Insertional tagging, cloning, and expression of the Toxoplasma gondii hypoxanthine-xanthine-guanine phosphoribosyltransferase gene. Use as a selectable marker for stable transformation.

A nonhomologous integration vector was used to identify the Toxoplasma gondii hypoxanthine-xanthine-guanine phosphoribosyl transferase (HXGPRT) gene by insertional mutagenesis. Parasite mutants resistant to 6-thioxanthine arose at a frequency of approximately3 x 10(-7). Genomic DNA flanking the insertion sites was retrieved by marker rescue and used to identify molecular clones exhibiting unambiguous homology to H(X)GPRT genes from other species. Sequence analysis of vector/genome junction sites reveals that integration of the linearized vector occurred with minimal rearrangement of either vector or target sequences, although the addition of filler DNA and small duplications or deletions of genomic sequences at the transgene termini was observed. Two differentially spliced classes of cDNA clones were identified, both of which complement hpt and gpt mutations in Escherichia coli. Kinetic analysis of purified recombinant enzyme revealed no significant differences between the two isoforms. Internally deleted clones spanning the genomic locus were used to create "knock-out" parasites, which lack all detectable HXGPRT activity. Complete activity could be restored to these knock-out mutants by transient transformation with either genomic DNA or cDNA-derived minigenes encoding both enzyme isoforms. Stable HXGPRT+ transformants were isolated under selection with mycophenolic acid, demonstrating the feasibility of HXGPRT as both a positive and negative selectable marker for stable transformation of T. gondii.

A nonhomologous integration vector was used to identify the Toxoplasma gondii hypoxanthine-xanthine-guanine phosphoribosyl transferase (HXGPRT) gene by insertional mutagenesis. Parasite mutants resistant to 6-thioxanthine arose at a frequency of ϳ3 ؋ 10 ؊7 . Genomic DNA flanking the insertion sites was retrieved by marker rescue and used to identify molecular clones exhibiting unambiguous homology to H(X)GPRT genes from other species. Sequence analysis of vector/genome junction sites reveals that integration of the linearized vector occurred with minimal rearrangement of either vector or target sequences, although the addition of filler DNA and small duplications or deletions of genomic sequences at the transgene termini was observed. Two differentially spliced classes of cDNA clones were identified, both of which complement hpt and gpt mutations in Escherichia coli. Kinetic analysis of purified recombinant enzyme revealed no significant differences between the two isoforms. Internally deleted clones spanning the genomic locus were used to create "knock-out" parasites, which lack all detectable HXGPRT activity. Complete activity could be restored to these knock-out mutants by transient transformation with either genomic DNA or cDNA-derived minigenes encoding both enzyme isoforms. Stable HXGPRT ؉ transformants were isolated under selection with mycophenolic acid, demonstrating the feasibility of HXGPRT as both a positive and negative selectable marker for stable transformation of T. gondii.
The protozoan parasite Toxoplasma gondii is a ubiquitous intracellular pathogen that has recently received considerable attention as a leading opportunistic infection associated with AIDS (Luft and Remington, 1992). Aside from its clinical significance, T. gondii is also a useful model for the study of intracellular protozoan parasites, due to its genetic accessibility and ease of culture (Pfefferkorn, 1990;Roos et al., 1994;Boothroyd et al., 1994). Molecular genetic tools for analyzing the haploid tachyzoite form of T. gondii parasites include transient and stable transformation vectors (Donald and Roos, 1993;Soldati and Boothroyd, 1993;Sibley et al., 1994;Messina et al., 1995;Soldati et al., 1996), gene knock-out strategies (Kim et al., 1993;Donald and Roos, 1994;Roos et al., 1994), and an insertional mutagenesis system (Donald and Roos, 1995) functionally analogous to the transposon tagging of genes in Drosophila, yeast, and bacteria (MacKay, 1993;Garfinkel et al., 1988;Berg et al., 1984) or Ti-plasmid gene tagging in plants (Feldman, 1991;Koncz et al., 1992).
Insertional mutagenesis in Toxoplasma is based on the parasite's dihydrofolate reductase-thymidylate synthase (DHFR-TS) 1 gene, mutated to encode a highly pyrimethamine-resistant enzyme Roos, 1993, 1994). Using this selectable marker, drug-resistant parasites containing transgenes integrated throughout the T. gondii genome (ϳ1-2 copies/parasite) are obtained at sufficiently high frequency to make insertional mutagenesis feasible Roos, 1994, 1995). The availability of genetic selection schemes for mutations in nonessential nucleotide salvage pathway enzymes (Pfefferkorn, 1978;Pfefferkorn and Pfefferkorn, 1978;Schwartzman and Pfefferkorn, 1981;Pfefferkorn and Borotz, 1994) has provided suitable targets to directly assess the frequency and randomness of chromosomal integration events. In the first validation of this system, insertional mutants defective in uracil phosphoribosyl transferase (UPRT) were obtained by 5-fluorodeoxyuridine (FUDR) selection, and the UPRT gene was subsequently cloned by the direct rescue of genomic DNA flanking a single-copy transgene insertion (Donald and Roos, 1995).
We were also interested in exploiting this approach to identify the parasite hypoxanthine-xanthine-guanine phosphoribosyl transferase (HXGPRT) gene, both as a possible target for antiparasitic chemotherapy (Ullman and Carter, 1995) and because of its potential as a safe, versatile selectable marker for T. gondii transformation. Toxoplasma tachyzoites deficient in HXGPRT activity can be selected in the presence of 6-thioxanthine (6-TX), whereas strains expressing HXGPRT activity can be selected in mycophenolic acid (MPA; Pfefferkorn and Borotz, 1994). In this report we describe (i) cloning of the T. gondii HXGPRT gene by insertional mutagenesis, (ii) mapping and analysis of transgene insertion sites, (iii) identification of differentially spliced cDNAs predicted to encode two HXGPRT isoforms, (iv) bacterial overexpression, purification, and preliminary characterization of the recombinant enzymes, (v) production of defined HXGPRT knock-out parasites, and (vi) the feasibility of both positive and negative selection for/against HXGPRT activity in transgenic parasites.

EXPERIMENTAL PROCEDURES
Parasite Reagents and General Procedures-Vectors, libraries, and parasite strains described herein are available free of charge through the National Institutes of Health AIDS Research and Reference Reagent Program (Ogden BioServices Corp, Rockville MD; obcaids@netcom.com). All experiments were carried out using a clonal isolate of the virulent T. gondii strain RH. Procedures for routine parasite maintenance, cloning, transfection, and insertional mutagenesis have been described elsewhere Donald and Roos, 1995). Negative selections for HXGPRT-deficient parasites were carried out using 80 g/ml 6-TX. Positive selections for HXGPRT ϩ parasites were carried out in 25 g/ml MPA supplemented with 50 g/ml xanthine to ensure sufficient substrate for the enzyme. 6-TX, MPA, and xanthine were obtained from Sigma.
Insertional Mutagenesis-Parasites were transfected with plasmid pTgDHFR-TSc3/M2M3 (Ogden Bioservices catalog number 2854), which harbors a cDNA-derived T. gondii DHFR-TS minigene mutated to produce a pyrimethamine-resistant allele Roos, 1993, 1994). This vector integrates throughout the parasite genome by nonhomologous recombination, conferring resistance to Ͼ1 M pyrimethamine Roos, 1993, 1995). ϳ2 ϫ 10 7 freshly harvested tachyzoites were electroporated with 50 g of undigested or HindIII-linearized plasmid and inoculated onto a monolayer of human foreskin fibroblast fibroblasts in 25-cm 2 T-flasks containing growth medium supplemented with 1 M pyrimethamine to enrich for parasites expressing the DHFR-TS transgene. After 2-3 days, parasites emerging spontaneously from the lysed host cell monolayer were inoculated into multiple 175-cm 2 T-flasks containing 80 g/ml 6-TX in addition to pyrimethamine. Flasks containing healthy parasites after 1-2 weeks in drug were passaged under continued selection for two further cycles of growth to ensure stable drug resistance and parasite tachyzoites cloned by limiting dilution .
Rescue of Genomic DNA Flanking Vector Insertion Sites-Genomic DNA downstream of the pDHFR-TSc3 transgene insertion points in 6-TX-resistant clones 2 and 3 was recovered by EcoRI digestion of total DNA, intramolecular ligation under dilute conditions, and direct transformation into Escherichia coli as described previously (Donald and Roos, 1995). The upstream junction from 6-TX R mutant 2 was subsequently cloned in the pCR-II vector (Invitrogen) as a 760-bp fragment by polymerase chain reaction (PCR) using an HXGPRT-specific sensestrand primer H3 (5Ј-GCCTACCAACTTTTCCCTAACGT-3Ј) upstream of the insertion point and an antisense primer homologous to the 5Ј end of the HindIII-linearized pDHFR-TSc3 transformation vector (TS49; 5Ј-GTTACACGTGACCACGCCAAAGT-3Ј). A similar PCR-based strategy was used to recover the upstream vector junction in 6-TX R clone 3, where multiple copies of the transgene integrated at the insertion site. Because the terminal transgenes in this array are in opposite orientation, pKS Reverse Primer was used to amplify both a 1.2-kb upstream and a 2.5-kb downstream PCR fragment using primers H3 (see above) or H11 (5Ј-GTGTTGTTCTCACCTGCAGTGTC-3Ј), respectively. Flanking sequences from both sides of the insertion junction in FUDR R mutant 3 (Donald and Roos, 1995) were recovered by marker rescue using the restriction enzyme NcoI, which does not cut within the insertional mutagenesis plasmid.
Genomic HXGPRT clones used in expression studies were constructed using restriction fragments isolated from overlapping clones and subcloned into pKS (Stratagene). Clone pTgHXGPRTg8 contains a 6.7-kb BamHI fragment including the entire known HXGPRT transcription unit (3.0 kb, including introns), along with 2.7-and 1.0-kb of 5Ј-and 3Ј-flanking DNA, respectively. The complete sequence of this region has been deposited in GenBank with accession number U10247. A gene targeting construct was produced by ligating a 5.8-kb 5Ј genomic fragment (extending from the 5Ј end of a clone upstream of the gene through a SalI restriction site within the HXGPRT coding sequence) to a 5.5-kb 3Ј SalI genomic fragment from downstream of the HXGPRT gene. The resulting 11.3-kb construct, pTgHXGPRTg11⌬Sal (Ogden Bioservices catalog number 2856), lacks an internal 1.5-kb SalI fragment, which is predicted to encode the COOH-terminal 82 amino acids of the HXGPRT enzyme.
pSK phagemids were recovered from purified ZAP-II cDNA clones encoding T. gondii HXGPRT, using the ExAssist/SOLR system (Stratagene). Two differentially spliced classes of cDNA predicted to encode two HXGPRT isoforms were identified. Minigene constructs pTgHXG-PRTcI and pTgHXGPRTcII were created by adding 554 bp of genomic sequence from upstream of the first known cDNA 5Ј end to clones from each cDNA class using a common KpnI site located 156 bp 5Ј to the presumed translational start site.
To facilitate the construction of heterologous promoter fusions, restriction sites were engineered into cDNA clones at the 5Ј end of the HXGPRT coding region by site-directed mutagenesis (Kunkel, 1985;McClary et al., 1989). For bacterial expression, oligonucleotide HM2 (5Ј-CTAACGTGAAGCCATATGGCGTCCAAACCC-3Ј) was employed to introduce an NdeI site (bold) at the initiation codon (italicized). Taking advantage of a fortuitous NdeI restriction site 7 bp downstream of the termination codon, a 702-bp NdeI fragment encoding HXGPRT isoform I and a 849-bp NdeI fragment encoding HXGPRT isoform II were cloned into the E. coli expression vector pBAce (Craig et al., 1991). For expression in transgenic parasites, oligonucleotide HM3 (5Ј-ACCAACTTTTC-CCTAGGATCCAGCAAAATGGCGTCCAAACCC-3Ј) was used to introduce a BamHI site (bold), preserving the immediate context of the initiation codon (italicized). BamHI to NdeI fragments encoding HXG-PRT-I or HXGPRT-II were placed under control of minimal promoter and 3Ј-untranslated regions derived from the T. gondii DHFR-TS gene 2 by replacing the chloramphenicol acetyltransferase cassette in pmin-CAT (Ogden Bioservices Corp catalog number 2850).
HXGPRT Complementation and Enzyme Assays in T. gondii-HXG-PRT activity was assayed by measuring the incorporation of [2,8-3 H]hypoxanthine, [8-3 H]xanthine, or [8-3 H]guanine (Moravek Biochemicals) by T. gondii-infected cultures of HGPRT-deficient human fibroblasts (Mi-Ten cells, from a patient with Lesch-Nyhan syndrome; American Type Culture Collection, CRL 1113). Transiently or stably transfected HXGPRT Ϫ knock-out parasites (see below) were inoculated into confluent host cell monolayers grown in 24-well tissue culture dishes at 37°C (10 5 parasite tachyzoites/well). 24 h post-infection, 1 Ci of radiolabel was added to each well, and incubation was continued for 4 h before the monolayers were fixed with trichloroacetic acid, rinsed, and counted as described (Pfefferkorn and Guyre, 1984;Roos et al., 1994). Experiments were carried out in triplicate in parallel with mock-infected and wildtype parasite controls. Crude sonicated extracts from transfected parasites were also assayed for HXGPRT activity by incubating labeled hypoxanthine, xanthine, and guanine with PrPP and quantitating the reaction products separated by ascending paper chromatography (Pfefferkorn and Borotz, 1994).

H(X)GPRT Sequence Alignment and Phylogenetic Analysis-Pre-
dicted H(X)GPRT protein sequences were downloaded from GenBank, EMBL, SwissProt, and PIR data bases using the NCBI "retrieve" Email server (Henikoff, 1993) and Network Entrez (Schuler et al., 1996). Sequences were aligned using MACAW and ClustalW software (Schuler et al., 1991;Thompson et al., 1994) with manual adjustment. Unambiguously aligned loci were analyzed using heuristic search algorithms in PAUP 3.0s software (Swofford, 1991) and subjected to bootstrap analysis (Hillis and Bull, 1993) using 100 replicates. Pfefferkorn and Borotz (1994) have demonstrated that T. gondii parasites deficient in HXG-PRT activity are viable, presumably synthesizing guanine nucleotides via AMP deaminase. We therefore sought to clone the parasite's HXGPRT gene by insertional mutagenesis, followed by selection with 6-TX (this subversive substrate is specific for the parasite enzyme, because the mammalian host cells lack xanthine phosphoribosyl transferase activity). Freshly harvested tachyzoites were electroporated with insertional mutagenesis vector pDHFR-TSc3/M2M3 Roos, 1993, 1995) and mutant clones resistant to 6-TX isolated as described under "Experimental Procedures." At least four independent 6-TX-resistant mutants were obtained from two mutagenesis experiments. Clone 1 was capable of growth in MPA ϩ xanthine and exhibited wild-type HXGPRT activity, but clones 2, 3 and 4 were sensitive to MPA ϩ xanthine, consistent with the phenotype of a known HXGPRT mutant (previously isolated by direct selection of chemically mutagenized parasites; Pfefferkorn and Borotz (1994)). Assays of crude cell extracts confirmed that these clones lacked HXGPRT activity (not shown).

Identification of the Toxoplasma HXGPRT Gene by Insertional Mutagenesis-Studies by
Clones 2 and 3 were produced using linearized vector DNA, which typically integrates as a single unrearranged unit, facilitating retrieval of flanking DNA by "marker rescue" through restriction of total genomic DNA, self-ligation, and direct transformation of E. coli Donald and Roos, 1995). Hybridization analysis indicated that an EcoRI rescue strategy should be feasible, and recombinant plasmids containing genomic sequences flanking the insertion points for 6-TX R clones 2 and 3 were obtained by direct transformation of E. coli with self-ligated, EcoRI-digested DNA. These clones contained T. gondii genomic DNA exhibiting high homology to H(X)GPRT genes from other organisms (see below). Probing bacteriophage libraries identified overlapping genomic clones and two distinct classes of differentially spliced HXGPRT cDNA clones, as shown in Fig. 1A.
The first intron of class I cDNAs (represented by five clones) was not eliminated in the class II cDNA (represented by a single clone). This intron is predicted to lie within the 5Јuntranslated region and hence should not affect the enzyme's amino acid sequence. (Introns within the 5Ј-untranslated region have been reported in T. gondii previously (Parmley et al., 1994).) Class II cDNA also contains an additional 147-nt exon within the second intron of class I. This exon is predicted to encode a 49-amino acid insertion near the amino terminus of the mature protein (asterisk in Fig. 2). All cDNA clones (including five shorter clones consistent with either class) were in perfect agreement with the genomic sequence except for polyadenylylation at the 3Ј terminus and elimination of intervening sequences; Southern blotting indicates the presence of only a single gene (not shown).
Conceptual translation of these cDNAs reveals that the T. gondii locus identified by insertional mutagenesis is predicted to encode proteins with molecular weights of 26,369 and 31,464 (for HXGPRT-I and HXGPRT-II, respectively). The predicted sequence is similar to H(X)GPRT enzymes from other species, as shown in Fig. 2A. Sequence conservation is highest within the putative PrPP-binding domain (Hershey and Taylor, 1986). Interestingly, the T. gondii sequence differs substantially from other H(X)GPRT enzymes in a signature sequence near the COOH terminus that is thought to be part of the purine binding domain (Eads et al., 1994): the eukaryotic consensus sequence Overlapping recombinant plasmids containing T. gondii genomic DNA clones p6TXR2 and p6TXR3 were recovered by EcoRI digestion of total DNA from insertional mutants 6-TX 2 and 3 and plasmid rescue of self-ligated fragments in E. coli Donald and Roos, 1995). These sequences were employed as hybridization probes to obtain cDNA and genomic DNA clones from libraries. Annotated genomic sequence of the entire region shown is available from GenBank with accession number U10247. Transcript splicing patterns deduced from comparisons of genomic and cDNA sequence data reveal two classes of differentially spliced mRNA predicted to encode enzyme isoforms differing by the presence or the absence of a single exon encoding 49 amino acids (striped box). Solid boxes indicate predicted coding sequence common to both isoforms, shaded boxes indicate 5Ј-and 3Ј-untranslated regions present in both cDNA classes, and introns are indicated by open boxes (intron I not excised in class II cDNA). Restriction sites: B, BamHI; K, KpnI; N, NdeI; P, PstI; R, EcoRI; S, SalI. After double-digestion of wild-type DNA with EcoRI and PstI, the 2.6-kb genomic KpnI-NdeI probe hybridizes with three restriction fragments and permits mapping of integration sites for 6-TX-resistant insertional mutants 2, 3, and 4 (B). A deletion lacking the 1.5-kb SalI fragment was used to knock out the endogenous HXGPRT locus by homologous recombination (see text). Open arrowheads indicate the location of PCR primers used to screen for deletions (H8 and H11) and to sequence insertion junctions (H3 and H11; used in conjunction with vector-specific primers).
with the sequence WIVGCCYDF in T. gondii. The loss of this otherwise well conserved region may explain the difficulty several groups have had in identifying the T. gondii HXGPRT gene by PCR. The novel exon present only in cDNA class II introduces 49 amino acids near the poorly conserved NH 2 terminus and shows no significant homology to other H(X)GPRTs (or any other sequence in the data base). Phylogenetic analysis of unambiguously aligned regions demonstrates that the predicted T. gondii HXGPRT sequence is most closely related to Plasmodium falciparum (Fig. 2B), also a member of the phylum Apicomplexan (Wolters, 1991;Gagnon et al., 1993;Cavalier-Smith, 1993). The unrooted tree shown in Fig. 2B is the single most parsimonious tree found and is identical to the consensus of the 114 shortest trees. All groupings were supported by bootstrap analysis, although the clustering of Trypanosoma brucei and Trypanosoma cruzi is relatively weak. The shortest tree in which the T. brucei branch is deep to a clade containing T. cruzi, Leishmania, and Crithidia (Fernandes et al., 1993;Vickerman, 1994) is only one step shorter than the tree shown, and this alternative topology is supported by 47% of bootstrap replicates.
Location and Fine Structure of Integration Sites-In order to map the location of transgene insertions, DNA from wild-type parasites and each of the four 6-TX R mutants was hybridized with a probe spanning the HXGPRT coding sequence (Fig. 1B). Digestion with both EcoRI and PstI cuts the wild-type HXG-FIG. 2. Toxoplasma HXGPRT alignments and phylogeny. A, protein sequences encoding H(X)GPRT enzymes from various species were downloaded from the SwissProt and GenBank data bases and aligned with the predicted amino acid sequence of T. gondii HXGPRT isoform I (bold type) as described under "Experimental Procedures." Lowercase letters in the human sequence indicate amino acids that are altered in Lesch-Nyhan syndrome or gout patients (references noted under SwissProt listing P00492); lowercase letters in the P. falciparum sequence indicate known strain differences. T. gondii residues found in at least one other sequence are underlined; residues conserved in the majority of species (or where only three alternative residues are observed) are indicated by apostrophes; carets indicate residues where only two alternative residues are observed; and asterisks indicate invariant residues. Putative PrPP and purine binding sites are based on sequence conservation and the human crystal structure (Hershey and Taylor, 1986;Eads et al., 1994). The former is highly conserved in all eukaryotic species, whereas the latter is divergent in the predicted T. gondii sequence. B, unambiguously aligned amino acids were employed to determine the single most parsimonious tree (for a larger data set than that shown in A) using the heuristic search algorithm of PAUP 3.0 s (Swofford, 1991). Branch lengths are proportional to distance in this unrooted tree; the scale bar indicates 20 steps (total tree length ϭ 639 steps). The numbers indicate the percentage of trees that support individual clades ( PRT gene into three restriction fragments of 1.38, 0.77, and 0.73 kb. Mutant 1 (which showed normal HXGPRT activity) retains the wild-type hybridization pattern. Each of the three HXGPRT-deficient mutants displays a disrupted band; however, transgene integration occurred within the middle (0.73kb) fragment in 6-TX R clone 2 and within the 3Ј (1.38-kb) fragment in 6-TX R clones 3 and 4. Thus the three transgene integrations occurred in at least two different locations within the HXGPRT locus.
To examine the structure of transgene integrations more precisely, junction regions from 6-TX R clones 2 and 3 and an unrelated integration at the UPRT locus (FUDR R clone 3; Donald and Roos (1995)) were cloned and sequenced, as shown in Fig. 3. All of these insertions were produced using HindIIIlinearized vector in order to facilitate cloning (see "Experimental Procedures"); no junctions have been sequenced from insertional mutants produced using circular plasmids (e.g., 6-TX 4). Insertions in 6-TX R clones 2 and 3 map to intron III and exon IV of the T. gondii HXGPRT gene, respectively (Fig. 1A). FUDR R clone 3 occurred within UPRT exon IV, near the intron IV boundary; as at the HXGPRT locus, other mutants (not precisely mapped) harbor integrations at different sites throughout the UPRT gene (Donald and Roos, 1995).
Small deletions (4 -31 nt; 4 nt of which were single-stranded DNA produced by HindIII digestion) were observed at vector termini, presumably the result of exonuclease activity prior to insertional ligation. Deletions of 3 or 6 nt were observed at the genomic integration site in clones 6-TX 2 and 3, respectively. Integration at the UPRT locus occurred at an endogenous Hin-dIII site and was accompanied by a 5-nt duplication. It is unlikely that this duplication was produced by filling in the ends of the transfected plasmid, because both ends of the integrated plasmid were truncated. We also consider it unlikely that this integration event was catalyzed by exogenous restriction enzyme, because HindIII was removed from digested plasmid by ethanol precipitation and washing prior to electroporation. The observed duplication in the UPRT locus insertion FUDR 3 may reflect the involvement of a Toxoplasma recombinase/polymerase system with the ability to match DNA substrate with partially homologous target sequence (Roth and Wilson, 1986;Merrihew et al., 1996). This is consistent with a small region of identity (10 out of 12 bp match) between the 3Ј end of the vector and the insertion point in 6-TX R mutant 3 (boxed regions in Fig. 3). No similarity was detected between the transfecting plasmid and the integration site in clone 6-TX 2.
Although neither the integrating plasmid nor the target locus were rearranged in any of the transgene integrations examined to date, all three transgenics incorporated an additional DNA sequence at the integration site, as has been observed in other systems (Brenner et al., 1984;Roth et al., 1991;Merrihew et al., 1996). Clone 6-TX 2 contains an addi- FIG. 3. Structure of pDHFR-TSc3 insertion junctions in three independent insertional mutants at two genetic loci. Mutagenesis vector pDHFR-TSc3 was recovered from parasite clone FUDR 3 (located at the T. gondii UPRT locus; Donald and Roos, (1995)) along with flanking genomic sequence on both sides, by plasmid rescue using NcoI (which does not cut within the vector). DNA fragments containing vector-genome junctions were recovered from insertion sites in 6-TX R mutants 2 and 3 (at the HXGPRT locus) by EcoRI-mediated marker rescue or by PCR amplification (see Fig. 1A and "Experimental Procedures"). For each insertion, the pDHFR-TSc3 vector is indicated on the upper line (in italics); the terminal HindIII site is enclosed in parentheses to reflect the single-stranded overhang produced by linearization at this site. For improved legibility only the HXGPRT sense DNA strand is shown, but 5 nt of the vector HindIII site is shown at each end. The native locus is shown on the bottom line in bold letters; uppercase letters indicate exon sequences, and lowercase letters indicate introns. The middle line presents sequence across the junctional domains of each insertion (the dotted line indicates the central portion of the 6.5 kb pDHFR-TSc3 vector, which is completely intact). Vertical lines (͉) indicate identity between the insertional mutant and either the native locus or the vector. Note that 4 -31 nt are lost from each end of the vector prior to integration. 5 nt at an endogenous HindIII site were duplicated at the integration site in mutant FUDR 3 (indicated by exclamation points), whereas 3-6 nt were lost at the genomic locus in mutants 6-TX 2 and 6-TX 3 (indicated by ⌬). Sources for short fragments of filler DNA (underlined) are indicated. Boxes indicate regions of similarity between the integration vector and the genomic insertion point in clone 6-TX 3. See text for further discussion.
tional CC dinucleotide of unknown provenance. Clone FUDR 3 contains a 30-nt fragment from the ampicillin resistance gene of the transfecting plasmid, integrated in the opposite orientation with respect to the transgene. Note that this DNA fragment is not derived from the particular plasmid molecule that integrated, because the integrated vector is completely intact (except for the loss of terminal nucleotides, as noted above). Clone 6-TX 3 includes an additional TTT and four fragments (63, 32, 70, and 38 nt in size) derived from various regions of the transfecting plasmid, integrated on both sides of the intact vector and in both orientations. All five of these integrated vector fragments (one in FUDR 3 and four in 6-TX 3) are perfect copies of sequence found in the transfected plasmid except for a single nucleotide mismatch in 6-TX 3 (indicated by a question mark in Fig. 3). No interesting sequence similarities were noted between the various fragment termini and the integration site.
Bacterial Complementation, Overexpression, and Purification of H(X)GPRT Isoforms-Plasmids encoding both of the observed classes of T. gondii HXGPRT cDNAs under control of a bacterial phoA promoter were transformed into both S606 and S609 E. coli lacking the corresponding bacterial phosphoribosyltransferase activities (Jochimsen et al., 1975). As shown in Fig. 4, after induction in low phosphate medium (Allen and Ullman, 1993), HXGPRT-I and HXGPRT-II were the predominant proteins expressed in S606 cells transformed with pBAce-HXGPRT-I and pBAce-HXGPRT-II, respectively. In S609 cells, both HXGPRT-expressing plasmids (but not the pBAce vector alone) conferred the ability to grow in adeninecontaining medium supplemented with either hypoxanthine, guanine, or xanthine. Neither the pBAce nor pBAce-HXGPRT transformants grew in adenine alone (without additional pu-rine), because exogenous adenine cannot serve as a source of guanylate nucleotides for E. coli (Jochimsen et al., 1975). However, all transformants grew when guanosine was added as a source of guanylate nucleotides, because adenine and guanosine can be salvaged to the nucleotide level by E. coli independent of HXGPRT activity. Expression of either T. gondii HXGPRT-I or HXGPRT-II complemented the hpt and gpt lesions in S609 E. coli, demonstrating enzyme activity and suggesting that hypoxanthine, guanine, and xanthine are all substrates of T. gondii HXGPRT.
Both recombinant T. gondii HXGPRT isoforms were purified by affinity chromatography over GTP-agarose columns, although yields were quite low (Fig. 4A). The HXGPRT-I isoform was further purified to virtual homogeneity by a combination of ion exchange chromatography, salt precipitation, and size exclusion chromatography (Fig. 4B). Both isoforms were completely stable for Ͼ2 months in the absence of PrPP at temperatures ranging from 4 to Ϫ80°C, unlike the T. brucei and T. cruzi counterparts Ullman, 1993, 1994). T. gondii HXGPRT isoforms I and II both catalyzed phosphoribosylation of hypoxanthine, guanine, and xanthine, but neither isoform recognized adenine as a substrate. Apparent K m and V max values were determined from Hanes plots, as shown in Table I. A K m value of 58.7 M for magnesium PrPP was also determined for isoform I. It is interesting to note that the ϳ10-fold higher K m toward xanthine as a substrate (relative to hypoxanthine or guanine) is partially compensated for by ϳ3-fold higher V max (or K cat ) values; overall enzyme efficiency is therefore comparable for all three substrates. Isoforms I and II exhibit very similar kinetics (the slightly lower V max and K cat values measured for isoform II may be attributable to difficulties in protein quantification due to the low yield of fractions eluted from GTP-agarose).
Transient Expression of HXGPRT in T. gondii Tachyzoites-To permit complementation and in vivo expression studies, an HXGPRT Ϫ knockout strain was produced by targeted homologous recombination. Previous studies have shown that homologous recombination occurs at high frequency when gene targeting constructs containing long stretches of contiguous genomic DNA are employed . A targeting plasmid was prepared using 11.3 kb of genomic HXGPRT sequence flanking a (deleted) 1.5-kb SalI fragment, as shown in Fig. 5 (pHXGPRTg11⌬Sal). The deleted fragment contains 650 bp from the 3Ј terminal exon, including sequences predicted to encode the NH 2 -terminal 80 amino acids. After transfection with this targeting construct and selection in 6-TX, 13 parasite clones were isolated and screened for the presence of the 1.5-kb SalI fragment using PCR primers flanking the deletion. 10 of the 13 clones yielded only the 0.5-kb fragment derived from the knock-out construct and not the 2-kb fragment observed in controls, suggesting that the wild-type locus had been deleted by allelic replacement (double cross-over). Four clones were chosen for further anal-  ysis. No hypoxanthine, guanine, or xanthine phosphoribosyl activity was detected in any of these clones (see below), and in each case the mutant allele was found at the wild-type locus by hybridization analysis (Fig. 6). Probing with the 1.5-kb SalI fragment deleted in the targeting construct confirms that this DNA has been completely lost from the genome (Fig. 6B). Recombinant HXGPRT plasmids used for complementation experiments in T. gondii are shown in Fig. 5. Four HXGPRT minigenes were constructed in which either the native HXG-PRT promoter or a heterologous promoter derived from the T. gondii DHFR-TS gene was fused to cDNAs encoding either HXGPRT-I or HXGPRT-II. As described under "Experimental Procedures," fusions were constructed so that constructs driven by the HXGPRT promoter retained the differentially spliced 5Ј-untranslated region of class I and II cDNAs, whereas DHFR-TS promoters were ligated immediately upstream of the putative coding regions. A genomic construct was also prepared, including the entire coding sequence plus 2.9 and 1.3 kb of 5Ј-and 3Ј-flanking DNA, respectively.
Twenty-four hours after inoculation of HGPRT-deficient host cells with electroporated HXGPRT Ϫ knock-out parasites, enzyme activity was assayed by measuring the incorporation of labeled hypoxanthine, guanine, and xanthine, as shown in Fig.  7 (left-hand panels). Uninfected host cells and cultures infected with HXGPRT knock-out parasites (or transfected with the HXGPRT knock-out construct) showed no detectable activity (Ͻ5% of parallel infections with wild-type parasites). In contrast, parasites transfected with any of the HXGPRT constructs yielded 18 -40% of wild-type incorporation of hypoxanthine, guanine, and xanthine. Taking into consideration decreased survival following electroporation (ϳ10%) and the ϳ50% efficiency of transient transformation (Donald and Roos, 1993;Roos et al., 1994), these levels of expression provide virtually complete restoration of wild-type HXGPRT activity (although we cannot assess cell-to-cell variability within the transiently transfected population).
Use of HXGPRT as a Positive Selectable Marker for Stable Transformation-Mycophenolic acid is a potent inhibitor of IMP dehydrogenase and therefore blocks guanine nucleotide synthesis from AMP, providing a potential positive selection for the reversion of HXGPRT Ϫ mutants to prototrophy. Because wild-type parasites are readily selectable from among 2 ϫ 10 6 HXGPRT Ϫ mutants in the presence of MPA ϩ xanthine (Pfefferkorn and Borotz, 1994), the potency of this selection is probably comparable with the use of antifolate selection for pyrimethamine-resistant DHFR-TS markers (Donald and Roos, 1993).
Transient resistance to 25 g/ml MPA ϩ 50 g/ml xanthine was observed in the first few days of growth after transfecting the knock-out mutants with HXGPRT constructs. In contrast to mock-transfected parasites, which showed no detectable growth, cultures inoculated with HXGPRT-transfected parasites completely lysed the host cell monolayer within 3-4 days. Plaque assays were used to measure the survival of transfected parasites under MPA ϩ xanthine selection (and to permit comparison with available DHFR-TS markers). Although transient expression levels decline in transfected parasites over a period of days , transient drug resistance can nevertheless be measured when strong selection is available by serially diluting transfected parasites directly into medium with and without drug and scoring the minute plaques visible after ϳ5 days (Donald and Roos, 1993). Stable drug resistance is measured by scoring the percentage of parasites emerging from heavily infected cell monolayers after primary inoculation for their ability to form mature plaques under Parasites were transfected with pHXGPRTg11⌬Sal, an 11.3-kb genomic subclone that spans the wild-type HXGPRT locus but lacks an essential 1.5-kb internal SalI fragment (see Fig. 5). A, genomic DNAs from four putative 6-TX-resistant knock-out clones (initially screened by PCR as described in text) were probed with a 6.5-kb BamHI fragment spanning the locus (insert from pHXGPRTg8; Fig. 5). B, blots were probed with the 1.5-kb SalI fragment lacking in the targeting construct. All four clones have replaced the wild-type BamHI fragment with a smaller fragment of the expected size and are missing the 1.5-kb SalI fragment, demonstrating perfect allelic replacement with the knock-out mutant.
continued drug pressure. The results from plaque assays of drug resistance frequencies are shown in Table II. All five of the HXGPRT constructs tested conferred ϳ5% survival frequency to transfected parasites in the first few days after electroporation. Stable transformation frequencies were considerably lower; ϳ8 ϫ 10 Ϫ4 of those parasites that exhibited transient resistance yielded an overall frequency of ϳ4 ϫ 10 Ϫ5 . Minigene constructs driven by the DHFR-TS promoter consistently conferred stable resistance at ϳ4-fold higher frequencies than those with the HXGPRT promoter. The genomic HXGPRT construct conferred stable resistance at intermediate frequency. No differences in transformation frequency were observed between cDNAs I and II. Overall, transformation frequencies using HXGPRT as a selectable marker are substantially below those observed using DHFR-TS mutants, which confer high level resistance to pyrimethamine Roos, 1993, 1994).
After a second passage in MPA ϩ xanthine, parasites were cloned by limiting dilution and subjected to hybridization analysis to assess HXGPRT gene copy number organization. From a total of 25 independent clones isolated following transfection with the various minigene constructs, 10 contained single copy transgenes (40%), 13 contained 2 copies (52%) and 2 contained multiple copies (8%). As previously observed at the DHFR-TS locus Roos, 1993, 1994), minigene transfectants integrated at various sites throughout the genome in contrast to parasites transfected with large segments of contiguous genomic DNA, which integrate by homologous recombination (cf. Fig. 6). All transgenic parasite lines exhibited the same level of resistance to MPA ϩ xanthine. Uptake of labeled purines was examined for individual clones bearing 1-2 HXGPRT transgenes in order to assess the variation in complementing HXGPRT activity levels among different transgenic parasite lines. As shown in the right-hand panels of Fig. 7, transgenic parasites expressing HXGPRT minigenes restored wild-type activity levels to HXGPRT Ϫ knock-out parasites. Enzyme assays were also performed on crude extracts from a subset of the HXGPRT transgenic parasite lines, with similar results (not shown). Differences observed between transgenic lines were not statistically significant, although the possibility that chromosomal position may affect expression levels is not critically tested by these experiments, because the selection process requires sufficient expression levels to confer MPA resistance.

DISCUSSION
The feasibility of insertional mutagenesis in Toxoplasma was first demonstrated in studies targeting the parasite's UPRT locus (Donald and Roos, 1995). Identification of the parasite's HXGPRT gene as described above serves to validate the general applicability of transformation vector-based insertional mutagenesis. The HXGPRT enzyme itself provides a possible target for rational drug design and a new selectable marker for molecular genetic studies in T. gondii.
The frequency of HXGPRT mutants obtained by insertional mutagenesis was ϳ3 ϫ 10 Ϫ7 , comparable with the frequency of 10 Ϫ6 previously reported for the UPRT locus. It is possible that the slightly lower frequency at the HXGPRT locus may be related to the smaller size of the HXGPRT gene: although neither promoter region has been mapped carefully, HXGPRT cDNAs span 3.0 kb of genomic sequence versus 5.5 kb at the UPRT locus. The apparent ability to tag the entire parasite genome using nonhomologous integrating vectors based on pyrimethamine-resistant DHFR-TS is a direct confirmation of both the high frequency of stable transformation previously reported  and of the essentially random nature of transgene integration. The observed frequency of insertional tagging in T. gondii is similar to that achieved with bacterial transposons such as Tn10, which transposes at a frequency of 10 Ϫ6 -10 Ϫ7 per recipient cell (Kleckner, 1983).
Because the plasmids employed for insertional mutagenesis in Toxoplasma do not encode transposase/integrase enzymes, it must be that (i) T. gondii possesses a highly active nonhomologous recombination system producing genomic integration of foreign DNA, (ii) sequences associated with the DHFR-TS transformation vectors stimulate or facilitate high frequency integration, or (iii) pyrimethamine selection promotes the high frequencies of integration observed (perhaps by inhibiting DNA replication). It may be possible to enhance integration frequencies in Toxoplasma tachyzoites by inducing some sort of recombination-repair mechanism (Black et al., 1995), but it is certainly not the case that all transfected plasmids spontaneously . Solid bar, wild-type parasites (untransfected); striped bars, HXGPRT minigene constructs; shaded bar, HXGPRT genomic construct. SW-to-NE stripes indicate isoform I; NW-to-SE stripes indicates isoform II; light stripes indicate HXGPRT promoter; and heavy stripes indicate DHFR-TS promoter. All HXGPRT constructs (but not the deletion construct pHXGPRTg11⌬Sal) confer significant levels of hypoxanthine, xanthine, or guanine PRT (HPRT, XPRT, and GPRT, respectively) activity. Right-hand panels, stable clones. Parasite clones expressing HXGPRT transgenes were selected with MPA ϩ xanthine (see text), inoculated into cultures of Lesch-Nyhan fibroblasts, and assayed as described above. Purine uptake activity was averaged for independent transformants harboring transgenic plasmids pminiHXG-PRT-I (n ϭ 4), pminiHXGPRT-II (n ϭ 6), pdhfrHXGPRT-I (n ϭ 6), or pdhfrHXGPRT-II (n ϭ 5). Error bars indicate standard deviation of the mean. Transgenic HXGPRT plasmids restore hypoxanthine, xanthine, and guanine PRT activity to wild-type levels, although there is significant variation between clones (presumably reflecting differences in chromosomal integration sites).
integrate into the T. gondii genome at high frequency (Donald and Roos, 1993;Kim et al., 1993;Sibley et al., 1994). For example, although it is clear that a single copy of the HXGPRT gene is sufficient to confer resistance to MPA, HXGPRT-based plasmids produce stable transformants ϳ1000-fold less efficiently than DHFR-TS-based plasmids (Table II). With respect to possibility (iii), it seems unlikely that inhibition of DNA replication alone can explain the high frequency of integration observed, because we have previously shown that transgene integration is independent of pyrimethamine treatment . Moreover, mycophenolic acid selection might be expected to cause inhibition of DNA replication comparable with pyrimethamine treatment. By process of elimination, this suggests that the pyrimethamine-resistance plasmids themselves may be responsible for the high transformation frequencies observed. In this regard, it is intriguing that the DHFR-TS promoter enhances stable HXGPRT transformation frequencies severalfold (Table II). Sequences known to stimulate nonhomologous recombination in bacteria include regions of high transcriptional activity or regions of stalled DNA replication (Ehrlich et al., 1993).
The mechanism by which DHFR-TS transformation vectors insert into the T. gondii chromosome remains unclear. Examination of three cloned insertion sites shows surprisingly little modification of either the integrating vector (exonucleolytic removal of 4 -31 nt) or the genomic target (duplication or deletion of 3-6 bp). This contrasts with certain aspects of nonhomologous genomic integration in some mammalian systems, which can be accompanied by more extensive deletion or rearrangement of vector sequences, and may occur preferentially within repetitive DNA (Brenner et al., 1984;Höglund et al., 1992). The presence of short "filler" segments between target sequences and vector termini (Fig. 3) is commonly observed in other systems however (Brenner et al., 1984;Roth et al., 1991;Merrihew et al., 1996). In mammalian cells these extra bits of DNA are thought to be generated by a template-dependent DNA polymerase within the recombination enzyme complex to fill in gaps between sterically hindered ends. Finally, it is interesting to note the similarity between the end of one transgene and its integration point (10 out of 12 nt; 8 nt precise match) and the integration of another HindIII-linearized transgene precisely at a genomic HindIII site. Short sequences of identity between the integrating plasmid and the target site have also been reported in mammalian systems (Roth and Wilson, 1986;Merrihew et al., 1996).
The presence of differentially spliced cDNAs from the T. gondii HXGPRT gene remains to be confirmed by direct transcript analysis, but it is unlikely that the single HXGPRT-II cDNA identified is an incompletely spliced mRNA precursor. Removal of introns flanking the novel exon characteristic of HXGPRT-II does not regenerate plausible splice sites, which might lead to maturation into HXGPRT-I cDNA. Both exon skipping (e.g. removal of exon III in HXGPRT-I cDNA) and intron exclusion (e.g. retention of HXGPRT intron I in HXG-PRT-II cDNA) are common forms of alternative splicing in other systems (McKeown, 1992). The novel exon in HXGPRT-II is predicted to produce a 49-amino acid insertion close to the NH 2 terminus, a region that has been poorly conserved over the course of evolution. Codon usage within this exon is consistent with the remainder of the HXGPRT sequence and other known T. gondii genes (Roos, 1993;Ellis et al., 1993). HXGPRT enzyme expressed from both class I and class II cDNAs are active in both E. coli and Toxoplasma and essentially indistinguishable in their ability to phosphoribosylate hypoxanthine, xanthine, or guanine. Translation from the first methionine downstream of the differentially spliced exon (resulting in a protein that lacks only 18 amino acids at the NH 2 terminus of HXG-PRT-I) produces a highly unstable enzyme. 3 The possibility that HXGPRT-I and HXGPRT-II isoforms may be developmentally regulated or targeted to distinct intracellular organelles (Opperdoes, 1987) remains to be explored.
The HXGPRT gene targeted in these studies encodes a parasite-specific enzyme with potential as a target for chemotherapy (Ullman and Allen, 1995). Successful overexpression and purification of recombinant T. gondii HXGPRT paves the way for future structural studies, with an eye toward structurebased drug design. Although this enzyme is not essential for the survival of intracellular tachyzoites, subversive substrates such as allopurinol have been used effectively to treat other parasitic diseases such as Chagas' disease and leishmaniasis (Gallerano et al., 1990;Martinez and Marr, 1992). It may also be possible to develop therapeutic strategies involving inhibition of multiple purine salvage enzymes.
The HXGPRT gene provides a versatile genetic marker, adding to the growing arsenal of safe alternatives for routine transformation of T. gondii (Kim et al., 1993;Sibley et al., 1994;Messina et al., 1995;Soldati et al., 1995). The knock-out mutants isolated in the course of this study provide appropriate hosts for such experiments. One particular advantage of the HXGPRT system is that this enzyme is suitable for both positive and negative selection (using MPA or 6-TX selection, respectively). A further attraction is the ability to apply selection immediately following transfection, yielding relatively rapid selections with low background. The frequency of transient transformation using HXGPRT vectors is ϳ5%, and ϳ0.08% of these transiently resistant parasites go on to produce stably resistant HXGPRT ϩ clones. The combined frequency of ϳ4 ϫ 10 Ϫ5 is adequate for many applications, although DHFR-TSbased vectors will probably remain the method of choice for 3 R. G. K. Donald, D. Carter, B. Ullman, and D. S. Roos, unpublished observations.  Fig. 5 for HXGPRT plasmid maps. See Roos (1993, 1994) for DHFR-TS plasmids. b After first passage in drug. c Pyrimethamine resistance; data from Donald and Roos (1994).
insertional mutagenesis and complementation cloning, due to their higher frequency of stable integration (ϳ5%; Roos (1993, 1994)). The observed difference in frequency between DHFR-TS and HXGPRT vectors remains unexplained. Use of HXGPRT as a negative selectable marker should permit a variety of experiments, such as genetic selection schemes to screen for transcriptional control mutants with altered ability to trans-activate promoter sequence-dependent gene expression (Pelham, 1984;Parker-Thornburg and Bonner, 1987)). It should also be possible to develop combined positive/ negative selection vectors suitable for allelic replacements ("hit and run" mutagenesis) at loci where no direct selection is available (Mortensen, 1993;Roos et al., 1994).