Arabidopsis polyamine biosynthesis: absence of ornithine decarboxylase and the mechanism of arginine decarboxylase activity

Authors


For correspondence (fax +44 1603 507723; e-mail tony.michael@bbsrc.ac.uk)

Summary

Unlike other eukaryotes, which can synthesize polyamines only from ornithine, plants possess an additional pathway from arginine. Occasionally non-enzymatic decarboxylation of ornithine could be detected in Arabidopsis extracts; however, we could not detect ornithine decarboxylase (ODC; EC 4. 1.1.17) enzymatic activity or any activity inhibitory to the ODC assay. There are no intact or degraded ODC sequences in the Arabidopsis genome and no ODC expressed sequence tags. Arabidopsis is therefore the only plant and one of only two eukaryotic organisms (the other being the protozoan Trypanosoma cruzi) that have been demonstrated to lack ODC activity. As ODC is a key enzyme in polyamine biosynthesis, Arabidopsis is reliant on the additional arginine decarboxylase (ADC; EC 4.1.1.9) pathway, found only in plants and some bacteria, to synthesize putrescine. By using site-directed mutants of the Arabidopsis ADC1 and heterologous expression in yeast, we show that ADC, like ODC, is a head-to-tail homodimer with two active sites acting in trans across the interface of the dimer. Amino acids K136 and C524 of Arabidopsis ADC1 are essential for activity and participate in separate active sites. Maximal activity of Arabidopsis ADC1 in yeast requires the presence of general protease genes, and it is likely that dimer formation precedes proteolytic processing of the ADC pre-protein monomer.

Introduction

Polyamines are small, evolutionarily ancient polycations found in all cells. The diamine putrescine and the triamine spermidine are found in most cells, but the tetraamine spermine is predominantly found in eukaryotes (Cohen, 1998). This very short primary metabolic pathway is of interest because increased polyamine levels are associated with tumour formation (Pegg, 1988), it is a target for chemotherapeutic intervention in cancer (Davidson et al., 1999) and protozoan diseases (Grishin et al., 1999), and spermine is known to regulate a number of ion channels and receptors (Igarashi and Kashiwagi, 2000). The role of polyamines in the cell cycle is becoming clearer with the discovery of an internal ribosome entry site active only in mitosis in the human ODC mRNA (Pyronnet et al., 2000). In Arabidopsis, a severe growth/cell elongation defect is seen in mutants of a spermine synthase gene (Hanzawa et al., 2000). Polyamines affect many cellular processes such as translation, transcription and chromatin structure (Marton and Morris, 1987; Pollard et al., 1999). In addition, the polyamine pathway is of intrinsic interest because of the unusual and sophisticated nature of its regulation by the polyamine products. This regulation includes antizyme-induced degradation of ODC (Ivanov et al., 2000), manifested through polyamine-mediated ribosomal frame-shifting of the antizyme message, and also polyamine-mediated translational regulation of S-adenosylmethionine decarboxylase (AdoMetDC) via the activity of an inhibitory upstream small open reading frame (Mize et al., 1998; Raney et al., 2000).

Eukaryotic cells synthesize putrescine directly from ornithine through the activity of ODC. An aminopropyl group derived from decarboxylated S-adenosylmethionine (dcSAM) is transferred to putrescine by spermidine synthase to form spermidine, and another aminopropyl group is added to spermidine by spermine synthase to form spermine (Figure 1a). The dcSAM is formed by the activity of AdoMetDC (also known as SAMDC). Plants and some bacteria possess an additional route to putrescine, indirectly from arginine, by the activity of the arginine decarboxylase (ADC) pathway. In plants, the product of ADC activity is agmatine, which is subsequently converted to N-carbamoylputrescine and putrescine by the successive activity of agmatine iminohydrolase and N-carbamoylputrescine amidohydrolase.

Figure 1.

The plant polyamine biosynthetic pathway and alignment of the Arabidopsis ADC1 and mouse ODC.

(a) The polyamine biosynthetic pathway from putrescine to spermine is found in all eukaryotes. Plants, uniquely amongst eukaryotes, possess the additional arginine pathway to putrescine. The aminopropyl group is obtained from decarboxylated S-adenosylmethionine produced by the activity of AdoMetDC from S-adenosylmethionine. (b) Alignment of the Arabidopsis ADC1 with the mouse ODC amino acid sequence. Identical conserved residues are shown in red. The mouse ODC K69 and Arabidopsis ADC1 K136 and the mouse C360 and Arabidopsis ADC1 C524 are shown in green; the residues flanking the putative ADC1 cleavage site are shown in green and underlined.

ODC is the first committed step to polyamine biosynthesis in animals and fungi. In mammalian cells it is recognized as a proto-oncogene (Auvinen et al., 1992), is a direct target of c-Myc (Bello-Fernandez et al., 1993) and is a mediator of c-Myc-induced apoptosis (Packham and Cleveland, 1994). The plant ODC is very similar in sequence to animal ODCs, and its expression is correlated with cell proliferation (Michael et al., 1996). An Arabidopsis ODC gene has not been cloned or identified, although we have screened four different cDNA libraries and a genomic library using the Datura stramonium ODC cDNA as a probe. We have been unable to detect enzyme activity in rapidly dividing transgenic root cultures, leaves and suspension cultures.

ADC is of agronomic interest in plants because fungi, nematodes and insects do not possess this pathway to putrescine, providing the possibility of differential inhibition of host and pathogen/pest polyamine pathways (Bailey et al., 2000). The reaction mechanism of ADC is poorly understood, but the pioneering work of Malmberg and colleagues in isolating the oat ADC cDNA (Bell and Malmberg, 1990) indicated that the oat ADC pro-enzyme was synthesized as a 66 kDa preprotein and subsequently cleaved into 42 kDa N-terminal and 24 kDa C-terminal domain polypetides (Malmberg et al., 1992). It was suggested that that the two processed domains of the peptide were held together in the enzyme by a disulphide bridge. Only the processed form of ADC is detectable in vivo in Arabidopsis plants (Watson and Malmberg, 1996). The processing activity necessary for cleavage of the oat pre-protein was shown to be distinct from ADC itself, sensitive to high concentrations of zinc and insensitive to polyamines (Malmberg and Cellino, 1994). By contrast, Borrell et al. (1996) found that spermine inhibited the processing of the oat ADC pre-protein. It was also tentatively suggested after extensive dialysis of samples that the oat ADC did not require pyridoxal-5′-phosphate (PLP) as a co-factor (Malmberg and Cellino, 1994).

The plant ADC belongs to the group IV class of PLP-dependent decarboxylases (Sandmeier et al., 1994), and there are limited, discrete regions of homology between ODC, ADC and the bacterial meso-diaminopimelate decarboxylase (DAPDC) (Grishin et al., 1995; Michael et al., 1996; Sandmeier et al., 1994). Structures of the mouse, human and trypanosome ODC were solved recently (Almrud et al., 2000; Grishin et al., 1999; Kern et al., 1999), and due to extensive site-directed mutant analysis of enzyme activity, the reaction mechanism of the homodimeric ODC enzyme is well-characterized (Coleman et al., 1993; Coleman et al., 1994; Osterman et al., 1994; Osterman et al., 1995a; Osterman et al., 1995b; Osterman et al., 1997; Osterman et al., 1999; Tobias and Kahana, 1993; Tobias et al., 1993; Tsirka and Coffino, 1992). In particular, eukaryotic ODC is a PLP-dependent homodimer with two active sites formed in trans across the interface of the homodimer subunits. The lysine residue represented by the mouse K69 forms a Schiff base with PLP, and the cysteine represented by C360 forms part of a substrate specificity loop. The K69 belongs to the N-terminal α/β barrel domain, and the C360 belongs to the C-terminal β-structure domain. Each residue participates in separate active sites on the monomer structure, but K69 and C360, on opposite sides of the homodimer interface, belong to the same active site due to the head-to-tail arrangement of the homodimer. Arabidopsis has two ADC genes: spe1 (ADC1) on chromosome 2 and spe2 (ADC2) on chromosome 4 (accession numbers AC007195 and AL023094, respectively). Comparison of the mouse ODC and ArabidopsisADC1amino acid sequences facilitated identification of the ADC equivalents of the mouse ODC K69and C360residues (Figure 1b). We have used site-directed mutagenesis of the ArabidopsisADC1and expression of mutant ADC enzymes in yeast to investigate the mechanism of ADC enzymatic activity.

Results

Absence of ornithine decarboxylase in Arabidopsis

Activities of ODC, ADC and AdoMetDC were simultaneously determined in fast-growing suspension cultures and rapidly dividing transformed root cultures of Arabidopsis. In suspension cultures 6 days after subculturing, both ADC and AdoMetDC activities were readily detectable but ODC activity was absent (Table 1). In the presence of 5 mmα-difluoromethylornithine and α-difluoromethylarginine (DFMO and DFMA, specific enzyme-activated suicide inhibitors of ODC and ADC, respectively), ADC activity was reduced by 96% whereas AdoMetDC activity was relatively unaffected. In transformed root cultures, the activities of ADC and AdoMetDC were prominent 16 h after subculturing and were repressed by 26% for ADC and 72% for AdoMetDC by the presence of 500 μm spermidine, whereas ODC activity was always undetectable (Table 1).

Table 1.  Decarboxylase activities in Arabidopsis suspension cultures and transformed root cultures
 Suspension culturesTransformed root cultures
H2OMA + MOH2OSpermidine
  1. Activities are presented as pmol CO2 mg−1 h−1. Results are the mean of duplicate assays of triplicate 200 ml cultures for the suspension cultures and the mean of duplicate assays of three pooled 100 ml cultures for the transformed root cultures. DFMA and DFMO (MA + MO) were present in suspension cultures at 5 mm and cultures were harvested six days after subculture with growth at 25°C. Transformed root cultures were grown in the presence or absence of 500 µm spermidine and roots were harvested 16 h after subculturing. n.d., not detectable

ODCn.d.n.d.n.d.n.d.
ADC78 ± 173 ± 32150 ± 1501590 ± 20
AdoMetDC95 ± 5121 ± 321015 ± 25280 ± 10

Occasionally, low levels of ornithine decarboxylation were found in Arabidopsis suspension culture extracts 1 day and 8 days after subculturing. When the extracts were boiled, the ornithine decarboxylation increased whereas the AdoMetDC activity, and at least for day 1 the ADC activity, were greatly decreased (Figure 2a). There is evidently non-enzymatic decarboxylation of ornithine and to a lesser extent of arginine in Arabidopsis, as found in other plant species (Birecka et al., 1985). One day after subculture, the Arabidopsis suspension cultures are actively dividing; however, at 8 days, they are in stationary phase. It is of interest to note that, whereas the activities of both ADC and AdoMetDC are reduced by approximately 80% at 8 days compared to 1 day after subculture, the level of ornithine decarboxylation was unchanged. It is possible that a substance is present in the Arabidopsis extracts that might interfere with the ODC assay, for example something that might oxidize the essential sulphydryl groups of ODC. Therefore, we mixed equal amounts of extract from Arabidopsis leaves and tobacco leaves and also from Arabidopsis and tobacco stationary-phase suspension cultures and determined the effect on tobacco ODC activity (Figure 2b). The tobacco ODC activity was reduced by approximately half, reflecting the degree of dilution of the tobacco extract by the Arabidopsis extract. This demonstrates that observed absence of ODC activity in Arabidopsis is not due to interference with the ODC assay.

Figure 2.

Decarboxylase activity of untreated and boiled Arabidopsis suspension cultures and comparison of decarboxylase activities of Arabidopsis and tobacco leaf and suspension culture extracts.

(a) Arabidopsis suspension culture extracts were prepared at 1 and 8 days after subculturing. Extracts were boiled for 2 min and each result represents the mean of two assays of five pooled 200 ml cultures. Decarboxylase activity was determined by 14C CO2 release. (b) Decarboxylase activities in tobacco leaves (tob leaf) from cultivar XHFD8 (Burtin and Michael, 1997) and Arabidopsis thaliana Landsberg erecta rosette leaves (A.t. leaf) and in tobacco BY2 and Arabidopsis suspension cultures (tob cells and A.t. cells, respectively) at stationary phase. Extracts were analysed individually or mixed in equal portions. The control Arabidopsis extract was boiled before assaying. The tobacco suspension culture ODC activities are off the scale of the graph and the activities are marked next to the appropriate bar. Results represent the mean of duplicate assays.

Growth of Arabidopsis seedlings in vitro in the presence of 2.5 mm DFMO resulted in severe stunting of the plants due to reduced stem elongation; cauline leaves were wrinkled and darker green, and flowering was greatly delayed (results not shown). Growth in the presence of 2.5 mm DFMO and either 1.0 mm putrescine or 0.5 mm spermidine did not reverse the effect of DFMO in Arabidopsis on growth and morphology, indicating that the effect of DFMO was not specific to the polyamine pathway and therefore not due to ODC inhibition (results not shown).

We have repeatedly failed to detect ODC sequences in diverse cDNA libraries of Arabidopsis (seedlings, flowers, mixed tissues, transformed root cultures) and in a λZAP genomic library (Stratagene) using the Datura stamonium ODC cDNA (Michael et al., 1996) as a probe. In addition, the Arabidopsis genome is essentially complete except for a few centromeric regions, and we could not detect either intact or degraded ODC sequences or any expressed sequence tags (ESTs). The Arabidopsis genome sequence and Arabidopsis ESTs were interrogated with the amino acid sequence of the D. stramonium ODC (Michael et al., 1996). No ODC sequences were detected by TBLASTN, but the C-terminus of DAPDC was detected on chromosome V. DAPDC is also a group IV PLP-dependent decarboxylase involved in lysine metabolism, related to ODC and ADC (Michael et al., 1996). Considering the lack of bona fide ODC enzyme activity and of genomic and EST sequences, we conclude that ODC is absent from Arabidopsis.

K136 and C524 are essential for Arabidopsis ADC1 function, and co-expression of the K136A and C524A mutants restores activity

The absence of ODC in Arabidopsis means that ADC is probably the sole route for putrescine biosynthesis. Understanding the mechanism of ADC activity therefore has particular importance in Arabidopsis polyamine metabolism. Previously the pea and oat ADC were expressed in yeast and shown to be active (Klein et al., 1999; Pérez-Amador et al., 1995). There is no endogenous ADC activity in yeast, which uses ODC as the only pathway for producing putrescine; thus yeast provides a convenient system for expressing and assaying ADC activity. To assess the efficiency of galactose-induced expression of the Arabidopsis ADC1 ORF in the yeast wild-type strain YW5-1B at 25 and 30°C, the full-length ADC1 ORF was expressed in the galactose-inducible vector pYX243 as a translational fusion. After 24 h induction, the mean ADC activity of duplicate cultures at 25°C was 33.9 nmol CO2 h−1 mg−1 protein, whereas the mean uninduced activity was 0.24 nmol CO2 h−1 mg−1 protein. At 30°C, the mean induced activity was 29.8 nmol CO2 h−1 mg−1 protein, therefore yeast cultures used for ADC expression were subsequently grown at 25°C.

Comparison of the Arabidopsis ADC1 and the mouse ODC amino acid sequences (Figure 1b) shows there is only a low level of sequence conservation. However, we were able to identify tentatively the ADC1 amino acids K136 and C524 as the equivalents of the mouse ODC residues K69 and C360, previously shown to be involved in Schiff base formation with the co-factor PLP and substrate binding, respectively (Poulin et al., 1992). Using site-directed mutagenesis, the ADC1 N-terminal domain K136 and C-terminal domain C524 residues were converted to alanine to form the K136A and C524A mutant forms of ADC1. Expression of the K136A/pYES2 or C524A/pYX243 mutant forms of ADC resulted in a large reduction of ADC activity: a 97.7% mean reduction for K136A over three experiments and a 91.8% mean reduction for C524A over two experiments (Figure 3a). This is in general agreement with the activities found for the K69A and C360A mouse ODC mutants: 99.5% reduction for K69A and 97.9% reduction for C360A (Coleman et al., 1994).

Figure 3.

Galactose-induced activity of Arabidopsis ADC1 site-directed mutants in yeast wild-type strain YW5-1B.

(a) C524A, C524A mutant ADC1 in pYX243; K136A, K136A mutant ADC1 in pYES2; ADC/pYX, wild-type ADC1 in pYX243; ADC/pYES, wild-type ADC1 in pYES2; Cys/Lys1-4, four independent cultures of YW5-1B co-expressing the K136A and C524A mutant ADC1 plasmids. (b) Schematic representation of the possible head-to-tail dimer conformations of the co-expressed K136A and C524A Arabidopsis ADC1 site-directed mutants. The N-terminal domain is shown in red, the C-terminal domain in green, and functional active sites are represented by the yellow ellipsoid. Non-functional active sites are indicated by the presence of the K and/or C mutated residues.

When the K136A/pYES2 and C524A/pYX243 ADC1 plasmids were co-expressed in the same yeast cells (Figure 3a), there was functional rescue of ADC activity (approximately 25% of the mean of the wild-type ADC1/pYES2 and ADC1/pYX243 activities). The relative activities of the wild-type ADC1 ORF in the pYX243 and pYES2 vectors varied to some extent between experiments. Our premise was that the ADC mechanism is similar to ODC, and that a head-to-tail homodimer will be formed with two active sites acting in trans across the interface of the dimer. The K136 residue is found in the N-terminal domain and C524 is found in the C-terminal domain of the monomer. If the ADC mechanism is similar to ODC, then one active site will be formed in trans between the N-terminal domain of one monomer and the C-terminal domain of the other monomer making up the dimer. In this way there will be two active sites per homodimer. When K136A and C524A ADC1 mutants are co-expressed in the same cell, four different types of dimer can be formed: K136A + K136A, C524A + C524A, K136A + C524A and C524A + K136A (Figure 3b). K136A homodimers and C524A homodimers will have both active sites disabled, whereas the mixed dimers containing K136A and C524A will possess one functional active site per dimer. Thus, from the four dimer combinations, representing eight active sites, only K136A + C524A and C524A + K136A will have one functional active site per dimer, representing two functioning active sites out of a possible eight or 25% of the wild-type homodimer activity. This is the activity found when K136A and C524A were co-expressed in yeast cells (Figures 3a and 4a), suggesting that the mechanism of ADC activity is as a head-to-tail homodimer.

Figure 4.

Galactose-induced expression of the Arabidopsis ADC1 in wild-type and a multiply-deficient protease mutant of yeast.

(a) Further independent experiments with the strains shown in Figure 3(a) with, in addition: Nterm, the N-terminal domain of ADC1 up to I472 in pYX243; Cterm, the C-terminal domain of ADC1 from G473 in pYES2; C/N term1/2, two independent cultures of YW5-1B co-expressing the N- and C-terminal domains of ADC1. (b) Activity of Arabidopsis wild-type ADC1 in a multiply-deficient yeast protease mutant SC1216. YWpYX, empty expression vector pYX243 in wild-type yeast strain YW5-1B; YW ADC, wild-type ADC1 in pYX243 in yeast wild-type strain YW5-1B; SC ADC1/2, two independent cultures of ADC1 in pYX243 in multiply-deficient yeast protease mutant SC1216.

Co-expression of the separated N- and C-terminal domains of ADC1 does not restore ADC activity

The oat ADC pre-protein monomer is cleaved to produce an N-terminal domain of 42 kDa and a C-terminal domain of 24 kDa (Malmberg et al., 1992). Malmberg and colleagues identified (Bell and Malmberg, 1990) the N-terminus of the 24 kDa C-terminal processed peptide by peptide sequencing, and by aligning the Arabidopsis ADC1 sequence with the oat ADC sequence we were able to identify isoleucine 472 (I472) as the putative equivalent N-terminus of the Arabidopsis ADC1 C-terminal domain. An artifically processed form of the Arabidopsis ADC1 protein was produced by introducing a haemagglutinin epitope tag and stop codon immediately after I472 to generate a 485 amino acid N-terminal fragment as a translational fusion in pYX243, and by introducing an ATG start codon immediately before glycine 473 to generate a 231 amino acid C-terminal fragment as a transcriptional fusion in pYES2 and then co-expressing the fragments in yeast. Figure 4(a) shows that the individually expressed ADC1 N-terminal domain/pYX243 and ADC1 C-terminal domain/pYES2 plasmids possessed no ADC activity. The co-expressed ADC1 N-terminal and C-terminal plasmids, in principle encoding the two final processed domains of ADC, did not restore ADC activity. This suggests that the separated processed products are not sufficient for ADC activity, although we cannot rule out the possibility that differences in the stoichiometry of the two separated domains might interfere with activity.

General protease activity is required for Arabidopsis ADC1 activity in yeast

It has been proposed that a specific plant zinc-sensitive protease activity might be required for oat ADC processing and activation (Malmberg and Cellino, 1994). Activity of the pea (Perez-Amador et al., 1995), oat (Klein et al., 1999) and Arabidopsis (this paper) ADC enzymes in yeast suggests that a more general phenomenon is operating. We introduced the full-length wild-type Arabidopsis ADC1 ORF in pYX243 into a multiply-deficient protease mutant of yeast (SC1216), deficient in four different general proteases. Figure 4(b) shows that while the full-length ADC1 ORF was active in the wild-type strain YW5-1B, there was a mean 95% decrease in ADC activity in the protease-deficient strain. However, it is interesting to note that there was not a complete abolition of activity, suggesting either that another protease activity is present or that there is some basal ADC activity with unprocessed ADC.

The activities we obtained for the expression of the Arabidopsis ADC1 in yeast ranged from 3 to 56 nmol CO2 h−1 mg−1 protein, depending on the vector and growth conditions. An approximately similar figure of 7.3 nmol CO2 h−1 mg−1 protein was obtained for the activity of the oat ADC in yeast (Klein et al., 1999). However, expression of the pea ADC in yeast resulted in an activity of 2700 nmol CO2 h−1 mg−1 protein (Pérez-Amador et al., 1995), similar to the endogenous ADC activity of E. coli, which was found to be 2580 nmol CO2 h−1 mg−1 protein (Klein et al., 1999).

Discussion

Until now, the only known example of a eukaryotic organism lacking ODC activity was the epimastigote stage (insect stage) of the human protozoan pathogen Trypanosoma cruzi, the causative agent of Chagas' disease. All attempts to detect enzyme activity in this organism have failed (Ariyanayagam and Fairlamb, 1997; Hunter et al.I, 1994); however, the epimastigotes take up exogenous putrescine and cadaverine very efficiently in a saturable manner, with a Km for putrescine of 2 μm (Le Quesne and Fairlamb, 1996). T. cruzi possesses an active AdoMetDC gene and can synthesize spermidine from exogenously supplied putrescine (Kinch et al., 1998; Persson et al., 1998). Our finding of the absence of ODC activity and the absence of intact or degraded ODC gene sequences in the Arabidopsis genome is surprising. It is possible that an ODC gene may be found in a difficult-to-sequence centromeric region, but the absence of any ODC EST sequences suggests that ODC is absent from this organism. Plant AdoMetDC mRNAs contain two highly conserved overlapping upstream open reading frames in the 5′ leader sequence (Franceschetti et al., 2001). The upstream tiny upstream ORF consists of three codons in angiosperms and Pinus taeda, but in both Arabidopsis AdoMetDC mRNAs, the tiny upstream ORF contains four codons as do the closely related Brassica juncea AdoMetDC mRNAs. It would be of interest to determine whether the change in the tiny upstream ORF is a consequence of loss of ODC. A salient question must be why Arabidopsis, as a free living organism, has lost ODC, as this enzyme activity is usually essential for cellular proliferation. There is a previous report of ODC activity in detached Arabidopsis leaves, but this activity was only 22 pmol CO2 h−1 mg−1 protein (Feirer et al., 1997), whereas we detected ADC activity in Arabidopsis suspension cultures of 900 pmol CO2 h−1 mg−1 protein but no detectable ODC activity. In Arabidopsis transformed root cultures, we did not detect any ODC activity but ADC activity was present at approximately 2000 pmol CO2 h−1 mg−1 protein.

Ethylmethane sulphonate-induced mutants affecting both Arabidopsis ADC enzyme activities have been described (Watson et al., 1998), and an En-1 transposon mutant of ADC2 revealed differential inducibility of the two ADC genes by osmotic stress (Soyka and Heyer, 1999). Two ADC genes have been found in all 12 species of the Brassicaceae examined to date, except for the basal genus Aethionema (Galloway et al., 1998), whereas in three outgroup taxa examined, only one ADC gene was found. In this respect, it would be interesting to determine whether the ADC gene duplication in the Brassicaceae coincides with loss of ODC activity. Much of the difference between the ADC1 and ADC2 protein amino acid sequences is at the N-terminus, indicating that the subcellular location of the two proteins could be different. Although ODC is a cytoplasmic and perinuclear protein in mammalian cells (Cohen, 1998), oat ADC was suggested to be localized to the chloroplast (Borrell et al., 1995).

Considering the unusual pivotal role of ADC in Arabidopsis polyamine biosynthesis, it is apt to revisit the mechanism of ADC activity. The previous model for ADC was put forward without knowledge of the ODC crystal structure. This proposed that the oat ADC pre-protein was cleaved into 42 and 24 kDa N- and C-terminal domains, respectively (Malmberg et al., 1992), held together by a disulphide bridge. It was thought that the increased molecular weight of the cleaved ADC products in the absence of thiol-reducing agents indicated that the enzyme might use such a disulphide bond between the N- and C-terminal domains of the monomer. Furthermore, it was tentatively proposed that the PLP co-factor was not necessary for function, as extensively dialysed ADC preparations did not require PLP for activity (Malmberg and Cellino, 1994). Proteolytic processing of ADC was proposed to be due to a protease activity distinct from ADC itself. Our results with in vitro-generated site-directed mutants of the Arabidopsis ADC1 cDNA, based on alignment with critical amino acids of the mouse ODC, cause us to suggest an alternative model for the mechanism of ADC activity. In particular we have shown that ADC is probably a homodimer in a head-to-tail arrangement with two active sites per homodimer operating in trans across the subunit interface. The behaviour of the single mutants, K136A and C524A, reflects the activity of the equivalent mouse ODC mutants. Mutation of the Arabidopsis ADC1 K136 and mouse ODC K69 has a severe effect on enzyme activity (Coleman et al., 1994). In the mouse ODC, this lysine binds the PLP co-factor via a Schiff base linkage and is critical for the mechansim of enzymatic activity (Kern et al., 1999; Osterman et al., 1999). The importance of the Arabidopsis ADC1 K136 residue in ADC activity and the known role of the equivalent mouse ODC K69 in PLP co-factor binding, which is critical to the enzymatic mechanism of decarboxylation (Kern et al., 1999), suggests that PLP is essential for ADC activity. The Arabidopsis ADC1 C524A and mouse ODC C360A mutants had a less critical effect. Although the mouse ODC C360 was identified as part of the substrate recognition component of the active site through its binding to radiolabelled DFMO (Poulin et al., 1992), the mouse ODC D361 residue is now recognized as the primary substrate binding determinant (Osterman et al., 1995a), and an equivalent aspartate is present in the Arabidopsis ADC1 (D525). The less critical effect of the C524A/C360A mutation is likely to be due to other residues such as D525 binding the substrate sufficiently for a low level of activity. It is unlikely that the two domains of the oat ADC monomer are held tgether by a disulphide bridge, as the oat ADC equivalent of the Arabidopsis ADC1 C524 will be required for substrate binding and there is only one cysteine residue in the oat ADC C-terminal domain (Bell and Malmberg, 1990).

The pea, oat and now Arabidopsis ADC enzymes have been shown to be active in yeast (Klein et al., 1999; Pérez-Amador et al., 1995). Yeast does not possess ADC and so is unlikely to have an ADC-specific cleavase/processing factor. Activity of the Arabidopsis ADC1 was greatly reduced in a multiply-deficient yeast protease mutant. This suggests that the ADC pre-protein may be processed by a general protease rather than a specific cleavase factor. The processing site of the oat ADC (Bell and Malmberg, 1990) corresponds exactly to a region of ODC containing a protease-sensitive loop (Osterman et al., 1995b). This protease-sensitive loop corresponds closely but not exactly with the interdomain region of the mouse ODC (Kern et al., 1999). It is not unreasonable to assume that an equivalent protease-sensitive loop exists in the same position in ADC and that this loop may be more accessible in ADC to proteases. Co-expression of two separated N- and C-terminal domains of the Trypanosoma brucei ODC (Osterman et al., 1995b) and expression of a mouse ODC circular permutation (Li and Coffino, 1993) resulted in ODC activity. However, co-expression of the separated N- and C-terminal domains of ADC did not produce ADC activity, suggesting that ADC homodimer formation must take place before processing or that ADC is more sensitive than ODC to changes in structure. Our model for the ADC activity mechanism underlines the similarity of the ODC and ADC mechanisms although ADC is 50% larger than ODC. Recently, Chang et al. (2000) demonstrated carnation ADC activity in vitro and with a maltose-binding protein fusion from E. coli without processing of the pre-protein. It is difficult to interpret the significance of the activity because of the different systems used, but one possible interpretation is that without processing there is some basal ADC activity, as was found in the yeast protease mutant reported here.

It will be interesting to compare the behaviour of the Arabidopsis polyamine pathway with the pathways in plants that possess ODC. Current thinking about the relative roles of ODC and ADC in plant polyamine metabolism cannot be applied to Arabidopsis, and in this context caution will be required in trying to extrapolate results obtained from Arabidopsis biosynthetic mutants to plants in general. However, it is likely that the regulation of spermidine and spermine biosynthesis as distinct from putrescine biosynthesis is similar to that in other plants.

Experimental procedures

Plant material

Transformed hairy root cultures of Arabidopsis thaliana were obtained from J. D. Hamill (Monash University, Melbourne, Australia) and grown at 25°C with gentle shaking as described previously for Datura stramonium hairy root cultures (Michael et al., 1996). Arabidopsis suspension culture was obtained from J. Murray (University of Cambridge) and grown at 25°C in the dark.

Yeast strains and transformation

The wild-type yeast strain YW5-1B (MATα, trp1, ura3-52, leu2-3, 112) was used for ADC expression studies. A multiply-deficient protease mutant SC1216 (MATα, leu2, his, pra1, prb1, prc1, cps1) was obtained from the National Yeast Culture Collection (Norwich, UK). All yeast cultures were grown aerobically at 25°C in minimal SD medium (0.67% yeast nitrogen base without amino acids (Difco Laboratories) and 2% glucose) and supplemented with amino acids as required. Strains were transformed using a modified lithium acetate procedure (Elbe, 1992).

Arabidopsis ADC1 plasmid construction

The Arabidopsis ADC1 (SPE1) cDNA was obtained as an EST (clone 180K7T7; accession number H36915) from the Arabidopsis Biological Resource Center, Ohio State University, USA (courtesy of Doreen Ware). This EST corresponded exactly to the ADC1cDNA in Genbank (accession number U52851) and contains approximately 360 bp of untranslated 5′ leader sequence. The open reading frame of the ArabidopsisADC1was amplified by PCR using Pfu polymerase (Stratagene) from the pBluescript-based EST plasmid using the sense primer 5′-ATA AGA TCT AGC TCT AGC TTT TGT TG-3′ and the antisense primer 5′-ATA GTC GAC CGA AAT AAG ACC AAT TCT CA-3′. The PCR product was then ligated into the EcoRV site of pBluescript and the junction sequences of the PCR product were verified by sequencing. A BglII site replaced the first two amino acids of the ADC1ORF and a SalI site was incorporated after the stop codon (the sites are underlined in the primer sequences). The N-terminus of the ADC1ORF was therefore changed from MPALAFV to MDLALAFV when the modified ADC ORF was digested with BglII and Sal1 and ligated into the compatible BamHI/SalI polylinker sites of the galactose-inducible yeast expression vector pYX243 (R & D Systems) as a translational fusion. The ADC1ORF was also amplified by PCR using Pwo polymerase (Boehringer Mannheim) using the sense primer 5′-GGT ACC ATG GAT CTA GCT CTA GCT TTT-3′ incorporating a Kpn1 site (underlined) and the antisense primer 5′-TCT AGA ACC GAA ATA AGA CCA A-3′ incorporating an Xba1 site (underlined) and ligated into the EcoRV site of pBluescript. Junction sequences of the ADC1ORF insert were confirmed by sequencing, and the modified ADC1cDNA was digested with Kpn1 and Xba1 and inserted into the same sites of the galactose-inducible yeast expression vector pYES2 (Invitrogen) as a 5′-Kpn1, 3′-Xba1 transcriptional fusion. To construct an expression plasmid containing only the N-terminal domain of the ADC1ORF, the 5′-BglII, 3′-Sal1 modified ADC1ORF was used as a PCR template employing sense primer 5′-ATA AGA TCT AGC TCT AGC TTT TGT TG-3′ and the antisense primer 5′-ATA GTC GAC GAA TCG CCT TAA TCA CCC ACT C-3′ incorporating BglII and Sal1 sites, respectively (underlined). The PCR product was ligated into pGEM-T Easy (Promega) and the resulting plasmid was then digested with BglII and Sal1 and ligated into the compatible BamH1/Sal1 sites of pYX243 to create a translational fusion of the ADC1N-terminal domain containing the first 472 amino acids of ADC1with the N-terminal modification described above for the full-length ADC1in pYX243. This N-terminal domain fragment contained up to and including isoleucine 472 plus an additional C-terminal 13 amino acids (GAPGYPYDVPDYA*) containing a haemagglutinin epitope tag. The C-terminal domain of ADC1was PCR-amplified from the full-length cDNA clone using the sense primer 5′-GGT ACC ATG GGT GCA TCG GAT CCG GTT C-3′ and the antisense primer 5′-TCT AGA ACC GAA ATA AGA CCA ATT CTC-3′ incorporating BamH1and Xba1 sites, respectively (underlined), and the PCR product was ligated into pGEM-T Easy. The C-terminal domain was then excised with BamH1and Xba1, and the fragment was ligated into the same sites of the pYES2 polylinker to form a 5′-BamH1, 3′-Xba1 transcriptional fusion containing the 231 C-terminal residues of the ADC1ORF including a new ATG start codon immediately 5′ to the glycine 473 beginning the C-terminal domain N-terminus.

Site-directed mutagenesis

The Arabidopsis ADC1 amino acids corresponding to the mouse ODC lysine 69 (K69) and cysteine 360 (C360) were separately mutated to alanines. Lysine 137 of the Arabidopsis ADC1 was mutated to alanine using the mutagenic primer 5′-ATC TTG ATT ACA TGC CAC AGG ATA CAC ACC-3′ with the Chameleon Double-Stranded, Site-Directed Mutagenesis kit from Stratagene (altered bases are underlined). The K167A mutation was confirmed by sequencing and the full-length K167A mutated form of ADC1 was excised as a 5′BglII, 3′Sal1 fragment and ligated into the BamH1, Sal1 sites of pYX243 as a translational fusion to form the plasmid K167A/pYX243. Cysteine 523 of the Arabidopsis ADC1 corresponding to the mouse C360 was mutated to alanine using the mutagenic primer 5′-TCT TTC CGT CGC TGT CAG CCG TCA AAT CCG-3′ with the Chameleon kit (altered bases underlined). The mutation was confirmed by sequencing and the C523A mutant form of ADC1 was digested with BglII and Sal1 and ligated into the compatible BamH1, Sal1 sites of pYX243 as a translational fusion to form plasmid C523A/pYX243. The K167A/pYX243 plasmid was digested with Sac1 and an approximately 1.7 kbp Sac1 fragment of the ADC1 K167A was used to replace the corresponding wild-type ADC1 Sac1 fragment in the wtADC1/pYES2 plasmid to form K167A/pYES2.

Induction and assay of Arabidopsis ADC1 activity in yeast

Single yeast colonies were picked and used to inoculate 10 ml of SD medium lacking leucine (pYX243) or uracil (pYES2), and grown overnight at 25°C. An aliquot of this culture was used to inoculate 100 ml of the same medium, and the culture was again grown overnight at 25°C. Cells from this culture were centrifuged 1580 g and washed twice with 15 ml sterile distilled water and then resuspended in the same SD medium minus glucose and with 2% galactose. Plasmids were induced by incubating the cells for up to 24 h. After induction, yeast cells were centrifuged 1580 g and resuspended in 1.0 ml of HEPES buffer pH 7.4 (0.1 m HEPES, 10 mm EDTA, 0.5% ascorbic acid, 100 µm pyridoxal 5′ phosphate, 5% polyvinylpolypyrrolidine) in a 2.0 ml screw top tube with 0.5 ml of glass beads (diameter 425–600 µm). Tubes were vortexed for 1 min five times, with cooling on ice between each agitation. Cellular debris was pelleted by centrifugation 1580 g at 4°C, and the supernatant was used for enzyme activity analysis. Activity of ADC was analysed as previously described (Burtin and Michael, 1997) and was measured for 45 min at 37°C followed by 30 min of CO2 absorption.

Assay of ODC, ADC and AdoMetDC activities in Arabidopsis suspension cultures

Enzymes assays were performed as previously described (Michael et al., 1996) on suspension culture cells filtered and dried on sterile paper towels, then frozen in liquid nitrogen and ground to a fine powder with a mortar and pestle.

Acknowledgements

We thank J. D. Hamill for screening the Arabidopsis transformed root cDNA library and J. Hofer for critical reading of the manuscript. This work was supported in part by a German Deutscher Akademischer Austauschdienst (DAAD) fellowship to S.S.

Ancillary