Phosphorylation of Geranyl and Farnesyl Pyrophosphates by Nm23 Proteins/Nucleoside Diphosphate Kinases*

The biochemical mechanism(s) by which Nm23 proteins/nucleoside diphosphate kinases suppress tumor metastasis, inhibit cell motility, and affect cellular differentiation are not known. Here we report that Nm23 proteins can phosphorylate geranyl and farnesyl pyrophosphates to give triphosphates. Wild type Nm23-H1 had higher geranyl and farnesyl pyrophosphate kinase activities than did mutants of Nm23-H1 that do not inhibit cell motility. The phosphorylation of farnesyl pyrophosphate appears to occur in vivo as cells with an elevated level of Nm23-H1 contained more farnesyl triphosphate than did control cells. To our knowledge, this is the first report that farnesyl triphosphate exists in cells. The phosphorylation of farnesyl pyrophosphate by Nm23 proteins could alter isoprenoid metabolism, and cells with an elevated level of Nm23 proteins were found to contain more farnesylated 46- and 24-kDa proteins than did control cells. The phosphorylation of geranyl and farnesyl pyrophosphates by Nm23 proteins provides a novel mechanism by which these proteins might exert their biological effects.

Nm23 proteins are nucleoside diphosphate kinases (NDP kinases) 1 (15) and catalyze the phosphorylation of nucleoside 5Ј-diphosphates to triphosphates by a ping-pong mechanism involving a high energy phosphohistidine intermediate. NDP kinases are thought to be responsible for maintaining nucleoside triphosphate pools; however, this activity does not account for all of the biological effects of Nm23 proteins. Transfection experiments have shown that different Nm23 proteins, although they have similar NDP kinase activities, have very different effects on the metastatic potential of rat mammaryadenocarcinoma cells (12), on the motility of human breast carcinoma cells (4), and on Drosophila development (16). It has been suggested that Nm23 proteins might phosphorylate other kinds of substrates.
Nm23 proteins have protein kinase or phosphotransferase activities (17)(18)(19)(20)(21). However, with the exception of the transfer of phosphate from Escherichia coli NDP kinase to a histidine on a Tar/EnvZ chimera (19), these phosphotransferase reactions have not been shown to occur in vivo, and it has not been shown under which physiological conditions E. coli NDP kinase phosphorylates proteins.
The binding of ADP to Nm23 proteins is primarily through the sugar and pyrophosphate; the base lies in a hydrophobic cleft and forms no specific polar interactions with the enzyme (22). The lack of specific interaction between the base and the enzyme accounts for the ability of NDP kinase to phosphorylate a variety of nucleoside diphosphates and suggests that Nm23 proteins might also phosphorylate other kinds of diphosphates. Here we report that Nm23 proteins phosphorylate geranyl pyrophosphate (GPP) and farnesyl pyrophosphate (FPP) to give the corresponding triphosphates in vitro and that cells with an elevated level of Nm23-H1 contain an elevated level of farnesyl triphosphate (FTP).
Isoprenoid pyrophosphates are important metabolic intermediates. FPP is a branch point in isoprenoid metabolism and is a precursor of geranylgeranyl pyrophosphate (GGPP), cholesterol, heme A, ubiquinone, and dolichol phosphate (23,24). FPP and GGPP are utilized for the isoprenylation of a number of proteins (25), many of which play critical roles in signal transduction and cell growth regulation. We have found that cells with an elevated level of Nm23-H1 contained an increased level of protein farnesylation. Purification and Phosphorylation of Rat Liver NDP Kinase and Nm23-H1-E. coli strain BL21(DE3) containing expression plasmids for Nm23-H1, Nm23-H1 P96S , Nm23-H1 S120G , Nm23-H1 S120A , and Nm23-H1 H118F (18) were gifts from Dr. Patricia Steeg. Recombinant human Nm23-H1 proteins and rat liver NDP kinase were purified as described in Refs. 20 and 21. Coomassie Brilliant Blue-stained SDSpolyacrylamide gels showed that the isolated rat liver NDP kinase and the recombinant human Nm23 proteins were greater than 99% pure. Autophosphorylated Nm23 proteins or NDP kinases were made by incubating these proteins with [␥-32 P]ATP (2 ϫ 10 15 cpm/mol) in 100 mM NaCl, 5 mM MgCl 2 , and 20 mM Tris, pH 7.5, for 5 min, and phosphorylated Nm23 proteins were separated from [␥-32 P]ATP by gel filtration (20). After gel filtration, phosphorylated Nm23 proteins contained 0.4 -0.5 mol of 32 P/mol of 18-kDa subunit. The concentrations of [ 32 P]Nm23 proteins or [ 32 P]NDP kinase given in this paper are the concentrations of 32 P bound to the proteins after gel filtration.

EXPERIMENTAL PROCEDURES
Phosphorylation of Isoprenoid Pyrophosphates-GPP, FPP, and GGPP were incubated in 100 mM NaCl, 4 mM MgCl 2 , and 50 mM Tris, pH 7.5, with 0.5-1 M [ 32 P]Nm23 proteins or 1 M Nm23 proteins and 10 M [␥-32 P]ATP. The reactions were stopped by the addition of 1/2 volume of 18 mM EDTA. Products of these reactions were analyzed by TLC on polyethyleneimine cellulose (PEI) sheets developed with either 0.2 M ammonium bicarbonate/isopropyl alcohol/acetonitrile (1:1:1) or 0.5 M NaCl and 0.75 M Tris, pH 7.5. The labeled products were detected by autoradiography and quantified using a Fuji Bio-Image Analyzer.
Phosphorylation of GPP and FPP by Cell Extracts-C100 and H1-177 cells were lysed in 0.3 M sucrose, 5 mM EGTA, 2 mM MgCl 2 , 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 20 mM PIPES, pH 6.7, and clarified by centrifugation at 100,000 ϫ g for 30 min. These extracts (3.5 mg of protein/ml) were incubated at room temperature with 40 M [␥-32 P]ATP and either GPP or FPP. The reactions were stopped by the addition of 1/2 volume of 18 mM EDTA. Products of these reactions were analyzed by TLC on PEI sheets developed with 0.2 M ammonium bicarbonate/isopropyl alcohol/acetonitrile (1:1:1).
Depletion of Nm23 Proteins from Cell Extracts-H1-177 cell extracts were incubated for 16 h at 4°C with 80 g/ml affinity-purified anti-rat NDP kinase antibodies (26) or preimmune IgG. These extracts were then incubated with Omnisorb Cells (Pierce) for 2 h at 4°C and centrifuged for 5 min at 10,000 ϫ g. The level of Nm23 proteins in these extracts were determined by phosphorylation with [␥-32 P]ATP in the presence of 20 mM EDTA. Under these conditions, Nm23 proteins are specifically phosphorylated (20).
Determination of Isoprenoid Pyrophosphate and Triphosphate Levels in Cells-C100 and H1-177 cells were incubated with 25 M lovastatin, an inhibitor of hydroxymethylglutaryl-CoA reductase, in culture medium for 8 h to reduce endogenous levels of mevalonate. The cells were then incubated for 16 h in culture medium containing 100 Ci/ml [ 3 H]mevalonolactone (60 Ci/mmol) in the presence of 25 M lovastatin. The cells were washed two times with phosphate-buffered saline and lysed with 0.5 M NH 4 OH in 75% ethanol at 75°C (27). The denatured cell suspensions were extracted three times with equal volumes of petroleum ether. The ethanolic phases were centrifuged at 10,000 ϫ g for 15 min to remove denatured proteins. The supernatants were saved, and the pellets were reextracted with 0.5 M NH 4 OH in 75% ethanol at 75°C. The supernatants were combined and clarified by centrifugation at 100,000 ϫ g for 30 min. The isoprenoid phosphates in these extracts were fractionated by reverse phase, ion pair HPLC (28) on a Hamilton PRP-1 (0.4 ϫ 15-cm) HPLC column (29) equilibrated in 75% buffer A (5 mM tetrabutylammonium phosphate, pH 8.0) and 25% buffer B (1 mM tetrabutylammonium phosphate, pH 8.0, and 60% acetonitrile). The column was eluted using the following gradient: 0 min, 25% B; 2 min, 45% B; 36 min, 65% B; 50 min, 80% B. The flow rate was at 1 ml/min, and 0.5-min fractions were collected. The radioactivity in 100-l aliquots was determined by scintillation counting.
Determination of Protein Prenylation-C100 and H1-177 cells were labeled with 100 Ci/ml [ 3 H]mevalonolactone in the presence of 25 M lovastatin. After washing, the cells were lysed in SDS sample. Aliquots containing 100 g of protein were separated on 10% SDS-polyacrylamide gel electrophoresis. The gels were stained with Coomassie Brilliant Blue, destained, equilibrated in EN 3 HANCE (DuPont), and dried. 3

Phosphorylation of GPP and FPP by [ 32 P]NDP Kinase-To
examine whether pyrophosphates, other than nucleoside diphosphates, are substrates for Nm23 proteins, autophosphorylated rat liver NDP kinase, [ 32 P]NDP kinase, was incubated with 1 mM sodium pyrophosphate, 1 mM 5Ј-phosphorylribose pyrophosphate, 1 mM thiamine pyrophosphate, or 100 M GPP for 10 min. Only the incubation with GPP resulted in the removal of 32 P from the NDP kinase (data not shown).
Thin layer chromatography was used to examine the phosphorylation of the isoprenoid pyrophosphates by NDP kinases. Incubation of [ 32 P]NDP kinase with GPP, FPP, and GGPP resulted in the formation of new 32 P-labeled compounds (Fig.  1 These results indicate that GPP and FPP, and to a lesser extent GGPP, were phosphorylated by [ 32 P]NDP kinase to give the corresponding triphosphates. These kinase reactions showed specificity as incubation of [ 32 P]NDP kinase with isopentenyl pyrophosphate, dimethylallyl pyrophosphate, geranyl monophosphate, or FMP did not result in the formation of any new 32 P-labeled compounds. The time courses of 32 P transfer from [ 32 P]NDP kinase to GPP and FPP are shown in Fig. 2, A and B. The half-lives of transfer of 32 P to GPP and FPP were 9 and 28 min, respectively (Fig. 2C). After 60 min, almost all of the 32  GPP and FPP were also phosphorylated when incubated with NDP kinase and [␥-32 P]ATP. The rates of phosphorylation of GPP and FPP by rat liver NDP kinase at room temperature were 1.2 Ϯ 0.2 and 0.72 Ϯ 0.07 nmol/min⅐mg, respectively (Table I), and at 37°C the rates of phosphorylation of GPP and FPP were 4.5 Ϯ 0.3 and 2.1 Ϯ 0.3 nmol/min⅐mg, respectively. No phosphorylation was observed when GPP and FPP were incubated with [␥-32 P]ATP and no added NDP kinase. The phosphorylation of farnesyl and geranyl pyrophosphates by NDP kinase was catalytic; one molecule of NDP kinase transferred phosphates to many molecules of geranyl or farnesyl pyrophosphate.
Wild Type Nm23-H1 Has Higher GPP and FPP Kinase Activities than Nm23-H1 Mutants That Do Not Inhibit Cell Motility-Transfection of human breast carcinoma cells with wild type nm23-H1 suppresses the motility of these cells, but transfections with nm23-H1 P96S or nm23-H1 S120G do not suppress motility (4). The serine 120 to glycine mutation of Nm23-H1 S120G was found in 6 of 28 aggressive childhood neuroblastomas (32). The Nm23-H1 P96S contains a proline to serine substitution at residue 96 and is homologous to the Drosophila awd kpn mutation (4,15). Transfection with nm23-H1 S120A gave mixed results, inhibiting migration toward fetal calf serum but not toward autotaxin. Extracts from these different cells all contain similar levels of NDP kinase activity (4).
The nucleoside diphosphate kinase activities of Nm23-H1, Nm23-H1 S120G , and Nm23-H1 S120A , assayed by measuring the transfer of phosphate from ATP to TDP, were found to be within 5% of each other, and the nucleoside diphosphate activity of Nm23-H1 P96S was 80 -90% that of Nm23-H1 (data not shown). Others (18,33) have also reported little differences in the nucleoside diphosphate kinase activities of these different forms of Nm23-H1.
The half-life for the transfer of 32 P from [ 32 P]Nm23-H1 (autophosphorylated Nm23-H1) to GPP was 3 min, and the half-  Phosphorylation of GPP and FPP by Cell Extracts-C100 and H1-177 cell lines are stable transfectants of the highly metastatic human MDA-MB-435 breast carcinoma cell line. The H1-177 line was transfected with nm23-H1, and C100 is a control transfectant (34). H1-177 cells contain 5 times as much nm23-H1 as do C100 cells and have a lower metastatic potential. Extracts from these two cell lines were incubated with [␥-32 P]ATP and either GPP or FPP, and the products were examined by TLC (Fig. 4). During the first 20 min of incubation, the rate of phosphorylation of GPP by H1-177 cell extracts was 1.95 Ϯ 0.12 times (mean Ϯ S.D. for three experiments) that by C100 extracts (Fig. 4A). A similar result was obtained with FPP; the rate of phosphorylation by H1-177 cell extracts was 1.8 Ϯ 0.3 times that by C100 extracts (Fig. 4B). After 15-20 min, the amounts of [ 32 P]GerTP and [ 32 P]FTP in these extracts decreased; endogenous phosphatases dephosphorylated these triphosphates. During these incubations, some 32 P was also transferred from [␥-32 P]ATP to nucleoside diphosphates present in the extracts. After 15 min, most of the [␥-32 P]ATP was hydrolyzed to give 32 P i (data not shown). GerTP and FTP were formed when these cell extracts were incubated with as little as 5 M GPP or FPP.
Extracts of H1-177 cells were partially depleted of Nm23 proteins by incubation with anti-rat liver NDP kinase antibodies. The depleted extracts gave less phosphorylation of GPP and FPP (Fig. 5, A and B) than did the control extracts (treated with prebleed IgG). The rate of FPP phosphorylation by the depleted extracts was 45 Ϯ 2% (mean Ϯ S.D. for four experiments) that of the control extracts, and the rate of GPP phosphorylation by the depleted extracts was 54 Ϯ 5% that of the control extracts. In three of the four experiments, the depleted extracts contained about half as much Nm23 proteins as did the control extracts (Fig. 5C). In one trial, the depleted extract contained only 8% as many Nm23 proteins as did the control extract, but it still had about half the GPP and FPP kinase activities of the control extract.
H1-177 Cells Contain More FTP than Do C100 Cells-C100 and H1-177 cells were labeled with [ 3 H]mevalonolactone in the presence of lovastatin to inhibit the synthesis of endogenous mevalonate. C100 and H1-177 cells took up the same amounts of [ 3 H]mevalonolactone. Isoprenoid phosphates were extracted from these cells and analyzed by reverse phase, ion-pair HPLC (Fig. 6A) (Fig. 6A).
Four major peaks of 3 H were obtained for both C100 and H1-177 cells (Fig. 6A). Three of these eluted at the same times as the FPP, FTP, and GGPP standards. The fourth peak of 3 H did not elute with any of the standards. The 3 H-labeled compound that eluted from the HPLC column at the same time as the FTP standard had the same mobility on TLC (PEI sheet developed with acetic acid/acetonitrile/NaCl) as the [ 3 H]FTP standard (Fig. 6B, lane TP). Most of the 3 H-labeled compounds that eluted from the HPLC column at the same time as the FPP standard had the same mobility as did the [ 3 H]FPP standard (Fig. 6B, lane PP), but some had the same mobility as the [ 3 H]FTP standard. The unidentified 3 H peak did not contain either FPP or FTP (Fig. 6B, lane UN). Farnesol and FMP moved with the solvent front.
Dephosphorylation was used to show that the 3 H-labeled compound that eluted from the HPLC column at the same time as the FTP standard contained farnesyl. Extracts from [ 3 H]mevalonolactone-labeled H1-177 cells were separated on this reverse phase, ion pair HPLC column, and fractions that eluted at the same times as the FPP and FTP standards were treated with alkaline phosphatase. After dephosphorylation, both fractions contained [ 3 H]farnesol (Fig. 6C, lanes PP and TP), the expected product for the dephosphorylation of both [ 3

H]FPP and [ 3 H]FTP.
These TLCs indicate that the 3 H-labeled compounds in these cell extracts that eluted from the HPLC column at the same time as the FPP and FTP standards were FPP and FTP, respectively. As shown in Fig. 6A, the C100 cells contained more [ 3 H]FPP than did H1-177 cells, and conversely, H1-177 cells contained more [ 3 H]FTP than did the C100 cells. The ratio of [ 3 H]FPP to [ 3 H]FTP in C100 cells was about 2:1, and the ratio in H1-177 cells was about 1:1. Similar differences were observed with two other sets of cells. The unidentified 3 H peak and that for GGPP were about the same for C100 and H1-177 cells.
Additional evidence that these cells contain FTP was obtained using a different HPLC system and a different TLC system. When the reverse phase HPLC column was eluted with a gradient of acetonitrile in NH 4 CHO 3 (35), the elution of the [ 3 H]FTP and [ 3 H]FPP standards usually overlapped, but [ 3 H]FTP started to elute slightly before [ 3 H]FPP. The 3 H-labeled compounds in extracts of H1-177 cells that eluted from this HPLC column at the same times as the FPP and FTP standards were examined by TLC (silica gel G developed with 1-propanol/NH 4 OH/H 2 O). Fraction 31 contained a 3 H-labeled compound with the same mobility as the FTP standard (Fig.  6D). Fraction 33 contained a 3 H-labeled compound with the same mobility as the FPP standard. The advantage of this TLC system is that mobility of FTP is much less than those of FPP and GGPP. Farnesol and FMP moved with the solvent front. This TLC system could not be used when the HPLC column (Fig. 6A) was eluted with buffers containing tetrabutylammonium phosphate. Under basic conditions, tetrabutylammonium binds to isoprenoid pyrophosphates and triphosphates, causing them to remain at the origin.
Protein Prenylation-Prenyl transferases use FPP and GGPP to covalently attach farnesyl and geranylgeranyl groups to cysteines near the C termini of certain intracellular proteins (25). To assess the effect of Nm23 proteins on protein prenylation, C100 and H1-177 cells were labeled with [ 3 H]mevalonolactone in the presence of lovastatin (Fig. 7). Although the overall pattern of protein prenylation was the same in these two cell lines, the relative levels of prenylation differed. C100 and H1-177 cells contained comparable amounts of prenylated proteins of about 27 and 25 kDa, but H1-177 contained 80 -100% more prenylated 46-kDa proteins and 40 -50% more prenylated 24-kDa proteins than did the C100 cells.
Farnesylated or geranylgeranylated proteins can be specifically labeled by incubating cells with either [ 3 H]FPP or [ 3 H]G-GPP (31). When H1-177 and C100 cells were incubated with [ 3 H]GGPP, both the pattern and extent of protein geranylgeranylation were comparable (Fig. 7). However, when incubated with [ 3 H]FPP, more [ 3 H]farnesyl was incorporated into proteins in H1-177 cells than into proteins in C100 cells. The greatest differences (50 -100%) were in the amounts of [ 3 H]farnesyl incorporated into proteins of 46 and 24 kDa, but there were also smaller differences in several other proteins. Western blots and enzymatic assays showed that C100 and H1-177 cells contain comparable levels of farnesyl transferase (data not shown). The 46-and 24-kDa farnesylated proteins have not yet been identified. The 24-kDa protein does not appear to be Ras, nor does the 46-kDa protein appear to be 2Ј,3Ј-cyclic nucleotide 3Ј phosphodiesterase, since under these conditions the fractions of these proteins farnesylated in H1-177 cells were the same as those in C100 cells (data not shown).

DISCUSSION
The phosphorylation of GPP and FPP by Nm23 proteins provides a novel mechanism by which these proteins might exert their effects on metastasis, motility, or differentiation. In this paper, we report 1) the phosphorylation of GPP and FPP to the corresponding triphosphate by purified Nm23 proteins and by Nm23 proteins in crude cell extracts, 2) an increase in the concentration of FTP in cells that have an elevated level of Nm23-H1, and 3) an altered pattern of protein prenylation in cells that have an elevated level of Nm23-H1. We also show that Nm23-H1 S120G , a mutant found in some aggressive neuroblastomas (32), has reduced GPP and FPP kinase activities and that Nm23 proteins that inhibit cell motility have higher GPP and FPP kinase activities than do Nm23 proteins that do not inhibit cell motility. In contrast, NDP kinase activities and histidine to histidine protein phosphotransferase activities of Nm23 proteins (18,21) and binding of Nm23 proteins to the c-myc promoter (36) do not correlate with inhibition of cell motility. A histidine to aspartate protein phosphotransferase activity of Nm23 proteins does appear to correlate with inhibition of motility (21); however, the stoichiometry of this phosphotransferase reaction is very low. Only 2-4% of the phosphate bound to NDP kinase is transferred (37). All of the phosphate bound to NDP kinase can be transferred to GPP or FPP.
The rates of phosphorylation of GPP and FPP by Nm23 proteins (1-4 nmol/min/mg) in vitro are much less than rates of phosphorylation of nucleoside diphosphates (500 mol/min/ mg). However, the K m values for the phosphorylation of GPP and FPP by Nm23 proteins (60 -125 M) compare favorably with K m values for the phosphorylation of nucleoside diphos-phates (0.2-1 mM), and GPP and FPP effectively competed with nucleoside diphosphates for phosphorylation by Nm23 proteins in crude cell extracts. More importantly, cells with an elevated level of Nm23-H1 contained more FTP than did control cells.
Isoprenoid pyrophosphates were detected by labeling cells with [ 3 H]mevalonolactone in the presence of lovastatin, an inhibitor of endogenous mevalonate synthesis. Under these conditions, the concentrations of GPP and FPP are lower than in the untreated cells. Our inability to detect GPP in these cells probably results from it being rapidly metabolized. NIH3T3 cells labeled using a similar protocol contained one-tenth as much GPP as FPP (29).
To our knowledge, neither FTP nor GerTP has been previously reported to be present in any type of cell. A possible explanation for this is that these triphosphates have not been directly looked for and that the published procedures used to fractionate cellular isoprenoid pyrophosphates do not separate triphosphates from pyrophosphates. When phosphorylated isoprenoids extracted from [ 14 C]mevalonolactone-labeled NIH3T3 cells were fractionated by HPLC, a peak of 14 C-labeled material eluted immediately before the FPP (29). This peak was not identified, but perillyl alcohol, a monoterpene, caused an increase in both FPP and this unidentified peak. Depending on the solvent system used, FTP elutes immediately before or after FPP.
Two observations suggest that mammalian cells may contain kinases other than Nm23 proteins that can phosphorylate GPP and FPP. 1) H1-177 cell extracts contain 5 times as many Nm23 proteins as do C100 cell extracts but only twice as much GPP and FPP kinase activity. 2) Immunodepleted H1-177 cell extracts contained as little as 8% as much Nm23 protein as did the control extract, but they retained half the GPP and FPP kinase activities. Extracts of the fungus Gibberella fujikuroi contain a kinase that phosphorylates FPP to give FTP (38). Unlike NDP kinases, this kinase is relatively specific for ATP, and it does not phosphorylate GPP.
When labeled with [ 3 H]mevalonolactone, H1-177 cells contained about twice as much FTP as did the C100 cells, and several prenylated proteins were almost twice as abundant in H1-177 cells as in C100 cells. These proteins appear to be farnesylated, since a similar result was obtained when the cells were labeled with [ 3 H]FPP. Thus, at least under this condition, cells with an elevated level of Nm23 proteins have an increased level of protein farnesylation. These results predict that FTP is substrate for farnesyl transferase, and we have found that farnesyl transferase isolated from bovine brain will use FTP to farnesylate recombinant Ras. 2 The labeling of prenylated proteins was performed in cells treated with lovastatin to inhibit the synthesis of endogenous mevalonate. In cells not treated with lovastatin, the amount of FPP may be sufficient for complete protein farnesylation. When cells not treated with lovastatin were examined, the fractions of Ras and lamin B farnesylated in H1-177 cells appeared to be the same as in C100 cells (data not shown). However, there did appear to be differences in the extent of prenylation of several low molecular weight GTP-binding proteins.
FPP is a branch point in isoprenoid metabolism and is a precursor of GGPP, cholesterol, heme A, ubiquinone, and dolichol phosphate (23,24). FPP and GGPP are used by prenyl transferases to prenylate proteins, many of which play critical roles in signal transduction, growth regulation, and cell motility. Farnesylation of oncogenic forms of Ras is essential for their transforming activity, and several small GTP-binding proteins involved in regulation of the cytoskeleton are preny- lated. Cholesterol is involved in membrane structure and affects endocytosis and cell signaling. Dolichol phosphate functions as a saccharide carrier for protein glycosylation and for synthesis of glycosylphosphatidylinositol protein anchors. The conversion of GPP or FPP to triphosphates could alter their flows through these pathways, resulting in changes in cellular metabolism that affect metastatic potential, cell motility, or differentiation (i.e. altered protein prenylation, membrane structure, or protein glycosylation). The observation that cells with an elevated level of FTP had an increased level of protein farnesylation suggests that cellular FTP is a substrate for farnesyl transferase but not for one or more of the other enzymes that use FPP as a substrate. GerTP and FTP may also have metabolic roles distinct from those of the pyrophosphates.