The CaaX specificities of Arabidopsis protein prenyltransferases explain era1 and ggb phenotypes

Background Protein prenylation is a common post-translational modification in metazoans, protozoans, fungi, and plants. This modification, which mediates protein-membrane and protein-protein interactions, is characterized by the covalent attachment of a fifteen-carbon farnesyl or twenty-carbon geranylgeranyl group to the cysteine residue of a carboxyl terminal CaaX motif. In Arabidopsis, era1 mutants lacking protein farnesyltransferase exhibit enlarged meristems, supernumerary floral organs, an enhanced response to abscisic acid (ABA), and drought tolerance. In contrast, ggb mutants lacking protein geranylgeranyltransferase type 1 exhibit subtle changes in ABA and auxin responsiveness, but develop normally. Results We have expressed recombinant Arabidopsis protein farnesyltransferase (PFT) and protein geranylgeranyltransferase type 1 (PGGT1) in E. coli and characterized purified enzymes with respect to kinetic constants and substrate specificities. Our results indicate that, whereas PFT exhibits little specificity for the terminal amino acid of the CaaX motif, PGGT1 exclusively prenylates CaaX proteins with a leucine in the terminal position. Moreover, we found that different substrates exhibit similar Km but different kcat values in the presence of PFT and PGGT1, indicating that substrate specificities are determined primarily by reactivity rather than binding affinity. Conclusions The data presented here potentially explain the relatively strong phenotype of era1 mutants and weak phenotype of ggb mutants. Specifically, the substrate specificities of PFT and PGGT1 suggest that PFT can compensate for loss of PGGT1 in ggb mutants more effectively than PGGT1 can compensate for loss of PFT in era1 mutants. Moreover, our results indicate that PFT and PGGT1 substrate specificities are primarily due to differences in catalysis, rather than differences in substrate binding.


Background
Protein farnesylation is the process by which proteins bearing a carboxyl terminal CaaX motif (C = Cys; a = aliphatic; X = Ser, Cys, Met, Gln, Ala) are post-translationally modified by the covalent attachment of a fifteencarbon farnesyl group [1][2][3][4]. This modification results in the formation of a stable thioether bond between the cysteine of the CaaX motif and the farnesyl moiety, with farnesyl diphosphate serving as the farnesyl donor (Figure 1). This lipidation reaction is catalyzed by protein farnesyltransferase (PFT), which is a cytosolic enzyme consisting of α-and β-subunits [1][2][3][4]. In a similar process, proteins bearing a carboxyl terminal CaaX motif with Leu, Ile, Met, or Phe in the terminal position are modified by the covalent attachment of a twenty-carbon geranylgeranyl group to the cysteine of the CaaX motif. This modification is catalyzed by protein geranylgeranyltransferase type I (PGGT1), which is a cytosolic enzyme consisting of an α-subunit identical to that of PFT and a distinct β-subunit [1][2][3][4][5]. A third enzyme, protein geranylgeranyltransferase type II (PGGT II), also called RAB geranylgeranyltransferase (RAB GGT), catalyzes the geranylgeranylation of RAB proteins bound to the RAB ESCORT PROTEIN (REP). All three enzymes have been found in protozoans, metazoans, fungi, and plants, including peas [6,7], tomato [8,9], and Arabidopsis [10][11][12][13][14].
In Arabidopsis, a single gene encodes the common αsubunit of PFT and PGGT1 (PLURIPETALA, PLP, At3g59380) [14], a second gene encodes the β-subunit of PFT (ENHANCED RESPONSE TO ABA1, ERA1, At5g40280) [11,13], and a third gene encodes the β-subunit of PGGT1 (GERANYLGERANYLTRANSFERASE BETA, GGB, At2g39550) [10,12]. The ERA1 gene was so named because knockout mutations in this gene cause an enhanced response to abscisic acid (ABA) in both seed germination and stomatal closure assays. Consequently, era1 mutants exhibit increased seed dormancy and stomatal closure in response to ABA, and are drought tolerant [11,13,[15][16][17]. These observations suggest that at least one farnesylated protein functions as a negative regulator of ABA signaling. However, to date, a farnesylated negative regulator of ABA signaling has not been definitively identified. era1 plants also exhibit enlarged meristems and supernumerary floral organs, especially petals, and this phenotype is greatly exaggerated in plp mutants lacking the common α-subunit of PFT and PGGT1 [14,[18][19][20][21]. The more severe developmental phenotype of plp mutants compared to era1 mutants suggests that PGGT1 partially compensates for loss of PFT in era1 mutants [14]. Plants with defects in the GGB gene exhibit increased ABA-induced stomatal closure and auxininduced lateral root formation [12], but without significant developmental phenotypes. These observations suggest that at least one geranylgeranylated protein functions as a negative regulator of ABA signaling and at least one functions as a negative regulator of auxin signaling. Indeed, ROP2 and ROP6, which are geranylgeranylated small GTPases [22,23], have been shown to function as negative regulators of ABA signaling, and ROP2 and AUX 2-11 (a geranylgeranylated member of the AUX/ IAA family) have been shown to function as negative regulators of auxin signaling [10,24]. Moreover, Arabidopsis plants possess two genes encoding G protein γ-subunits, both of which are geranylgeranylated, and mutants lacking either of these genes exhibit an enhanced response to auxin-induced lateral root formation [25]. Prenylated proteins have also been implicated in a plethora of other processes, including calcium signal transduction [26,27], response to heat and heavy metal stress [28][29][30], cytokinin biosynthesis [31], and regulation of the cell division cycle [6,32,33]. Given these multiple roles, it is surprising that, unlike other organisms, Arabidopsis plants survive without the shared α-subunit of PFT and PGGT1 [14].
Proteins that are prenylated by either PFT or PGGT1 are further modified. First, the aaX portion of the CaaX motif is proteolytically removed by specific CaaX proteases (AtSTE24, At4g01320 and AtFACE-2, At2g36305 in Arabidopsis) [34][35][36] and, second, the isoprenylcysteine at the newly formed carboxyl terminus is methylated ( Figure 1) [37][38][39][40][41][42][43]. Two distinct isoprenylcysteine methyltransferase (ICMT) enzymes, encoded by the AtSTE14A (At5g23320) and AtSTE14B (ICMT, At5g08335) genes, catalyze the methylation of carboxyl terminal isoprenylcysteines in Arabidopsis [41,[43][44][45][46]. Demethylation of isoprenylcysteine methyl esters is catalyzed by isoprenylcysteine methylesterase (ICME), which is encoded by the ICME gene (At5g15860) [46,47]. As described above, two geranylgeranylated proteins (ROP2 and ROP6) and at least one farnesylated protein negatively regulate ABA signaling in Arabidopsis. However, it is not clear at the present time how these proteins function in ABA signaling. The stomata of ggb plants were found to exhibit an enhanced response to ABA, consistent with the known role of ROP6 in negative regulation of ABA-induced stomatal closure [12,23], but the response of ggb seeds to ABA was normal, despite a report that ROP2 is involved in negative regulation of ABA signaling in seeds [22]. While this may seem like a contradiction, it is possible that PFT activity in ggb plants is sufficient for the prenylation and function of certain prenylated proteins, such as ROP2 (i.e., PFT compensates for loss of PGGT1 in ggb mutants). Indeed, numerous reports exist of prenylated proteins that are substrates of both PFT and PGGT1 and others that are substrates of either PFT or PGGT1 [48,49]. Given this heterogeneity in the specificity of PFT and PGGT1 for certain CaaX proteins, deconvoluting the complex roles of protein prenylation in negative regulation of ABA signaling, meristem development, and other fundamental processes poses a significant challenge. Nevertheless, to address this problem, we characterized Arabidopsis PFT and PGGT1 with respect to substrate specificity and catalysis. These studies were aimed at answering the following questions: 1) What distinguishes plant CaaX prenyltransferases from animal and fungal prenyltransferases and what gives them their unique substrate specificities? 2) Do the substrate specificities of Arabidopsis PFT and PGGT1 potentially explain the phenotypes of era1 and ggb mutants? The results reported here indicate that Arabidopsis PFT exhibits less specificity for the terminal position of the CaaX motif than PFT enzymes from metazoans and yeast and Arabidopsis PGGT1 exhibits greater specificity for CaaX motifs with leucine in the terminal position than PGGT1 enzymes from metazoans and yeast. These results potentially explain the phenotypes of era1 and ggb mutants. Moreover, we show that different CaaX substrates exhibit differences in reactivity rather than differences in affinity in the presence of Arabidopsis PFT and PGGT1.

Results
Recombinant Arabidopsis PFT is more specific for isoprenoid substrates than PGGT1, whereas PGGT1 is more specific for CaaX substrates To functionally characterize Arabidopsis PFT and PGGT1, we co-expressed the PLP and ERA1 coding sequences in E. coli using the pETDuet-1 vector (PLP was expressed with an amino terminal FLAG tag and ERA1 was expressed with an amino terminal 6 × His tag). We also co-expressed the PLP and GGB coding sequences in E. coli (PLP was expressed with an amino terminal FLAG tag and GGB was expressed with an amino terminal 6 × His tag). IPTG-inducible PFT and PGGT1 activities were detected in E. coli extracts and analyzed for substrate specificity using [1-3 H]FPP, [1-3 H]GGPP, and 32 distinct GFP-BD-CaaX protein substrates, which were generated by site-directed mutagenesis of the GFP-BD-CaaX constructs recently reported by Gerber et al. (each protein substrate consists of GFP fused to the carboxyl terminal basic domain of the rice CaM61 protein and one of 32 different CaaX motifs) [50]. As shown in Figures 2 and 3, recombinant Arabidopsis PFT exhibited modest selectively for the terminal amino acid of the Ca 1 a 2 X motif. GFP-BD-CaaX substrates with glutamine, methionine, serine, cysteine, alanine, isoleucine, and even leucine (in descending order) were appreciably farnesylated by Arabidopsis PFT. As previously reported, PFT exhibited low selectivity for the a 1 position of the Ca 1 a 2 X motif, consistent with the observation that the a 1 position is solvent exposed and not constrained by active site amino acids [51][52][53][54][55]. In contrast, PFT exhibited high selectivity for the a 2 position, with charged amino acids (basic as well as acidic) strongly excluded. GFP-BD-CaaX substrates that were efficiently farnesylated were also weakly geranylgeranylated by Arabidopsis PFT, but GFP-BD-CaaX farnesylation was 50-fold greater than geranylgeranylation (i.e., the y-axes in the two graphs of Figure 2 are not the same).
As shown in Figures 4 and 5, recombinant Arabidopsis PGGT1 exhibited high selectivity for the terminal amino acid of the Ca 1 a 2 X motif. Only GFP-BD-CaaX substrates ending in leucine were significantly prenylated by this enzyme (not even CVII, which is a good substrate for mammalian PGGT1, was appreciably prenylated by Arabidopsis PGGT1). As with PFT, PGGT1 exhibited low selectivity for the a 1 position of the Ca 1 a 2 X motif, consistent with the observation that the a 1 position is solvent exposed and not constrained by active site amino acids. On the other hand, PGGT1 exhibited extremely high selectivity for the a 2 position, and only GFP-BD-CaaX substrates with a hydrophobic amino acid at the a 2 position (CVIL and CVFL) were prenylated. GFP-BD-CaaX substrates that were efficiently geranylgeranylated were also farnesylated by PGGT1. Indeed, GFP-BD-CaaX geranylgeranylation was only 4-fold greater than farnesylation in the presence of recombinant Arabidopsis PGGT1.

Purified recombinant PFT is more active than purified recombinant PGGT1
The next step in the characterization of Arabidopsis PFT and PGGT1 was to examine the activity and substrate specificity of purified recombinant enzymes. The enzymes described above were purified by immobilized metal affinity chromatography (IMAC) using Talon ®; Co 2+ -based resin. As shown in Figure 6, IMAC-purified enzymes (80-90% pure) were assayed and found to exhibit the same substrate specificities as described above (PFT exhibits low specificity for the terminal amino acid of the Ca 1 a 2 X motif, whereas PGGT1 is highly selective for GFP-BD-CaaX substrates ending in leucine). Comparing purified recombinant PFT and PGGT1 allowed us to make the following conclusion: purified recombinant PFT is 30-to 100-fold more active than purified recombinant PGGT1. This is not due to errors in the expressed sequences, nor is it likely to be due to differential effects of the FLAG tag on the alpha subunit or the 6 × His-tags on the two β-subunits because the amino termini of prenyltransferase α-and β-subunits are solvent exposed and not involved in the formation or stabilization of active heterodimers [56,57]. Moreover, plant extracts (tobacco BY2 as well as Arabidopsis extracts) consistently exhibit 30-100 fold higher PFT activity compared with PGGT1 activity [12].

Kinetic Analysis of Recombinant Arabidopsis PFT and PGGT1
Purified recombinant Arabidopsis PFT and PGGT1 were subjected to kinetic analyses under Michaelis-Menten conditions (product formation was linear with time and substrate conversion was less than 10%). The results of these experiments were interpreted by Lineweaver-Burk analysis and are shown in Figures 7 and 8. The catalytic constants (k cat /K m ) shown in Table 1 for Arabidopsis PFT confirm the results shown in Figures 2 and 3. Both data sets demonstrate the following substrate preferences for Arabidopsis PFT (normalized to 1.0 for GFP-BD-CVIQ): GFP-BD-CVIQ (1.0), GFP-BD-CVIM (0.60), GFP-BD-CVII (0.21) and GFP-BD-CVIL (0.06). Moreover, the results in Table 1 show that, while different CaaX substrates (CVIQ, CVIM, CVII, and CVIL) have similar K m values, they have markedly different k cat values in the presence of Arabidopsis PFT. Thus, PFT substrate specificities reflect differences in catalytic turnover rate rather than differences in binding affinity. The catalytic constants (k cat /K m ) for Arabidopsis PGGT1, which are shown in Table 2, confirm the results shown in Figures 4 and 5. Moreover, while different CaaX substrates have slightly different K m values, they have dramatically different k cat values. Indeed, the significantly higher k cat value for GFP-BD-CVIL is the primary determinant of substrate specificity for Arabidopsis PGGT1. While these k cat values can be compared, they are nevertheless low, suggesting that only a fraction of the purified PGGT1 protein was catalytically active. Purified recombinant Arabidopsis PFT and PGGT1 were also analyzed with respect to isoprenyl diphosphate specificity. As shown in Table 3, Km values for FPP and GGPP are an order of magnitude lower than Km values for CaaX substrates in the presence of recombinant Arabidopsis PFT and PGGT1. However, as with CaaX substrates, the different specificities of Arabidopsis PFT and PGGT1 for isoprenyl diphosphates cannot be explained by differences in Km. Thus, the preferences of PFT and PGGT1 for different isoprenyl diphosphate substrates is primarily determined by reactivity rather than binding affinity.

Discussion
In this report, it is shown that recombinant Arabidopsis PFT exhibits broad specificity for CaaX substrates with Gln, Met, Ser, Cys, Ala, Ile, or Leu in the terminal 'X' position, whereas PGGT1 exhibits strict specificity for CaaX substrates ending in Leu. Both PFT and PGGT1 exhibit little or no specificity for the a 1 position of the Ca 1 a 2 X motif, which is consistent with previous observations using mammalian prenyltransferases that the a 1 position is solvent exposed and not constrained by active site amino acids [51][52][53][54][55]. In contrast, both prenyltransferases exhibit specificity for the a 2 position of the Ca 1 a 2 X motif. While the mechanism for CaaX specificity remains unknown for Arabidopsis PFT and PGGT1, it is clear that  the substrate specificities of both prenyltransferases reflect differences in catalytic turnover rates rather than differences in K m values (Tables 1 and 2). This finding suggests that, while binding affinities of GFP-BD-CVIQ, GFP-BD-CVIM, GFP-BD-CVII, and GFP-BD-CVIL to the active sites of PFT and PGGT1 are similar, the terminal amino acid of the CaaX motif dramatically affects catalysis. Moreover, the data described above provide an explanation for era1 and ggb phenotypes. PFT prenylates a wide range of CaaX substrates and compensates almost fully for loss of PGGT1 in ggb plants. However, PGGT1 specifically prenylates CaaL substrates and only partially compensates for loss of PFT in era1 plants. These biochemical differences potentially account for the mild phenotype of ggb mutants and the dramatic phenotype of era1 mutants. The different isoprenoid specificities of Arabidopsis PFT and PGGT1 cannot be explained by differences in K m . Indeed, the K m for FPP was only slightly lower than that for GGPP in the presence of recombinant Arabidop-sis PFT, despite the fact that CaaX farnesylation was 50fold greater than geranylgeranylation in the presence of this enzyme (Figures 2 and 3). Moreover, the K m values for FPP and GGPP were almost identical in the presence of recombinant Arabidopsis PGGT1, despite the fact that PGGT1 catalyzed CaaX geranylgeranylation 4-fold more efficiently than farnesylation. Thus, the primary determinant of isoprenoid substrate specificity is reactivity rather than binding affinity.
The results in Figure 6 raise an interesting question. Given the higher specific activity and lower CaaX substrate specificity of PFT, why are CaaX substrates with leucine in the terminal position predominantly geranylgeranylated rather than farnesylated in planta [12,50]? The results in Figure 6 suggest that a CAIL (or CVIL) protein should be predominantly farnesylated in planta because farnesylation of CAIL (or CVIL) by Arabidopsis PFT is approximately 20% as efficient as farnesylation of CAIM (or CVIM), which greatly exceeds the efficiency of CAIL (or CVIL) geranylgeranylation by PGGT1. We pro- pose that PFT is regulated in planta, perhaps by posttranslational modifications or protein-protein interactions, to reduce recognition and farnesylation of CaaX substrates with leucine in the terminal position. Nevertheless, it is likely that CaaX substrates with leucine in the terminal position are, to some extent, farnesylated by PFT and that these aberrantly farnesylated proteins retain full or partial function. This explains why ggb mutants, which were expected to exhibit severe meristem and tip-growth defects due to loss of ROP function, do not exhibit these phenotypes [12,22,23]. We propose that PGGT1 activity is higher in planta than the purified recombinant, E. coli-expressed PGGT1 activity we have characterized. The activity observed with purified recombinant PGGT1 was low, suggesting that only a portion of the purified PGGT1 enzyme was active.

Conclusions
In this report, recombinant Arabidopsis PFT is shown to prenylate CaaX substrates with little specificity for the terminal amino acid. In contrast, recombinant Arabidopsis PGGT1 is shown to exclusively prenylate CaaX substrates with leucine in the terminal position. These different substrate specificities provide a straightforward explanation for the phenotypes of era1 and ggb mutant plants. In addition, substrate specificities for PFT and PGGT1 are shown to reflect differences in catalytic turnover rates rather than differences in substrate binding.

RNA isolation
Total RNA was isolated from wild type Arabidopsis plants (ecotype Col-0) using TRIzol ® Reagent according to the manufacturer's instructions (Invitrogen/Life Technologies Corp., Carlsbad, CA).