lysed. Growth factor in the cytoplasmic fraction and in material extracted with 0.5 M NaCl from the sonicated nuclei was precipitated with heparin-Sepharose (upper panel) or with anti-aFGF-protein A-Sepharose (lower panel). In all cases, the material was analyzed by SDS-PAGE and fluorography.

Entry of CAAX-tagged aFGF
factor was precipitated with heparin-Sepharose (upper panel) or with immobilized anti-aFGF (lower panel) either from the cytoplasmic (lanes 2 and 3) or from the nuclear fractions (lanes 8 and 9). When aFGF-CAAX was used instead, labeled growth factor was precipitated from both fractions, both with immobilized heparin and immobilized anti-aFGF (lanes 4, 5 and 10, 11). The relative amount of growth factor in cytoplasm and nucleus varied somewhat between experiments. In the presence of the farnesylation inhibitor B581, no labeling was observed ( lanes 6 and 12). The results indicate that externally added aFGF-CAAX is indeed farnesylated by the cells and that part of the modified protein migrates to the nucleus.

Requirement of FGF Receptors for in Vivo Farnesylation of aFGF-CAAX-To test if binding to FGF receptors is required
for farnesylation, NIH3T3 cells were incubated with aFGF-CAAX in the absence and presence of excess aFGF, which competes out binding of aFGF-CAAX to the specific receptors but not to surface heparans. Fig. 4A, lane 2, shows that aFGF-CAAX was not labeled in the presence of excess aFGF. Binding of aFGF-CAAX to specific receptors therefore appears to be required for translocation of the growth factor to the cytosol.
The possibility existed that the labeling obtained was derived from a subpopulation of cells that, although gently treated, could have been damaged but not killed, and that the growth factor could have entered the cells by a mechanism similar to that in scrape loading (Frankel and Pabo, 1988;Leevers and Marshall, 1992;Morris et al., 1993). To test this, we first measured the ability of NIH3T3 cells to modify a protein that does not bind to the cells, viz. diphtheria toxin A fragment with a C-terminal CAAX motif (dtA-CAAX). This protein is efficiently farnesylated in vitro (data not shown). The cells were incubated with increasing concentrations of dtA-CAAX and subjected to the process of scrape loading. A faint 20-kDa band was observed when 100 g/ml of the construct had been present in the medium (Fig. 4B, lane 8). This is 10 4 times the concentration required to detect labeled aFGF-CAAX in these cells (lane 2).
U2OS Dr1 cells lack specific aFGF receptors, but the growth factor binds extensively and with high affinity to surface heparans on the cells (Sakaguchi et al., 1991;Wiȩdłocha et al., 1994). In these cells farnesylation of the growth factor was observed only after exposure to 100 g/ml aFGF-CAAX and scrape loading (Fig. 4B, lanes 10 -13). These data indicate that the farnesylation of aFGF-CAAX in lanes 2-4 was not due to "leakage" of the growth factor into wounded cells, and they support the finding above that specific FGF receptors are required for translocation.
When the U2OS Dr1 cells were transfected with FGF receptor 4 (FGFR4), they bound aFGF specifically in the sense that excess unlabeled aFGF in the presence of heparin prevented the binding (Fig. 5A). The untransfected cells bound very little aFGF in the presence of heparin. When aFGF-CAAX was added to these cells in the presence of labeled mevalonic acid, a strong band corresponding to the growth factor appeared (Fig. 5B,  lane 2), which was absent in the untransfected cells (lane 1). In the presence of excess unlabeled growth factor, only a weak band appeared (lane 3). Clearly, farnesylation of externally added aFGF-CAAX was dependent upon binding to the specific receptor in U2OS Dr1 R4 cells.
Partitioning into Triton X-114 of Farnesylated aFGF-CAAX-To estimate how much of the cell-associated aFGF-CAAX becomes farnesylated, we used Triton X-114 partitioning. Farnesylated Ras has been shown to enter the detergent phase when submitted to Triton X-114 partitioning (Gutierrez et al., 1989). This was also the case with [ 35 S]methioninelabeled aFGF-CAAX, which had been farnesylated in vitro (Fig. 6A, lanes 1 and 2), but not when the farnesylation had been prevented by B581 (lanes 3 and 4).
When NIH3T3 cells were incubated with [ 35 S]methioninelabeled aFGF and then treated with Triton X-114, all the labeled protein partitioned into the water phase (Fig. 6B, lanes  1 and, 2). On the other hand, when aFGF-CAAX was used, part of the protein partitioned into the detergent phase (lanes 3 and  4). This was not the case when the cells had been incubated with labeled aFGF-CAAX in the presence of the farnesylation inhibitor B581 (lanes 5 and 6). Densitometric analysis indicated that approximately one-third of the cell-associated aFGF-CAAX was present in the detergent phase and therefore farnesylated.
Experiments with 125 I-aFGF-CAAX indicated that ϳ200,000 molecules were associated with each cell after overnight incubation under conditions as in Fig. 6B. Since not all aFGF-CAAX molecules that enter the cytosol may be farnesylated (see "Discussion"), this indicates that at least 70,000 molecules of aFGF- FIG. 4. Requirement of specific FGF receptors for farnesylation of aFGF-CAAX. A, serum-starved NIH3T3 cells were incubated overnight in serum-free medium containing lovastatin and labeled mevalonic acid. Then, aFGF-CAAX (10 ng/ml) was added in the absence and presence of 1 g/ml aFGF, and the incubation was continued for 24 h more. The cells were then lysed, and the cytoplasmic fraction and material extracted with 0.5 M NaCl from the sonicated nuclei was mixed and submitted to immunoprecipitation with anti-aFGF-protein A-Sepharose. B, NIH3T3 (lanes 1-9) and U2OS Dr1 cells (lanes 10 -13) were incubated overnight with lovastatin and labeled mevalonic acid as above. Then, 10 units/ml heparin and increasing amounts of aFGF-CAAX, dtA-CAAX, or dtA were added. When indicated, the cells were exposed to the process of scrape loading and, without change in medium, incubated further overnight. Finally, the cells were lysed and treated as above and immunoprecipitated with anti-aFGF (lanes 1-4 and 10 -14) or anti-diphtheria toxin adsorbed to protein A-Sepharose (lanes 5-9). The immunoprecipitated material was analyzed by SDS-PAGE and fluorography. Lane 14, aFGF-CAAX labeled in vitro with [ 3 H]farnesyl pyrophosphate in the presence of microsomes.

Entry of CAAX-tagged aFGF
CAAX per cell entered the cytosol.
Kinetics of Prenylation of Externally Added aFGF-CAAX-When cells were incubated with aFGF-CAAX and labeled mevalonic acid for increasing periods of time, the extent of labeling increased considerably between 8 and 12 h, but further incubation for an additional 12 h did not increase the labeling detectably (Fig. 7). This is consistent with previous observations that the entry into the cells occurs mainly in the late G 1 stage (Friedman et al., 1994;Imamura et al., 1994). DISCUSSION The main conclusion from the present work is that externally added aFGF-CAAX (and therefore also aFGF) is able to cross cellular membranes to gain access to the cytosol and/or the nucleoplasm. FGF receptors appear to be required for this, as aFGF-CAAX was not farnesylated when added to U2OS Dr1, a human osteosarcoma cell line, which does not express measurable amounts of functional FGF receptors and which does not respond to addition of aFGF with increased growth or with increased thymidine incorporation (Wiȩdłocha et al., 1994). While these cells bind aFGF extensively in the absence of heparin, no binding was detected in the presence of heparin (Wiȩdłocha et al., 1994), indicating that binding occurs only to surface heparans. The data therefore indicate that binding to specific FGF receptors is required for translocation to the cytosol.
Scrape loading has been used to introduce exogenous proteins into the cytosol, such as the Tat-encoded protein from human immunodeficiency virus (Frankel and Pabo, 1988), human papillomavirus E7 protein (Morris et al., 1993), and p21 ras (Leevers and Marshall, 1992). In these cases, extracellular protein concentrations in the range of 0.1-1 mg/ml were required to obtain measurable biological effects. This is similar to the concentration of aFGF-CAAX here found to be required to detect farnesylation in cells lacking FGF receptors, as well as of dtA-CAAX, which does not bind to cells. We therefore consider it unlikely that the farnesylation detectable in NIH3T3 cells at 10 ng/ml of extracellular aFGF-CAAX is due to growth factor that enters through occasional and transient membrane defects in the cells.
The Triton X-114 partitioning experiments indicated that about one-third of the total cell-associated aFGF-CAAX is farnesylated, which may amount to ϳ70,000 molecules per cell. The rest may either still be bound to cell surface receptors or it could be located in endocytic vesicles and other membranebounded compartments. Also, some of the CAAX-tagged growth factor present in the cytosol or in the nucleus may for some reason not be farnesylated. If farnesylation of aFGF-CAAX can FIG. 5. Farnesylation of aFGF-CAAX in U2OS Dr1 cells transfected with FGFR4. A, to untransfected and FGFR4-transfected U2OS Dr1 cells growing in 24-well microtiter plates, which had been coated with gelatine, were added increasing amounts of 125 I-aFGF (specific activity, 15,000 counts/min/ng) in the presence of 10 units/ml heparin. After 4 h at 4°C, the cells were washed and dissolved in 0.5 M NaOH, and the cell-associated radioactivity was measured. B, serumstarved untransfected and transfected U2OS Dr1 cells were incubated overnight in serum-free medium containing lovastatin and labeled mevalonic acid. Then, aFGF-CAAX (100 ng/ml) was added in the absence and presence of 1 g/ml unlabeled aFGF, as indicated, and the incubation was continued for further 24 h. The cells were then lysed, and the cytoplasmic fraction and material extracted with 0.5 M NaCl from the sonicated nuclei was mixed and submitted to immunoprecipitation with anti-aFGF-protein A-Sepharose.
FIG. 6. Ability of aFGF-CAAX farnesylated in vitro and in vivo to partition into Triton X-114. A, aFGF-CAAX was synthesized in a rabbit reticulocyte lysate in the presence of [ 35 S]methionine and then incubated with farnesyl pyrophosphate under conditions as in Fig. 1, in the absence and presence of 50 M B581. 2 l were subjected to Triton X-114 partitioning. The water phase (W) and the detergent phase (D) were treated with anti-aFGF adsorbed to protein A-Sepharose, and the immunoprecipitated material was analyzed by SDS-PAGE and fluorography. B, NIH3T3 cells were incubated overnight with 10 units/ml heparin and 5 ng/ml [ 35 S]methionine-labeled aFGF (lanes 1 and 2) or aFGF-CAAX in the absence (lanes 3 and 4) and presence of B581 (lanes 5 and 6). The cells were lysed, and the cytoplasmic fraction and material extracted with 0.5 M NaCl from the sonicated nuclei were mixed and submitted to Triton X-114 partitioning. Then, the water phase and the detergent phase were treated with heparin-Sepharose, and the adsorbed material was analyzed by SDS-PAGE and fluorography.
FIG. 7. Time course of labeling of aFGF-CAAX after addition to cells. Serum-starved U205R4 cells were incubated with 100 ng/ml aFGF-CAAX and 10 units/ml heparin for the indicated periods of time in the presence of lovastatin and labeled mevalonic acid. The cells were lysed, and the postnuclear supernatant and material extracted from the nuclei with 0.5 M NaCl was treated with heparin-Sepharose, which was subsequently washed with 0.7 M NaCl. The adsorbed material was analyzed by SDS-PAGE and fluorography. only occur in the cytosol, transport of the unmodified growth factor to the nucleus could prevent farnesylation from reaching completion. Our recent data demonstrate that farnesylation of CAAX-modified diphtheria toxin A fragment is completed only 1 h after translocation to the cytosol (Falnes et al., 1995).
Although farnesylation can only be taken as evidence for transport to the cytosol, once present in the cytosol the growth factor can migrate into the nucleus (Imamura et al., 1992;. Also, the farnesylated growth factor is partly found in the nucleus. In the experiment shown in Fig.  3B, the major part is in the cytoplasmic fraction, whereas in other experiments more was found in the nuclear fraction. The reason for this variation is not clear. The ability of aFGF to locate to the nucleus has been found to depend upon the presence of a positively charged sequence near the N terminus, and removal of this sequence rendered the growth factor unable to stimulate DNA synthesis but not to bind to and activate the specific FGF receptors (Imamura et al., 1990;Wiȩdłocha et al., 1994). Unmodified aFGF could migrate more efficiently into the nucleus than the farnesylated growth factor, which may adhere to cellular membranes by its hydrophobic tail. It should be noted, however, that a secondary signal, such as a polybasic motif or palmitoylation, in addition to prenylation is usually required to confer membrane association of the protein (Hancock et al., 1991a).
The use of a CAAX tag to study translocation of a protein to the cytosol should be useful for many purposes. We have recently shown that dtA-CAAX reconstituted with diphtheria toxin B fragment becomes farnesylated upon translocation to the cytosol by the toxin pathway and that the appearing modification can be used to monitor toxin entry under various conditions (Falnes et al., 1995). We are currently studying if CAAX-tagged growth factors other than aFGF become farnesylated when added to cells.
aFGF is synthesized as a cytoplasmic protein without a signal sequence, and it is not clear how it is transported out of the cells (Muesch et al., 1990;Jackson et al., 1995). The present finding that it also enters into the cytosol when added to the medium suggests that there must be transport mechanisms in both directions. The nature of these mechanisms is not known, but it is tempting to speculate that proteins involved are related to proteins of the ABC transporter family, such as the multidrug resistance protein (Endicott and Ling, 1989;Sharma et al., 1992), the peptide-transporting Tap proteins in the endoplasmic reticulum (Roelse et al., 1994), and to the product of the STE6 gene, transporting a-factor in Saccharomyces cerevisiae (McGrath and Varshavsky, 1989).
The location of the transport in the cells is also not known. It could occur at the level of the cell surface membrane or across membranes of intracellular vesicles. It has recently been demonstrated that binding of aFGF induces translocation of the receptor from the cell surface to a juxtanuclear location (Prudovsky et al., 1994), which could be the site where membrane translocation of the growth factor takes place. Certain protein toxins, such as diphtheria toxin, ricin, Shigella toxin, and others , are able to cross membranes of intracellular vesicles to gain access to the cytosol. In some cases, such as diphtheria toxin, the receptorbinding B fragment of the toxin facilitates translocation of the A fragment into the cytosol. However, this may not be the case with all toxins, and in the case of ricin and Shigella toxin there is no evidence that the receptor binding subunits have a function beyond binding to the relevant receptors. As we have earlier speculated (Olsnes et al., 1974), the toxins may take advantage of transport mechanisms developed by the cells for other purposes, such as transport of certain growth factors.