Plasmodium falciparum Apicomplexan-Specific Glucosamine-6-Phosphate N-Acetyltransferase Is Key for Amino Sugar Metabolism and Asexual Blood Stage Development

Apicomplexan parasites cause a major burden on global health and economy. The absence of treatments, the emergence of resistances against available therapies, and the parasite’s ability to manipulate host cells and evade immune systems highlight the urgent need to characterize new drug targets to treat infections caused by these parasites. We demonstrate that glucosamine-6-phosphate N-acetyltransferase (GNA1), required for the biosynthesis of UDP-N-acetylglucosamine (UDP-GlcNAc), is essential for P. falciparum asexual blood stage development and that the disruption of the gene encoding this enzyme quickly causes the death of the parasite within a life cycle. The high-resolution crystal structure of the GNA1 ortholog from the apicomplexan parasite C. parvum, used here as a surrogate, highlights significant differences from human GNA1. These divergences can be exploited for the design of specific inhibitors against the malaria parasite.

mediated excision of the floxed sequence could be induced by adding rapamycin into the media. The ablation disrupts a region comprising 669 bp of the gna1 gene, including the Gcn5-related N-acetyltransferase (GNAT) domain, conserved in enzyme families that use acetyl-CoA to acetylate substrates (22,23). The genomic modification was engineered in the DiCre-expressing P. falciparum line 3D7 II.3 (19).
Confirmation of the appropriate modification of the gna1 gene after transfection was assessed by diagnostic PCR. Subsequent limiting dilution cloning of the gna1 locus-modified parasites resulted in the isolation of gna1-loxP-C4 and gna1-loxP-G8 clones, in which the modification of the locus was reconfirmed again by PCR (Fig. 1B). Efficiency of the conditional excision of the floxed gna1-loxP clones was assessed by the addition of rapamycin or dimethyl sulfoxide (DMSO; vehicle control) in ring stage synchronized cultures. Cells were treated for 1 h, followed by washing and incubation for 29 h to allow parasite maturation. Genomic DNA from the clones was used for diagnostic PCR using primers annealing in the homology regions that demonstrated the efficient excision of gna1-loxP (Fig. 1C). Both clones were used in all subsequent experiments. Integration of this sequence by double crossover homologous recombination was promoted by (i) Cas9 endonuclease-mediated double-strand DNA breaks, guided by a single guide RNA (sgRNA) targeting a 20-nucleotide region of Pfgna1 (green arrows); and (ii) the addition of homologous sequences (red box at the 5= region and thick red line at the 3= region) to either side of the loxP sites. Closed colored arrows indicate the primer binding sites. Rapamycin-induced site-specific recombination between the loxP sites removes the recodonized part of the gene (green box) encompassing 223 amino acids of the 291 amino acids of the PfGNA1 enzyme. (B) PCR analysis of gna1-loxP clones C4 and G8 confirms successful gene editing. Genomic DNA from parental 3D7 (WT) parasites or C4 and G8 clones was used as the template for PCR using the primers indicated in panel A. The numbers between the colored arrows indicate the expected amplicon size. (C) Truncation of the rapamycininduced gna1 gene. gna1-loxP C4 and G8 clones were analyzed by PCR ϳ29 h after treatment with DMSO (D) or rapamycin (R) using the primers identified in panel A. Excision reduces the amplicon from 1,552 bp to 733 bp, disrupting gna1. P. falciparum GNA1 Is Essential for Parasite Survival Disruption of the gna1 gene leads to the inhibition of parasite growth and reduction of UDP-GlcNAc pools. GNA1 catalyzes the acetylation of glucosamine-6phosphate (GlcN6P) to generate N-acetylglucosamine-6-phosphate (GlcNAc6P), which is needed for the biosynthesis of UDP-GlcNAc, a required precursor for both GPIanchors and N-glycans (10). These glycoconjugates are critical for parasite viability and, hence, a defect in their synthesis may largely affect P. falciparum development. To ascertain the effect of gna1 loss in asexual parasites, rapamycin-treated cells were grown and monitored by flow cytometry over ϳ144 h (i.e., three intraerythrocytic developmental cycles [IDCs]). Upon rapamycin-induced gna1-loxP excision, parasites failed to expand over three IDCs, whereas DMSO-treated parasites grew normally. Neither rapamycin nor DMSO affect parasite growth at the incubation times and concentrations used (19). Interestingly, parasite growth in rapamycin-treated cultures was rescued by supplementing the medium with 10 mM GlcNAc, suggesting the presence of a metabolic pathway for the uptake and activation of free GlcNAc ( Fig. 2A). Rapamycin-treated cultures showed an ϳ2.5-fold invasion rate reduction in the first IDC, indicating the expansion of a small proportion of viable parasites that likely reflects the kinetics of DiCre-mediated excision (19), GNA1 turnover, and/or the dynamics of sugar nucleotide pools (24) (Fig. 2B). Nevertheless, invasion rates were reduced to 1 (i.e., Synchronous ring-stage gna1-loxP growth curves of C4 and G8 clones showing parasite replication over the course of three erythrocytic cycles when grown in the presence of DMSO, rapamycin, or rapamycin ϩ 10 mM GlcNAc. Data were averaged from three biological replicates and presented as the mean Ϯ standard deviation (SD). (B) Invasion rates of DMSO-or rapamycin-treated cultures (clones C4 and G8). The graphs show mean Ϯ SD values of three replicates and significance was assessed using unpaired t tests. (C) Quantification of UDP-GlcNAc and other sugar nucleotide pools in late stages (mature trophozoites) of gna1-loxP parasites treated with rapamycin. Values indicate the sugar nucleotide concentration Ϯ SD of six different biological replicates. Unpaired t tests were used to compare differences; ***, P Ͻ 0.05; ns, not significant. absence of growth) in rapamycin-treated cultures at cycles 2 and 3. Parasite morphology along the IDCs was normal in DMSO treated-cultures, gradually losing synchronization after 2 to 3 cycles. In contrast, in rapamycin-treated cultures, growth-arrested parasites accumulated as trophozoites and schizonts, indicating that gna1 is needed for parasite development through mature stages (Fig. S1 in the supplemental material). To precisely assess the effect of PfGNA1 loss on UDP-GlcNAc levels, we measured and compared the sugar nucleotide dynamic pools of DMSO-treated parasites with those of rapamycin-treated parasites. Upon rapamycin treatment, the UDP-GlcNAc pool was significantly reduced, despite that levels of other sugar nucleotides remained similar (Fig. 2C). Therefore, without discarding potential PfGNA1 moonlighting activities essential for parasite development, the results strongly suggest that gna1 disruption prevents GlcN6P acetylation by PfGNA1 and reduces the pool of UDP-GlcNAc, hindering P. falciparum asexual intraerythrocytic growth.
C. parvum GNA1 structural analysis. PfGNA1 activity is crucial for the development of the malaria parasite and, therefore, understanding the structural differences with human GNA1 (HsGNA1) may contribute to the discovery of PfGNA1-selective inhibitors. To further characterize the enzyme and, given our inability to express and purify sufficient amounts of active PfGNA1 (10), we used CpGNA1 as a surrogate, considering that it belongs to the same Apicomplexa-specific GNA1 family and can be purified fairly well with a high yield (10). The percentage of pairwise identity within the ca. 90-amino acid conserved region of the CpGNA1 and PfGNA1 GNAT domains is 48.8% (similarity 61.9%). The first CpGNA1 structure, determined from crystals belonging to the space group P2 1 2 1 2, with one monomer in the asymmetric unit, was solved by using the multisolution parallel phasing software ARCIMBOLDO (25). Extension of the placed fragments with SEQUENCE SLIDER (26) was required. The dimeric structure of CpGNA1 in complex with acetyl-CoA and glucose-6-phosphate (Glc6P) from data at 1.95 Å in space group P2 1 2 1 2 1 was solved by molecular replacement with Phaser (27) using the monomeric structure ( Fig. 3A and B), and further assessing ligand binding by representing unbiased sim-omit maps (Fig. S2). Note that Glc6P was used previously as a pseudosubstrate to get insights into GNA1-GlcN6P recognition and catalysis ( Fig. 3A and B) (28). CpGNA1 is found as a dimer, as is the case for other reported GNA1 crystal structures (29)(30)(31). CpGNA1 adopts the typical GCN5-related N-acetyltransferase (GNAT) fold previously described in GNA1 enzymes, though with some variations (see below; Fig. 3A) (29). The CpGNA1 structure consists of a central twisted four-stranded antiparallel ␤-sheet surrounded by four ␣ helices (Fig. 3A). This feature is conserved with HsGNA1, though this enzyme also contains additional ␤-strands and ␣-helices ( Fig. 3C and Fig. S3) (32). The superposition of CpGNA1 and HsGNA1 monomers gives a root mean square deviation (RMSD) of 1.68 Å on 132 C␣ atoms, suggesting some conformational differences (Fig. 3C). Additionally, HsGNA1 resembles the more typical signature of the GNAT superfamily than does CpGNA1, supporting our previous data that suggest an independent evolutionary origin and highlight the presence of significant differences between apicomplexan and nonapicomplexan GNA1s, as discussed in reference 10. Actually, the Apicomplexa GNA1 family shows a taxonomic distribution restricted to apicomplexan organisms (10). In agreement, the typical topology for GNAT-fold-adopting members (␤1-␣1-␣2-␤2-␤3-␤4-␣3-␤5-␣4-␤6-␤7) (32-34) is more consistent with that found for HsGNA1 (␣1-␤1-␣2-␣3-␤2-␤3-␣4-␤4-␣5-␤5-␣6-␤6) and more distant to that found for CpGNA1 (␤1-␣1-␣2-␤2-␤3-␤4-␣3-␣4), which contains significantly fewer secondary structure elements (Fig. S3). Remarkably, the highly conserved tyrosine of the active site, also present in CpGNA1 (Tyr127), establishes a hydrogen bond between its hydroxyl group and the thioester sulfur, leading to an increase in the electrophilic character of the carbonyl group of the acetyl-CoA molecule during catalysis (28,30) (Fig. 3A). This is key for favoring a direct nucleophilic attack by the GlcN6P amine (28) that promotes catalysis. Note the proximity of the Glc6P OH 2 that mimics the amine group of GlcN6P to the carbonyl group of acetyl-CoA (distance  Fig. 3B), which clearly supports that the same mechanism might take place for CpGNA1.
CpGNA1 active site. A close inspection of the active site of CpGNA1 and its comparison with the HsGNA1 active site showed that both the GlcN6P and the acetyl-CoA binding sites have different shapes (Fig. 3D), and are also different at the sequence level ( Fig. 3E and Fig. 4). This is more evident for the acetyl-CoA binding site, in which significant differences in the acetyl-CoA/CoA phosphorylate ADP moiety are found. For example, the angle between both adenine moieties is close to perpendicular (74°) and the distance between the free amines is large (5.4 Å) (Fig. 3D). Note that the angle is formed between the C1B of the phosphorylated ribose and the two C4A of the adenine moieties. However, the CpGNA1 active site and the predicted PfGNA1 active of conservation of amino acid residues surrounding the binding site are color coded. CpGNA1 was compared with HsGNA1 (upper panel) and PfGNA1 (lower panel). Note that no models were generated for either HsGNA1 or PfGNA1 and that the differences were exclusively based on amino acid sequence alignments. Orange and white represent identity/conservative substitutions and nonconservative substitutions, respectively. The ligands are represented as sticks with green carbon atoms, except for acetyl-CoA and glucose-6P in lower panel of (D), in which they are colored as brown carbon atoms.

FIG 4
Multiple sequence alignment of CpGNA1, HsGNA1, and PfGNA1. The CpGNA1 residues that interact with acetyl-CoA and Glc6P and the grade of conservation with other CpGNA1 orthologs based on these interactions are highlighted as follows: yellow (full conservation); green (conserved residues between CpGNA1 and PfGNA1); magenta (conserved residues between CpGNA1 and the HsGNA1); and black/gray colors that indicate the degree of conservation between the three orthologs. Residues interacting with acetyl-CoA and Glc6P are also displayed as magenta diamonds and inverted orange triangles, respectively. Shown above the sequence for CpGNA1, in red and blue, is displayed the secondary structure elements (␣-helices and ␤-strands, respectively). Note that residues 90 to 93 of CpGNA1 align well at the structural level with the equivalent residues in HsGNA1 but not at the sequence level. This is evidenced for CpGNA1 Asn90 and HsGNA1 Gln132, and CpGNA1 Gly92 and HsGNA1 Gly130. The four different motifs present in the GNAT family of higher eukaryotes are represented as open red boxes (32). The secondary structures are mostly conserved between CpGNA1 and HsGNA1 in motifs D and A, and to a lesser extent in motif B (see Fig. S3 for a comparison of the secondary structure elements between both enzymes), exemplifying the large differences between the enzymes from higher eukaryotes and the Apicomplexa phylum.
site are more similar (Fig. 3B), as anticipated in our previous work (10). There are 19 residues of CpGNA1 engaged in the recognition of acetyl-CoA and Glc6P ( Fig. 4 and Fig. 5). Some of these residues forming the GlcN6P binding site come from both monomers (see Fig. 5). Out of these 19 residues, 11 residues are conserved between HsGNA1 and CpGNA1, whereas 15 residues are conserved between CpGNA1 and PfGNA1. Two residues are unique in CpGNA1, though these residues interact with the ligands through their backbones ( Fig. 4 and Fig. 5).
C. parvum GNA1 site-directed mutagenesis. Based on the structural information, nine CpGNA1 site-specific mutants were created to study the contribution of specific residues to CpGNA1 binding and activity. All mutants were designed to disrupt the binding of either acetyl-CoA (N90A, I122G, and Y127A) or GlcN6P (R73A, R141A, R113A, T117A, E119A, and H81A) (Fig. 5). Note that in our structure, the critical Tyr127 and Arg141 were very close to the ligands but did not show any interaction (Fig. 5). The protein backbone is shown as a gray ribbon. Ligands are shown as sticks with green carbon atoms. HsGNA1 residues interacting with the ligands are shown for comparison and illustrative purposes but will not be discussed in detail. Protein-ligand hydrogen bonds are shown as dotted green lines. Identical, homologous, and dissimilar residues between both enzymes are colored in gray, black, and brown carbon atoms, respectively. Note that the superscript B in several residues defines residues that come from chain B.
However, and considering the importance of the Tyr counterparts in other orthologs (28), it would be expected that the mutation of Tyr127 to Ala would diminish or abolish the activity, as shown below. The selected amino acids were replaced by alanine (A) in all cases except for Ile122, for which isoleucine was replaced by glycine (I122G).
The steady-state kinetic parameters of wild type (WT) and CpGNA1 mutants were studied by an in vitro colorimetric assay by monitoring CoA release, with increasing concentrations of acetyl-CoA at a fixed concentration of GlcN6P and vice versa (Fig. 6). The R73A, R113A, R141A, and Y127A activities were almost negligible, hampering the calculation of reliable kinetic parameters (Table 1).
CpGNA1 mutant kinetics data showed a general decrease in catalytic efficiency (k cat /K m ) and enzyme maximal specific activity (defined here as V max ) ( Table 1). The catalytic efficiency of CpGNA1 mutants decreased between 2.6-and 5.2-fold and 18.6to 1,754-fold for acetyl-CoA and GlcN6P, respectively. V max also decreased between 8and 83-fold, depending on the mutant enzyme. With respect to the K m values, all mutants displayed lower values for acetyl-CoA, indicating a higher affinity for this substrate than WT CpGNA1. Regarding GlcN6P, all mutants displayed higher K m values, suggesting a lower affinity for GlcN6P compared to WT. T117A showed the lowest K m for acetyl-CoA (K m ϭ 33.41 Ϯ 11.82 M, which is a 15.9-fold K m decrease compared to  WT) and H81A the highest K m for GlcN6P (K m ϭ 4,007 Ϯ 1,425 M, which is a 26-fold K m increase compared to WT). Nevertheless, T117A and H81A showed remarkably low catalytic efficiencies for both acetyl-CoA and GlcN6P. Although it was expected to see a weak affinity for GlcN6P in the T117A mutant, the higher affinity for acetyl-CoA was not predicted because Thr117 shows only interactions with the pseudosubstrate Glc6P (Fig. 5). The relevance of key residues for substrate interaction, especially those related to GlcN6P binding (i.e., Arg73, Arg113, and Thr117), which are not conserved in HsGNA1, strongly prompts further exploration of the enzyme as a selective drug target against P. falciparum and, possibly, other apicomplexan parasites. Overall, and except for Y127A, our results suggest that most of the mutants with alterations around the GlcN6P binding site show more significant effects on the kinetic properties under variable GlcN6P concentrations than the corresponding mutants show with acetyl-CoA. This implies a more significant role of the native residues around the GlcN6P binding site than the residues around acetyl-CoA.

DISCUSSION
In this work we use CRISPR-Cas9-based engineering techniques (20), combined with a DiCre-recombinase system, for conditional P. falciparum gna1 gene disruption upon rapamycin induction (19,21,35). This is a widely used and versatile strategy for rapid conditional gene ablation and functional analysis of essential genes in the malaria parasite. The data demonstrate that PfGNA1, which acetylates GlcN6P in the amino sugar pathway, is required for the synthesis of adequate pools of UDP-GlcNAc and for parasite development during the asexual intraerythrocytic stages of the parasite. This confirms results from large-scale genetic screenings recently carried out in P. falciparum, underlining the relevance of the enzymes in the amino sugar pathway for the survival of asexual parasites (11). Remarkably, works performed in a P. berghei murine model of malaria also highlight the critical role of UDP-GlcNAc and the amino sugar metabolic route in mosquito and liver stages (12). However, contrasting to P. falciparum (11), P. berghei mutants in the UDP-GlcNAc biosynthetic route showed only slowgrowth phenotypes in asexual intraerythrocytic parasites, indicating the nonessentiality of the pathway at these stages of the life cycle (12,36). Despite that a gna1 mutant has not been generated and UDP-GlcNAc levels have not been characterized in P. berghei (12,36), this suggests that murine parasites can obtain UDP-GlcNAc from host cells or alternative sources along intraerythrocytic asexual development (12). Indeed, there exist other examples in which the alteration of a de novo sugar nucleotide metabolic route does not cause a significant reduction of the pathway's product (37), and our own data show that P. falciparum gna1 mutant growth is rescued in vitro by GlcNAc medium supplementation. Thus, P. falciparum parasites seem to be able to take up and activate GlcNAc when the amino sugar pathway is disrupted (38), in agreement with the wide range of sugars that can be imported through its hexose transporter 1 (39). Nevertheless, P. falciparum potential GlcNAc uptake is deemed to be negligible in vivo, as this is not an abundant free sugar in the parasite host (40), opening the door to explore the selective inhibition of PfGNA1 as a new approach to treat malaria.
To identify structural differences between PfGNA1 and HsGNA1 that could be exploited for selective drug design, we intended to solve the P. falciparum enzyme 3D structure. However, the deficient expression of PfGNA1 prompted us to explore CpGNA1 as a surrogate model for Apicomplexa-specific GNA1 (10). Fortunately, CpGNA1 is expressed at high levels in Escherichia coli, allowing us to obtain structural and enzymatic insights. Remarkably, despite the low identity at sequence level between CpGNA1 and HsGNA1, the crystal structures are similar and share the typical acetyltransferase fold (23). Although this had been already described in members of the GNAT superfamily, which show well-conserved structures and catalytic mechanisms despite extensive divergence at the sequence level (29,32,33), the independently evolved Apicomplexa GNA1 family still presents different catalytic residues crucial for enzymatic activity (10). Though both the CpGNA1 acetyl-CoA and GlcN6P binding sites are different, some features are conserved, such as the highly conserved and critical tyrosine residue at position 127 (Tyr127), also present in nonapicomplexan GNA1 enzymes (29,32,33). At the level of the GlcN6P binding site, the main differences are localized in two arginine residues at positions 73 and 113, together with histidine, threonine, and glutamic acid residues at positions 81, 117, and 119, respectively. At the level of the acetyl-CoA binding site, the main differences are localized in histidine, asparagine, isoleucine, lysine, and leucine at positions 81, 90, 122, 125, and 126, respectively. Site-directed mutagenesis and kinetic assays confirmed the essential function of Tyr127. Likewise, GNA1 activity was not detected in our R141A mutant, suggesting that the positive charge of the arginine, which is also conserved in HsGNA1, might interact with the negatively charged atoms on the phosphate group of GlcN6P or facilitate the interaction of Arg113 with the phosphate group. Furthermore, our analyses showed that Arg73 and Arg113 residues are relevant for GlcN6P affinity, whereas His81 is important for both GlcN6P and acetyl-CoA binding, in agreement with our structural data showing that His81 interacts with both ligands. Although our data reveal that Thr117 is important for GlcN6P and acetyl-CoA binding, the latest is not completely understood because this residue only interacts with the sugar moiety. Overall, there are remarkable differences between CpGNA1 and HsGNA1 at the level of GlcN6P and acetyl-CoA binding sites, which could be exploited for drug selectivity against parasites.
The amino sugar pathway has been explored in different organisms as a possible source of novel therapeutic targets, and inhibitors of the Aspergillus fumigatus GNA1 enzyme have been recently identified (32,33,41). The independent evolution and sequence divergence of apicomplexan GNA1, together with its essentiality in P. falciparum, highlight the potential of the enzyme as a selective therapeutic target against the malaria parasite (10). Furthermore, the recent identification of the amino sugar pathway as an important metabolic bottleneck in P. falciparum indicates that the inhibition of PfGNA1 could largely hamper the growth of the parasite (9). Remarkably, UDP-GlcNAc has also been reported as the most limiting sugar nucleotide in mosquito and liver stages, and, as stated above, several enzymes in the metabolic route have been proven to be essential for parasite transmission to mosquitoes in a malaria murine model (12). Altogether, the data underline the potential of PfGNA1, central to the amino sugar metabolic pathway, as a multistage drug target against P. falciparum. Interestingly, considering the particularities of GNA1 in the Apicomplexa and the importance of the UDP-GlcNAc donor for the biology of these organisms, the selective inhibition of the enzyme might also provide therapeutic opportunities against other apicomplexan pathogens such as C. parvum or Toxoplasma gondii (42,43).

MATERIALS AND METHODS
P. falciparum culture and transfection. P. falciparum asexual stages were cultured at 37°C in an atmosphere of 92% N 2 , 3% O 2 , and 5% CO 2 with washed Bϩ blood type red blood cells (RBCs) at 2 to 4% hematocrit in Albumax-containing RPMI 1640 medium. Human erythrocytes were purchased from the Banc de Sang i Teixits (Catalonia, Spain), after approval from the Comitè Ètic Investigació Clínica Hospital Clínic de Barcelona. Parasite growth was monitored by counting the infected erythrocytes in Giemsa-stained blood smears by light microscopy. Parasites harboring a floxed region of gna1 were generated by Cas9-mediated gene replacement using a pDC2 Cas9-and specific-single guide RNA (sgRNA)-expressing construct, and a linearized pUC19 plasmid containing the engineered gna1 sequence (19). P. falciparum 3D7 II-3 parasites, with the DiCre system inserted into the p230p genomic locus (a generous gift from Ellen Knuepfer [19]), were transfected during ring stages, as described elsewhere (37).
Parasites recovered after transfection were harvested for genotyping. Genomic DNA from parasites was extracted with the QIAmp DNA minikit (Qiagen) following the manufacturer's instructions. Purified samples were used as a template for PCR amplification of the inserted construct. Clonal parasite lines were then obtained from engineered populations by limiting dilution. P. falciparum growth analysis. For rapamycin treatment experiments, 0.1% DMSO or 10 nM rapamycin was added to the culture medium in ring-stage-synchronized parasites, incubated for 1 h, and washed away (19). This was done by adding 1:1,000 (vol/vol) stocks of 100% DMSO or 10 M rapamycin, respectively, to the culture medium. Genomic DNA was recovered 24 to 30 h later. For growth assessment after rapamycin treatment, 200 l of parasite culture at 3% hematocrit and with an initial parasitemia of 0.1% were transferred to 96-well plates. Parasitemia was assessed by flow cytometry after 48 h, 96 h, and 144 h, using SYTO11 as previously described (44).
Sugar nucleotide analysis. Asexual blood stage parasite cultures were sorbitol-synchronized (or combining Percoll and sorbitol treatments, for tight synchronization) and osmotic lysis was performed on trophozoite-infected red cells at 35 to 40 h postinvasion (5 to 10% parasitemia). Sugar nucleotides were P. falciparum GNA1 Is Essential for Parasite Survival ® September/October 2020 Volume 11 Issue 5 e02045-20 mbio.asm.org 11 analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) and multiple-reaction monitoring (MRM) in a QTRAP 6500 System (AB Sciex). The metabolites, separated using a Hypercarb PGC column (5 m, 2.1 ϫ 100 mm; Thermo Fisher Scientific) with a mobile phase composed of 0.1% formic acid (pH 9) and an acetonitrile gradient (24), were identified by their diagnostic MRM transitions (38). Expression and purification of CpGNA1. The pGEX6P1-CpGNA1 plasmid encoding the CpGNA1 protein C terminus fused to the glutathione S-transferase (GST), with a PreScission Protease (PP) cleavage site (LEVLFQGP) encoded in the linker between GST and CpGNA1, was transformed into the Escherichia coli expression strain BL21 Star (Thermo Fisher Scientific). Transformants were selected on LB (Sigma-Aldrich) plates containing 100 g/ml of ampicillin. Batch cultures were scaled up to 2 liters of LB medium, and cells were grown to an optical density of 0.5 to 0.6 at 37°C and 180 rpm. They were subsequently induced with 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) (Sigma-Aldrich). After 20 h of induction at 18°C, the cells were collected by centrifugation at 11,295 ϫ g for 20 min. The resulting pellet was suspended in 60 ml of binding buffer A (25 mM TRIS [pH 7.5], 150 mM NaCl) that was supplemented with 4 l of Universal Nuclease for Cell Lysis (Thermo Fisher Scientific, 250 U/l), lysozyme (Sigma-Aldrich, Ͼ95% pure) up to a final concentration of 1 mg/ml, and 500 l of a protease-inhibitor cocktail solution (1 M PMSF, 10 mM benzamidine, 0.5 mM leupeptine). This solution was incubated for 30 min at 37°C and then sonicated in ice for 15 cycles, each consisting of a 30-s pulse plus a 30-s cooling step. The lysate was afterward centrifuged at 39,086 ϫ g for 20 min.
The supernatant was filtered through a 0.45-m membrane and loaded at 1 ml/min into two connected GSTrap 4B (GE Healthcare) columns, previously equilibrated with buffer A. The columns were washed with buffer A until no 280-nm absorbance was observed, and the bound protein was eluted by a 25-min gradient from buffer A to buffer B (25 mM TRIS, 150 mM NaCl, 400 mM L-glutathione reduced [Sigma-Aldrich] [pH 7.5]). The recombinant fusion protein containing CpGNA1 was then exchanged into buffer C (25 mM Tris [pH 7.5], 150 mM NaCl) and incubated overnight at 4°C with PP at a ratio of 1 mg of PP per 50 mg of fusion protein, in order to remove the GST protein, leaving the GPLGS amino acids at the CpGNA1 N terminus. Once the cleavage was verified by SDS-PAGE, the mixture was loaded into three coupled GSTrap 4B columns, previously equilibrated in buffer A. Both GST and PP remained fixed to the column matrices, and the CpGNA1 was collected as the unbinding fraction. CpGNA1 was finally purified by size exclusion chromatography using a HiLoad 26/60 Superdex 75 Column (GE Healthcare), previously equilibrated with buffer C. Fractions containing CpGNA1 were dialyzed against buffer D (25 mM Tris pH 7.5), concentrated, and used for biophysical experiments. Quantification of the enzyme was carried out by absorbance at 280 nm using its theoretical extinction coefficient. A yield of 15 mg/liter of culture was obtained.
CpGNA1 crystallization, phasing, and refinement. Crystal forms for CpGNA1 were obtained when CpGNA1 (1.5 mg/ml) was cocrystallized with 10 mM acetyl-CoA. The sitting-drop vapor-diffusion method was used to produce crystals by mixing 0.5 l of the protein-ligand solution with an equal volume of mother liquor (0.2 M NH 4 Cl, 0.1 M HEPES [pH 7.5], 25% glycerolethoxilate) at 18°C. Bar-like-shaped crystals grew within 1 day. The crystals were soaked for 45 min with a mixture containing 50 mM acetyl-CoA and 50 mM Glc6P in the mother liquor described above. Then, crystals that were either cocrystallized with acetyl-CoA or further soaked with acetyl-CoA and Glc6P were cryoprotected using 25% glycerol in the mother liquor and frozen in nitrogen gas stream cooled to 100 K. X-ray diffraction data were collected at XALOC-13 beamline at ALBA Synchrotron. Data were processed and scaled using XDS (45) and CCP4 (46,47) software packages. Relevant statistics are given in Table S1 in the supplemental material.
A promising partial solution for the monomeric CpGNA1 was identified for the P2 1 2 1 2 crystal form (crystals cocrystallized with acetyl-CoA) from a data set diffracting anisotropically to 1.5 Å, using the multisolution parallel phasing software ARCIMBOLDO_LITE (48). ARCIMBOLDO uses PHASER to place individual ␣-helices by eLLG-guided molecular replacement (49) and then expand partial solutions with SHELXE (50) through density modification and autotracing. The search was set to locate 1 copy of an ensemble generated with an alpha version of the software ALEPH (51) The ensemble of two models represented the probable fold composed by 5 beta strands flanked by 2 alpha helices at each side and it contained the GNA1 of Saccharomyces cerevisiae (PDB id 1I21, 19% identity to target sequence) and the acetyltransferase from Agrobacterium tumefaciens (PDB 2DXQ, 26% identity to target sequence). The ARCIMBOLDO_LITE solution was completed by SEQUENCE SLIDER (26) using side chain modeling with Scwrl4 (52), refinement with REFMAC5 (53), and expansion with SHELXE, resulting in a best trace of 120 residues with CC 31.9. The structure of CpGNA1 with the substrates at 1.95 Å, in space group P2 1 2 1 2 1 , containing two monomers in the asymmetric unit was phased by molecular replacement (27) using the previous model. Initial phases were further improved by cycles of manual model building in Coot (54) and refinement with REFMAC5 (53) and Phenix (55). Further rounds of Coot and refinement with Phenix (55) were performed to obtain the final structure. The final model was validated with PROCHECK, and model statistics are given in Table S1. The Ramachandran plots for the CpGNA1-acetyl-CoA-Glc6P show that 89.0%, 10.6%, 0.4%, and 0% of the amino acids are in most favored, allowed, generously allowed, and disallowed regions, respectively. PyMOL (56) was used to generate images.
CpGNA1 colorimetric in vitro enzyme kinetic assays. Codon-optimized versions of CpGNA1 WT and site-specific mutant proteins were expressed and purified as previously described (10). Briefly, sequences were cloned in a pGEX 6P-1 vector containing an N-terminal GST-tag. After induction, fusion proteins were purified by affinity chromatography and eluted with reduced glutathione without cleaving the GST-tag. In vitro activity colorimetric assays were based on the detection of coenzyme A, generated during acetyl transfer, by reaction with the thiol reagent 5,5=-dithiobis(2-nitrobenzoic acid) (DTNB). Acetyl-CoA and GlcN-6P were used as the substrates and all the reactions were carried out at room temperature, in triplicate, in a 96-well-plate format. CpGNA1 was assayed in buffer solution (25 mM Tris-HCl and 150 mM NaCl, pH 7.2) using different acetyl-CoA concentrations and a fixed GlcN-6P concentration (2,000 M), or different GlcN-6P concentrations and a fixed acetyl-CoA concentration (500 M). CpGNA1 WT was assayed at a 5 g/ml concentration whereas CpGNA1 mutants were tested at a 25 g/ml concentration. The data were fitted to a nonlinear least-squares regression using GraphPad Prism.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.