The Arabidopsis thaliana FPS1 Gene Generates a Novel mRNA That Encodes a Mitochondrial Farnesyl-diphosphate Synthase Isoform*

The enzyme farnesyl-diphosphate synthase (FPS; EC2.5.1.1./EC 2.5.1.10) catalyzes the synthesis of farnesyl diphosphate from isopentenyl diphosphate and dimethylallyl diphosphate. FPS is considered to play a key role in isoprenoid biosynthesis. We have reported previously that Arabidopsis thaliana contains two differentially expressed genes, FPS1 and FPS2, encoding two highly similar FPS isoforms, FPS1 and FPS2, (Cunillera, N., Arró, M., Delourme, D., Karst, F., Boronat, A., and Ferrer, A. (1996) J. Biol. Chem. 271, 7774–7780). In this paper we report the characterization of a novel ArabidopsisFPS mRNA (FPS1L mRNA) derived from the FPS1 gene. A cDNA corresponding to the FPS1L mRNA was cloned using a reverse transcription-polymerase chain reaction strategy. Northern blot analysis showed that the two FPS1-derived mRNAs are differentially expressed. The FPS1L mRNA accumulates preferentially in inflorescences, whereas the previously reported FPS1 mRNA (FPS1S mRNA) is predominantly expressed in roots and inflorescences. FPS1L mRNA contains an in-frame AUG start codon located 123 nucleotides upstream of the AUG codon used in the translation of the FPS1S isoform. Translation of the FPS1L mRNA from the upstream AUG codon generates a novel FPS1 isoform (FPS1L) with an NH2-terminal extension of 41 amino acid residues, which has all the characteristics of a mitochondrial transit peptide. The functionality of the FPS1L NH2-terminal extension as a mitochondrial transit peptide was demonstrated by its ability to direct a passenger protein to yeast mitochondria in vivo and by in vitro import experiments using purified plant mitochondria. TheArabidopsis FPS1L isoform is the first FPS reported to contain a mitochondrial transit peptide.

The enzyme farnesyl-diphosphate synthase (FPS; EC 2.5.1.1./EC 2.5.1.10) catalyzes the synthesis of farnesyl diphosphate from isopentenyl diphosphate and dimethylallyl diphosphate. FPS is considered to play a key role in isoprenoid biosynthesis. We have reported previously that Arabidopsis thaliana contains two differentially expressed genes, FPS1 and FPS2, encoding two highly similar FPS isoforms, FPS1 and FPS2, (Cunillera, N., Arró , M., Delourme, D., Karst, F., Boronat, A., and Ferrer, A. (1996) J. Biol. Chem. 271, 7774 -7780). In this paper we report the characterization of a novel Arabidopsis FPS mRNA (FPS1L mRNA) derived from the FPS1 gene. A cDNA corresponding to the FPS1L mRNA was cloned using a reverse transcription-polymerase chain reaction strategy. Northern blot analysis showed that the two FPS1-derived mRNAs are differentially expressed. The FPS1L mRNA accumulates preferentially in inflorescences, whereas the previously reported FPS1 mRNA (FPS1S mRNA) is predominantly expressed in roots and inflorescences. FPS1L mRNA contains an inframe AUG start codon located 123 nucleotides upstream of the AUG codon used in the translation of the FPS1S isoform. Translation of the FPS1L mRNA from the upstream AUG codon generates a novel FPS1 isoform (FPS1L) with an NH 2 -terminal extension of 41 amino acid residues, which has all the characteristics of a mitochondrial transit peptide. The functionality of the FPS1L NH 2 -terminal extension as a mitochondrial transit peptide was demonstrated by its ability to direct a passenger protein to yeast mitochondria in vivo and by in vitro import experiments using purified plant mitochondria. The Arabidopsis FPS1L isoform is the first FPS reported to contain a mitochondrial transit peptide.
The enzyme farnesyl-diphosphate synthase (FPS 1 ; EC 2.5.1.1/EC 2.5.1.10) catalyzes the sequential 1Ј-4 condensation of two molecules of isopentenyl diphosphate with both dimethylallyl diphosphate and the resultant 10-carbon compound geranyl diphosphate to produce the 15-carbon compound farnesyl diphosphate (1). Because of the central branch point location of farnesyl diphosphate in the isoprenoid pathway, FPS is considered to play a key role in isoprenoid biosynthesis. It has been shown that in mammals FPS is a highly regulated enzyme involved in the control of the sterol biosynthetic pathway (2)(3)(4)(5). FPS is also considered to play an important role in the control of plant isoprenoid biosynthesis. In plants, farnesyl diphosphate is the starting point of different branches of the isoprenoid pathway leading to the synthesis of key end products that are required for normal growth and development, such as phytosterols, dolichols, ubiquinone, plastoquinone, sesquiterpenoid phytoalexins, and prenylated proteins. Therefore, changes in FPS activity could alter the flux of isoprenoid compounds down the different branches of the pathway in competition for the available farnesyl diphosphate and, hence, play a central role in the regulation of a number of essential functions in plant cells (6). However, more experimental data are still required before obtaining a clear picture of the regulatory significance of FPS in the overall control of isoprenoid biosynthesis in plants.
Genomic and cDNA sequences encoding FPS have been isolated and characterized from a variety of organisms, ranging from bacteria to higher eukaryotes (7,8). Recently, the enzyme has been cloned from several plant species such as Arabidopsis thaliana (9,10), Lupinus albus (11), Zea mays (12), Artemisia annua (13), Hevea brasiliensis (14), and Parthenium argentatum (15). Although the complexity of the FPS gene families has been studied in a limited number of plants, it seems that plant FPS is encoded by multigene families like other key enzymes of the isoprenoid biosynthetic pathway, such as 3-hydroxy-3methylglutaryl-CoA reductase (16 -18), and geranylgeranyldiphosphate synthase (19). We have recently shown that Arabidopsis contains a small FPS gene family consisting of at least two genes, FPS1 and FPS2, which have a very similar structure and share a high level of sequence similarity (10). At least two FPS gene copies have been detected in the maize genome (12), and cDNA sequences encoding two highly similar FPS isoforms have also been reported in L. albus (11) and P. argentatum (15). The pattern of expression of individual genes encoding FPS has been reported only in Arabidopsis, where it has been shown that the two currently characterized FPS genes (FPS1 and FPS2) are expressed differentially at both quantitative and qualitative levels (10). At present, the biological significance of the occurrence of highly similar FPS isoforms in plants is still unclear, although the differential pattern of expression of the two FPS Arabidopsis genes suggests that each FPS isoform might have a specialized function in directing the flux of pathway intermediates into specific classes of isoprenoid end products. FPS has long been considered to be a cytoplasmic enzyme. However, it has recently been demonstrated that in mammals FPS is mainly localized within the peroxisomes (20), although significant levels of FPS activity have also been detected in rat liver mitochondria (21). In plants, the only cell compartment where plant FPS has been detected is the cytosol (6,22,23), although it is widely accepted that the enzymes utilizing isopentenyl diphosphate are distributed in three subcellular compartments, namely cytosol, mitochondria, and plastids. All of the FPS reported to date show high amino acid sequence similarity, and all contain several conserved domains, including the two aspartate-rich domains involved in substrate binding and enzyme catalysis (24,25). However, none of the FPS characterized so far from a number of eukaryotic organisms (including fungi, plants, and animals) contains NH 2 -terminal sequences that could represent transit peptides for targeting into plastids or mitochondria. In this paper we report that the expression of the Arabidopsis FPS1 gene generates a novel mRNA that encodes a mitochondrial FPS isoform.
Cloning of Arabidopsis FPS1L cDNA-The cDNA encoding FPS1L was cloned by means of a reverse transcription-polymerase chain reaction (PCR) strategy similar to that used previously for the cloning of the Arabidopsis FPS2 cDNA (10). The cDNA was amplified from a singlestranded cDNA pool, obtained by reverse transcription of 5 g of poly(A) ϩ RNA from Arabidopsis inflorescences, using an upstream primer specific for the leader region of the FPS1L mRNA (5Ј-GGCGTTTTCGGGAGAAGAAGG-3Ј, nucleotides Ϫ97 to Ϫ77 in Fig.  1A) and an adaptor oligonucleotide (5Ј-GACTCGAGTCGACATCGGG-3Ј) as a downstream primer. The resulting PCR product (approximately 1.4 kilobases) was gel purified and ligated into plasmid pGEM-T (Promega) prior to sequencing. The resulting plasmid was designated pcNC3.
DNA Sequencing-Appropriate restriction fragments were subcloned into plasmid pBluescript. Both strands of DNA were sequenced by the dideoxynucleotide chain termination method (26) using an automated fluorescence-based system (Applied Biosystems).
RNase Protection Analysis-To obtain the desired RNase protection probe a DNA fragment extending from nucleotide positions Ϫ201 to ϩ141 in the FPS1 gene was generated by PCR. The amplified fragment was cloned into the EcoRI and HindIII sites of plasmid pSP65 (Promega). To rule out the existence of PCR artifacts, the cloned fragment was sequenced. The resulting plasmid was linearized by digestion with HindIII and used as a template for in vitro transcription using SP6 RNA polymerase (Promega) and [␣-32 P]rUTP. The 351-nucleotide RNA antisense probe contained 342 nucleotides corresponding to the FPS1 gene and 9 additional nucleotides derived from plasmid pSP65. RNase protection experiments were performed as described (27), except that the RNA probe was purified on a 5% polyacrylamide, 8 M urea gel before use. The probe was eluted by diffusion at room temperature in 600 l of 0.5 M ammonium acetate, 1 mM EDTA, and 0.1% SDS. After precipitation, the antisense RNA probe (2.5 ϫ 10 5 cpm) was hybridized overnight at 42°C with 8 g of poly(A) ϩ RNA from Arabidopsis inflorescences or yeast tRNA. Digestion was performed for 2 h with 15 units of RNase ONE (Promega) according to the manufacturer's recommendations. Analysis of the protected fragments was performed by electrophoresis on a 5% polyacrylamide, 8 M urea gel. The gel was dried and exposed to x-ray film. Variation of the quantity of RNA or the digestion conditions did not alter the pattern of protected bands. A known DNA sequencing reaction was included as a marker. The size of the RNA protected fragments was initially calculated according to the sequencing reaction and corrected as described (27).
Northern Blot Analysis-Total RNA from different tissues of Arabidopsis was isolated as described (28). Thirty g of RNA from each sample was fractionated by electrophoresis in 1% (w/v) agarose gels containing 2.2 M formaldehyde and blotted to Hybond-N nylon membranes (Amersham). The 32 P-labeled (random primed) probes used were a PCR amplification fragment of 289 bp extending from nucleotides Ϫ309 to Ϫ20 in the FPS1 gene, and a 370-bp BglII-HindIII fragment from the 3Ј-flanking region of the FPS1 gene (10). The conditions of hybridization and washing of the filters were as described previously (10). Filters were reprobed with a 32 P-labeled 900-bp BamHI-EcoRI fragment of the gene for the wheat 25 S cytoplasmic rRNA (29).
In Vitro Transcription/Translation-A SacII-SalI fragment, containing the FPS1L cDNA, was excised from plasmid pcNC3 and cloned into the corresponding sites of plasmid pBluescript to create plasmid pcBNC3. By using site-directed mutagenesis (30), the ATG start codon of the FPS1S isoform was converted to an ATC codon (encoding Ile). The resulting plasmid was designated pcBNC3Mut. The two cDNA sequences, FPS1L and FPS1LMut, were cut out as SacI-SalI fragments from plasmids pcBNC3 and pcBNC3Mut, respectively, and cloned into the corresponding sites of plasmid pSP65 (Promega). The resulting plasmids, pcSPNC3 and pcSPNC3Mut, were used as templates for in vitro transcription/translation using [ 35 S]Met and the TNT TM Coupled Wheat Germ Extract System (Promega). The 35 S-labeled proteins were separated by SDS-polyacrylamide gel electrophoresis (12% acrylamide) and detected by fluorography.
Functional Complementation in Yeast-To construct plasmid pFPS1Ltp-Y⌬COX, a cDNA fragment encoding the 41 NH 2 -terminal amino acid residues of the FPS1L isoform was amplified by PCR using a forward primer (5Ј-GGGAATTCAAAAATGTCTGTGAGTTGTT-GTTGTAGG-3Ј) extending from nucleotide positions Ϫ74 to Ϫ47 in the Arabidopsis FPS1 gene (Fig. 1A), a reverse primer (5Ј-AGCTCTAGAT-GAAGAGCTTTGGATACG-3Ј) complementary to the nucleotide sequence ϩ36 to ϩ53 in the FPS1 gene (Fig. 1A), and plasmid pcSPNC3 (see above) as a template. EcoRI and XbaI sites (shown in italic) were added to the 5Ј-end of the primers, respectively. The 144-bp PCR product was digested with EcoRI and XbaI and cloned into the corresponding sites of plasmid pY⌬COX (kindly provided by Ian D. Small; Institut National de la Recherche Agronomique, Versailles, France), which contains the presequence-less yeast COXIV gene under the control of the alcohol dehydrogenase gene promoter (31). To optimize the synthesis of the chimeric FPS1Ltp-Y⌬COX protein in yeast, four changes (underlined) were introduced in the sequence of the forward primer to convert the nucleotide sequences surrounding the ATG start codon of the Arabidopsis FPS1L cDNA (GAATATGAG) into the consensus reported for functional translational start codons in yeast (AAAAATGTC) (32). Changes in the second triplet of the Arabidopsis FPS1L cDNA coding sequence did not alter the amino acid residue encoded (Ser). Yeast strain WSR was transformed with plasmids pYCOX, pY⌬COX, or pFPS1Ltp-Y⌬COX using the modified lithium acetate procedure described by Gietz et al. (33). Ura ϩ transformants were selected at 28°C on agar plates containing minimal medium (0.16% (w/v) yeast nitrogen base without amino acids and ammonium sulfate, 0.5% (w/v) ammonium sulfate and 1% (w/v) glucose) supplemented with histidine (20 g/ml), tryptophan (40 g/ml), and adenine (40 g/ml). After 4 days of growth, selected colonies were subcultured on N3 medium (1% (w/v) yeast extract, 1% (w/v) Bacto-peptone, 2% (v/v) glycerol, and 50 mM potassium phosphate (pH 6.25)) to test for complementation of respiratory deficiency.
Import into Purified Potato Mitochondria-Mitochondria were isolated from potato tubers (Solanum tuberosum var. Bintje) as described (34). For in vitro import studies the purified mitochondria were washed and resuspended in a buffer containing 400 mM mannitol, 10 mM potassium phosphate (pH 7.2), and 0.1% (w/v) bovine serum albumin. The FPS1L protein was synthesized by in vitro transcription/translation of plasmid pcSPNC3 using [ 35 S]Met and the TNT TM Coupled Reticulocyte Lysate System. The import reaction contained 35 l of purified mitochondria (10 mg of protein/ml), 160 l of import buffer (250 mM mannitol, 20 mM HEPES (pH 7.5), 80 mM KCl, 1 mM K 2 HPO 4 , 1 mM ATP, 1 mM malate, 2 mM NADH, and 1 mM dithiothreitol), and 15 l of the reticulocyte lysate translation mixture. After incubation for 30 min at 20°C, 70-l aliquots of the import reaction were treated with proteinase K (20 g/ml), either in the presence or absence of 0.5% (v/v) Triton X-100. These aliquots and the remainder of the import reaction were then incubated for 20 min at 20°C. After the addition of 1 mM phenyl-methylsulfonyl fluoride, samples were incubated for further 15 min. For inhibition of mitochondrial import, purified mitochondria were incubated for 5 min in the presence of 1 M valinomycin prior to the import reaction. After the different treatments, mitochondria were repurified by centrifugation through a 25% (w/v) sucrose cushion. The resulting pellets were subjected to SDS-polyacrylamide gel electrophoresis (10% acrylamide), and the radiolabeled products were detected by fluorography.

RESULTS
Identification and Cloning of a Novel FPS mRNA Derived from the Arabidopsis FPS1 Gene-We have recently reported that the expression of the Arabidopsis FPS1 gene generates an mRNA of approximately 1.3 kilobases which encodes the isoform FPS1 (10). A cDNA encoding this isoform had previously been cloned by functional complementation of a mutant yeast strain defective in FPS activity (9). A detailed analysis of the nucleotide sequence of the 5Ј-flanking region of the FPS1 gene (Fig. 1A) revealed the presence of an in-frame ATG codon located 123 bp upstream of the ATG start codon used in the translation of the reported FPS1 isoform (9). The utilization of the upstream ATG codon as a translation start site would generate a different FPS1 isoform containing an NH 2 -terminal extension of 41 amino acids with respect to the previously described FPS1 isoform.
The occurrence of an mRNA containing the upstream ATG was first detected by using a reverse transcription-PCR strategy (data not shown). To confirm the existence of this novel FPS1 mRNA we isolated the corresponding cDNA by reverse transcription-PCR using poly(A) ϩ RNA from Arabidopsis inflorescences (for details, see "Experimental Procedures"). A cDNA fragment of approximately 1.4 kilobases was cloned and sequenced. The cDNA insert was found to have a nucleotide sequence of 1,396 bp, excluding a polyadenylate tail of 30 bases, which contained the complete sequence of the FPS1 cDNA previously characterized as well as 99 and 89 additional nucleotides at the 5Ј-and 3Ј-ends, respectively. The open reading frame starting at the most 5Ј ATG triplet encodes a protein of 384 amino acid residues, with a predicted molecular mass of 44,254 Da, which differs from the previously reported FPS1 isoform in having an NH 2 -terminal extension of 41 amino acids. The cDNA clone also contains a 5Ј-untranslated region of 27 bp and a 3Ј-untranslated region of 214 bp. The polyadenylate tail was located 89 bp downstream of the polyadenylation site of the FPS1 cDNA (9), thus indicating that different polyadenylation sites are used in the FPS1 gene. These two FPS1 mRNAs will be hereafter referred to as FPS1S and FPS1L, with FPS1S mRNA encoding the previously reported FPS1 isoform (9, 10) (FPS1S) and FPS1L mRNA encoding the novel FPS1 isoform (FPS1L) containing the NH 2 -terminal extension.
To define the 5Ј-ends of the two FPS1 mRNAs (FPS1S and FPS1L) we performed RNase protection analysis using poly(A) ϩ RNA isolated from Arabidopsis inflorescences. The results obtained are shown in Fig. 1B. The estimated size of the most intense protected bands indicated the occurrence of a major FPS1L transcript (band of 269 nucleotides) with a 5Ј-end located 62 nucleotides upstream of the FPS1L ATG start codon (TS3, Fig. 1A) and two major FPS1S mRNAs (bands of 137 and 154 nucleotides) with 5Ј-ends located 53 and 70 nucleotides upstream of the FPS1S ATG start codon (TS1 and TS2, Fig.  1A). Taken together, these results indicated that the expression of the Arabidopsis FPS1 gene generates two mRNAs encoding two FPS1 isoforms (FPS1S and FPS1L) that differ only in their NH 2 terminus.
Expression Analysis of the Two FPS1-derived mRNAs-The expression pattern of the Arabidopsis FPS1S and FPS1L mRNAs was analyzed by Northern blot analysis using total RNA isolated from roots, stems, leaves and inflorescences (Fig.  2). A PCR fragment of 289 bp extending from nucleotides Ϫ309 to Ϫ20 in the FPS1 gene was used as an FPS1L mRNA-specific probe. A 370-bp BglII-KpnI genomic fragment corresponding to the 3Ј-end of the FPS1 gene (10) was used as an FPS1 genespecific probe that recognizes both FPS1L and FPS1S mRNAs simultaneously. Similar amounts of RNA were present in each lane, as confirmed by hybridization of the filters with a probe derived from the wheat 25 S rRNA gene, and radiolabeled probes of equivalent specific activity were used. Comparison of the results indicated that FPS1S and FPS1L mRNAs are detected in all tissues analyzed although they have a different pattern of expression. FPS1L mRNA accumulates preferentially in inflorescences, whereas FPS1S mRNA is detected mainly in roots and inflorescences. No significant change in the levels of FPS1L mRNA was detected when RNA samples were prepared from light-or dark-grown seedlings.
Functional in Vitro Analysis of the Upstream AUG Codon in the FPS1L mRNA-To analyze whether the upstream in-frame AUG codon in the FPS1L mRNA is used as translational start codon, the FPS1L transcript was synthesized in vitro from plasmid pcSPNC3 (Fig. 3A) and translated in a wheat germ cell-free system. The products obtained were separated by SDS-polyacrylamide gel electrophoresis and analyzed by fluo-rography. Two proteins of 44 and 40 kDa resulted from the translation of the FPS1L mRNA (Fig. 3B, lane 1). The estimated size of the two proteins is in good agreement with the molecular mass predicted for proteins initiated at the first (44,254 Da) and the second (39,689 Da) AUG codons. When a similar experiment was performed using transcript FPS1LMut (plasmid pcSPNC3Mut, Fig. 3A), in which the second AUG codon was converted to an AUC codon (encoding Ile) by sitedirected mutagenesis, only the 44-kDa protein was synthesized (Fig. 3B, lane 2). Taken together, these results show that in the wheat germ lysate system, the most 5Ј AUG codon in the FPS1L mRNA is used preferentially to initiate translation, although the second AUG codon is also used, giving rise to significant levels of the FPS1S protein.
Functional in Vivo Analysis of the FPS1L NH 2 -terminal Extension-Analysis of the amino acid composition and the predicted secondary structure of the NH 2 -terminal extension of the Arabidopsis FPS1L isoform (Fig. 4) revealed that it has all of the features characteristic of a mitochondrial transit peptide (35,36). There is a complete absence of acidic residues and an enrichment in basic (arginine, lysine, and histidine), hydroxylated (serine), and hydrophobic (leucine) residues. It contains the sequence RIQS (amino acid residues 36 -39 in Fig. 4A), which fits the most commonly reported mitochondrial targeting peptide cleavage motif RX/XS (where X represents any amino acid) (36), preceded by two arginine residues located at positions Ϫ2 and Ϫ12 relative to the predicted cleavage site RI/QS. Finally, analysis of the FPS1L NH 2 -terminal extension using an improved method of protein structure prediction (37) shows that amino acid residues 8 -17 can form a positively charged amphiphilic ␣-helix, in which hydrophobic residues are clustered on one face of the helix while the basic and polar residues are on the other face (Fig. 4B).
To study the possible role of the FPS1L NH 2 -terminal extension as a mitochondrial targeting sequence, we analyzed the ability of this predicted transit peptide to direct the CoxIV subunit of cytochrome c oxidase into yeast mitochondria in vivo. This experimental approach was based on previous observations demonstrating the interchangeability of mitochondrial targeting sequences between plants and yeast (31, 38 -41). For this purpose, an in-frame fusion was made between the fragment of the Arabidopsis FPS1L cDNA coding for the 41 NH 2terminal residues of the FPS1L isoform (FPS1Ltp) and the S. cerevisiae COXIV gene lacking the region coding for its own mitochondrial transit peptide (⌬COXIV). The resulting plasmid (pFPS1Ltp-Y⌬COX, Fig. 5A) was used to transform the yeast strain WSR, which contains a disrupted copy of the COXIV gene and, consequently, is unable to grow on medium containing glycerol as energy source. The results shown in Fig. 5B indicate that plasmid pFPS1Ltp-Y⌬COX efficiently complements strain WSR. The same result was obtained when strain WSR was transformed with plasmid pYCOX, which encodes the complete amino acid sequence of the S. cerevisiae CoxIV subunit. In contrast, plasmid pY⌬COX, which encodes a truncated form of the CoxIV subunit lacking the mitochondrial transit peptide, does not complement the respiratory-deficient strain WSR. These results clearly demonstrate that the NH 2terminal extension of the Arabidopsis FPS1L isoform is able to replace the mitochondrial targeting sequence of the CoxIV subunit of yeast cytochrome c oxidase, thus demonstrating that it is a functional mitochondrial transit peptide.
In Vitro Import of FPS1L into Purified Plant Mitochondria-To confirm further the functional role of the FPS1L NH 2terminal extension as a mitochondrial targeting sequence we analyzed the import of the Arabidopsis FPS1L isoform into plant mitochondria (Fig. 6). The FPS1L mRNA was in vitro translated using a rabbit reticulocyte lysate, and the resulting products were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. The translation mixture contained similar amounts of radiolabeled FPS1L (44 kDa) and FPS1S (40 kDa) isoforms (Fig. 6, lane 2), which were completely digested after incubation with proteinase K (Fig. 6, lane 1). When the translation mixture was incubated with purified potato mitochondria followed by treatment with proteinase K, a protected polypeptide of approximately 40 kDa was detected (Fig.  6, lane 4). The protection was abolished when mitochondria were solubilized with Triton X-100 prior to proteinase K treatment (Fig. 6, lane 5). When the import experiment was performed in the presence of valinomycin, which collapses the membrane potential required for protein import into mitochondria, the protected 40-kDa polypeptide was not detected (Fig. 6,  lane 6). No import into mitochondria was observed when similar experiments were performed using as a precursor the 40-kDa FPS1S isoform generated from the FPS1S mRNA (data not shown). Taken together, these results demonstrate that the 44-kDa Arabidopsis FPS1L isoform is targeted into plant mitochondria and processed to a 40-kDa mature FPS1 isoform by cleavage of a transit peptide of approximately 4 kDa.

DISCUSSION
The great complexity of the plant isoprenoid biosynthetic pathway has led to the suggestion that its regulation requires the coordinated activity of several key enzymes (6,42,43). FPS, a prenyltransferase that catalyzes the synthesis of farnesyl diphosphate from isopentenyl diphosphate and dimethylallyl diphosphate, is considered to play a central role in the overall control of the plant isoprenoid pathway. This assumption has been primarily made on the basis that farnesyl diphosphate is the starting point of different branches of the plant isoprenoid pathway leading to the synthesis of essential end products (6,42,43). The proposed regulatory role of plant FPS is further supported by the observation that in mammals FPS is a highly regulated enzyme that plays a relevant role in the control of sterol biosynthesis (2)(3)(4)(5). We have recently shown that Arabidopsis contains a small FPS gene family consisting of at least two genes, FPS1 and FPS2. These genes are differentially expressed and encode two highly similar FPS isoforms (10). At least two FPS isoforms have also been reported to occur in other plant species (11,15). Thus, the occurrence of FPS isoforms is a general feature of higher plants, although the specific role of each individual FPS isoform in the isoprenoid biosynthetic pathway is currently unknown.
In this paper we report that the expression of the Arabidopsis FPS1 gene generates a previously undetected mRNA that encodes a novel FPS1 isoform (FPS1L) with an NH 2 -terminal extension of 41 amino acids with respect to the FPS1 isoform (FPS1S) previously characterized (9,10). The occurrence of the FPS1L mRNA was demonstrated by the cloning of its corresponding cDNA. RNA blot analysis using a FPS1L mRNA-specific probe and an FPS1-derived probe that recognizes both FPS1S and FPS1L mRNAs simultaneously showed that FPS1S and FPS1L mRNAs are present in all tissues analyzed, although FPS1L mRNA accumulates preferentially in inflorescences, and FPS1S mRNA is mainly expressed in roots and inflorescences. The fact that the FPS1 gene generates two mRNAs that are also differentially expressed suggests that the two transcripts are under the control of alternative promoters. The results obtained in the RNase protection analysis show that FPS1S and FPS1L mRNAs have heterogeneous 5Ј-ends, which correlates with the lack of consensus TATA box located at appropriate distances from the corresponding transcription start sites.
At the protein level, the novel Arabidopsis FPS1L isoform is identical in sequence to FPS1S but is extended at its NH 2 terminus by an additional sequence of 41 amino acids. This NH 2 -terminal extension has no counterpart among the FPS characterized so far from a number of eukaryotic organisms (7)(8)(9)(10)(11)(12)(13)(14)(15). This observation raised the question about its possible function as a transit peptide for targeting the enzyme into subcellular organelles. The accumulation of sequence data on organellar transit peptides has revealed that these targeting sequences share a very low level of sequence conservation, although they have a number of common features in terms of amino acid composition, positional amino acid preferences, and secondary structure (35,36). Interestingly, the FPS1L NH 2terminal extension nicely fits all known requirements to be a mitochondrial transit peptide. First, there is a lack of acidic amino acid residues and an enrichment in hydroxylated and hydrophobic residues. Second, the sequence motif RIQS, which fits the consensus mitochondrial targeting peptide cleavage motif RXXS (where X represents any amino acid), is found just two residues upstream of the NH 2 -terminal methionine residue of FPS1S isoform. This motif is preceded by two arginine residues located at positions Ϫ2 and Ϫ12 from the putative cleavage site (RI/QS). It has been proposed that arginine residues located around positions Ϫ2 and Ϫ10 from the processing site play a role in defining the precise cleavage site of targeting peptides by mitochondrial matrix proteases (35). Third, secondary structure analysis indicates that a sequence of 10 amino acid residues located at the NH 2 -terminal part of the FPS1L NH 2 -terminal extension can potentially fold into a positively charged amphiphilic ␣-helix. Taken together, all these observations strongly suggested that the FPS1L NH 2 -terminal extension was a mitochondrial transit peptide.
To analyze whether the FPS1L NH 2 -terminal extension could function as a mitochondrial transit peptide we first studied its ability to direct a passenger protein into yeast mitochondria in vivo. To this end, a construct expressing a chimeric protein containing the 41 NH 2 -terminal amino acid residues of FPS1L fused to the CoxIV subunit of yeast cytochrome c oxidase lacking its own mitochondrial transit peptide (plasmid pFPS1Ltp-⌬COXIV) was assayed for its ability to complement the mutant yeast strain WSR. This strain is unable to grow in a medium containing glycerol as energy source because of a disruption of the COXIV gene, which encodes the CoxIV subunit of mitochondrial cytochrome c oxidase. Plasmid pFPS1Ltp-⌬COXIV completely restored the respiratory activity of strain WSR, thus indicating that the yeast CoxIV subunit was properly targeted into mitochondria. Since the interchangeability of mitochondrial transit peptides between plants and yeast is well documented (31, 38 -41), the observation that the NH 2 -terminal extension of the Arabidopsis FPS1L isoform acts as a mitochondrial transit peptide in yeast indicates that Arabidopsis FPS1L is the precursor of a mitochondrial FPS isoform. This fact was confirmed further by in vitro import studies of the Arabidopsis FPS1L into plant mitochondria. In vitro translated FPS1L protein was imported into purified potato mitochondria and processed to a mature mitochondrial FPS1 isoform of 40 kDa by cleavage of a transit peptide of approximately 4 kDa. From this result and assuming that FPS1L is processed by mitochondrial matrix proteases at the predicted processing site (see above), the mature mitochondrial FPS1 isoform would contain four additional amino acid resi- dues at its NH 2 terminus (QSSS) with respect to the FPS1S isoform.
Our results indicate that Arabidopsis FPS1 is a bifunctional gene encoding cytosolic and mitochondrial FPS isoforms. Consequently, FPS1 belongs to the increasing group of eukaryotic genes encoding isozymes that perform analogous functions at different intracellular locations (44). These genes usually have more than one ATG start codon in their 5Ј-region and, depending on the translation start codon used to initiate translation of the corresponding mRNAs, have the potential to encode enzymes differing only at their NH 2 terminus. These isozymes can be targeted to different intracellular compartments. Genes of this type have recently been reported in plants. For example, the Arabidopsis alanyl-tRNA synthetase gene encodes both mitochondrial and cytosolic forms of the enzyme (31), and the Arabidopsis glutathione reductase gene encodes both mitochondrial and chloroplastic isoforms and possibly also a cytosolic isoform (45). In the case of the Arabidopsis FPS1 gene, the strategy used to generate the cytosolic and the mitochondrial FPS1 isoforms is the use of alternative transcription start sites, resulting in the synthesis of mRNAs that differ in the presence or absence in the 5Ј-region of the sequence encoding the mitochondrial transit peptide. However, the in vitro translation results showed that FPS1L mRNA can give rise not only to the FPS1L isoform but also to significant levels of the FPS1S isoform. It is thus possible that the choice of cytosolic or mitochondrial location of FPS might involve not only the use of alternative transcription start sites but also a mechanism of AUG selection during the translation initiation process. It is remarkable that a similar situation has been reported to occur in the Arabidopsis HMG1 gene, which encodes another key regulatory enzyme of the isoprenoid pathway (46). This gene contains two alternative promoters that direct the expression of different mRNAs encoding two 3-hydroxy-3-methylglutaryl-CoA reductase isoforms (HMGR1L and HMGR1S), which differ only in the presence or absence of an NH 2 -terminal extension of 50 amino acids. In this case both 3-hydroxy-3-methylglutaryl-CoA reductase isoforms are targeted primarily to the endoplasmic reticulum membrane, although it has been proposed that the different NH 2 -terminal sequences might direct each 3-hydroxy-3-methylglutaryl-CoA reductase isoform to specific subdomains of the endoplasmic reticulum (46,47). These observations give further support to the view that the regulation of plant isoprenoid biosynthesis is under the control of complex regulatory mechanisms operating at both transcriptional and post-transcriptional level.
A general feature of the enzymes involved in plant isoprenoid biosynthesis is their occurrence in multiple isoforms (10, 16 -19). Although the biological significance of this fact has not yet been fully evaluated, it is likely that the multiplicity of isoforms reflects the great complexity of plant isoprenoid biosynthesis, concerning not only the regulation of the overall pathway but also its subcellular compartmentalization. In contrast to the controversy surrounding the subcellular location of the enzymes involved in the synthesis of isopentenyl diphosphate in plants, it is widely accepted that the enzymes utilizing isopentenyl diphosphate are distributed in at least three cellular compartments: the cytosol, mitochondria, and plastids. However, the cytosol is the only cell compartment in which plant FPS has been detected (6,22,23). It is worth noting that a significant level of FPS activity (approximately 13% of the total cellular FPS activity) has been detected in rat liver mitochondria (21). Interestingly, none of the previously reported FPS from a number of eukaryotic organisms (including fungi, plants, and animals) (7-15) contains NH 2terminal sequences that could represent transit peptides for targeting to mitochondria. The Arabidopsis FPS1L isoform described here represents the first reported eukaryotic FPS that contains a mitochondrial transit peptide. This finding reinforces the view that plant mitochondria can use isopentenyl diphosphate as a precursor for the synthesis of farnesyl diphosphate, which in turn could be utilized for the production of mitochondrial isoprenoid compounds such as ubiquinone, heme a, tRNA species, and prenylated proteins. On the basis of these observations, it is reasonable to speculate that FPS isoforms with mitochondrial targeting peptides might be found not only in other plant species but also in eukaryotic organisms other than plants. We are currently applying different molecular and cellular approaches to assess the possible regulatory role of FPS1L in the synthesis of mitochondrial isoprenoid compounds in plants.