The nomenclature used for enzymes of the GA biosynthesis pathway is based on that suggested by Coles et al. (1999) for the dioxygenase enzymes, as follows: AtCPS1, Arabidopsis copalyl diphosphate synthase; AtKS1, ent-kaurene synthase; AtKO1, ent-kaurene oxidase; AtKAO1 or AtKAO2, ent-kaurenoic acid oxidase.
We have used fusions of gibberellin biosynthesis enzymes to green fluorescent protein (GFP) to determine the subcellular localization of the early steps of the pathway. Gibberellin biosynthesis from geranylgeranyl diphosphate is catalysed by enzymes of the terpene cyclase, cytochrome P450 mono-oxygenase and 2-oxoglutarate-dependent dioxygenase classes. We show that the N-terminal pre-sequences of the Arabidopsis thaliana terpene cyclases copalyl diphosphate synthase (AtCPS1) and ent-kaurene synthase (AtKS1) direct GFP to chloroplasts in transient assays following microprojectile bombardment of tobacco leaves. The AtKS1–GFP fusion is also imported by isolated pea chloroplasts. The N-terminal portion of the cytochrome P450 protein ent-kaurene oxidase (AtKO1) directs GFP to chloroplasts in tobacco leaf transient assays. Chloroplast import assays with 35S-labelled AtKO1 protein show that it is targeted to the outer face of the chloroplast envelope. The leader sequences of the two ent-kaurenoic acid oxidases (AtKAO1 and AtKAO2) from Arabidopsis direct GFP to the endoplasmic reticulum. These data suggest that the AtKO1 protein links the plastid- and endoplasmic reticulum-located steps of the gibberellin biosynthesis pathway by association with the outer envelope of the plastid.
The gibberellins (GAs) are an important class of plant hormones, involved in many growth and developmental processes in plants (Hooley, 1994). The genes encoding many of the enzymes that catalyse the biosynthesis of GAs have been isolated (reviewed in Hedden and Kamiya, 1997; Hedden and Proebsting, 1999). In Arabidopsis thaliana, genes encoding the enzymes catalysing each step of the GA biosynthesis pathway from geranylgeranyl diphosphate to GA4 and GA1 have now been isolated (Helliwell et al., 2001). Copalyl diphosphate synthase (AtCPS1) and ent-kaurene synthase (AtKS1), both terpene cyclases, catalyse the first two steps of GA biosynthesis from geranylgeranyl diphosphate to ent-kaurene. The AtCPS1 protein has been shown to be imported by isolated chloroplasts with cleavage of its leader sequence (Sun and Kamiya, 1994). The amino acid sequence of AtKS1 suggests that this protein is also plastid-located (Yamaguchi et al., 1998), although this has not yet been demonstrated experimentally. The product of these cyclization reactions, ent-kaurene, is non-polar and likely to partition into membranes rather than remain in solution.
The oxidation steps from ent-kaurene to GA12 are catalysed by cytochrome P450 mono-oxygenases, which are generally localized in the endoplasmic reticulum. The first three of the P450-mediated steps of the pathway, from ent-kaurene to ent-kaurenoic acid, are catalysed by ent-kaurene oxidase (AtKO1) (Helliwell et al., 1999), which is encoded by the GA3 gene (Helliwell et al., 1998). The next three steps of the GA biosynthesis pathway, from ent-kaurenoic acid to GA12, are catalysed by ent-kaurenoic acid oxidase (AtKAO1 and AtKAO2 in Arabidopsis) (Helliwell et al., 2001). The final stages of GA biosynthesis are catalysed by soluble 2-oxoglutarate-dependent dioxygenase enzymes, which catalyse 20-oxidation (AtGA20ox1–3) and 3β-hydroxylation (AtGA3ox1, AtGA3ox2) reactions to produce bioactive GAs, and 2-oxidation reactions (AtGA2ox1–3), which inactivate GAs.
The aim of this study was to determine the subcellular localization of the enzymes of the GA biosynthesis pathway. Two complementary experimental approaches were used: confocal microscopy to detect protein fusions with green fluorescent protein (GFP) in tobacco leaf cells following microprojectile bombardment, and import of 35S-labelled proteins into isolated pea chloroplasts. We show that AtKS1 is targeted to the chloroplast stroma, AtKO1 is targeted to the outer envelope of the chloroplast, and AtKAO1 and AtKAO2 are targeted to the endoplasmic reticulum. AtKO1 therefore appears to be spatially positioned to link the plastid and endoplasmic reticulum stages of the GA biosynthesis pathway.
The protein sequences of AtCPS1, AtKS1, AtKO1, AtKAO1, AtKAO2, the dioxygenases of the GA biosynthesis pathway and pea RbcS were analysed using the ChloroP program (http://www.cbs.dtu.dk/services/ChloroP; Emanuelsson et al., 1999) to predict whether they contained plastid transit peptides (Table 1). ChloroP uses a neural network method to identify chloroplast transit peptides and cleavage sites based on a training set of proteins with known subcellular localizations. A transit peptide score of greater than 0.5 predicts a chloroplast transit peptide (cTP). For proteins with a predicted cTP, a cleavage site position and cleavage site score are also generated; a more positive cleavage site score indicates a higher probability of a cleavage site.
Table 1. Summary of ChloroP predictions for enzymes of the GA biosynthesis pathway
Genbank accession number
The chloroplast transit peptide (cTP) score, the prediction of whether the protein contains a cTP, the cleavage site (CS) score and the location of the predicted cleavage site is shown for each of the protein sequences analysed.
The AtCPS1 protein is known to be plastid-located and to have a cleavage site after the leader sequence (Sun and Kamiya, 1994), agreeing with the ChloroP prediction. AtKS1 was not predicted to have a plastid transit peptide by ChloroP, although Yamaguchi et al. (1998) suggested the presence of a transit peptide based on the N-terminal 44 amino acid residues of the protein being rich in hydroxylated amino acids and having a predicted pI of 10.3. The program predicted the AtKO1 protein to have a plastid transit peptide and a cleavage site between amino acid residues 28 and 29. The score for the AtKO1 cleavage site (−0.89) is low compared to that of the AtCPS1 protein (3.24) and RbcS (11.6). The AtKAO1 and AtKAO2 cytochrome P450s were both predicted not to have a chloroplast transit peptide, as would be expected for cytochrome P450 enzymes which are predominantly located in the endoplasmic reticulum. Finally, none of the dioxygenases was predicted to have a transit peptide, as expected for soluble enzymes.
The N-terminal sequences of AtCPS1, AtKS1 and AtKO1 target GFP to plastids
To test the ChloroP predictions for subcellular localization of this set of proteins, we used microprojectile bombardment of tobacco leaves with plasmid constructs encoding protein fusions to GFP (Figure 1). The GFP variant smGFP (Davis and Vierstra, 1998), which localizes to the nucleus and cytoplasm, was used for all protein fusions. For AtCPS1, AtKS1, AtKO1, AtKAO1 and AtKAO2, the first 100 amino acid residues of the protein were fused to GFP (giving TPCPS–GFP, TPKS–GFP, TPKO–GFP, TPKAO1–GFP and TPKAO2–GFP). The N-termini were used for these fusions as both plastid-targeting peptides and the membrane-anchoring portion of P450s are located at N-termini. Fusions of the full-length AtKO1 and AtGA20ox2 to GFP were also made (giving KO–GFP and 20ox2–GFP). The control constructs were GFP (smGFP as used in the test constructs without a leader sequence), TPRbcS–GFP (pea RbcS transit peptide (Anderson and Smith, 1986) fused to GFP), which is chloroplast-targeted, and mGFP5 (Haseloff et al., 1997), which is targeted to the endoplasmic reticulum.
The plasmid constructs were precipitated on tungsten particles and used to bombard tobacco leaves. After 24 h, the bombarded leaves were examined for GFP fluorescence using a confocal microscope. Both the TPCPS–GFP and TPKS–GFP constructs gave fluorescence associated with chloroplasts in epidermal cells (Figure 2A and B), similar to the chloroplast-targeted TPRbcS–GFP control construct (Figure 2C). The control smGFP construct gave fluorescence in the nucleus and cytoplasm, but not in chloroplasts (Figure 2D). The images shown are representative of many transformed cells and show superimposed chlorophyll (red) and GFP (green) channels for a series of optical sections. Where the GFP and chlorophyll signals overlap, the image colour is shifted to yellow. The chlorophyll signal from underlying mesophyll cells can also be seen; this signal is unavoidable due to the uneven nature of the leaf epidermis and the strength of the chlorophyll signal from the underlying mesophyll layer. However, the presence of chloroplasts in epidermal cells of most higher plants, including tobacco, is well documented (see Dupree et al., 1991), and these cells have been used previously as an experimental system to study the targeting of GFP fusion proteins (Takechi et al., 2000; Millen et al., 2001). There can be no doubt that the N-terminal 100 amino acid residues of AtCPS1 and AtKS1 are sufficient to target GFP to chloroplasts.
The TPKO–GFP and KO–GFP constructs both gave GFP fluorescence associated with chloroplasts (Figure 2E,F). The images show the separate chlorophyll (Figure 2Ei and Fi) and GFP (Figure 2Eiii and Fiii) channels, together with the merged channels (Figure 2Eii and Fii). A single mesophyll cell is shown for the TPKO–GFP construct (Figure 2E), clearly demonstrating that the GFP signal is associated solely with chloroplasts. Expression in a single epidermal cell containing a few large chloroplasts is shown for the KO–GFP construct (Figure 2F). The large chlorophyll-containing structures in the plane of this section all show GFP fluorescence associated with them. However, in addition, the KO–GFP construct produced some GFP signal in the cytoplasm (Figure 2F). This cytoplasmic signal may be in the endoplasmic reticulum, but the images obtained did not allow a conclusive location to be determined.
The TPKAO1–GFP and TPKAO2–GFP proteins (Figure 2G and H) both gave fluorescence with a pattern similar to the mGFP5 protein, which is targeted to the endoplasmic reticulum (Figure 2I). The proteins appear to be located in the endoplasmic reticulum but were excluded from the chloroplasts. There was no overlap of the GFP and chlorophyll signals from the transfected cells for these constructs. This suggests that the AtKAO1 and AtKAO2 cytochrome P450s are endoplasmic reticulum-located. The 20ox2–GFP protein showed the same localization as smGFP (Figure 2D), giving fluorescence in the nucleus and cytoplasm (Figure 2J), without any indication of association with the endoplasmic reticulum. From this result, it is unlikely that the AtGA20ox2 protein contains any targeting sequences, resulting in a localization predominantly in the cytoplasm.
The TPKO–GFP and KO–GFP constructs used to bombard tobacco leaves were also used to transform Arabidopsis plants. The TPKO–GFP and smGFP constructs (in the pBin19 vector) were used to transform the wild-type Arabidopsis Landsberg erecta ecotype. The KO–GFP construct was used to transform a ga3-2 ttg1-1 (transparent testa glabra) double mutant line (Helliwell et al., 1998). The transformed plants obtained with the KO–GFP construct were complemented with respect to the ga3-2 phenotype, i.e. they did not require exogenous gibberellin for normal development, but they still showed the ttg1 phenotypes of glabrous leaves and stems and yellow seeds. The transformed plants were indistinguishable from ga3-2 ttg1-1 transformed with 35S::GA3 (Helliwell et al., 2000) indicating that the KO–GFP fusion protein is a functional ent-kaurene oxidase. Western blot analysis of transgenic plants expressing TPKO–GFP and ga3-2 ttg1-1 plants expressing KO–GFP using antibodies to GFP showed that both lines contained fusion proteins of the expected sizes (data not shown). Only plants expressing the smGFP construct gave GFP fluorescence that could be clearly distinguished from chlorophyll fluorescence. GFP fluorescence was visible in nuclei and cytoplasm (data not shown) as described previously for the same construct in bombarded tobacco leaves (Figure 2D). In plants expressing the KO–GFP and TPKO–GFP constructs, the GFP fluorescence was too weak to detect, although the GFP fusion proteins were detectable in these plants based on Western blot analysis with antibodies to GFP.
ent-Kaurene oxidase is targeted to the plastid envelope
To determine the subchloroplast localization of the AtKS1 and AtKO1 proteins, we carried out protein import assays with isolated pea chloroplasts using 35S-labelled KO, TPKO–GFP, KO–GFP, TPKS–GFP and smGFP proteins, produced by transcription and translation of constructs in pSP72 plasmids (Figure 1b). The major translation products were approximately 55, 37, 80, 41 and 28 kDa, respectively, in good agreement with the expected molecular masses of the fusion proteins. The 35S-labelled translation products were incubated with intact pea chloroplasts in the presence of ATP, and subsequently treated with the protease thermolysin to remove any translation products that had not been imported into the chloroplasts. The chloroplasts were then re-isolated and fractionated into stroma (Figure 3, lane St), washed thylakoids (lane Th), thermolysin-treated thylakoids (lane pTh), inner envelope membranes (lane IE) and outer envelope membranes (lane OE) to determine the subchloroplast localization of the imported proteins.
35S-labelled KO, TPKO–GFP and KO–GFP were all associated with the washed chloroplast fraction (Figure 3A, B and C, lane C) but were not resistant to exogenously added thermolysin (lane C+), indicating that the proteins had not been translocated across the envelope membranes. There was no reduction in the size of any of these proteins, indicating that the N-terminal leader is not processed. In contrast, 35S-labelled smGFP was not detectably associated with re-isolated pea chloroplasts (Figure 3D), suggesting that the association of the KO proteins with the washed chloroplast fraction is due to binding to the outer envelope membrane. The chloroplasts used for these import experiments efficiently imported a stromal β-amylase to yield a protease-protected form (Lao et al., 1999) and a thylakoid lumen protein (R.M.M., unpublished data), indicating that the chloroplasts were import- competent and the protease treatment was successful.
35S-labelled TPKS–GFP was associated with the washed chloroplast fraction (Figure 3E, lane C) and was resistant to degradation by exogenously added thermolysin (Figure 3E, lane C+) indicating that the fusion protein was located within the chloroplast. Fractionation of the intact thermolysin-treated chloroplasts demonstrated that the 35S-labelled protein was present in the stromal fraction (Figure 3E, lane St). There was no change in the size of the imported protein, suggesting that the AtKS1 protein is not processed following import into chloroplasts.
In this paper, we describe experiments to determine the subcellular localization of the enzymes of the GA biosynthesis pathway in Arabidopsis. We initially used the ChloroP program to predict whether the proteins were likely to be located in chloroplasts. The ChloroP predictions were shown to be correct for AtCPS1 and AtKO1, which were targeted to chloroplasts in vivo and in vitro, but not for the AtKS1 protein, which was not predicted to have a plastid transit peptide, but which was targeted to chloroplasts in vivo and in vitro. This may reflect deficiencies in the algorithms of ChloroP.
GA biosynthesis starts from geranylgeranyl diphosphate synthesized in the plastid. The first two enzymes of the pathway in Arabidopsis, AtCPS1 and AtKS1, were shown to be located in the plastid stroma. AtCPS1 had previously been shown to be imported into isolated chloroplasts with cleavage of an N-terminal targeting peptide (Sun and Kamiya, 1994). We have shown that the N-terminal 100 amino acid residues of the AtKS1 protein are sufficient to target smGFP to the stroma of chloroplasts. We did not observe any change in size of the protein following import, suggesting that the AtKS1 leader sequence is not cleaved following import into plastids. We cannot rule out the possibility that the fusion to GFP affects the processing of the AtKS1 leader sequence. Most nuclear-encoded chloroplast proteins examined to date have a cleavable N-terminal transit peptide (Bruce, 2000; De Boer and Weisbeek, 1991), although some proteins targeted to the chloroplast outer envelope membrane lack a cleavable transit peptide (Keegstra and Cline, 1999). The product of the sequential reactions catalysed by AtCPS1 and AtKS1 is ent-kaurene, which is non-polar and has a low solubility in aqueous media; it is therefore likely to partition into membranes.
The oxidation of ent-kaurene to ent-kaurenoic acid is catalysed by AtKO1. We have used two independent approaches to show that the N-terminal 100 amino acids of the ent-kaurene oxidase protein are sufficient to target GFP to the plastid envelope. In tobacco leaf cells bombarded with the TPKO–GFP plasmid construct, the protein localized to the chloroplasts. The AtKO1 N-terminus also targets GFP to the outer surface of isolated chloroplasts in in vitro import experiments. The full-length AtKO1 peptide and the KO–GFP fusion also associate with isolated chloroplasts and are not protease-resistant, consistent with a location on the outer surface of the outer envelope. In tobacco leaf cells bombarded with TPKO–GFP, GFP fluorescence is associated with chloroplasts. This construct also gave some GFP fluorescence which may be associated with the endoplasmic reticulum. The other two P450s tested (AtKAO1 and AtKAO2) did not show any evidence of chloroplast targeting.
Overall, our results suggest that AtKO1 is targeted to the outer envelope membrane of the plastid. Following synthesis in the chloroplast stroma, ent-kaurene is likely to partition into chloroplast membranes. The location of AtKO1 on the plastid envelope suggests that ent-kaurene export from the plastid is linked to its further oxidation (Figure 4). This is the first report of a cytochrome P450 protein located on the outer face of the chloroplast, although the P450 allene oxide synthase in barley has been shown to be located within chloroplasts (Maucher et al., 2000). We also have some evidence suggesting that some KO–GFP is located in the endoplasmic reticulum. In yeast assays of AtKO1, the intermediate products ent-kaurenol and ent-kaurenal were detected (Helliwell et al., 1999), indicating that AtKO1 may not channel substrate straight to ent-kaurenoic acid. ent-kaurenol and ent-kaurenal are more water-soluble than ent-kaurene, and their further metabolism could be catalysed by endoplasmic reticulum-localized AtKO1.
The N-termini of both AtKAO proteins targeted smGFP to the endoplasmic reticulum in bombarded tobacco leaves, indicating that the cytochrome P450-mediated steps of the GA pathway from ent-kaurenoic acid to GA12 are located in this compartment of the cell. These results confirm that the stages of the GA pathway are compartmentalized within the plant cell, with AtKO1 providing a crucial link between the plastid- and endoplasmic reticulum-located steps of the pathway.
Arabidopsis experiments were carried out with the Landsberg erecta ecotype; the ga3-2 ttg1-1 double mutant line was as described in Helliwell et al. (1998). Arabidopsis transformation was by the floral dip method (Clough and Bent, 1998). Seeds were spread on MS plates supplemented with kanamycin (50 mg l−1) to select for transformed plants.
The plasmid constructs for microprojectile bombardment and Arabidopsis transformation were assembled in the cassette illustrated in Figure 1(a). The CaMV 35S promoter from pBI101.2 (Jefferson et al., 1987) directs expression of the GFP fusion protein, consisting of the N-terminal 100 amino acid residues or the full-length protein fused in-frame to smGFP (Davis and Vierstra, 1998). The plasmid backbone was pUC18 for the microprojectile bombardments and pBin19 for Arabidopsis transformation. DNA fragments encoding leader sequences or proteins were amplified by PCR using Pfu DNA polymerase (Stratagene, La Jolla, California, USA) with either a BamHI or XbaI site incorporated in the 5′ primer and a StuI site incorporated in the 3′ primer. Cloned PCR products were sequenced to check the accuracy of PCR amplification. These constructs were named TPCPS–GFP, N-terminal 100 amino acid residues of AtCPS1 fused to GFP; TPKS–GFP, N-terminal 100 amino acid residues of AtKS1 fused to GFP; TPKO–GFP, N-terminal 100 amino acid residues of AtKO1 fused to GFP; KO–GFP, full-length AtKO1 fused to GFP; TPKAO1–GFP, N-terminal 100 amino acid residues of AtKAO1 fused to GFP; TPKAO2–GFP, N-terminal 100 amino acid residues of AtKAO2 fused to GFP; 20ox2–GFP, full-length AtGA20ox2 fused to GFP.
The plasmid constructs for transcription and translation in vitro encoded identical proteins to those used for microprojectile bombardment. The GFP fusions were inserted between the BamHI and SacI sites of pSP72 (Figure 1b); the construct names are prefixed with pSP72. The pSP72-KO construct contained the complete AtKO1 open reading frame inserted between the XhoI and BamHI sites of pSP72.
Production of 35S-labelled precursor proteins
Plasmids pSP72-KO, pSP72-TPKO-GFP, pSP72-KO-GFP, pSP72-TPKS-GFP and pSP72-GFP were linearized by restriction digestion with BamHI, EcoRI, EcoRV, EcoRI and SacI, respectively, and transcribed in vitro with SP6 RNA polymerase. 35S-labelled precursor proteins were synthesized by incubation of the transcription products in a wheatgerm translation system, with 35S-Promix (Amersham Pharmacia, Amersham, UK), essentially as described by Mould and Gray (1998b).
Chloroplast protein import assays
Intact chloroplasts were prepared from approximately 100 g of pea shoots (Pisum sativum cv. Progress no. 9) using Percoll-step gradients as described by Mould & Gray (1998a). Protein import assays contained intact pea chloroplasts (1 mg chlorophyll), 5 mm methionine, 5 mm cysteine and 10 mm MgATP in a final volume of 1.0 ml of import buffer (50 mm HEPES–KOH, pH 8.0, 0.33 m sorbitol) with 95 µl 35S-labelled translation products. Assays were incubated in the light (100 µmol photons m−2 sec−1) for 60 min at 25°C. After incubation, chloroplasts were treated with thermolysin (to give a final concentration of 0.2 mg ml−1 in import buffer) for 30 min on ice and then the protease reaction was stopped by the addition of EDTA to 50 mm in import buffer. Chloroplasts were re-isolated by centrifugation through a cushion of 40% Percoll in import buffer and then washed in import buffer (Mould and Gray, 1998b). An aliquot (1/10) of the thermolysin-treated chloroplast sample was taken for analysis and the remainder was fractionated essentially as described by Schnell and Blobel (1993). The fractionation of the chloroplasts was confirmed by SDS–PAGE (Laemmli, 1970) followed by Coomassie blue staining.
Samples of thermolysin-treated chloroplasts, stromal fraction, thylakoids and thermolysin-treated thylakoids were quantified by SDS–PAGE (Laemmli, 1970) followed by Coomassie blue staining and scanning densitometry of stained protein bands (Rubisco subunits and LHCP). Equivalent amounts of these fractions (approximately equal to 2% of the chloroplasts recovered from the Percoll gradient), and 50% of the inner and outer envelope fractions recovered were analysed by electrophoresis on 12.5% polyacrylamide gels in the presence of SDS, followed by fluorography.
Microprojectile bombardment and GFP imaging
Microprojectile bombardment of tobacco (Nicotiana tabacum cv Samsun) leaves was performed with a BioRad PDS-1000 (BioRad, Hercules, California, USA), as previously described by Hibberd et al. (1998). Imaging of GFP and chlorophyll fluorescence was performed on a Leica DMRXA confocal laser scanning microscope 24 h after bombardment. Serial optical sections of 0.5–2 µm were obtained using an excitation wavelength of 488 nm, with GFP and chlorophyll images being collected through FITC and TRITC filters, respectively. A combined extended focus image was then produced from the confocal series using Leica TCS-NT software.
We would like to thank Jane Pulford, Anna Wielopolska and Helen Adams for technical assistance, Frank Gubler for Western blot analysis and Jim Haseloff for the mGFP5 clone. J.A.S. was supported by a research grant from Biotechnology and Biological Sciences Research Council (BBSRC). R.M.M. was supported by a Royal Society University Research Fellowship.