Plasmodium Apicoplast Gln-tRNAGln Biosynthesis Utilizes a Unique GatAB Amidotransferase Essential for Erythrocytic Stage Parasites*

Background: Plasmodium apicoplast protein synthesis is essential for parasite survival, yet few of the enzymes involved have been biochemically characterized. Results: Nucleus-encoded apicoplast GatAB glutamyl-tRNA amidotransferase forms Gln-tRNAGln in concert with a non-discriminating glutamyl-tRNA synthetase. Conclusion: Formation of apicoplast Gln-tRNAGln is via indirect aminoacylation. Significance: The apicoplast indirect aminoacylation pathway is a potential drug target. The malaria parasite Plasmodium falciparum apicoplast indirect aminoacylation pathway utilizes a non-discriminating glutamyl-tRNA synthetase to synthesize Glu-tRNAGln and a glutaminyl-tRNA amidotransferase to convert Glu-tRNAGln to Gln-tRNAGln. Here, we show that Plasmodium falciparum and other apicomplexans possess a unique heterodimeric glutamyl-tRNA amidotransferase consisting of GatA and GatB subunits (GatAB). We localized the P. falciparum GatA and GatB subunits to the apicoplast in blood stage parasites and demonstrated that recombinant GatAB converts Glu-tRNAGln to Gln-tRNAGln in vitro. We demonstrate that the apicoplast GatAB-catalyzed reaction is essential to the parasite blood stages because we could not delete the Plasmodium berghei gene encoding GatA in blood stage parasites in vivo. A phylogenetic analysis placed the split between Plasmodium GatB, archaeal GatE, and bacterial GatB prior to the phylogenetic divide between bacteria and archaea. Moreover, Plasmodium GatA also appears to have emerged prior to the bacterial-archaeal phylogenetic divide. Thus, although GatAB is found in Plasmodium, it emerged prior to the phylogenetic separation of archaea and bacteria.

The human malaria parasite Plasmodium falciparum is responsible for 124 -283 million cases of malaria and an estimated 0.6 million deaths every year (1). It contains a relict plastid, the remnant of an ancient secondary endosymbiotic event in which the eukaryotic progenitor of the malaria parasite engulfed a photosynthetic eukaryote known as the apicoplast (2). The Plasmodium apicoplast possesses a 35-kb circular genome with 60 genes (3) that encode components of the apicoplast transcriptional and translational apparatus such as RNA polymerase subunits, the elongation factor EF-Tu, several ribosomal proteins, rRNAs, and tRNAs (4 -8), as well as the SufB protein thought to play a role in FeS cluster formation (9). Most apicoplast proteins, however, are encoded by the nuclear genome and are imported into the organelle post-translationally (10). Over 500 apicoplast-targeted proteins were identified in P. falciparum (11,12), revealing apicoplast biosynthetic pathways for fatty acids (13,14), isoprenoid precursors (15), and heme (16), as well as enzymes for tRNA modification (12) and lipoylation (17). Several of these pathways exhibit prokaryote-like features and are potential drug targets (12,15,18). Recent studies have shown that apicoplast isoprenoid precursor biosynthesis is essential in P. falciparum asexual stages (19), indicating that the pathway cannot be bypassed by salvaging lipids from the host and may be a good drug target in asexual stages. The type II fatty acid and heme biosynthetic pathways, however, are not essential in the asexual stages (18), and although not good targets for asexual stage chemotherapy, they may prove to be valuable drug targets in liver stages (2).
Translational accuracy is required to properly decipher the genetic code during protein synthesis. The fidelity of protein synthesis largely depends on the formation of correct aminoacyl-tRNAs by aminoacyl-tRNA synthetases (aaRSs). 2 In the classic model, each species of aaRS strictly discriminates one amino acid from among the 20 canonical amino acids, as well as

Experimental Procedures
Bioinformatics-A list of putative apicoplast-targeted proteins conserved in P. falciparum, Theileria parva, and Toxoplasma gondii was obtained from the supplementary informa-tion in Gardner et al. (24). Nucleotide or amino acid sequences of Plasmodium genes or proteins (11,25) and those from other species were obtained from PlasmoDB (26), the Wellcome Trust Sanger Institute GeneDB website, or UniProt (27).
Homology Modeling-The structure of Plasmodium GatA and GatB was modeled according to the crystal structures of Staphylococcus aureus GatCAB (2g5h and 3ip4) (28,29) using the Swiss-Model automated comparative protein-modeling server (30). In the modeled range, the sequence identity of PfGatA and 2g5h was 32.2%, and the sequence identity of PfGatB and 3ip4 was 24.8%. No suitable structural template could be found for amino acid regions 1-180, 425-539 and 708 -744 in PfGatA and amino acid regions 1-350 and 461-530 in PfGatB. The spatial arrangement of PfGatA and PfGatB was determined based on the crystal structures of S. aureus Gat-CAB. Attempts to model GatA and GatB according to the crystal structure of Pyrococcus abyssi archaeal GatDE (1zq1) (31) resulted in unusable models. All figures were prepared with Chimera (32).
Phylogenetic Analysis-The phylogenetic trees were constructed using the tools provided on line (33). The "one-click" mode was used, employing ClustalW (34) for sequence alignment, and Gblocks (35,36), PhyML (36), and TreeDyn (37) for curation of the multiple sequence alignment, tree construction, and rendering, respectively. The final tree was constructed using 100 bootstraps.
P. falciparum Cell Culture, Plasmid Constructs, and Parasite Transfection-P. falciparum clone 3D7 parasites (MRA-102, MR4, BEI Resources, Manassas, VA) were grown in human O ϩ red blood cells at 4% hematocrit in RPMI 1640 medium supplemented with Albumax (Life Technologies, Inc.) to a final concentration of 0.5% and gassed with 5% CO 2 and 0.5% O 2 in N 2 at 37°C as described previously (38). To generate an episomal transfection vector, the bipartite apicoplast targeting sequences of PfGatA (amino acids 1-102) and PfGatB (amino acids 1-98) were amplified from P. falciparum 3D7 genomic DNA and cloned into the BglII and AvrII sites of a pCHD-GFP vector (39). Synchronized P. falciparum 3D7 ring-stage parasites were transfected with 100 g of purified plasmid DNA (plasmid maxi kit, Qiagen) via electroporation (40,41), and the transfected parasites were selected using 2.5 nM WR99210. The transfection experiments were performed in duplicate and repeated once.
Subcellular Localization-Subcellular localization of cloned P. falciparum transfectants expressing GFP-tagged PfGatA, PfGatB, or Myc-tagged PbGatA was performed as outlined in Ref. 42. Double staining was performed using a rabbit polyclonal anti-acyl carrier protein (ACP) primary antibody (diluted 1:500) (10) as an apicoplast marker and a mouse monoclonal anti-Myc antibody (Santa Cruz Biotechnology, diluted 1:500) to detect PbGluRS-Myc. Fluorescent staining was achieved using Alexa Fluor-conjugated secondary antibodies (Invitrogen) specific to rabbit (Alexa Fluor 594, red) or mouse (Alexa Fluor 488, green) IgGs. DAPI was used to stain nucleic acids, and the mitochondrion was stained by incubating the parasites in culture media for 30 min with 20 nM MitoTracker Red (Invitrogen) and fixing the cells as outlined above. Images were acquired using an Olympus Delta Vision imaging system (Applied Precision) with a ϫ100 objective and deconvolved using the SoftWoRx package (Applied Precision) with the default parameters.
Myc Tagging and Attempted Deletion of the Endogenous Plasmodium berghei GatA Gene-A 4ϫ myc tag was appended to the 3Ј end of the gene encoding the putative apicoplast-targeted P. berghei GatA (PlasmoDB ID PBANKA_071810) as described previously (18). A 3.5-kb fragment of the 3Ј end of the gene without the stop codon was amplified from P. berghei ANKA genomic DNA using primers PbGatA_F2 (TACCGCGGATA-ATATACAACCAATAACATTATAG) and PbGatA_R (ATA-CTAGTACTAGCCTTATTTTCCAAATTGTGAAC); the SacII and SpeI restriction sites are underlined. Polymerase chain reactions (PCR) (50 l) contained 50 ng of genomic DNA, 0.1 M of each primer, 5 l of buffer, and 1 l of Advantage 2 polymerase (Clontech). The PCR product was digested with SacII and SpeI and cloned into the b3D myc vector (18). P. berghei ANKA parasites were transfected, and parasites with myc insertions in the gatA gene were selected and cloned as described (18). Plasmid integration at the 5Ј-and 3Ј-insertion sites was confirmed via PCR using primer pairs PbGatA_For2 (TACCGCGGATAATATACAACCAATAACATTATAG) and PbGatA_int5_Rev2 (GAGACAGCTCAATTCTTTATG-TCC) for the 5Ј integration test and PbGatA_int3_For2 (CCT-CTTCGCTATTACGCCAGCT) and PbGatA_Rev4 (GAACC-ACCAGATGACCCACCACATG) for the 3Ј integration test. The strategy described previously (42) was used to attempt deletion of the PbGatA gene via double-crossover recombination. The primers used to amplify the genomic regions were as follows: PbGatA_KO.Pr1For (GCCCGCGGGCATGAGTTGTTAAAA-GTTGCC) and PbGatA_KO.Pr2Rev (AGTTCTACTGGGCCC-AAATTTAAGCATACAGAAAGTGAC); PbGatA_KO.Pr3For (CTTAAATTTGGGCCCAGTAGAACTAGAACATGAGGG) and PbGluRS_KO.Pr4Rev (GCCCGCGGTTTGTCCTTACAA-CTTTCTTACC). Primers 1 and 2 were designed to amplify a 942-bp fragment containing the last 102 bp of the PbGatA coding sequence and 840 bp of the 3Ј-UTR. Primers 3 and 4 were designed to amplify a 734-nucleotide fragment containing the first 62 bp of the PbGluRS coding sequence and 672 bp of the 5Ј-UTR. Primer 1 contained an added 5Ј-terminal GC dinucleotide and a SacII site. Primer 4 contained an added 5Ј-terminal GC dinucleotide and a SacII site. Primers 2 and 3 contained an ApaI site flanked by complementary sequences (underlined) for recombinatorial PCR. The two genomic fragments were first amplified in separate reactions using the cycling parameters described above, and the resulting products were combined in a second PCR to form a single product, which was digested with SacII cloned into the B3D KO Red vector. This construct was linearized with ApaI, and P. berghei ANKA parasites were transfected as described (43). The transfection experiments were performed in duplicate and repeated once.
Preparation and Aminoacylation of tRNA Substrates-Synthetic genes encoding P. falciparum apicoplast tRNA Glu and tRNA Gln (both from GenBank TM accession number X95276) were expressed and purified from E. coli and 32 P-labeled on their 3Ј-OH termini using the E. coli CCA-adding enzyme as described (23). The aminoacylation assay using recombinant PfGluRS was performed and quantified as described (23). Glutamylated tRNAs were phenol (Tris-buffered, pH 7.9)/chloroform-extracted, and unincorporated [␣-32 P]ATP was removed using Bio-Spin 30 columns (Bio-Rad).
Amidotransferase Assay-Transamidation assays were carried out in 1ϫ AdT buffer (100 mM Hepes⅐KOH, pH 7.2, 30 mM KCl, 12 mM MgCl 2 , and 5 mM DTT) with 2.6 mM L-glutamine, 4 mM ATP, 500 nM 32 P-labeled Glu-tRNA Gln , and 50 nM each of GatA and GatB. Reactions were carried out at 37°C for 5 min. Aliquots (4 l) were quenched on ice with 4 l of 100 mM sodium citrate, pH 4.74, and 0.66 mg/ml of nuclease P1 (Sigma) and incubated at room temperature for 35 min. To separate glutaminyl-AMP (Gln-AMP) from Glu-AMP and AMP, 2.0 l of the digested samples were separated on 20 ϫ 20-cm PEI cellulose TLC plates. The plates were then developed in 100 mM ammonium acetate, 5% acetic acid and air-dried. Spot positions and intensities were measured by phosphorimaging, as described (44). To test whether PfGluRS and PfGluAdT together could sequentially form Gln-tRNA Gln from precursors in a single reaction, we initiated some reactions by adding premixed PfGluRS, PfGatA, and GatB (50 nM each) in the presence of 2.6 mM L-Glu and L-Gln at 37°C for 5 min and quantified as described above. To test whether PfGatAB could use an alternative amide donor, 2.6 mM L-asparagine was used instead of L-glutamine.

Subunit Composition of the Apicoplast Glu-tRNA Gln
Amidotransferase-The formation of aminoacyl-tRNAs is a crucial step in protein synthesis. Despite the central importance of this process in all living organisms, it has been unclear how Plasmodium synthesizes Gln-tRNA Gln in the apicoplast. We recently reported that Plasmodium apicoplast glutamyl-tRNA synthetase is a non-discriminating enzyme that forms both Glu-tRNA Glu and Glu-tRNA Gln and is essential in erythrocytic stages of the parasite life cycle (23). Amidation of Glu-tRNA Gln to Gln-tRNA Gln requires a tRNA-dependent amidotransferase (AdT). The P. falciparum genome contains two single-exon genes that encode putative orthologs of the GatA and GatB subunits of the bacterial glutamyl-tRNA amidotransferase (Glu-AdT) as follows: GatA (PF3D7_0416100, 96 kDa, 826 amino acids) and GatB (PF3D7_0628800, 102 kDa, 882 amino acids) ( Table 1). Both P. falciparum proteins possess predicted N-terminal bipartite apicoplast targeting sequences, suggesting that they are the subunits of an amidotransferase that participates in an apicoplast indirect aminoacylation pathway. Orthologs of PfGatA and PfGatB are conserved in apicomplexans that possess an apicoplast (e.g. Theileria (24, 45)) but not in apicomplexans that lack an apicoplast (e.g. Cryptospo-ridium). Although bacterial Glu-AdTs have the subunit composition GatCAB (28), we did not find a Plasmodium gene that encoded a GatC homolog nor did we find a GatF ortholog of the GatFAB yeast mitochondrial Glu-AdT (46,47). In archaea, GatDE is used for Gln-tRNA Gln formation. It is notable that the cradle domain in archaeal GatE and bacterial GatB share a similar fold, but the overall architectures of archaeal GatD and bacterial GatA are completely different.
PfGatB and PfGatA Possess Broadly Conserved Elements and Unique Inserts of Unknown Function-To predict the tertiary structure and identify conserved and divergent features of PfGatAB, homology modeling was performed using the S. aureus GatCAB crystal structure as a template (28,29).
Bacterial GatCAB includes three subunits as follows: the glutaminase GatA; the transamidase GatB; and GatC, which is a small protein (12 kDa) that appears to perform the role of a structural stabilizer at the interface between the GatA and GatB subunits (28,29). GatB contains a transamidase domain called the "cradle" and a helical domain that binds to tRNA.
Plasmodium GatB ( Fig. 1) includes an N-terminal apicoplast targeting sequence (residues 1-183), two unique inserts (residues 184 -362 and 461-524) of unknown functions, and two domains typical for bacterial GatB, the cradle domain (residues 350 -460 and 526 -700), and a helical domain (residues 704 -880). Instead of GatB, archaea use GatE, which contains an AspRS-like insertion in its cradle domain (31). No similarity could be found between the Plasmodium and archaeal insertions. Furthermore, we could not find a structural template for the Plasmodium insertions, which is the reason why they were not modeled. Most of the amino acids that are important for substrate recognition, Mg 2ϩ /Mn 2ϩ coordination, and the ammonia channel are either conserved in PfGatB, bacterial GatB (28,29), and archaeal GatE (31) proteins or replaced by conservative substitutions (Fig. 1).
In S. aureus GatB (28,29), the conserved residues His 12 , Glu 124 , and Glu 150 and three water molecules coordinate the Mg 2ϩ ion; the equivalent residues in PfGatB are Glu 451 and Glu 539 . The corresponding residue to His 12 of S. aureus is probably situated in the histidine-rich unmodeled N-terminal region of PfGatB. Most likely residues Asp 579 and Glu 595 are part of the second transient binding site of Mg 2ϩ /Mn 2ϩ ions; the corresponding residues in S. aureus GatB are Glu 10 , Asp 192 , and Glu 210 (28,29). Furthermore all residues forming the ADPbinding site are conserved or functionally replaced in Plasmodium. In P. falciparum, the phosphate moiety could interact with the conserved residues Asn 581 , Ser 583 , and Arg 593 , which is also an arginine in archaeal GatE (Arg 239 , P. abyssi) (31) but a lysine in bacterial GatB (Lys 208 , S. aureus) (28,29). The hydrophobic adenosine binding pocket is less conserved between the species; in S. aureus it is formed by Val 152 , Pro 155 , and Phe 205 (28,29) and in P. abyssi by Ser 191 , Pro 194 , and Gly 236 (31). The corresponding residues in Plasmodium are Val 541 and two charged residues Lys 544 and Lys 590 . However, the lysine residues interact with their hydrophobic side chains as well as with the adenosine, so that the plasmodial adenosine binding pocket is also hydrophobic. Inspection of the different GatB structures revealed that the C-terminal region in PfGatB (amino acids 805-880, Fig. 1) is highly mobile and could be involved in tRNA recognition, as seen in bacterial GatBs (28,29).
Bacterial GatA consists of a single domain (28,29), which is homologous to other amidases in its catalytic core (22,28,29,48), whereas archaeal GatD consists of three domains as follows: an N-terminal domain and two AnsA-like domains (31). The predicted PfGatA ( Fig. 1) contains an apicoplast targeting sequence (amino acids 1-110) and two inserts of unknown function, which are 115 and 37 amino acids long and found only in Plasmodium (residues 425-539 and 708 -744, respectively). The active site of bacterial GatA (28, 29) includes a conserved Arg and a conserved Asp residue, which interact with the carboxyl and amide groups of bound glutamine. The corresponding residues in PfGatA are Arg 647 and Glu 755 . Furthermore the Ser-cis-Ser-Lys catalytic scissors (Ser 322 , Ser 346 ,and Lys 251 ) that are involved in glutaminase-and glutamine-dependent transamidase activity in bacterial GatA (28,29,49) and the oxy- anion hole (28,29) that stabilizes the tetrahedral covalent intermediate (Thr 343 , Gly 344 , Gly 345 , and Ser 346 ) are conserved in PfGatA (Fig. 1). In contrast to PfGatB, which has only 21% sequence identity to a potential human analog, PfGatA exhibited 40% identity to the human homolog, primarily due to the conservation of residues in the vicinity of the glutaminase domain ( Fig. 1). Nakamura et al. (28) propose that a hydrophilic tunnel channels NH 3 from the glutaminase site in GatA to the transamidase site of GatB. Almost all residues that line this tunnel in S. aureus GatCAB (21 residues) (28) are strictly conserved in Plasmodium (17 residues), including a conserved Thr (343 PfGatA and 175 S. aureus, GatA ) at the entrance and a conserved Lys (410 PfGatB and 79 S. aureus, GatB ) at the tunnel exit.
GatC, a small protein, is the third subunit in bacterial Glu-AdTs (28,29) and performs a structural stabilizer role at the interface between the GatA and GatB subunits. A homolog of the gatC gene could not be identified in the P. falciparum genome, but components of the insertions found in GatA (residues 425-539 and 708 -744) and in GatB (residues 1-350 and 461-525) could perform the task of bacterial GatC. This hypothesis is supported by the presence of two conserved residues (Asn 52, S. aureus and Arg 268, S. aureus ), which form hydrogen bonds between the bacterial GatB and GatC subunits are also present in PfGatB (Asn 383, PfGatB and Asn 669, PfGatB ) (Fig. 1).
Phylogenetic Analysis of PfGatB and PfGatA-To determine the evolutionary origin of plasmodial GatB and GatA proteins, we constructed phylogenetic trees with bacterial GatB and archaeal GatE proteins and with bacterial GatA and archaeal GatD proteins, respectively. If P. falciparum GatB or GatA evolved from either the bacterial GatB/GatA or archaeal GatE/ GatD lineages, one would expect P. falciparum GatB/GatA to appear in either the GatB/GatA or GatE/GatD clades, much like GlnRS enzymes grouping with the eukaryotic glutamyl-tRNA synthetases (GluRS) (50,51). Instead, our phylogenetic trees had three branches representing distinct groupings, one for the P. falciparum GatB or GatA subunits, a second containing the bacterial GatB/GatA subunits, and a third for the archaeal GatE/GatD subunits. Suspecting an artifact because of long-branch attraction (52), we reanalyzed our dataset by including GatA and GatB sequences from other Plasmodium species, the cyanobacterium Anabaena variabilis, the vascular plant Arabidopsis thaliana, and the chromerids Chromera velia and Vitrella brassicaformis. This phylogenetic tree also had three branches representing distinct groupings, one for the apicomplexan/Plasmodium GatB or GatA, a second of the bacterial and green plastid GatB/GatA, and a third of the archaeal GatE/GatD subunits (Fig. 2, A and B) with the newly added sequences of chromerids (C. velia and V. brassicaformis) and plants (A. variabilis and A. thaliana) nesting deeply among the bacterial sequences, away from the Plasmodium branch. The Plasmodium branch of the tree was also consistent with the following previously established phylogenetic relationships within Plasmodium spp.: the monophyly of rodent parasites FIGURE 2. Phylogenetic analysis of the Plasmodium apicoplast GatB and GatA proteins. A, maximum likelihood phylogenetic relationship between all Plasmodium GatB sequences and GatB and GatE sequences from bacteria and archaea, respectively. If Plasmodium GatB evolved from GatB or GatE, one would expect the phylogenetic pattern to show these enzymes to be specifically related to a particular GatB or GatE clade; instead, the resulting phylogenetic trees had three branches representing three distinct subfamilies with Plasmodium GatB being highly distinct and separated from the other plastid GatB, bacterial GatB and archaeal GatE clades. Scale bar, 0.5 changes/site B, maximum likelihood phylogenetic tree of Plasmodium GatA sequences and GatA and GatD sequences from bacteria and archaea. The resulting phylogenetic tree also had three branches representing three distinct subfamilies with Plasmodium GatA placed on a separate branch away from bacterial GatA, archaeal GatD, or the plastid GatA clades. Scale bar, 0.9 changes/site. (53); the close affinity of Plasmodium vivax with Plasmodium cynomolgi and Plasmodium knowlesi, each with bootstrap values Ͼ ϭ 0.79; and the separate clustering of P. falciparum from the clades containing the rodent plasmodial parasites or the clade that contains P. vivax (53,54). Furthermore, the GatA and GatB sequences from two other Apicomplexa, the piroplasms Theileria (24,45) and Babesia (55), which like Plasmodium lack gatC or gatF in their nuclear genomes, were placed in the same clade as the Plasmodium spp. (Fig. 2, A and B), indicating that the placement of Plasmodium GatA and GatB proteins on a separate branch was unlikely to be an artifact.
These results indicate that Plasmodium GatA and GatB belong to subfamilies that are distinctly different from those of known GatA, GatB, GatD, or GatE subunits. These findings imply that Plasmodium GatA and GatB co-evolved and that Plasmodium GatAB is a paralog of GatCAB, GatFAB, and GatDE. Subcellular Localization of PfGatA and PfGatB by Tagging the Bipartite Apicoplast Targeting Sequence-The P. falciparum GatA (PF3D7_0416100) and GatB (PF3D7_0628800) contain a predicted apicoplast bipartite targeting sequence (11,25,56), but their subcellular localization has never been established experimentally. To determine whether the enzymes are targeted to the apicoplast, P. falciparum parasites were transfected with episomal constructs in which the predicted apicoplast targeting sequences of either PfGatA and PfGatB were fused to the N terminus of green fluorescent protein (GFP) and expressed in transfected parasites under the control of the P. falciparum calmodulin promoter (39). Fixed blood stage parasites expressing PfGatA-GFP or PfGatB-GFP were stained with anti-GFP and anti-ACP antibodies (14) to detect the fusion proteins and mark the apicoplast, respectively. Deconvolution fluorescence microscopic examination revealed distinct subcellular GFP localization for all constructs. As with other cell lines expressing GFP in the apicoplast (39), PfGatA-GFP and PfGatB-GFP were immunolocalized within a characteristically small and round compartment in ring stage parasites (Fig. 3, A, panel i, and B, panel i), which then elongated and developed into a complex branched form at the trophozoite stage (Fig. 3, A, panel ii, and B, panel ii) prior to splitting into numerous individual structures in schizonts, one for each daughter merozoite. Furthermore, the anti-GFP and anti-ACP signals were colocalized, confirming that the GatA-and GatB-GFP fusion proteins were present in the apicoplast.
To investigate potential dual localization to the mitochondrion, the same transgenic PfGatA-and PfGatB-GFP P. falciparum clones were incubated with MitoTracker Red and then fixed, stained with anti-GFP monoclonal antibody, and examined via deconvolution fluorescence microscopy. In ring stages transfected with PfGatA-GFP, the anti-GFP antibody and MitoTracker Red marked closely apposed organelles that were clearly distinct (Fig. 3C, panel i) and which began to enlarge in early trophozoites (Fig. 3C, panel ii). Little co-localization between the mitochondrion and PfGatB-GFP was observed in late trophozoites (Fig. 3D, panel i) or late schizont stages with FIGURE 3. PfGatA and PfGatB localizes to the apicoplast. Transgenic PfGatA-GFP and PfGatB-GFP parasites were generated in which an episomal construct contained the bipartite apicoplast targeting sequences from PfGatA and PfGatB were cloned in front of a GFP, and the expression was controlled by the P. falciparum calmodulin 5Ј promoter (35). The transfection experiments were performed in duplicate and repeated once. Differential interference contrast and fluorescent images were captured and processed using deconvolution microscopy; a merge of the images is presented on the far right column (overlay). A and B, PfGatA-GFP and PfGatB-GFP apicoplast localization was monitored via immunofluorescence assay using an anti-GFP antibody (green), and the apicoplast was detected by staining it with anti-ACP antibody (red). Nucleic acid was stained with DAPI (blue). PfGatA-GFP and PfGatB-GFP form characteristically small and round compartments early in the infection cycle (A, panel i, and B, panel i), which then elongate and develop into complex and multiply branched forms at the trophozoite stage prior to splitting into individual spots, one for each merozoite, in the schizont stage (A, panel ii, and B, panel ii). PfGatA-GFP and PfGatB-GFP co-localized with ACP (␣-GFP/␣-ACP overlay), confirming localization to the plastid. C and D, PfGatA-GFP and PfGatB-GFP do not localize to the mitochondrion in erythrocytic stages. The mitochondrion was labeled using MitoTracker Red, whereas PfGatA-GFP and PfGatB-GFP were detected via immunofluorescence assay using ␣-GFP antibody (green). Nucleic acid was stained with DAPI. Panels Ci and Cii show rings and early trophozoites, respectively, expressing PfGatA-GFP. In rings (C, panel i), the mitochondrion and apicoplast are clearly distinct organelles that enlarge and elongate during trophozoite development (C, panel ii). Panels Di and Dii show early and late schizonts, respectively. Nuclear division is underway in early schizonts (panel Di), and the apicoplast and mitochondrion are beginning to elongate. In late schizogony (panel Dii), daughter merozoites have formed, and the apicoplast (green) and the mitochondrion (red) have segregated to daughter merozoites. the mitochondrial staining largely distinct from that of PfGatB-GFP (Fig. 3D, panel ii).
Subcellular Localization in P. berghei by Myc Tagging the Endogenous GatA Gene-To confirm apicoplast localization that was determined via transfection of the episomal constructs in P. falciparum, we tagged the endogenous gene encoding the P. berghei ortholog of PfGatA (PBANKA_071810) with a quadruple Myc tag (Fig. 4). We tagged the endogenous PbGatA coding sequence to reduce the possibility that the fusion protein would be mistargeted due to inappropriate timing or intensity of expression. Fixed blood stage parasites were stained with anti-Myc antibody to detect PbGatA and ACP antisera (14) to detect the apicoplast, and the samples were observed using Deltavision deconvolution fluorescence microscopy. Structures containing the Myc-tagged PbGatA (PbGatA-myc) exhibited a typical apicoplast appearance (Fig. 4C) similar to that observed using the episosomal PfGatA-GFP and PfGatB-GFP constructs in P. falciparum in vitro.
Attempted Deletion of PbGatA Gene-Bioinformatic analyses strongly suggest that Plasmodium lacks an apicoplast-targeted GlnRS (57,58), implying that indirect aminoacylation is probably the sole route for Gln-tRNA Gln formation in the apicoplast and that GatA and GatB are essential components of the protein biosynthetic pathway in Plasmodium. To test whether PbGatA was required for blood stage growth, we transfected P. berghei parasites with a construct (Fig. 5) designed to delete the endogenous PbGatA gene via double-crossover recombination. As a control, we transfected parasites from the same batch with the construct used previously to generate the PbGatA-myc transgenic parasites (Fig. 4). In three independent experiments, the PbGatA genomic locus was refractory to gene deletion via double-crossover recombination. Transgenic PbGatA-myc parasites, however, were readily obtained (data not shown). These two experiments indicate that although the PbGatA locus is accessible to recombination, the PbGatA gene is refractory to deletion, strongly suggesting that the apicoplast-targeted GatA is essential in blood stage parasites.
PfGatAB Is a Glutamyl-tRNA Amidotransferase-To show biochemically that PfGatAB encodes the apicoplast glutamyl-tRNA amidotransferase, we independently expressed PfGatA and PfGatB in E. coli. A two-step purification procedure combining Ni-NTA affinity and size exclusion chromatography allowed purification of 10 mg of PfGatA and PfGatB per liter of culture. SDS-PAGE analysis of the purified enzymes corroborated the predicted molecular masses of the two open reading frames, GatA and GatB (85.0 and 94.5 kDa, respectively, Fig. 6,  A and B), indicating that both P. falciparum subunits could be expressed and purified independently. This was also the case with Helicobacter pylori GatCAB (22), but not with Bacillus subtilis GatCAB, in which the GatA subunit could not be expressed in E. coli in the absence of GatB (59).
Before assaying PfGatAB for amidotransferase activity, we first performed control reactions to verify that it would not glutamylate 32 P-labeled apicoplast tRNA Glu or tRNA Gln . The control reactions showed that PfGatAB did not exhibit any glutamylation activity (Fig. 6C, lanes 1 and 2). We then tested  (18). Following linearization of the construct with PacI, it was transfected into P. berghei blood stage schizonts that were subsequently injected into mice. The vector contains a mutated T. gondii DHFR/TS gene as a pyrimethamine selectable marker (TgDHFR). Transgenic parasites were selected via pyrimethamine treatment and cloned by limiting dilution. B, PCR analysis of the integration site in a cloned PbGatA-myc transgenic (tr) parasite. Ethidium bromidestained agarose gels showing the integration of the PbGatA-myc vector into the P. berghei genome. Only PbGatA-myc is positive in the 3Ј and 5Ј integration (int) tests, whereas both PbGatA-myc and wild type (wt) parasite genomic DNAs are positive for the PbGatA open reading frame test (ORF test). M indicates the size ladder. C, PbGatA-Myc apicoplast localization was monitored via immunofluorescence assay using an anti-Myc antibody (green), and the apicoplast was detected by staining with anti-ACP antibody (red). Nucleic acid was stained with DAPI (blue). PbGatA-Myc forms a characteristically small and round compartment early in the infection cycle (top row), which then elongates and develops into a complex and multiply branched form at the trophozoite stage prior to splitting into individual spots, one for each merozoite, in the schizont stage (bottom row). PfGatA-Myc is co-localized with ACP (␣-Myc/␣-ACP overlay), confirming localization to the plastid. D, PbGatA-Myc did not localize to the mitochondrion in erythrocytic stages. The mitochondrion was labeled using MitoTracker Red, and PbGatA-Myc was detected via immunofluorescence assay using an anti-Myc antibody (green). Nucleic acid was stained with DAPI. In the early stages of parasite development (top row), PbGatA-Myc and mitochondria are discrete single organelles. In the late trophozoite stages, the mitochondria and PbGatA-Myc are heavily branched but mostly distinct with a few points of overlapping signal (bottom row). Scale bar 1 m. DECEMBER 4, 2015 • VOLUME 290 • NUMBER 49 FIGURE 5. Attempted knock-out of the PbGatA gene via double-crossover recombination. Two fragments containing 5Ј-and 3Ј-UTRs of the PbGluRS gene, with Ϸ100 nucleotides of the start and end, respectively, of the PbGatA coding sequence were cloned into the B3D KO Red vector (38). After linearization with ApaI, the construct was transfected into P. berghei ANKA parasites (39). In three independent experiments, we were unable to delete the endogenous PbGatA gene by integrating the deletion construct via double-crossover recombination.  1 and 2 contained PfGatA, PfGatB, and either 32 P-labeled apicoplast tRNA Gln or tRNA Glu substrates in the presence of L-glutamate. No Glu-tRNA Gln or Glu-tRNA Glu was observed, implying that PfGatAB does not possess aminoacylation activity. Lanes 3 and 4 contained PfGatA, PfGatB, and either 32 P-labeled Glu-tRNA Glu or Glu-tRNA Gln in the presence of L-glutamate. No reaction was observed, implying that PfGatAB does not utilize L-Glu as a substrate. Lanes 5 and 6 contained PfGatA, PfGatB, and either 32 P-labeled Glu-tRNA Glu or Glu-tRNA Gln in the presence of L-glutamine. PfGatAB converted Glu-tRNA Gln to Gln-tRNA Gln but did not utilize Glu-tRNA Glu as a substrate, indicating that PfGatAB has a high specificity for Gln-tRNA Gln . D, representative phosphorimage of the effect of different amide donors on amidation activity. Lane 1 contained 50 nM each of PfGluRS, PfGatA, PfGatB, and 32 P-labeled apicoplast tRNA Gln . Lane 2 contained 50 nM each of PfGluRS, PfGatA, PfGatB, 32 P-labeled apicoplast tRNA Gln , and L-glutamate. Lanes 3 and 4 contained 50 nM each of PfGluRS, PfGatA, PfGatB, 32 P-labeled apicoplast tRNA Gln , and L-glutamate, plus either L-glutamine or L-asparagine as amide donors. PfGatAB utilizes both L-glutamine and L-asparagine as amide donors.

P. falciparum Apicoplast Glutamyl-tRNA Amidotransferase
PfGatAB for transamidation activity in a reaction containing L-glutamate as the amide donor and either P. falciparum 32 Plabeled Glu-tRNA Gln or Glu-tRNA Glu that had been produced using recombinant PfGluRS (23). PfGatAB transamidated neither of these glutamylated tRNA substrates (Fig. 6C, lanes 3 and  4), implying that PfGatAB does not utilize L-glutamate as an amide donor. However, PfGatAB transamidated Glu-tRNA Gln to form Gln-tRNA Gln in the presence of L-glutamine (Fig. 6C, lane 6) but as expected did not utilize Glu-tRNA Glu as an amide acceptor (Fig. 6C, lane 5). These data show that PfGatAB has a high specificity for Glu-tRNA Gln and utilizes L-glutamine as the amide donor in the transamidation reaction.
Next, we explored the possibility that PfGluRS and PfGlu-AdT could act together in a single reaction to sequentially form Gln-tRNA Gln from precursors. We performed an experiment that allowed the transamidation reaction to take place immediately after aminoacylation by premixing equimolar amounts of PfGluRS, PfGatA, and PfGatB for 3 min on ice and then incubating them at 37°C with all substrates required for the glutamylation and amidation reactions. Formation of Glu-tRNA-Gln and its conversion to Gln-tRNA Gln were observed (Fig. 6D,  lane 3). Similar results were obtained when we used L-asparagine as the amide donor for the transamidation reaction (Fig.  6D, lane 4), implying that PfGatAB can utilize both L-asparagine and L-glutamine as amide donors. We also noted that the Glu-AMP spot in the TLC assay was less intense than the Gln-AMP (Fig. 6D, lanes 3 and 4) and control Glu-AMP spots (Fig.  6D, lane 2), indicating that the sequential glutamylation and amidotransferase reactions proceeded rapidly. This could be due to the sequestration and immediate substrate channeling (60) of the misacylated Glu-tRNA Gln from the ND-GluRS to GatAB to prevent the release of misacylated Glu-tRNA Gln and subsequent misincorporation of L-Glu in place of L-Gln during protein synthesis (61). The different amidotransferase assay approaches used here and the results obtained demonstrate that PfGatAB does not require the GatC subunit for transamidation reaction.
We did not test whether PfGatAB possesses Asp-AdT activity as bacterial GatCABs do because the Plasmodium genome possesses two predicted AspRSs, one cytoplasmic and the other apicoplast-targeted (11,25,56,62), as well as an apicoplasttargeted AsnRS. Thus, the apicoplast appears to possess all components required for asparaginylation of apicoplast tRNA Asn via direct aminoacylation. Together, these data are consistent with the conclusion that PfGatAB is a Glu-AdT.
We previously demonstrated that the P. falciparum nuclear genome (23) encodes two putative GluRS enzymes. One, Pf3D7_1349200, appears to be cytoplasmic (58). The other, Pf3D7_1357200, possesses a predicted N-terminal apicoplasttargeting sequence (56,62), was localized to the apicoplast, and exhibits non-discriminating glutamylation activity in vitro, producing both Glu-tRNA Glu and Glu-tRNA Gln . It is the first enzyme in the apicoplast's indirect aminoacylation pathway (23). In this study we have further dissected the apicoplast's indirect aminoacylation pathway by identifying the PfGatA and PfGatB subunits of the apicoplast aminoacyl-tRNA AdT. Using episomal constructs in which GFP was fused to the predicted PfGatA and GatB apicoplast targeting sequences, we demonstrated that GFP was trafficked to the apicoplast in erythrocytic stage parasites. Minor overlaps between the anti-GFP and MitoTracker Red signals were observed where the apicoplast and mitochondrion appeared to contact one another (Fig. 3, C, panel ii, and D, panel ii), a phenomenon observed with other apicoplast-targeted proteins (81,82). Furthermore, we myctagged the endogenous P. berghei chromosomal gatA gene, and we observed that the tagged protein was trafficked to the apicoplast, confirming the results obtained in P. falciparum with episomal constructs. Localization of P. falciparum GatAB solely in the apicoplast differs from the situation in Arabidopsis, where the GatCAB amidotransferase is targeted to both the plastid and the mitochondrion (83). Dual targeting of PfGatAB to the plastid and mitochondrion is also unlikely because, as in the related parasite T. gondii, the Plasmodium mitochondrion probably imports aminoacylated tRNAs from the cytoplasm (58).
We expressed recombinant PfGatA and PfGatB independently in E. coli (Fig. 6, A and B), and we combined them in vitro to examine their ability to transamidate apicoplast Glu-tRNA Glu or Glu-tRNA Gln that had been previously glutamylated using recombinant PfGluRS (23). PfGatAB demonstrated a remarkable tRNA substrate specificity by converting Glu-tRNA Gln to Gln-tRNA Gln but did not transamidate Glu-tRNA Glu (Fig. 6C). The Plasmodium apicoplast possesses an ND-GluRS (23) but lacks an ND-AspRS (57, 58), and therefore it almost certainly utilizes the PfGatAB as a Glu-AdT. As in Plasmodium, bacteria such as B. subtilis (70) that possess an ND-GluRS but lack an ND-AspRS use their GatCAB only as a Glu-AdT (64). Because genes encoding GatA and GatB are present in all known Plasmodium genomes but not in any known organism in the other two domains, we concluded that GatAB is the Plasmodium glutamyl-tRNA Gln amidotransferase (GatAB/Glu-AdT).
We also showed that PfGluRS, PfGatA, and PfGatB, when briefly pre-mixed, can glutamylate apicoplast tRNA Gln and transamidate it to form Gln-tRNA Gln in vitro (Fig. 6D). This suggests that the Plasmodium apicoplast may contain a tRNA Gln ⅐ND-GluRS⅐GatA⅐GatB complex, akin to the transamidosomes described in archaea (61) and the eubacterium T. thermophilus (84). Such a complex would prevent challenging the genetic code integrity as demonstrated for tRNA-dependent Asn formation (61).
We used different approaches to test for the PfGatAB-catalyzed amidotransferase reaction. Our findings consistently showed that the Plasmodium parasite has established an evolved glutamyl-tRNA amidotransferase reaction that takes into account the absence of the gatC gene in the parasite genome and therefore does not require it for the transamidation reaction, contrary to the amidotransferase activity of the GatCAB paralog where the three subunits are required for enzyme activity (22). This is the first biochemical evidence for Glu-tRNA Gln transamidation by a GatAB in the absence of a GatC subunit.
We investigated the evolution of Plasmodium GatA and GatB subunits in comparison with the human GatA and GatB, the GatA and GatB subunits of bacterial GatCABs, the GatA and GatB subunits of other plastids, and the GatD and GatE subunits of archaeal GatDE. In the unrooted phylogeny of GatB and GatE proteins, Plasmodium GatBs were not placed within the bacterial GatB, archaeal GatE, or the plastid GatB clades but were placed on a separate branch (Fig. 3A). The divide is well supported with a bootstrap value of 100. In a similar fashion, the unrooted phylogeny of GatA and GatD proteins, Plasmodium GatA was not placed within the bacterial GatA, the plastid GatA, or the archaeal GatD clades but was placed on a separate branch (bootstrap value of 100) (Fig. 3B). Additionally the GatB and GatA sequences of T. parva and Babesia bovis, which like Plasmodium lack gatC in their genomes, were placed in the same clade as the Plasmodium orthologs. Taken in total, the results strongly suggest that GatAB, which is uniquely found in Apicomplexa that possess an apicoplast, is a paralog to Gat-CAB, GatFAB, and GatDE.
To identify conserved and divergent features of PfGatB and PfGatA, homology modeling was performed in comparison with the known S. aureus GatCAB structure (Fig. 1) (28,29). All functional features of bacterial GatCAB and archaeal GatDE are conserved in PfGatAB, but we found that PfGatAB had unique inserts that could not be fitted into the GatCAB model (Fig. 1). In archaeal GatDE, the unstructured N-terminal insert of GatD has been reported to play a structural role that gives it the ability to associate to GatE (31). GatF found only in fungal genomes (46,47) and GatC found in bacterial, plant, and mammalian AdTs have similar unstructured characteristics and play a similar structural role where they reinforce the interaction between the GatA and GatB subunits by encircling their inter-face (28,29). We were not able to identify a homolog of the gatC or gatF genes in the P. falciparum genome, leading us to speculate that the unique inserts present in PfGatAB could be playing a yet undetermined structural role of holding PfGatA and PfGatB together, similar to that of the insertion found in GatD. Crystallization studies on the PfGatAB enzyme will help determine the roles of these inserts in catalysis.
Here, we have shown that besides possessing an apicoplasttargeted, non-discriminating GluRS (23), malaria parasites contain a unique apicoplast-targeted glutamyl-tRNA amidotransferase that amidates Glu-tRNA Gln in the absence of GatC. The apicoplast is an attractive drug target because it is essential to both blood and liver stage parasites (18,19), harbors several metabolic pathways absent in the host, and its transcriptional and translational machinery are of bacterial origin. The apicoplast's indirect aminoacylation pathway is probably essential in malaria parasites because the parasite genome does not encode an apicoplast-targeted GlnRS (57,58,85). PfGluRS, the first enzyme in the apicoplast's indirect aminoacylation pathway, was refractory to gene deletion and is therefore essential for blood stage development (23). In this study, the gene encoding for apicoplast PbGatA was refractory to gene deletion, despite the fact that it was accessible to myc tagging. Thus, both PfGluRS and PfGlu-AdT are potentially good drug targets.
The emergence of resistant strains of P. falciparum continues to fuel an urgent need to develop new antimalarials. Besides our findings in this study, the Plasmodium GatAB amidotransferase has also been identified to be an important antimalarial drug target using in silico molecular modeling approaches (86). However, despite the chemical diversity of compound collections, robust inhibitors that act specifically against Plasmodium GatAB amidotransferases and not other amidotransferases have yet to be identified. Further investigations examining the requirements for apicoplast indirect aminoacylation across the Plasmodium life cycle and screening of compounds to identify specific inhibitors of plasmodial GluRS and Glu-AdT enzymes may lead to novel ways to target this pathway for chemotherapy.