CRISPR/Cas9 and glycomics tools for Toxoplasma glycobiology

Infection with the protozoan parasite Toxoplasma gondii is a major health risk owing to birth defects, its chronic nature, ability to reactivate to cause blindness and encephalitis, and high prevalence in human populations. Unlike most eukaryotes, Toxoplasma propagates in intracellular parasitophorous vacuoles, but like nearly all other eukaryotes, Toxoplasma glycosylates many cellular proteins and lipids and assembles polysaccharides. Toxoplasma glycans resemble those of other eukaryotes, but species-specific variations have prohibited deeper investigations into their roles in parasite biology and virulence. The Toxoplasma genome encodes a suite of likely glycogenes expected to assemble N-glycans, O-glycans, a C-glycan, GPI-anchors, and polysaccharides, along with their precursors and membrane transporters. To investigate the roles of specific glycans in Toxoplasma, here we coupled genetic and glycomics approaches to map the connections between 67 glycogenes, their enzyme products, the glycans to which they contribute, and cellular functions. We applied a double-CRISPR/Cas9 strategy, in which two guide RNAs promote replacement of a candidate gene with a resistance gene; adapted MS-based glycomics workflows to test for effects on glycan formation; and infected fibroblast monolayers to assess cellular effects. By editing 17 glycogenes, we discovered novel Glc0–2-Man6-GlcNAc2–type N-glycans, a novel HexNAc-GalNAc-mucin–type O-glycan, and Tn-antigen; identified the glycosyltransferases for assembling novel nuclear O-Fuc–type and cell surface Glc-Fuc–type O-glycans; and showed that they are important for in vitro growth. The guide sequences, editing constructs, and mutant strains are freely available to researchers to investigate the roles of glycans in their favorite biological processes.

Glycosylation-associated drugs are beginning to make an impact in human health, and there is reason to anticipate success in combating pathogens with their distinctive glycan synthetic pathways (27)(28)(29)(30). Unique glycoconjugates are prime candidates for vaccine development and as markers for diagnostic monitoring. But before the full potential of these strategies can be realized, more structure-function studies are needed to guide selection of the most promising targets. Progress in exploiting these strategies has been inhibited by the uniqueness of the glycomes of Toxoplasma and other parasites and lack of information about the genetic basis of their nontemplate-driven mechanisms of assembly.
Further understanding of the roles of glycans will benefit from genetic approaches to manipulate their biosynthesis. Sufficient knowledge currently exists to allow informed predictions of genes associated with glycan assembly (31), and MS provides the necessary resolution and sensitivity to interpret glycomic consequences of glycogene perturbations. This report describes the generation of a collection of CRISPR/Cas9-associated guide DNAs that are suitable for disrupting glycogenes, using an approach that is applicable to strains regardless of their NHEJ (Ku80) status (32), and presents the results of trials using the type 1 RH strain. A streamlined MS workflow was implemented to benchmark the new studies, which led to the discovery of several novel N-and O-glycans and the enzymatic basis for the assembly of these and other known glycans. Finally, the glycogene mutants were screened in a cell culture growth assay for information about their physiological significance. It is expected that the availability of these findings and associated resources will help stimulate interrogation of the role of glycosylation in future functional studies.

Predicted glycogenes
Our algorithms predicted 33 GT-like genes and one glycophosphotransferase gene expected to contribute to the assembly of N-, O-, and C-glycans on proteins and the assembly of GPI-anchor precursors and other glycolipids. Four genes probably mediate N-glycan and GPI processing. An additional seven genes are likely to be involved with starch assembly and disassembly, and another contributes directly to synthesis of a cell wall polysaccharide. A total of 18 are involved in the generation of sugar nucleotides and Dol-P-sugars from simple sugars, and three probably mediate transport of sugar nucleotides into the secretory pathway. Our summary does not exclude the possibility of novel GT sequences yet to be discovered. Table 1 summarizes characteristics of the candidates, organized according to the class of glycoconjugates for which they are expected to contribute, along with findings from the literature and studies described below that provide support for the predicted functions. All but eight of the genes are correlated graphically with known or predicted glycosylation pathways in Figs. 4 -8. For an initial indication of the cellular roles of the glycogenes, we analyzed data from a published genome-wide CRISPR gRNAbased screen for genes that contribute to fitness in a fibroblast monolayer infection setting (26). The change of representation of guide DNA sequences in the population after three rounds of invasion and growth was translated into a fitness score based on a log 2 scale. An analysis of the 68 glycogenes in Table 1 revealed a range of fitness values ( Fig. 1) that roughly superimposed on the range of fitness values of all genes. 19 had fitness values ϽϪ2.5, suggesting that they may be important for viability. This group was enriched for predicted genes for N-glycosylation, GPI assembly, and precursor formation and deficient in genes associated with O-glycosylation in the secretory pathway and glycosylation in the nucleocytoplasmic compartment.

Glycogene disruption strategy
Gene editing was employed to associate individual glycogenes with assembly of the glycome and cell growth. As summarized in Fig. 2, we implemented a variation of the double-CRISPR/Cas9 approach (32), in which two gRNAs are employed with the intent of replacing most of the coding region with the pyrimethamine resistance marker in natively Ku80positive parasite strains (45). gRNA sequences were chosen using the Eukaryotic Pathogen CRISPR guide RNA/DNA Design Tool (46). We modified an existing transient expression plasmid designed to express a single gRNA sequence and the Cas9 endonuclease to express two gRNAs as shown in Fig. 2E. Transient expression of the plasmid, together with a universal DHFR amplicon prepared by a simple PCR, results in two double-stranded breaks that can be repaired via NHEJ by the insertion of the DHFR amplicon. Treatment with pyrimethamine recovers the parasites that have incorporated the amplicon. Editing of the target gene was examined by PCR experiments utilizing oligonucleotide primer pairs that flank both gRNA target sites and primer pairs that bridge from outside the expected deletion and either end of the DHFR cassette, as outlined in Fig.  2F. Clones were classified based on evidence for a double cut, in which the intervening coding region was deleted and replaced with the DHFR cassette and/or other DNA, and single cut at either the 5Ј-or 3Ј-gRNA site and insertion of the DHFR and/or other DNA, or only an improper repair with no introduced DNA, as summarized for each gene in Table S1. All clones showed evidence consistent with incorporation of the DHFR resistance cassette at one or both of the double-stranded break sites. An expanded analysis of the results from three represen- Table 1 Toxoplasma glycogenes and their characteristics Shown is a summary of documented and predicted genes contributing directly and indirectly to the parasite glycome. Genes are organized according to the glycan type to which they are expected to contribute. Characteristics and evidence regarding their functions from this project and others are summarized.
We applied the double-CRISPR/Cas9 disruption method to genes from each of the major glycan classes but whose fitness scores ( Fig. 1) did not suggest inviability, as summarized in Table 1. Of the 68 glycogene candidates, dual gRNA plasmids were constructed for 31, and single gRNA plasmids were generated for three others (summarized in Table S1). Of the 31 for which dual-guide plasmid (pDG) constructs were available, 17 successfully edited the target gene, and eight were unsuccessful, as determined by PCR studies (Table S1). Six genes were not tested, usually because mutants became available from other studies during this project. Of the eight unsuccessful trials, each was independently repeated, with the same outcome, and three were retried with new pairs of gRNAs, with the same negative outcome (Table S1). Although the efficiency of successful insertions in drug resistant clones was generally Ͼ80% (14 of 17), others ranged from 6 to 63% (as enumerated in Table  S1), suggesting that unknown factors can influence gene editing; these might include lower than expected fitness values for fibroblast egress (which was not probed by the original fitness screen), effects on nonprotein coding gene functions, or local differences in chromosomal organization that affect gRNA accessibility. Of the 17 successful disruptions, 10 were full deletions engaging both gRNA sites, and seven resulted in insertions at only one of the gRNA sites. Fifteen of these were examined glycomically and for cell level effects using a plaque assay.

Assessment of cellular consequences
To examine cellular effects of disrupting a glycogene, we employed a 6-day monolayer infection model, in which spreading infection from an initial single cell generates a lytic plaque, whose area is a measure of successful cycles of parasite invasion, proliferation, and egress. As shown in Fig. 3, nine of the 13 strains examined exhibited significant deficiency in this setting. These included mutations predicted to affect N-or O-glycosylation in the secretory pathway, O-glycosylation in the nucleocytoplasm, or sugar-nucleotide precursor supply, as detailed below. The effects of nine of the disruptions reasonably correlated with fitness scores from Sidik et al. (26) (Fig. 1), but they diverged for four genes. The effects of three disruptions were more severe than observed in the fitness assay. Although this might be attributable to unknown spurious genetic changes, it is notable that these three genes are associated with protein fucosylation: synthesis of GDP-Fuc or its utilization by the Spy or POFUT2 GTs (see below). Thus, the variations might be due to different parasite activities queried by the fitness and plaque assays. Although these correlations strongly implicate the importance of different glycosylation pathways for overall parasite success in cell culture, follow-up studies beyond the scope of this resource will be required to confirm causality for individual glycogenes. Fitness data for glycogenes for invasion of and growth in human fibroblasts were extracted from Ref. 26 and graphed according to rank order of fitness value for separate categories of glycosylation as indicated at the bottom. Fitness scores are based on a log 2 scale, and values below Ϫ2.5 (red dashed line) have increasing likelihood of being essential. Filled data points represent genes for which mutants are available: from this study (green) or from other studies (blue). Red filled circles, unsuccessful attempts from this study; unfilled circles, no mutants attempted. Genes for which pDGs are available or that have been mutated elsewhere are labeled in black; others are labeled in gray. Gene names correspond to listings in Table 1.

Assessment of glycomic consequences
We adapted established methods to evaluate glycomic consequences of glycogene editing. Because the parasite grows within human host cells, which are themselves reared in the presence of bovine serum, measures are required to minimize contamination with human and bovine glycans. Thus, we allowed tachyzoites to spontaneously lyse out of fibroblast monolayers in low (1%) serum medium, and washed the para-sites by centrifugation with serum-free medium. Because glycans are found in glycolipids and glycoproteins and as polysaccharides, which vary in their chemical characteristics, we adapted a classical scheme of Folch extraction to generate fractions enriched in glycolipids, GPI-anchored glycoproteins, and other glycoproteins and then applied chemical or enzymatic methods to release glycans from these fractions to be analyzed by MS (Fig. S5). The various chemical linkages of glycans to  (26), contains a BsaI site for convenient introduction of a synthetic duplex guide DNA (B). pU6 was modified by the addition of an XhoI site at the 5Ј-end of its guide DNA cassette to generate p2 (C) or at its 3Ј-end to generate p3 (D). Thus, the guide 2 cassette of p2 could be conveniently excised from and religated into p3 to generate the dual-guide plasmid pDG (E). F, gene replacement strategy. Co-transient transfection of pDG, expressing the gRNAs and Cas9, with the DHFR amplicon, results in the excision of the intervening genomic DNA (a) and ligation of the DHFR amplicon by nonhomologous end joining (b). The desired replacement clones are selected in the presence of pyrimethamine (c) and screened by PCR using gene-specific primers (GSP) flanking the excision sites.

Tools for Toxoplasma glycobiology
protein amino acids necessitate specialized methods for their release. In general, N-glycans are released enzymatically using PNGase F, O-glycans are released chemically using NaOH to induce ␤-elimination, and GPI-anchor glycans are released using aqueous HF that cleaves phosphodiester linkages. For monosaccharide glycans that are too short to recover by current methods, we Western blotted the carrier glycoproteins with antibodies and lectins; alternatively, glycans might be analyzed as glycopeptides by MS after trypsinization.

N-Glycans
N-Glycans are present on dozens if not hundreds of proteins that traverse the Toxoplasma secretory pathway (17,18,47), although the density of N-sequons to which they are linked is less than in most eukaryotes (13). We identified genes expected to assemble a Glc 3 Man 5 GlcNAc 2 -P-moiety on the P-Dol precursor (Table 1), in accord with previous studies (48). N-Glycans are typically extensively processed after transfer to the protein, but only enzymes that remove Glc residues were detected in Toxoplasma (Table 1). Notably, the processing ␣-mannosidases and UDP-Glc-dependent ␣-glucosyltransferase, found in most eukaryotes, are absent. Together with the absence of other genes likely due to secondary loss, N-glycans are not expected to have a role in chaperone-assisted protein folding and quality control (49,50). Studies indicate that N-glycosylation is important for parasite motility and invasion (47,51), consistent with the low fitness scores (ϽϪ2.5) for four of the five enzymes that assemble the Man 5 GlcNAc 2 -core ( Fig. 4A and Table 1A).
nLC-MS n -based glycomic profiling was employed to interrogate consequences of N-glycogene disruptions. Analysis of unfractionated tachyzoites identified three prominent N-glycan-like species, H8N2, H7N2, and H6N2 (Fig. 4B), and other minor species, including H9N2, H5N2, and H4N2 ( Fig. S6A), where H indicates the number of hexose and N the number of HexNAc residues. H8N2 was initially expected to represent Glc 3 Man 5 GlcNAc 2 , with H7N2 and H6N2 being processed forms lacking one or two Glc residues. The absence of complex, hybrid, and paucimannose N-glycans confirmed minimal contamination from host or serum glycans (Fig. S6B). However, the H6N2 species elutes at the same time both in RH and in a mutant disrupted in alg5 (Fig. 4), the enzyme that generates the Dol-P-Glc precursor required for glucosylation. H7N2 and H8N2 were absent from alg5⌬ cells, suggesting that they instead consisted of H6N2 with one or two Glc caps, respectively. To address the identity of the Hex residues in H6N2, the samples were treated with jack bean ␣-mannosidase, which cleaves ␣-1-2/3/6-linked mannoses. H6N2 from both RH (parental) and alg5⌬ cells, which are likely identical owing to co-elution of both of their anomers from the C18 column, were each converted to H1N2 (Fig. 4, B and C), indicating that the extra Hex residue is an ␣Man. Furthermore, ␣2-mannosidase converted the H6N2 to H3N2, indicating that the extra ␣Man is 2-linked. After treatment with either ␣-mannosidase, H8N2 appeared to convert to H7N2 (similar elution time as the original H7N2), and H7N2 appeared to convert to H6N2 (different elution time compared with original H6N2), which suggests that the extra ␣2Man occurs in both of these species. This conclusion assumes that jack bean ␣-mannosidase, at the enzyme concentration used, can only access the ␣6Man residue after the ␣3-arm is partially trimmed (52), as can only occur in the absence of the ␣Glc caps.
In this new model, we propose that Toxoplasma assembles an unusual Glc 3 Man 6 GlcNAc 2 -linked precursor (Fig. 4A), and Clones were analyzed using a plaque assay on confluent HFF monolayers, and plaque areas were measured 6 days after infection. A, image comparing plaques from RH (parental) and alg8⌬ strains. B, dot plots from pairwise comparisons of mutant and RH strains. Gene names are color-coded to indicate their glycosylation pathway association (red, N-glycans; green, other secretory pathway glycans; purple, cytoplasmic glycans; black, precursor assembly), and fitness scores from Fig. 1 are indicated in parentheses. To facilitate comparisons between trials, the mean areas of RH plaques were normalized to 1, and the respective mutant plaque areas were scaled accordingly (only one representative RH experiment is graphed for clarity). Mean values Ϯ 2 S.D. values (95% confidence interval) are shown with red bars. Statistical significance is shown at the top. Data were pooled from three independent trials, which each independently conformed to the same trend.

Tools for Toxoplasma glycobiology
Tools for Toxoplasma glycobiology that the 6th Man residue is ␣2-linked to the ␣6Man, although linkage to the ␣3Man or to the nonreducing end ␣2Man cannot be excluded. The responsible ␣ManT is unknown, but it might be Alg11, which applies the other two ␣2Man before the precursor flips from the cytoplasm to the lumen of the rER, or might be attached after flipping by the GPI-anchored PigB, which also applies an ␣2Man to an ␣6Man. This model is consistent with Hex 9 GlcNAc 2 being the largest and most abundant lipid-linked oligosaccharide in Toxoplasma (53). After transfer to protein by the oligosaccharyltransferase (33), partial trimming down to H6N2 is consistent with evidence for ␣-glucosidase-I and ␣-glucosidase-II genes ( Fig. 4A and Table 1A). The minor H9N2, which was inconsistently observed, is evidently converted to H8N2 by jack bean ␣-mannosidase, and the minor H5N2 and H4N2 species were eliminated by treatment with jack bean ␣-mannosidase, indicating that they are canonical Man 5 GlcNAc 2 and Man 4 GlcNAc 2 species that might have resulted from processing by lysosomal hydrolases, as described in Dictyostelium (54). We detected no effect of alg10⌬, expected to assemble the third, ␣2-linked, Glc, consistent with its efficient removal by ␣-glucosidase-1 (Fig. 4A), as expected. Our analysis of the putative alg8 mutant yielded a reduced level of H8N2 relative to H7N2 and H6N2 and a small plaque phenotype (Fig. 3). This difference suggested that the genetic lesion, which consisted of an insertion of DHFR in a poorly conserved region of the 5Ј-end of the protein-coding region, likely created a hypomorphic rather than null allele. We found no evidence for host-derived N-glycans (53), but our spontaneously lysed extracellular parasites could potentially have turned over former N-glycans after exiting host cells. We found no evidence for ␣3-linked core Fuc, because results using PNGase A (which cleaves N-glycans containing this modification) and PNGase F were indistinguishable (data not shown). Interestingly, the genome harbors a predicted ␤2GlcNAcT (TGGT1_266310; Table 1E) that initiates the processing of N-glycans to complex forms in metazoan (55), and this gene potentially shares a promoter with a predicted ␣HexT (TGGT1_266320), suggesting an associated function. Although disruption of these genes did not affect the N-glycans detected, their existence suggests the possibility of further processing of a minor population of N-glycans, which might occur during growth, as evidenced by the small plaque phenotype of a TGGT1_266310⌬ strain (Fig. 3).
Thus, Toxoplasma proteins appear to be decorated with just three major species of N-glycans, each with an unexpected and unusual structure: Glc 2 Man 6 GlcNAc 2 , Glc 1 Man 6 GlcNAc 2 , and Man 6 GlcNAc 2 (Fig. 4A). N-Glycans of Cryptosporidium, a related apicomplexan, include just two forms, Hex 6 HexNAc 2 and Hex 5 HexNAc 2 , each of which lacks an extension on the short mannose arm and so likely represents Glc 1 Man 5 GlcNAc 2 and Man 5 GlcNAc 2 , respectively (56). Failure to trim the Glc residues is consistent with the absence of the quality control mechanism (50). Attempts to genetically disrupt assembly of the Man 6 GlcNAc 2 portion of the N-glycan precursor were unsuccessful, consistent with essentiality of its highly conserved Man 5 GlcNAc 2 core. Disruption of assembly of the terminal ␣Glc residues, which are retained on many N-glycans, was deleterious for the parasite. Thus, the ␣Glc termini contribute functionality to Toxoplasma N-glycans, which can in the future be investigated using pDG plasmids described here.

O-Glycan glycogenes associated with the secretory pathway
O-Glycans are also found on dozens if not hundreds of glycoproteins that are processed via the parasite rER and Golgi, based mainly on lectin capture studies (17,19). The genome encodes five pp-␣GalNAcTs (Table 1C and Fig. 5A) that initiate mucin-type O-glycan assembly (35) (i.e. glycans that are linked via ␣GalNAc residues to Thr/Ser residues that are often clustered in hydrophilic amino acid repeats). In addition, two genes (Table 1C and Fig. 5B) are predicted to assemble a second type of O-glycan (Glc␤1,3Fuc1␣-) involved in quality control of thrombospondin type 1 repeat (TSR) domains and proteinprotein interactions in animals (59). Four Golgi-associated GTs of unknown function are also predicted (Table 1E and Fig. 5D), although as discussed above and below, these may contribute to N-glycosylation of GPI synthesis. With fitness scores ranging from Ϫ1.9 to 1.4 (Table 1), the contributions of these genes to tachyzoite biology is expected to be modulatory rather than essential for viability. Mutational analyses of the pp-␣GalNAcTs (gant genes) and a sugar nucleotide transporter that may selectively impact O-glycan assembly indicate functional roles in the integrity of semidormant bradyzoite cyst walls (22), which contain mucin-type glycoproteins, including Cst1 (60), and in tissue cyst persistence in mice (24).
Little is known about Toxoplasma O-glycans, so we initiated a global O-glycomic analysis of material released from total tachyzoite protein powder by classical reductive ␤-elimination (Fig. S7). Remarkably, only two O-glycan types could be confidently detected and assigned, with compositions of HexNAc 2 (N2) and Hex-deoxy-Hex (HdH). The assignments were verified by exact mass and diagnostic MS 2 and MS 3 fragment ions. These ions did not appear to be host-derived because (i) N2 was not detected in host cell samples, (ii) HdH is more abundant relative to other glycans in Toxoplasma versus host cell samples, and (iii) host O-glycans, including HN, H2N, H3N, and sialylated derivatives ( Fig. 5G) (data not shown), were not detected (Fig. S7). The N2 and HdH ions were observed at higher levels than the Man 6 GlcNAc 2 N-glycan that is inefficiently released by conditions of ␤-elimination. Monosaccharide modifications cannot be confidently assigned by this  (Table 1A). Glycans are represented according to the Consortium for Functional Glycomics system of nomenclature (57). B and C, N-glycans were prepared by treating tryptic digests of total parasite protein powders with PNGase F. The released glycans were permethylated and subjected to nLC on a C18 column followed by MS n in a linear ion-trap MS. Extracted ion chromatograms for all detected N-glycan-like, lithiated (Li ϩ ) composition (H1N2-H8N2) species are shown before and after digestion with jack bean ␣-mannosidase (␣1-2/3/6) or A. satoi (␣1-2) mannosidase to probe the identity of the hexose (H) species. The two peaks observed for each species are ␣and ␤-anomers. Minor related N-glycan-like species are shown in Fig. S6A. B, RH (parental). C, alg5⌬ mutant, which is unable to form Dol-P-Glc, the Glc donor for terminal ␣-glucosylation.

Tools for Toxoplasma glycobiology
method because their small size precludes their efficient recovery and chromatography. These findings indicate a limited repertoire of O-glycosylation at the proliferative tachyzoite stage and document the selectivity of the method for parasite glycans.
The N2 species is potentially representative of a mucin-type O-glycan, and, based on studies using the lectins jacalin, Vicia villosa agglutinin, Dolichos biflorus agglutinin, and Helix pomatia agglutinin, expression of a GalNAc-GalNAc-structure was Shown are Tn antigen (GalNAc␣-Ser/Thr) and the N2 disaccharide, consisting of a HexNAc of uncertain identity (potentially GalNAc) linked to ␣GalNAc, inferred from MS glycomic studies and metabolic and lectin labeling studies (see "Results and discussion"). Five pp-␣GalNAcTs are predicted to catalyze formation of the Tn antigen, and T2 and T3 are confirmed. T2 is required for accumulation of N2, but the enzyme responsible for the second sugar is unknown, except that genes in green in D are not required. B, TSR-type Glc-Fuc-disaccharide and its documented GT genes (see Fig. 7C). C, GT for C-mannosylation. D, four genes encoding predicted sugar nucleotide-dependent GT that appear to be type 2 membrane proteins in the secretory pathway. The glycans to which they contribute are yet to be identified. E, GT for assembly of the oocyst wall ␤3-glucan, whose priming mechanism is unknown. F-H, O-glycans were released by reductive ␤-elimination, permethylated, and subjected to nLC on a C18 column followed by MS n in a linear ion-trap MS as in Fig. 4. Extracted ion chromatograms for N2 (drawn as GalNAc-GalNAc-) and HdH (drawn as Glc-Fuc-) species (lithiated adducts Ϯ70 ppm) and the DP4 standard are shown. F, parental strain (type 2 Pru) and two derivative mutant strains (pp-␣GalNAcT2 and pp-␣GalNAcT3). G, the parental strain (type 1 RH) and the derivative galE⌬ strain and an extract of host hTERT cells showing HN, HNSa (Sa ϭ sialic acid), and HdH, but lacking N2. H, RH and two derivative mutants predicted to encode the sequential addition of ␣Fuc (POFUT2⌬) and ␤Glc (glt2⌬ or ␤3GlcT⌬). I, Western blot analysis of total RH strain tachyzoites for the Tn-antigen, probed with mAb 5F4. The Coomassie Blue-stained gel confirms equal sample loading. Similar results were obtained with mAb 1E3, except that mAb-reactive proteins were differentially emphasized.

Tools for Toxoplasma glycobiology
previously proposed (Fig. 5A). Analysis of the O-glycomes of pp-␣GalNAcT2 (gant2⌬) and pp-␣GalNAcT3 (gant3⌬) mutants showed that N2 depended on Gant2, with no effect observed on the expression of the HdH species (Fig. 5F). N2 was detected in gant3⌬ extracts, although its apparent abundance relative to the HdH ion was consistently reduced. Furthermore, N2 was also dependent on the expression of galE (see below), the epimerase that generates the sugar nucleotide UDP-GalNAc precursor for the pp-␣GalNAcTs from UDP-GlcNAc (Fig.  5G). These results indicate that the reducing end N residue is GalNAc, but do not reveal the identity of the second N. The involvement of two genes predicted to encode Golgi Hex-NAcTs, TGGT1_266310 and TGGT1_207750 (Table 1E), has been excluded based on analysis of their O-glycans ( Fig. S7B) (data not shown). Thus, tachyzoites express only one major mucin-type O-linked oligosaccharide, a disaccharide that may correspond to the previously predicted core 5 structure (GalNAc␣1,3GalNAc␣1-Thr) based on lectin studies of bradyzoites (22). This model is strengthened by findings from metabolic labeling of a disaccharide on the GRA2 glycoprotein (61), that can be interpreted as GalNAc-GalNAc-based on the absence of GlcNAc and more recently available information that the jacalin lectin used to enrich this O-glycan binds nonreducing terminal GalNAc as well as Gal (62). However, further studies are needed to determine the identity, anomericity, and linkage position of the nonreducing terminal HexNAc.
To investigate the possibility of nonextended ␣GalNAc, referred to as Tn-antigen, we utilized anti-Tn mAbs developed from studies on human cancer (63). As shown in Fig. 5I, using two validated (see "Experimental procedures") anti-Tn mAbs, we observed that both type 1 (RH) and type 2 (Pru) strains expressed at least six proteins (A-F) carrying the Tn-antigen in the M r range of 20,000 -60,000. Proteins in this M r range have also been detected with various GalNAc-preferring lectins expected to be specific for O-glycans (22,19). mAb recognition depended on both Gant2 and Gant3 as well as, as expected, GalE. Because only one pp-␣GalNAcT is required to express Tn-antigen, these findings support the model of Tomita et al. (22), consistent with precedent in animals, that Gant2 is a priming pp-␣GalNAcT, and Gant3 is a follow-on pp-␣GalNAcT that adds to the number of structures. Because N2 is more dependent on Gant2 than on Gant3, N2 is likely to be assembled on the amplified number of Tn-antigen structures generated by Gant3.
The HdH ion is a candidate for the Glc␤1,3Fuc␣1disaccharide that has been reported to be linked to Ser/Thr residues at consensus positions of TSR domains in animal glycoproteins (59) and of Plasmodium CSP and TRAP proteins that also have the consensus peptide motif (64). MS 2 and MS 3 fragmentation indicated that the order of the sugars is Hex-dHex-(data not shown). Disruption of the Plasmodium homolog of pofut2, which catalyzes the addition of ␣Fuc to Ser/Thr, prevents assembly of the glycan (65). Similarly, our pofut2⌬ strain also blocked accumulation of HdH (Fig. 5H), demonstrating by analogy that the core deoxy-Hex residue is ␣Fuc. Two recent preliminary reports confirm this result in Toxoplasma (36,37). The genome harbors a CAZy GT31 gene (TGGT1_239752 or glt2; Table 1) with homology to animal B3GlcT, which catalyzes the addition of the ␤Glc cap at the 3-position of Fuc. Our glt2⌬ strain also failed to express the disaccharide (Fig. 5H), confirming its role in its assembly, and indicating conservation of the Glc␤1,3Fuc␣1-disaccharide in Toxoplasma (Fig. 5B). The precursor for the HdH species, O-Fuc, was also recently detected on TSR5 of MIC2 in a glycopeptide analysis (36). Both assembly mutants exhibit a small plaque phenotype (Fig. 3), suggesting conservation of a role in quality control of folding (59) of the several GRA proteins that bear the consensus motif in their TSRs.

GPI-anchor and GIPL-glycogenes
Numerous parasite proteins are linked to the cell surface via GPI-anchors. The glycan portion of Toxoplasma GPI-anchors conforms to the conventional Man 3 GlcNH 2 inositol core formed by most eukaryotes but is reportedly modified at the core Man by a 3-linked ␤GalNAc residue (58). Some GPI-anchor precursors also remain free as GIPLs and are distinguished by an additional ␣Glc 4-linked to the ␤GalNAc (Fig. 6A). In addition, there is strain variation with regard to the presence of the phosphoethanolamine and ␣Glc termini in GIPLs (66). Besides serving a structural role, GPI-anchors and possibly GIPLs are recognized by host cell TLR2 and probably galectins as co-receptors (21). Studies targeting either the GPI-anchor or the galectin implicate their participation in the infection of host cells. The novel Glc␣1,4GalNAc␤1-branch has been the subject of vaccine candidate and antibody-based diagnostics (67).
We developed a new MS-based method to analyze GPI-anchor glycans released from an enriched pool of GPI-anchored proteins, by saponification and HF to break the phosphodiester linkages at both ends of the glycan. Unexpectedly, our analysis of strain RH tachyzoites yielded only the linear Man 3 GnIn core lacking a side chain (Fig. 6B).
The high degree of conservation of the enzymes that assemble the canonical Man 3 GnIn backbone enabled confident prediction of the enzyme coding genes (Table 1B). The phosphatidylinositol synthase (pis) that assembles the initial precursor from CDPdiacylglycerol and myoinositol has been described (34). pis, the pigA GlcNAcT, and the pigM, pigV, and pigB ManTs have very low fitness values, Ϫ3.1 to Ϫ5.1 (Fig. 1), indicating essentiality of the complete backbone. Our attempts to disrupt the final ␣ManT (pigB) of the pathway were unsuccessful (Fig. 6A and Table S1), consistent with the fitness data. With respect to the GTs that mediate the formation of the disaccharide side chain, the Toxoplasma genome encodes two ␤HexNAcT-like GTs that are predicted to be Golgi-associated type 2 membrane proteins (TGGT1_266310 and TGGT1_207750) and one predicted ␣HexT that is a type 2 membrane protein (TGGT1_ 266320). These candidates, which have modest fitness values, were each successfully disrupted in our RH strain, but their effects on the assembly of the GPI-anchor glycan branch could not be evaluated owing to its absence in our studies (Fig. 6B).
In this first examination of Toxoplasma GPI-anchor glycans by MS, we confirmed the core backbone, but did not detect a previously described unique glycan feature. Nevertheless, we have enumerated a plausible genetic basis for the assembly of the glycan part of the anchor and GIPLs with their branch Tools for Toxoplasma glycobiology chains and developed tools for their analysis in future studies in other strains and under other conditions.

Other glycogenes associated with the secretory pathway
In addition to N-and O-glycosylation of secretory proteins, and their C-terminal glycolipidation, Toxoplasma also modifies a carbon atom of the side chain of certain Trp residues with ␣Man (Fig. 5C). Originally found in association with a multi-Trp sequence motif in animal TSRs, this monosaccharide and the consensus motif (59) are conserved on CSP and TRAP proteins in Plasmodium, a related apicomplexan. Both Plasmodium and Toxoplasma genomes encode a high-scoring homolog of Dpy19 (Table 1D) that mediates formation of this linkage from Dol-P-Man in worms, and the Toxoplasma enzyme was recently shown to harbor the same activity (68). Using our double-CRISPR plasmid, we achieved a single insertion near the C terminus of the predicted protein (Fig. 5C), but it is unclear whether this affects C-␣ManT activity. The fitness score (Ϫ2.4) indicates importance for growth, potentially related to quality control of secretion of the proteins carrying TSRs, and possibly in concert with Glc-Fuc-modifications (69). Because there is no method known to release C-Man, monitoring the outcome will depend on glycopeptide analyses beyond the scope of this project.
The oocyst wall, formed during the sexual cycle in the gut of the cat host, contains disulfide-rich proteins and a ␤3-glucan polysaccharide (Fig. 5E), and the genome encodes a predicted ␤3-glucan synthase (38,70). Its disruption had no effect on proliferation of the wall-less tachyzoites (Fig. 3), as expected. Deletion of ␤3-glucan synthase also has no effect on bradyzoite and tissue cyst formation in vitro (38). A previously described hint for the presence of chitin in the bradyzoite cyst wall is not supported by evidence for a chitin synthase-like sequence in the genome.
In addition to the free GIPLs, there is evidence for a class of glycosphingolipid-like molecules that can be metabolically labeled with tritiated Ser and Gal (71). They are so classified based on the resistance of the radiolabeled material to saponification and their sensitivity to known inhibitors of sphingolipid synthesis. The tritiated Gal was evidently incorporated as . GIPL and GPI-anchor biosynthesis glycogenes. A, free GIPLs and GPI-anchors share a related glycan linked to a diacylglycerol on the right. Glycan assembly proceeds stepwise from right to left along the upper arm by the enzymes predicted, followed by the lower arm (58). The GPI-anchor precursor may receive the lower arm ␤4GalNAc, whereas the GIPL may also receive an ␣4Glc (dashed line). Phosphoethanolamine is then conjugated to the terminal ␣2Man, and the resulting NH 2 group of the GPI can be conjugated to the polypeptide C terminus by a multisubunit transamidase. Potential fatty acyl modifications are not shown. B, the glycan region of GPI-anchored proteins enriched from RH strain tachyzoites was released by HF (after saponification), which cleaves phosphodiester bonds, subjected to re-N-acetylation and permethylation, and analyzed by MS n . m/z values corresponding to glycans containing one or two of the sugars of the predicted lower arm are indicated in gray, but were not detected in this study.

Tools for Toxoplasma glycobiology
GalNAc, which might be mediated by one of the two predicted ␤HexNAcTs found in the genome (Table 1E).

Cytoplasmic glycosylation glycogenes
Proteins of the nucleocytoplasmic compartment of eukaryotic cells are also potentially subject to glycosylation (72). Because these involve different GT genes and often require detection by different methods, this class of glycosylation is often overlooked.
A subunit of the SCF class of E3 ubiquitin ligases, Skp1, has been shown to be modified by an O-linked pentasaccharide (Fig. 7A). A closely related glycan has been previously characterized in the social amoeba Dictyostelium (73), where it has been shown to mediate cellular O 2 -sensing owing to the O 2 dependence of the prolyl hydroxylase that forms the hydroxyproline residue on which it is assembled (74,75). Because the linkage to Hyp is alkali-resistant, the glycan was characterized on tryptic glycopeptides and appears to be restricted to a single protein. The Toxoplasma glycan is assembled by four GT genes (Table 1G). One (pgtA) consists of two GT domains, and two (glt1 and gat1) are unrelated to the Dictyostelium pathway GTs yet form a similar glycan (23,39). The Skp1 GT genes are not essential, but their disruptions by conventional double-crossover homologous recombination each generated small plaques. pDG plasmids and mutants for the first and final GTs in the pathway, gnt1 and gat1 (Table 1G), are available in this study.
Recently, Toxoplasma nuclei were shown to be enriched in a novel Fuc modification that could be detected cytologically with Aleuria aurantia lectin (AAL), which glycoproteomic analysis showed consisted of ␣Fucmodifications of Ser/Thr side chains (Fig. 7B) often in clusters (20). Our analysis pre-dicted four unknown cytoplasmic (or nuclear) GTs, one of which was homologous to the O-␤GlcNAcT (OGT) that modifies Ser/Thr residues of metazoan and higher plant nucleocytoplasmic proteins with O-␤GlcNAc (76). Disruption of TGGT1_273500 abolishes AAL labeling of nuclei (Fig. 7C) and of extracts analyzed by Western blotting (Fig. 7D), indicating that this gene is an ␣FucT (or OFT) that modifies nucleocytoplasmic proteins with O-Fuc, not O-GlcNAc. This family of related genes is deeply diverged into two groups that evidently emerged from an ancient prokaryotic gene duplication (77). OGTs reside in the Sec (Secret Agent) group, and the Toxoplasma OFT enzyme and Arabidopsis Spy (Spindly), which was recently also shown to be a bona fide OFT (78), reside in the Spy group. This suggests that members of the Spy clade, which are extensively distributed through protists and plants, mediate O-fucosylation of nucleocytoplasmic proteins, whereas members of the SEC clade, which is extensively distributed throughout plants and animals, mediate O-GlcNAcylation of nucleocytoplasmic proteins. The evolutionary and structural relatedness of the OFT to the OGT enzyme invites speculation of shared cellular roles. Spy appears to contribute functionality to the parasite according to the small plaque size phenotype (Fig. 3), although it did not contribute to fitness in the high-throughput assay (Fig. 1).
The cytoplasm is also the site of accumulation of amylopectin (Fig. 7E), a polymer of 4-linked ␣Glc with periodically spaced 6-linked ␣4-glucan chains (79). Amylopectin assembles into semicrystalline floridean starch granules that are highly induced during bradyzoite differentiation. The Toxoplasma genome contains sequences for UDP-Glc-dependent amylo-  Table 1I. F, other genes predicted to encode sugar nucleotide-dependent cytoplasmic (or nuclear) GTs.

Tools for Toxoplasma glycobiology
pectin assembly and disassembly that relate more to fungal and animal glycogen metabolism than to the ADP-Glc dependent pathways in plants (80). The branched polymer is assembled by a starch synthase and potentially two branching enzymes (Fig.  7E) and disassembled by a debranching enzyme, starch phosphorylase (40) (Table 1I), and exoglycosidases (not listed). A gene with sequence similarity to glycogenin, an ␣4GlcT that primes the synthesis of the related polymer glycogen in yeast and animals, is the Skp1 glycosyltransferase Gat1, which might have been the evolutionary precursor of glycogenin. 7 The starch assembly genes do not evidently contribute much to tachyzoite fitness, consistent with the induction of starch accumulation in bradyzoites and oocysts. Thus, this pathway was not targeted in this study.
The genome encodes a gene with an ␣,␣-trehalose-6-P synthase-like domain and a domain with a resemblance to the trehalose-6-phosphatase, glycohydrolase GH77 (␣4-glucanotransferase), and CBM20 starch binding domains (Table 1I). Genes for trehalose formation and turnover are widely distributed in eukaryotes (81), but the absence of a candidate for trehalase, or evidence for trehalose in Toxoplasma, suggests a potential role for this gene in starch metabolism. The function of this gene, which evidently contributes to fitness (Ϫ2.9), remains unassigned.
There are two reports of a glycolipid, characterized by a dihexosyl-moiety, possibly including a di-Gal epitope, on ceramide, diacylglycerol, and acylalkylglycerol lipids (82,83). This is reminiscent of galactolipid-type structures associated with plastids in plants, so they might potentially be associated with the apicoplast, the relict plastid that persists in Toxoplasma. The enzymatic basis for the addition of these sugars remains to be determined.
Three GT-like sequences predicted to reside in the cytoplasmic compartment (Table 1H and Fig. 7F) remain to be assigned. pDG plasmids (as well as mutants for TGGT1_288390 and TGGT1_217692) are available for their investigation, and disruption of the CAZy GT2 family sequence exhibits a small plaque phenotype (Fig. 3). Detection of their presumptive glycan products might require specialized approaches if their protein linkages are alkali-resistant as for the Skp1 glycan.

Sugar precursor glycogenes
Almost all glycosyltransferases rely on a sugar nucleotide or a Dol-P-sugar as their high-energy sugar phosphate donor for the transfer of the sugar to the acceptor molecule. The formation of these donors from metabolic precursors tends to be well-conserved. Fig. 8 (A-C) summarizes the metabolic pathways proposed to generate these compounds based on genomic analyses and experimental reports, and the genes are summarized in Table 1J. Based on information linking specific precursors to glycan types (Fig. 8D), genetic disruption of a precursor pathway offers an approach to investigate the functional significance of multiple glycosylation pathways simultaneously or a glycosylation pathway with redundant GTs.
Current information suggests that all sugars normally derive from Glc or Gal (Fig. 8B). Glc enters the cell via the plasma membrane-associated Glc transporter TgGT1 (84), and the ␣-anomer is converted to Glc-6-PO 4 (Glc6P) by hexokinase (41,44). Glc6P can be the source of all glycosylation precursors and is also the carbon source for central carbon metabolism via the classical Emden-Meyerhof glycolytic pathway and the pentose phosphate shunt. Alternatively, precursors for glucosylation and galactosylation may derive from the ␣-anomer of Gal via a salvage pathway (see below), as TgGT1 also transports Gal, Man, and GlcNAc (84). Finally, glutamine can contribute to sugar precursor assembly via Glc generated by gluconeogenesis (85).
The lipid-linked N-glycan donor contains GlcNAc, Man, and Glc, whose precursors derive from each of the three main branches of precursor assembly (Fig. 8, A-C). GlcNAc derives from UDP-GlcNAc, which is assembled from Glc6P using the classical hexosamine biosynthetic pathway involving five enzymes (Fig. 8A). The highly negative fitness score (Ϫ2.5 to Ϫ5.2) of several of the enzymes implies essentiality of the pathway, consistent with the importance of N-glycosylation. The ability to radiolabel GlcNAc-and GalNAc-containing glycans with tritiated GlcNH 2 (71,58) suggests transport and processing activities that are not explained by current knowledge. Owing to the likely necessity of the hexosamine biosynthetic pathway for viability (43), UDP-GlcNAc formation was not targeted in this study.
Man derives from GDP-Man, which is formed via three conventional enzymatic transformations from Glc6P (Fig. 8C). The strongly negative fitness scores (Ϫ2.7 to Ϫ5.0) also suggest essentiality, consistent with the fitness scores for the N-glycan ManTs. The very negative fitness score for pmi (Ϫ5.0) indicates that Glc6P is the only carbon source for this pathway. Thus, GDP-Man biosynthesis was not targeted in this study. The ability of parasites to be radiolabeled with tritiated Man is consistent with the activities of the Glc transporter and hexokinase to utilize this sugar in addition to Glc.
The GlcNAc and Man residues of the future N-glycan are incorporated while the lipid-linked precursor is oriented toward the cytoplasm and derive from sugar nucleotide donors (86). In contrast, the Glc residues are assembled after the precursor is flipped to the luminal face by a scramblase whose identity is unknown in any eukaryote. As is typical, the three ␣GlcTs, Alg6, Alg8, and Alg10, utilize Dol-P-Glc as donors. Dol-P-Glc is formed from UDP-Glc by Dol-P-Glc synthase (Alg5), whose Ϫ3.4 fitness score suggests importance of this modification, because there is no evidence that other GTs utilize this precursor. alg5 was successfully disrupted, which resulted in the absence of the Glc modifications and small plaque sizes as expected (see N-glycans above).
UDP-Glc is in turn assembled from Glc6P via two traditional enzymatic transformations (Fig. 8B). Conversion to Glc1P is expected to be mediated by a phosphoglucomutase (87), but a double-knockout of the two candidate genes does not affect viability (42). Therefore, this function might also be provided by phospho-GlcNAc mutase (pagm) or phosphomannomutase (pmm), as these enzymes are promiscuous in other parasites (88). There is a single UDP-sugar pyrophosphorylase (usp)

Tools for Toxoplasma glycobiology
candidate whose score (Ϫ0.9) suggests that it also does not contribute much to fitness. Because this varies from the negative fitness score for Dol-P-Glc synthase, the UDP-GlcNAc pyrophosphorylase may be promiscuous as in Arabidopsis (89), allowing for redundancy. Nevertheless, we have been unsuccessful in disrupting this gene using our double-CRISPR approach. Alternatively, Gal may be an alternate source of UDP-Glc via the so-called Isselbacher pathway found in plants and other eukaryotic parasites (90). The Glc transporter TgGT1 also transports Gal (84), and Toxoplasma has predicted galactokinase (GalK; fitness Ϫ2.2) and UDP-Glc/UDP-Gal epimerase (GalE; fitness ϩ1.7) enzymes. There is no good candidate for the Gal1P uridyltransferase of the classical Leloir pathway, suggesting that USP, the pyrophosphorylase that forms UDP-Glc from Glc1P, may promiscuously perform this role, as observed in a trypanosome and plants (91,92). The importance of this pathway was investigated by disrupting galE. This deletion resulted in a rather strong plaque defect (Fig. 3) despite a neutral fitness score, consistent with disruption of the Skp1 GalTs. However, GalE is also involved in the formation of UDP-GalNAc from UDP-GlcNAc, as demonstrated by effects on mucin-type O-glycosylation (Fig. 5G). A selective effect on UDP-Gal formation could be investigated by disrupting GalK, for which a double-CRISPR plasmid was generated. Thus, Toxoplasma may utilize redundancy in its phosphoglycomutases and UDP-sugar pyrophosphorylases, as observed in other eukaryotes, to ensure access to precursors within varied host cells. Alternatively, other genes may yet be identified, or a critical intermediate might derive directly from the host cell.
Assembly of GPIs and the C-Man of TSR domains depends on Dol-P-Man. The Dol-P-Man synthase gene was not targeted owing to its very negative fitness score (Ϫ4.3) and the likely essentiality of the GPI-anchor glycan (Fig. 6).
Fuc is known to be selectively applied to TSR domains in the secretory pathway, and to Skp1 and over 60 other proteins in the nucleus and cytoplasm. Toxoplasma lacks good candidate genes for the salvage pathway of GDP-Fuc assembly but does possess the two genes of the so-called conversion pathway. The second enzyme gene (fx), the epimerase (or GDP-Fuc synthase), was modified by an insertion near the 5Ј-end of its coding region, yielding a modest small plaque phenotype (Fig. 3). How-

Tools for Toxoplasma glycobiology
ever, the Glc-Fuc-structure was expressed normally (Fig. S7A), emphasizing the importance of biochemically confirming consequences of the genetic modification, especially when an insertion rather than the desired deletion is obtained.
Altogether, the metabolic pathways summarized in Fig. 8 are thus far consistent with the gene repertoire and biochemical studies. However, it is still not clear how parasites can be labeled with free tritiated Man and GlcNH 2 and how tritiated Gal can label GalNAc (71). Nevertheless, the gene assignments resemble those of the agent for malaria, the related apicomplexan Plasmodium (93). Two major differences are that (i) Toxoplasma also possesses enzymes for the utilization of Gal, consistent with the ability to metabolically label the parasite with tritiated Gal (71), and good candidate genes for GalK and GalE, and (ii) Toxoplasma has a Dol-P-Glc synthase to glucosylate its N-glycans within the rER lumen.
The sugar nucleotides described above are synthesized in the cytoplasm. Their access to GTs within the secretory pathway is afforded by specific transmembrane sugar nucleotide transporters (Fig. 8E), of which three are predicted in Toxoplasma (Table 1K). Nst1 has been demonstrated to be important for the availability of UDP-GalNAc for O-glycosylation and for UDP-GlcNAc (24). Recent studies demonstrated a role for Nst2 in the transport of GDP-Fuc (36). Nst3, which has a very negative fitness score (Ϫ4.9) and has yet to be studied, may contribute to the transport of UDP-Gal and/or UDP-Glc.

Conclusions
The Toxoplasma genome encodes at least 33 predicted GTs and one phospho-GT that contribute to the modifications of a variety of proteins and lipids and two that generate polysaccharides. 27 appear to be associated with the secretory pathway and nine with cytoplasmic glycosylation. An additional 21 genes appear to generate glycosylation precursors or mediate their transport into the secretory pathway. 10 other genes are associated with glycan or starch remodeling. This enzyme gene enumeration creates a new opportunity to explore gene-glycancell function relationships.
The range and abundance of Toxoplasma glycan classes, as profiled by MS and including findings from radiolabeling, mAb, and lectin studies, is more than previously appreciated but relatively limited compared with other protozoans, such as trypanosomatid parasites and the amoebozoan Dictyostelium (12, 94 -97). Furthermore, there is a notable lack of known anionic modifications of Toxoplasma glycans that are prominent elsewhere, although potential sulfation was not investigated in this study. These differences may be a consequence of the homogeneous environment experienced by this parasite, because when it is not within the parasitophorous vacuole of its host cell, it is enclosed within a protective cell wall.
N-Glycosylation is remarkably simple, with incomplete Glc trimming and little evidence for further remodeling of the novel Man 6 GlcNAc 2 -precursor, consistent with the absence of N-glycan-mediated quality control mechanisms in the parasite rER (50). O-Glycans of the Toxoplasma secretory pathway are also simple, consisting mainly (if not only) of HexNAc-GalNAc-and GalNAc-in the mucin class, and Glc-Fuc-and Fuc-in the TSR class. Toxoplasma pp-␣GalNAcTs are more related to animal pp-␣GalNAcTs (CAZy GT27) than typical protist pp-␣GlcNAcTs (CAZy GT60) from which they evolved (98), suggesting that they derived from a host by horizontal gene transfer. This interpretation is consistent with the uniquely apicomplexan presence of the TSR glycosylation genes relative to other protists and the evident absence of more diverse O-glycosylation. However, the four uncharacterized GT-like genes may allow for additional complexity at low abundance or at other stages of the parasite life cycle.
Distinct O-glycosylation pathways occur in the parasite nucleus and cytoplasm. Over 60 different nucleocytoplasmic proteins are modified by O-Fuc (20) in a pattern that is reminiscent of O-GlcNAc modifications that occur on thousands of animal nucleocytoplasmic proteins. This parallel is emphasized by the paralogous relationship, presented here, of their enzyme genes, which is certain to expose new mechanisms of protein regulation, as has unfolded from studies on O-GlcNAc. The cytoplasm also harbors five additional GT activities that assemble a pentasaccharide on a single protein (75), Skp1, which contrasts with the role of the O-Fuc (Spy) GT to add a single sugar to many proteins. The cytoplasm rather than the apicoplast plastid is the locus of the enzymes that process amylopectin (starch), and additional possible glycans are suggested by three unknown cytoplasmic GT-like genes.
Altogether, glycosylation has many clear roles in the parasite that remain to be explored. The double-CRISPR approach is designed to facilitate these studies in all WT strains of Toxoplasma (owing to NHEJ DNA repair). The validated guide DNAs are also adaptable to Ku80⌬ strains (lacking NHEJ) by appending 45-bp gene-specific homology arms to the ends of the DHFR amplicon, which may limit promiscuous insertions of the DHFR expression cassette elsewhere in the genome. All guide sequences, CRISPR/Cas9 plasmids, and disrupted strains as summarized in Tables S1 and S4 are available to the community; individuals may contact C. M. W. It is anticipated that these strains and tools will be useful for others to investigate contributions of glycosylation to their favorite cellular or protein functions.

Glycogene prediction
To search for candidate GTs that contribute to the Toxoplasma glycome, we expanded approaches previously implemented in Dictyostelium discoideum (95) and Dictyostelium purpureum (94). The predicted proteome (8460 proteins) from Toxodb.org was analyzed by (i) the SUPERFAMILY server, which assigns protein domains at the SCOP "superfamily" level using hidden Markov models; (ii) dbCAN, an automated carbohydrate-active enzyme annotation database that utilizes a CAZyme signature domain-based annotation based on a CDD (conserved domain database) search, literature curation, and a hidden Markov model; (iii) the Pfam database; and (iv) Bernard Henrissat (Marseille), who supervises the CAZy database (31). This yielded 39 GT-like coding regions. No gene candidates for DNA glucosylation as characterized in trypanosomatids (99) are apparent in apicomplexans. The search was expanded to include genes predicted to contribute to the synthesis of sugar Tools for Toxoplasma glycobiology nucleotide precursors and their transport into the secretory pathway, genes that contribute to N-glycan processing, and genes associated with processing of starch, yielding 28 additional glycogenes.
Parasite freezer stocks were prepared from a T25 flask culture of hTERT or HFF monolayer that was 30 -50% lysed. Scraped cells were collected and pelleted as described above and resuspended in 1-1.5 ml of ice-cold DMEM/FBS (1:1). After incubating for 10 min on ice, an equal volume of ice-cold DMEM/DMSO (Sigma, D2650) in a 4:1 ratio was added dropwise to the cell suspension. Cell suspensions were slowly frozen in 1-ml cryo-vials at Ϫ80°C and transferred to liquid N 2 .

Dual-guide plasmid construction and DHFR cassette amplification
Plasmid 2 and plasmid 3 were derived from pU6-Universal (26) (Addgene plasmid 52694) by the introduction of a novel XhoI site either upstream of the U6 promoter or between the guide RNA scaffold and the NsiI site, respectively. Plasmid 2 (p2) was constructed by replacing the PvuI-PciI fragment of pU6-Universal with a new PvuI-PciI fragment generated by PCR of pU6-Universal with plasmid 2 FOR and plasmid 2 REV (which contains an XhoI site) (Table S3), using pU6-Universal as the template. Similarly, plasmid 3 (p3) was constructed by replacing the HindIII-NsiI fragment with the HindIII-NsiI fragment of a PCR amplicon generated using plasmid 3 FOR and plasmid 3 XhoI (which contains an XhoI site). All enzymatic manipulations of plasmid DNA were performed as described by the manufacturer, and all PCR amplification reactions were performed with Q5 DNA polymerase (New England Biolabs, M0491L). p2 and p3 were prepared in dam Ϫ /dcm Ϫ competent Escherichia coli (New England Biolabs, C2925I). Guide sequences were selected using the Eukaryotic Pathogen CRISPR guide RNA/DNA Design Tool (EuPaGDT (46); http:// grna.ctegd.uga.edu) 8 based on the calculated efficiency score against GT1 genome (and/or Me49 genome) in the latest available version of ToxoDB (http://toxodb.org/toxo/) (111). 8 Guide sequences for each gene were cloned into p2 and p3, as described in the Addgene protocol with minor modifications. Briefly, annealed synthetic "top" and "bottom" ssDNA guide oligonucleotides flanked by BsaI-compatible 4-nucleotide overhangs were directionally cloned, and insertion was confirmed by PCRs using the "top" strand guide oligonucleotide and a downstream primer, Plasmid 1 REV. pDG plasmids were constructed by transferring the XhoI-NsiI fragment of p2 into p3 that was double-digested with XhoI and NsiI (Fig. 2). Plasmids were prepared in transformed Top10 cells (Thermo Fisher Scientific) using a ZR Plasmid Miniprep-Classic (Zymo Research, D4016) kit. Finally, the inserted guide sequences were confirmed by sequencing the pDG plasmid with primers gRNA For and gRNA Rev (Table S3). All available plasmids are referred to in Table S1 and defined in Table S4.

Transfection and cloning
Transfection was performed as described (100). Briefly, 10 million freshly prepared intracellular parasites were electroporated with 1 g of DHFR expression amplicon with or without 10 g pDG, or no DNA, in Cytomix Buffer plus 100 mM each of ATP and GSH in a final volume of 400 l in a 2-mm-wide cuvette (Bulldog BIO, 12358 -346), at a setting of 1.5 kV, 25 microfarads in a Gene Pulser Xcell (Bio-Rad). After 10 min of incubation at room temperature, treated parasites were diluted into 5 ml of Complete Medium with 10% FBS and inoculated onto confluent HFFs in a T25 flask. After standard incubation overnight, the medium was replaced with 5 ml of Complete Medium with 10% FBS containing 1 M pyrimethamine (Sigma, 46703) for selection. If the majority of HFF cells started lysing before the mock-transfected parasites died, the parasites were transferred to another T25 flask with hTERT monolayer for further selection in Complete Medium with 1% FBS containing 1 M pyrimethamine until no more visibly viable mock-transfected parasites were observed.
Pyrimethamine-resistant clones were isolated by limiting dilution on confluent HFFs in Complete Medium with 1% FBS containing 1 M pyrimethamine. Six serial 2-fold dilutions starting with 25 parasites/well were prepared in octuplet in a 96-well plate and grown for 6 -10 days. Single clones were expanded in hTERT monolayers in T25 flasks for genomic DNA isolation and freezer stock preparation.

Analysis of T. gondii glycogene loci
Genomic DNA was prepared from extracellular parasites using a Quick DNA Miniprep Plus kit (Zymo Research, D4069), eluted in 100 l of DNA Elution Buffer, and stored at 4°C in the DNA Elution Buffer. Q5 DNA polymerase was used to PCRamplify 10 -20 ng of genomic DNA with primer sets flanking expected double-stranded break sites (Table S2) and near the 8 Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.

Tools for Toxoplasma glycobiology
ends of the DHFR amplicon (Table S3) to characterize the disruption of glycogene loci.

Plaque assay
Plaque assays were performed on 100% confluent, tightly packed HFF monolayers under 3 ml of Complete Medium, 1% FBS without pyrimethamine in a 6-well plate. Freshly lysed-out intracellular parasites were inoculated at 100 cells/well, and the medium was replaced after 3 h by aspiration. On day 6, cultures were aspirated and rinsed three times with 3 ml of PBS (Corning, 21-040-CV) before fixation with Ϫ20°C 100% methanol for 5 min followed by air drying. The monolayer was stained with 1 ml of 2% (w/v) crystal violet, 0.8% (w/v) ammonium oxalate in 25% ethanol for 5 min at room temperature, rinsed three times with PBS, and air-dried. Mutant and control (RH) strains were always compared on the same plate. Monolayers were scanned by the Odyssey CLx Imaging System (LI-COR), and plaque areas were measured using ImageJ software (National Institutes of Health). Mutant plaque areas were referenced to control plaque areas from the same plate, and all control plaque areas were normalized to a value of 1 for presentation. Data were presented and statistically analyzed using GraphPad Prism version 6.

Preparation of parasites
The type 1 RH strain (WT) and mutant strains were grown on confluent hTERTs in Complete Medium supplemented with 1% FBS. The medium was collected after 36 -48 h, and spontaneously lysed-out tachyzoites were harvested by centrifugation (8 min, 2000 ϫ g, room temperature). The pellets were syringe-lysed (through a 27-gauge needle) three times, filtered through a 3-m Nuclepore filter (Whatman) followed by two rinses of DMEM to remove host cell debris. Pellets were resuspended and centrifuged three times with ice-cold PBS, frozen in liquid N 2 , and stored at Ϫ80°C. 125 mg wet weight of cells (ϳ5 ϫ 10 9 cells) were typically recovered from 14 T175 flasks.

Preparation of delipidated total cell pellets and Folch partition
Cell pellets (ϳ5 ϫ 10 9 cells) were delipidated, and total protein powder was prepared as described (101). Briefly, cell pellets were disrupted by Dounce homogenization in a mixture of icecold MeOH and H 2 O. The resulting suspension was transferred to a screw-top 8-ml glass tube and adjusted with water and CHCl 3 to a final ratio of CHCl 3 /MeOH/H 2 O (4:8:3, v/v/v). The extract was incubated for 3 h at room temperature with endover-end agitation. The insoluble proteinaceous material was collected by centrifugation and re-extracted three times in CHCl 3 /MeOH (2:1, v/v) and twice in CHCl 3 /MeOH/H 2 O (4:8:3, v/v/v). The final pellet was dried under a stream of N 2 , washed with acetone, dried, weighed (typically ϳ25-30 mg), and stored at Ϫ20°C until analysis. The clarified lipid extract was subjected to Folch partitioning (102). The resulting aqueous fraction (GIPLs) and organic fraction (other lipids) were dried under a stream of nitrogen and stored separately at Ϫ20°C.

Enrichment in GPI-anchored proteins and GPI-glycan core release
GPI-anchored proteins were enriched from delipidated pellets by partitioning in butan-1-ol (n-BuOH, 99%; Alfa Aesar, Tewksbury, MA) as described (103,104). Briefly, 20 mg of protein powder was extracted three times with 4 ml of H 2 O saturated with n-BuOH (ϳ9%) for 4 h at room temperature. The n-BuOH-saturated water fractions were pooled and extracted with an equal volume of n-BuOH. After centrifugation for 15 min at 3500 ϫ g at room temperature, upper butanolic and lower aqueous phases were separately dried and stored at Ϫ20°C. Butanolic fractions, enriched in GIPLs, were saved for future analysis. Aqueous fractions, containing GPI-anchored proteins, were treated with 100 l of 0.5 M NH 4 OH (in H 2 O) for 6 h at 4°C to release fatty acids. The sample was then dried under a N 2 stream, redissolved in H 2 O, and dried again twice. The glycan core was released by treatment with 50 l of 48% aqueous HF at 4°C for 12 h (105), resuspended in 500 l of 0.05% TFA (Thermo Fisher Scientific, 28904) in H 2 O, filtered through a C 18 -SepPak to capture protein and lipid, and dried under vacuum. The sample was subjected to N-acetylation by the addition of 10 l of pyridine (Sigma, 360570) and 50 l of acetic anhydride (Sigma, 539996) in 0.5 ml of MeOH, incubation for 30 min at room temperature (106), and drying under a stream of N 2 .

Preparation of N-linked glycans
N-Linked glycans were prepared as described (101). 2-5 mg of protein powder was resuspended in 450 l of trypsin digestion buffer (20 mM NH 4 HCO 3 , pH 7.9, 1 M urea) with brief bath sonication. 25 l each of 20 ng/l trypsin (Sigma, T8003) and 10 g/l chymotrypsin (Sigma, C4129) were added at a final ratio of ϳ1:20 protease/protein and incubated for 18 h at 37°C. The reaction was boiled for 5 min, cooled, and centrifuged to remove insoluble material. The supernatant was adjusted to 5% (v/v) acetic acid (HAc) and loaded onto a Sep-Pak C 18 cartridge (Sep-Pak Vac 1cc C 18 cartridge, Waters, Milford, MA), which had been pre-equilibrated by washing with Ͼ3 ml of 100% acetonitrile and Ͼ3 ml of 5% HAc, and washed with 10 column volumes of 5% HAc. Glycopeptides were eluted stepwise, as described (107), with 2 volumes of 20% 2-propanol in 5% HAc, 2 volumes of 40% 2-propanol in 5% HAc, and 3 volumes of 100% 2-propanol. The 2-propanol fractions were pooled, evaporated to dryness by vacuum centrifugation, resuspended in 50 l of 20 mM sodium phosphate buffer (pH 7.5), and digested with PNGase F (New England Biolabs, P0705S) for 18 h at 37°C. Reaction mixtures were evaporated to dryness, resuspended in 500 l of 5% HAc, and loaded onto a Sep-Pak C 18 cartridge column. The column run-through and an additional wash with 3 column volumes of 5% HAc, containing released oligosaccharides, were collected together and evaporated to dryness. N-Glycans were alternatively released with PNGase A (108) and processed similarly. Core ␣3-linked Fuc, which interferes with PNGase F action, has not been detected on Toxoplasma N-glycans, consistent the absence of candidate ␣3FucT-encoding genes in the genome.

Preparation of O-linked glycans
O-Linked glycans were obtained by reductive ␤-elimination as described previously (109) with minor modifications. Briefly, ϳ5 mg of protein powder was resuspended with bath sonication in 500 l of 50 mM NaOH (diluted from 50% stock; Fisher, SS254-500) containing 1 M NaBH 4 (Sigma, 213462-25G) and incubated for 18 h at 45°C in an 8-ml glass tube sealed with a Teflon-lined screw top. The reaction mixture was then neutralized with 10% HAc and desalted by applying to a 1-ml bed of Dowex 50W-X8 (H ϩ form; Sigma, 217506) and washing with 5 bed volumes of 5% HAc. The flow-through fraction, containing released oligosaccharide alditols, was evaporated to dryness. Borate was removed as an azeotrope with MeOH by adding 0.5 ml of 10% HAc in MeOH, drying under an N 2 stream at 42°C, and repeating three additional times. To remove residual peptide and reagent contaminants, the dried material was resuspended in 500 l of 5% HAc and loaded onto a C 18 -SepPak as above. Released oligosaccharide alditols were recovered by collecting the column run-through and an additional 2 ml of wash with 5% HAc. The run-through and wash were combined and evaporated to dryness.

Mass spectrometry of permethylated glycans
Direct infusion-The released glycans from the original 5-20 mg of protein powder were permethylated, as described (110), and the final CH 2 Cl 2 phase was dried down under N 2 and redissolved in 100 l of MeOH. 10 l of the glycan sample was combined with 35 l of MS Buffer (1 mM NaOH in 50% MeOH) and 5 l of 13 C-permethylated isomaltopentaose (DP5), as an internal standard at a final concentration of 0.2 pmol/l. Samples were directly infused (1 l/min) into an Orbitrap mass spectrometer (LTQ Orbitrap XL or Orbitrap Elite, Thermo Fisher Scientific) using nanospray ionization in positive ion mode, with the capillary temperature set to 200°C. The full Fourier transform mass spectrum was collected, and the top 10 most intense peaks were selected for MS 2 fragmentation using collision-induced dissociation (CID) at 40% collision energy. Once an ion was fragmented, the m/z was automatically put on an "exclusion list" to prevent reanalysis in the run. MS 2 product fragments matching a preselected list (Table S5) were selected for MS 3 CID fragmentation at the same collision energy.
Nano-LC-MS/MS-10 l of permethylated glycans from above were combined with 2 l of an internal standard mix ( 13 C-permethylated isomaltose series (DP4, DP5, DP6, and DP7) at a final concentration of 0.2 pmol/l each) and 38 l of LC-MS Buffer A (1 mM LiAc, 0.02% HAc). 5 l was injected into a PepMap Acclaim analytical C18 column (75 m ϫ 15 cm, 2-m pore size) maintained at 60°C in an Ultimate 3000 RSLC (Thermo Fisher Scientific/Dionex), which was coupled to a Thermo Scientific Velos Pro dual-pressure linear ion-trap mass spectrometer. For N-glycans, after equilibrating the column in 99% LC-MS Buffer A and 1.5-min ramp up to 45% LC-MS Buffer B (80% (v/v) acetonitrile, 1 mM LiAc, 0.02% HAc), separation was achieved using a linear gradient from 45 to 70% Buffer B over 150 min at a flow rate of 300 nl/min. For O-glycans, after equilibrating the column in 99% LC-MS Buffer A and a 1.5-min ramp to 30% Buffer B, separation was achieved using a linear gradient from 30 to 70% Buffer B over 150 min at a flow rate of 300 nl/min. The column was regenerated after each run by ramping to 99% Buffer B over 6 min and maintaining at 99% Buffer B for 14 min, before a 1-min ramp back down to 99% Buffer A. The effluent was introduced into the mass spectrometer by nanospray ionization in positive ion mode via a stainless steel emitter with spray voltage set to 1.8 kV and capillary temperature set at 210°C. The MS method consisted of first collecting a full ion-trap mass spectrometry (MS 1 ) survey scan, followed by MS 2 fragmentation of the top 3 most intense peaks using CID at 42% collision energy and an isolation window of 2 m/z. Dynamic exclusion parameters were set to exclude ions for fragmentation for 15 s if they were detected and fragmented five times within 15 s.

Glycan annotation
Structural assignments for the glycans detected at the reported m/z values were based on the compositions predicted by the exact mass (Ϯ0.1 atomic mass units) of the intact molecule, the presence of diagnostic MS 2 and MS 3 fragment ions that report specific glycan features, and the limitations imposed on structural diversity by known glycan biosynthetic pathways. The GRITS toolbox and its GELATO annotation tool (http:// www.grits-toolbox.org/) 8 were used to help identify glycan compositions using customized glycan databases and with the following tolerance settings: MS 2 accuracy 500 ppm, maximum number of adducts (charge states) ϭ 3.

Western blotting and immunofluorescence analysis
Whole-cell pellets (3 ϫ 10 6 cells/lane) were solubilized in Laemmli sample buffer containing 50 mM DTT, subjected to SDS-PAGE on 4 -12% preformed gels in MES buffer (NuPAGE Novex, Invitrogen), and Western blotted onto nitrocellulose membranes using an iBlot system (Invitrogen). Blots were blocked with 5 mg/ml BSA in PBS. For Tn antigen, blots were probed with anti-Tn mAbs 1E3 and 5F4 (63), which were validated for specificity by comparison of reactivity with WT and ␤3GalT (core 1)-knockout mouse colon samples generously provided by Dr. Lijun Xia (Oklahoma Medical Research Foundation) and a 1:10,000-fold dilution of Alexa-680 -labeled rabbit anti-mouse IgG secondary antibody (Invitrogen). For O-Fuc, blots were probed with biotinylated AAL and ExtrAvidin horseradish peroxidase (Sigma) and imaged on an ImageQuant LAS4000 imager (GE Healthcare) as described previously (20). Blots were imaged on a LI-COR Odyssey IR scanner and processed in Photoshop, with contrast maintained at ␥ ϭ 1. Parasites were analyzed by immunofluorescence with AAL as described previously (20).