Creation of homozygous mutants of Leishmania donovani with single targeting constructs.

Homozygous null mutants of the hypoxanthine-guanine phosphoribosyltransferase (hgprt) and adenine phosphoribosyltransferase (aprt) loci were created in Leishmania donovani in which both alleles were eliminated using only a single targeting construct. Functional heterozygotes were first generated by homologous recombination after transfection with vectors containing 5′- and 3′-flanking regions of either the hgprt or the aprt gene circumscribing drug resistance markers. Homozygous null mutants were then isolated from the heterozygotes by negative selection in media containing subversive substrates of the encoded proteins, i.e. allopurinol for HGPRT and 4-aminopyrazolopyrimidine for APRT. The novel alleles created by homologous recombination were verified by Southern blotting, and the effects of gene replacement upon gene expression in intact parasites were evaluated by direct enzymatic assay and by immunoblotting. All mutant strains were viable under the selection conditions and exhibited appropriate drug resistance phenotypes. The ability to generate homozygous knockouts with single targeting constructs greatly facilitates the genetic dissection and subsequent biochemical investigations of the purine pathway in Leishmania and has important general implications for the genetic manipulation and analysis of the leishmanial genome.

Homozygous null mutants of the hypoxanthine-guanine phosphoribosyltransferase (hgprt) and adenine phosphoribosyltransferase (aprt) loci were created in Leishmania donovani in which both alleles were eliminated using only a single targeting construct. Functional heterozygotes were first generated by homologous recombination after transfection with vectors containing 5-and 3-flanking regions of either the hgprt or the aprt gene circumscribing drug resistance markers. Homozygous null mutants were then isolated from the heterozygotes by negative selection in media containing subversive substrates of the encoded proteins, i.e. allopurinol for HGPRT and 4-aminopyrazolopyrimidine for APRT. The novel alleles created by homologous recombination were verified by Southern blotting, and the effects of gene replacement upon gene expression in intact parasites were evaluated by direct enzymatic assay and by immunoblotting. All mutant strains were viable under the selection conditions and exhibited appropriate drug resistance phenotypes. The ability to generate homozygous knockouts with single targeting constructs greatly facilitates the genetic dissection and subsequent biochemical investigations of the purine pathway in Leishmania and has important general implications for the genetic manipulation and analysis of the leishmanial genome.
Improved methods for the chemotherapeutic manipulation of parasitic diseases depends upon the exploitation of fundamental metabolic or biochemical differences between the parasite and the mammalian host. Basic investigations of metabolic pathways in parasitic protozoa have been hampered by the inability to select directly for mutants genetically defective in some component of the metabolic pathway. This impediment has been circumvented in part by the implementation of DNA transfection methods that now permit the genetic engineering of trypanosomatid chromosomes by targeted gene replacement (1)(2)(3)(4)(5)(6)(7). As parasites of the Leishmania and Trypanosoma genera are generally thought to be diploid at most loci (1, 8 -11) and do not readily undergo sexual crosses in the laboratory (12)(13)(14), the finding that genes can be efficiently targeted by homologous recombination has created new avenues for mutationally altering parasites that lack specific gene sequences, thereby enabling direct assessment of gene function. The protocols for the generation of knockout parasites have routinely involved the sequential replacement of both alleles of a genetic locus with two different targeting constructs, each containing flanking regions from the targeted gene encompassing independent selectable markers. Although double targeted gene replacement has permitted the construction of a variety of knockout mutants in both Leishmania (6,15,16) and Trypanosoma (11,17), the process for generating these strains is both time-consuming and laborious, and the restricted number of available drug resistance markers limits the introduction of multiple homozygous mutations within a single parasite population for biochemical and cellular studies. These barriers can be surmounted in part by the creation of homozygous knockout strains using only a single targeting construct. Such a strategy has been exploited in mammalian cell lines by selecting for loss-of-heterozygosity after a single round of gene targeting by increased drug pressure (18).
The purine salvage pathway offers an excellent opportunity to examine whether Leishmania express the metabolic machinery to undergo loss-of-heterozygosity, because several purine salvage enzymes metabolize cytotoxic purine analogs, thereby affording a powerful negative selection scheme to detect the phenomenon. These enzymes include hypoxanthine-guanine phosphoribosyltransferase (HGPRT) 1 and adenine phosphoribosyltransferase (APRT), which initiate the metabolism of 4-hydroxypyrazolopyrimidine (HPP, allopurinol) and 4-aminopyrazolopyrimidine (APP), respectively (19). Both HPP and APP are toxic to Leishmania spp. at relatively low concentrations (9,20). As both hgprt (21) and aprt (22) and their respective flanking sequences have been isolated from cosmid libraries, the molecular reagents to construct appropriate knockout vectors are now available.
We now describe the generation of homozygous null mutants of both hgprt and aprt in Leishmania donovani by targeted gene replacement using only a single targeting construct. The construct is used to target the first allele by transfection, and the loss of the wild type allele from the functional heterozygote is selected directly in a subversive substrate of the encoded gene product. Recently, Guerios-Filho and Beverley have demonstrated that loss-of-heterozygosity can also occur at the dihydrofolate reductase-thymidylate synthase (dhfr-ts) locus of Leishmania major (23 mmol) and [␣-35 S]dATP (1320 Ci/mmol) were acquired from DuPont NEN. All restriction and DNA modifying enzymes were bought from either Life Technologies, Inc. or New England Biolabs, and Thermus flavus DNA polymerase was obtained from Epicentre Technologies (Madison, WI). The pX63-NEO and pX63-HYG vectors that encompass the neomycin phosphotransferase (neo r ) and hygromycin phosphotransferase (hyg r ) genes, respectively, were generously provided by Dr. Stephen Beverley (Harvard Medical School). Probes encompassing neo r and hyg r were obtained by excision of pX63-NEO and pX63-HYG with BamHI/SpeI and SpeI alone, respectively. A probe for the ␣-tubulin gene (␣-tubulin) from Leishmania enriettii was kindly provided by Dr. Scott Landfear of the Oregon Health Sciences University (Portland, OR). Geneticin (G418), hygromycin, HPP, and APP were all obtained from Sigma. The sources of all other materials, chemicals, and reagents employed in these experiments have been reported previously and were of the highest purity commercially available.
Cell Culture-L. donovani promastigotes were cultivated in completely defined DME-L culture medium (24). Single cell cloning of parasites was carried out in semi-solid DME-L medium as reported (24). The DI700 cell line is a wild type clone of the Sudanese 1S strain of L. donovani that was used as the initial recipient strain for transfections. For the purposes of the genetic manipulations reported in this report the DI700 cell line is denoted H ϩ/ϩ A ϩ/ϩ , where H refers to HGPRT, A to APRT, and ϩ is the wild type allele. Sensitivities of wild type and mutant strains to G418, hygromycin, HPP, and APP were determined as described (9).
DNA Manipulations and Library Construction-L. donovani genomic DNA and plasmid DNA were isolated by standard protocols (25,26). Southern blotting and nucleotide sequencing were performed as reported (25,26).
Quantitative Densitometry-Quantitative densitometry was performed with a Molecular Dynamics Phosphoimager SI driven by Molecular Dynamics Scanner software (version 1.0). IP Lab Gel (version 1.5g) analysis software was used to calculate the density of each band. In all samples, the intensity of the hybridization signals obtained with probes from the hgprt and aprt loci and to neo r and hyg r were normalized to the intensity of the ␣-tubulin band of the same sample, and the results were expressed as corrected arbitrary units.
Molecular Constructs for Replacement of hgprt and aprt Alleles-The hgprt (21) and aprt (22) genes were originally isolated from a cosmid library of L. donovani DNA and subcloned as 3.5-kb EcoRI and 5.4-kb BamHI-XbaI fragments in pBluescript KS ϩ , respectively. 5Ј-and 3Јflanking regions of the hgprt and aprt genes were then identified by restriction mapping and Southern blotting and amplified by PCR for subcloning into the targeting vectors. PCR primers for amplifying the 3Ј-flanking regions of the hgprt were determined from nucleotide sequence information obtained from the 3.5-kb EcoRI fragment. The PCR primers for the hgprt 5Ј-flanking region were determined from a 5.2-kb EcoRI-SalI fragment that was isolated from the cosmid clone from which hgprt was cloned. PCR primers for subcloning both 5Ј-and 3Ј-flanking regions of the aprt were determined after mapping and sequencing appropriate segments of the 5.4-kb BamHI-XbaI fragment from the aprt-containing cosmid. Standard PCR conditions (10 -15 cycles; 94°C for 45 s, 50°C for 45 s, and 72°C for 90 s) for amplification of plasmid DNA sequences were employed (21). Sense and antisense primers used in the PCR were 5ЈH5Ј and 5ЈH3Ј for the 1.7-kb hgprt 5Ј-flanking region, 3ЈH5Ј and 3ЈH3Ј for the 1.8-kb hgprt 3Ј-flanking region, 5ЈA5Ј and 5ЈA3Ј for the 1.4-kb aprt 5Ј-flanking region, and 3ЈA5Ј and 3ЈA3Ј for the 1.5-kb aprt 3Ј-flanking region. PCR products for the flanking regions of hgprt and aprt were separated on agarose gels, purified using the GeneClean Kit (Intermountain Scientific Corp., Bountiful, UT), digested with appropriate restriction enzymes, and ligated into the pertinent sites of the pX63-NEO and pX63-HYG plasmids, respectively. Each of the two flanking regions of both loci were ligated sequentially into the appropriate vectors containing the drug resistance markers. The final vector constructs employed for allelic replacement were designated pX63-NEO-⌬hgprt and pX63-HYG-⌬aprt, respectively. All plasmid constructs were purified as described (25,26) and checked for identity and orientation by nucleotide sequencing.
Transfections-Parasites were transfected by electroporation using conditions fundamentally identical to those previously described (6). The deletion constructs were cleaved with HindIII and BglII prior to electroporation to liberate the linear targeting fragment from pX63-NEO-⌬hgprt and pX63-HYG-⌬aprt. Linearized targeting DNAs were designated on the basis of the nomenclature of the plasmid from which they were excised, without the initial letter; X63-NEO-⌬hgprt from pX63-NEO-⌬hgprt and X63-HYG-⌬aprt from pX63-HYG-⌬aprt. Cells after transfection were maintained in liquid medium for 24 h prior to selection.
Gene Replacements-To isolate hgprt heterozygotes, wild type cells were transfected with X63-NEO-⌬hgprt and selected on plates containing 16 g/ml G418. aprt heterozygotes were generated after transfection with X63-HYG-⌬aprt and isolated on plates containing 50 g/ml hygromycin B and 16 g/ml G418. Heterozygosity was established by Southern blotting using appropriate probes. To obtain null mutants of hgprt and aprt, the appropriate heterozygotes were selected in 1 mM HPP (containing 16 g/ml G418) or 100 M APP (containing 16 g/ml G418, 1 mM HPP, and 50 g/ml hygromycin). All mutant strains created by targeted gene replacement were maintained in medium containing the drugs from which it and its progenitors had been selected.
Preparation of Cell Extracts and Enzyme Assays-L. donovani cell pellets containing 2 ϫ 10 9 parasites were extracted by sonication in 50 mM Tris, pH 8.0. 0.5% Triton X-100 was added to the lysis buffer, as HGPRT activity in Leishmania is associated with a particulate fraction (27). Cell supernatants were obtained after centrifugation at 20,000 ϫ g for 35 min at 4°C and processed as described (19). HGPRT and APRT enzymatic activities were measured as described (28). Protein content was measured by the method of Bradford (29).
Western Blotting-Recombinant L. donovani HGPRT and APRT were purified by affinity chromatography as reported (21,22). Polyclonal antibodies to purified L. donovani HGPRT and APRT were raised in New Zealand White rabbits by conventional protocols (31). Promastigotes were lysed by in SDS sample buffer, and 20,000 ϫ g cell supernatants were prepared by centrifugation. Lysates from 2 ϫ 10 6 cells were loaded onto each lane, fractionated by SDS-polyacrylamide gel electrophoresis (31), blotted onto nitrocellulose membranes using a semi-dry electrophoretic transfer cell (Bio-Rad), and Western blot analysis was performed by standard methodologies (30).

Generation of hgprt and aprt Null Mutants-
The stratagem for the derivation of hgprt and aprt null mutants by a combination of targeted gene replacement and direct selection in cytotoxic substrate is displayed in Fig. 1. Starting with the DI700 wild type H ϩ/ϩ A ϩ/ϩ strain, a single wild type hgprt allele was disrupted by targeted gene replacement using the X63-NEO-⌬hgprt construct (shown in Fig. 2B) to generate the hgprt heterozygote, H ϩ/n A ϩ/ϩ . The homozygous null mutant of hgprt, H n/n A ϩ/ϩ , was then obtained from the H ϩ/n A ϩ/ϩ heterozygote line by direct negative selection in 1.0 mM HPP, a subversive substrate of the L. donovani HGPRT (19) with amply demonstrated antileishmanial properties (20). The H n/n A ϩ/ϩ cell line was then transfected with the aprt knockout construct X63-HYG-⌬aprt (shown in Fig. 2E) to produce the aprt heterozygote, H n/n A ϩ/h , within the hgprt Ϫ genetic back-ground. Finally, the aprt heterozygote was subjected to selective pressure in 100 M APP, a cytotoxic substrate of the L. donovani APRT (9,19), to generate the hgprt Ϫ , aprt Ϫ double knockout mutant H n/n A h/h . Southern Blot Analysis-Southern blot analysis of genomic DNA prepared from wild type and mutant strains confirmed the existence of the new alleles that had been created after a single round of gene targeting followed by direct selection for loss of the second wild type allele (Figs. 3-5). The maps of the genomic loci of hgprt and aprt and the novel alleles created by   FIG. 2. Restriction maps of the hgprt and aprt genomic loci, the plasmid constructs used in the targeted gene replacements, and the novel alleles encompassing the drug resistance cassettes. Figure shows restriction maps of the wild type hgprt locus (A), the X63-NEO-⌬hgprt targeting construct (B), the neo r -replaced hgprt locus after homologous recombination with the X63-NEO-⌬hgprt construct (C), the wild type aprt locus (D), the X63-HYG-⌬aprt targeting construct (E), and the hyg r -targeted aprt locus obtained after replacement by X63-HYG-⌬aprt (F). Expected restriction fragment sizes from the hgprt locus after digestion with EcoRI or from the aprt locus after digestion with BamHI/SalI are marked. The positions of probe A from the hgprt coding region, probe B from the 3Ј flank of hgprt, probe C from the aprt coding region, and probe D from the 5Ј flank of aprt are shown under the appropriate genomic loci. hgprt and aprt coding regions are indicated by the black rectangles, whereas corresponding 5Ј-and 3Ј-flanking regions that were amplified by PCR to create the gene targeting constructs are indicated by the lightly shaded rectangles. The antibiotic resistance markers, neo r and hyg r , are indicated by the white boxes and designated by neo and hyg, respectively, and the corresponding pX-based flanking regions from the L. major dhfr-ts are indicated by the gray shaded boxes.
homologous recombination of the X63-NEO-⌬hgprt and X63-HYG-⌬aprt constructs into the hgprt and aprt loci, as well as the restriction fragments employed in the Southern blot experiments, are shown in Fig. 2. The differences in the new hgprt alleles can be observed in Fig. 3. The disrupted hgprt allele could be distinguished from the wild type allele by an altered EcoRI restriction pattern (Fig. 3). After digestion of genomic DNA with EcoRI and hybridization to probe B (shown in Fig.  2A) corresponding to a portion of the 3Ј-flanking region of hgprt, only the expected 4.6-kb EcoRI fragment from the wild type allele was observed in the wild type H ϩ/ϩ A ϩ/ϩ cell line (Fig. 3A, probe B). It should be noted that the 4.6-kb EcoRI restriction fragment corresponds to the expected size of the wild type hgprt allele even though the original hgprt clone was isolated as a 3.5-kb EcoRI fragment from a cosmid (21). How-ever, fine mapping of the cosmid indicated that the 3Ј EcoRI restriction site of the original 3.5-kb EcoRI insert was located within the cosmid DNA sequence. In contrast, a band of 6.4-kb (Fig. 3A, probe B) was observed in all strains in which the wild type hgprt allele had been disrupted by X63-NEO-⌬hgprt. The cell line heterozygous for hgprt, H ϩ/n A ϩ/ϩ , also retained a copy of the wild type hgprt allele, as illustrated by the presence of the 4.6-kb EcoRI fragment derived from the wild type locus (Fig. 3A, probe B). The H n/n A ϩ/ϩ strain generated by selection of H ϩ/n A ϩ/ϩ cells in HPP (Fig. 1), as well as its progeny, i.e. the H n/n A ϩ/h and H n/n A h/h cell lines, did not display the wild type 4.6-kb EcoRI band and revealed only the expected 6.4-kb displacement by X63-NEO-⌬hgprt (Fig. 3A, probe B). The losses of the wild type hgprt alleles by targeting and then selection were verified further by hybridization of genomic DNA from the five cell lines to probe A encompassing the protein coding portion of hgprt (Fig. 3A, probe A). The 4.6-kb EcoRI fragment from the wild type allele was observed only in the H ϩ/ϩ A ϩ/ϩ and H ϩ/n A ϩ/ϩ cells and not in the H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h lines (Fig. 3, probe A).
Normalization of the signal intensities of the bands in the genomic DNA that hybridized to either probe A or probe B to that obtained with ␣-tubulin (Fig. 3A, ␣-tubulin probe) generally supported the idea that the first wild type hgprt allele was replaced by homologous recombination of the targeting construct used in the transfection, while loss-of-heterozygosity accounted for the elimination of the remaining wild type allele (Table I). Densitometric ratios of the signal intensities from the 4.6 EcoRI band hybridized to either probe A (A/T) or probe B (B u /T) to that obtained with genomic DNA hybridized to ␣-tubulin (see Fig. 3A) were 0.273 and 0.115 (A/T) and 0.105 and 0.044 (B u /T) for the H ϩ/ϩ A ϩ/ϩ and H ϩ/n A ϩ/ϩ cell lines, respectively.
Disruption of the aprt locus in the H n/n A ϩ/h and H n/n A h/h cell lines that were targeted by X63-HYG-⌬aprt was also authenticated by Southern blotting (Fig. 3B). Digestion of genomic DNA with SalI/BamHI and hybridization to either probe C from the aprt coding region or probe D derived from the aprt 5Ј-flanking region (Fig. 2D) revealed the presence of the 3.5-kb SalI/BamHI restriction fragment from the wild type allele in H ϩ/ϩ A ϩ/ϩ , H ϩ/n A ϩ/ϩ , H n/n A ϩ/ϩ , and H n/n A ϩ/h cells FIG. 3. Southern blot analysis of hgprt and aprt loci in wild type and mutant strains. Genomic DNAs from H ϩ/ϩ A ϩ/ϩ , H ϩ/n A ϩ/ϩ , H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h cells were digested with EcoRI and subjected to Southern blot hybridization using probes A and B from the hgprt locus (panel A). The same genomic DNAs were also digested with BamHI and SalI and hybridized to probes C and D from the aprt locus (panel B) and to the neo r and hyg r probes (panel C). Each set of Southern blots was normalized by hybridization to the L. enriettii ␣-tubulin. The locations of probes A, B, C, and D are shown in Fig. 2. The positions of molecular size markers are indicated on the left in base pairs.  1.5g) analysis software was employed to calculate the density of each hybridization signal. For each cell line, the density of the hybridization signal to the ␣-tubulin probe was arbitrarily set at 1.0, and the densitometric signals of the bands obtained with probes A and B from the hgprt locus, probes C and D from the aprt locus, and neo r and hyg r are reported relative to the ␣-tubulin value. Lettered abbreviations designate the densitometric ratio for specific hybridization bands as follows: A, the 4.6-kb band obtained with probe A; B l , the 4.6-kb band obtained with probe B; B u , the 6.4-kb band obtained with probe B; C, the 3.5-kb band obtained with probe C; D l , the 1.3-kb band obtained with probe D; D u , the 3.5-kb band obtained with probe B; N, the 5.9-kb band obtained with neo r ; and D, the 2.0-kb band obtained with hyg r .  Fig. 3B, probes C and D). A new allele corresponding to the integration of the X63-HYG-⌬aprt drug resistance cassette into the wild type aprt locus was detected in the H n/n A ϩ/h and H n/n A h/h cell lines (Fig. 3B, probe D). Only a 1.4-kb SalI fragment is detected in SalI/BamHI restricted DNA from H n/n A h/h cells after hybridization to probe D (Fig. 3B, probe D), while the H n/n A ϩ/h presumptive heterozygote contained both the 3.5-and 1.4-kb bands (Fig. 3B, probe D). The disruption of the wild type aprt alleles was confirmed by hybridization of the same blot with probe C (Fig. 2B) obtained from the aprt coding region. The 3.5-kb wild type SalI/BamHI fragment was observed in all cell lines, except H n/n A h/h (Fig. 3B, probe C). Quantitative densitometry of the autoradiograms probed with aprt fragments and with the ␣-tubulin probe (Fig. 3B, ␣-tubulin probe) confirmed the sequential allelic replacements ( Table I). The ratio of the signal intensities of the 3.5-kb SalI/ BamHI band in H n/n A ϩ/h cell DNA hybridized to either probe C (C/T) or D (D u /T) to that of the ␣-tubulin band was 64% of that obtained with DNA from the H n/n A ϩ/ϩ parental line. Similarly, the densitometric ratio obtained for the 1.4-kb SalI/ BamHI fragment (D l /T) of H n/n A ϩ/h cells was 41% of that of its H n/n A h/h derivative (Fig. 3B). The insertion of the drug resistance markers into the L. donovani genome was confirmed by Southern analysis using neo r and hyg r probes (Fig. 3C). The neo r and hyg r insertions were observed only in the appropriate selected strains. The successive replacement of each of the two wild type aprt alleles by X63-HYG-⌬aprt in the H n/n A ϩ/h and H n/n A h/h cell lines was supported by densitometric examination, although the data demonstrating sequential elimination of hgprt coding regions with X63-NEO-⌬hgprt were not supported by this analysis (see "Discussion").
Enzymatic Activities and Immunoblotting-To establish that elimination of wild type hgprt and aprt alleles conferred the appropriate enzymatic deficiencies, the activities of HGPRT and APRT were determined for the H ϩ/ϩ A ϩ/ϩ , H ϩ/n A ϩ/ϩ , H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h cell lines. All null mutants created by single gene targeting strategies expressed low activities of the appropriate enzyme, as expected, whereas heterozygous cell lines, i.e. H ϩ/n A ϩ/ϩ and H n/n A ϩ/h , expressed intermediate levels of HGPRT and APRT activity, respectively, when compared with wild type H ϩ/ϩ A ϩ/ϩ cells (Fig. 4). It is interesting to note that the H ϩ/n A ϩ/ϩ and H n/n A ϩ/ϩ cell lines both expressed greater levels of APRT activity than did H ϩ/ϩ A ϩ/ϩ cells. Overexpression of both APRT and xanthine phosphoribosyltransferase activities have been consistent findings among hgprt -/clones that have been generated in this laboratory by double targeted gene replacement, suggesting an operative compensatory mechanism for purine salvage in the parasite (data not shown).
To confirm that the double knockout mutants did not express their corresponding gene products, extracts of wild type and mutant L. donovani were evaluated by Western blotting with polyclonal antibodies prepared against the pure recombinant HGPRT and APRT proteins. As observed in Fig. 5A, HGPRT protein was detected in extracts of H ϩ/ϩ A ϩ/ϩ and H ϩ/n A ϩ/ϩ lysates in expected proportions but was not observed in extracts of H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h cells. In a parallel set of experiments, APRT protein was observed in all lysates except those prepared from H n/n A h/h cells, although H n/n A ϩ/h lysates appeared to contain relatively less APRT protein than H ϩ/ϩ A ϩ/ϩ , H ϩ/n A ϩ/ϩ , and H n/n A ϩ/ϩ cells (Fig. 5B).
Drug Resistance Phenotypes-No significant differences in growth rate or morphology were observed among the H ϩ/ϩ A ϩ/ϩ , H ϩ/n A ϩ/ϩ , H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h cell lines. However, comparisons of the sensitivities of the five lines to G418 and hygromycin reflected the incorporation of the neo r and hyg r genes into the hgprt and aprt loci, respectively (Table II). The effective concentration of G418 that inhibited wild type cell growth by 50% (EC 50 value) was 2.5 g/ml, whereas the EC 50 value of the H ϩ/n A ϩ/ϩ , H n/n A ϩ/ϩ , H n/n A ϩ/h , and H n/n A h/h cell line, all of which encompassed neo r within their genomes (Fig. 3C), for the drug was 100 -120 g/ml (Table II). Conversely, the EC 50 values of the H n/n A ϩ/h and H n/n A h/h cell lines for hygromycin were 500 g/ml and 600 g/ml, respectively, ϳ50 -60-fold greater than the EC 50 values obtained for the three progenitor cell lines that did not contain hyg r .
The five cell lines were also evaluated for their sensitivities to the subversive substrates of the two purine PRTs, HPP and APP. All mutants in which both copies of the relevant wild type gene have been obliterated expressed a high degree of growth resistance to the germane purine base analog, i.e. the EC 50 values obtained with the mutant strain were 2-3 orders of magnitude greater than the EC 50 values obtained with H ϩ/ϩ A ϩ/ϩ cells. The H ϩ/n A ϩ/ϩ cell line exhibited intermediate resistance to HPP, while the H n/n A ϩ/h line displayed the same modest increase in APP resistance previously observed for an L. donovani cell line obtained by direct selection that expressed only 50% of wild type APRT activity (9). DISCUSSION Homozygous null mutants of L. donovani in which both wild type alleles of a genetic locus have been replaced using only a single targeting construct have been created at two independent loci, hgprt and aprt. The strategy for the generation of homozygous null mutants of both genetic loci required elimination of the first wild type allele after transfection and integration of a linear drug resistance cassette encompassing 5Ј and 3Ј untranslated regions of the hgprt and aprt loci by homologous recombination, which Leishmania accomplish at high frequency (1,4). The second step in the generation of knockout mutants entailed selection for loss-of-heterozygosity by exposure of the heterozygote to selective pressure using subversive substrates of the encoded gene products. That Leishmania exhibit the capacity to undergo this loss-of-heterozygosity is a general phenomenon for any gene for which a strong negative selection can be devised is supported by the applicability of this strategy for creating null mutants to two different genetic loci. This approach is further substantiated by the recent observations of Guieros-Filho and Beverley that dhfr-ts heterozygotes of L. major can be selected for loss-of-heterozygosity at the dhfr-ts locus by plating in methotrexate and thymidine (23).
Our data do not, however, support a general mechanism for selecting or distinguishing homozygous null mutants from heterozygotes on the basis of positive selection, as no differences in the growth sensitivities of H ϩ/n A ϩ/ϩ and H n/n A ϩ/ϩ to G418 or of H n/n A ϩ/h and H n/n A h/h to hygromycin could be discerned (Table II). Conversely, it should be noted that loss-of-heterozygosity at multiple loci could be obtained by positive selection in murine ES cells by exposure to high concentrations of the selectable marker (18), and the frequency of loss-of-heterozygosity observed at the dhfr-ts locus of L. major could be increased by selection in hygromycin (23).
The mechanisms by which loss-of-heterozygosity occurs at the hgprt and aprt loci in Leishmania remain to be investigated. The genesis of H n/n A ϩ/ϩ and H n/n A h/h from H ϩ/n A ϩ/ϩ and H n/n A ϩ/h cells, respectively, could have arisen as a consequence of a number of genetic events including chromosome loss, gene deletion, chromosome nondisjunction, mitotic recombination, gene conversion, or homologous recombination of ep-isomal DNAs derived either from the chromosome or the original targeting construct. Although none of these mechanisms has been described for Leishmania, the karyotypic plasticity of the leishmanial genome under selective (10,32) or stress (33) conditions and the ability of the parasite to undergo conservative amplifications of chromosomal segments (34), to maintain these amplified segments episomally as extrachromosomal elements (34), and to perform homologous recombination (1,4) have been well documented. To definitively differentiate among the various models, a number of flanking markers that distinguish between the two homologous chromosomes will have to be generated and characterized. In mammalian cell systems in which loss-of-heterozygosity has been investigated, the predominant mechanisms appear to be simple chromosome loss (35), chromosome loss followed by duplication (35,36), or mitotic recombination (36,37).
In the two homozygous strains generated in this study, the mechanisms for loss-of-heterozygosity may very well be different. Normalization of the densitometric tracings of the Southern blot data imply that H ϩ/n A ϩ/ϩ and H n/n A ϩ/ϩ cells have a single copy of the neo r gene, while H n/n A ϩ/h and H n/n A h/h cells have one and two copies of hyg r , respectively. Thus, one could conjecture that loss-of-heterozygosity at the hgprt locus could be ascribed to a simple chromosome loss or deletion of DNA sequences encompassing hpgrt, while the mechanism of allelic replacement in the H n/n A h/h aprt homozygote appears to involve a duplication of the hyg r allele in the heterozygote. In the L. major selected for loss-of-heterozygosity at the dhfr-ts locus, Southern and slot blot analyses indicated that the knockouts contained two copies of the hyg r marker, suggesting that the null mutants did not originate as a consequence of simple chromosome loss or dhfr-ts deletion (23). As no differences in hygromycin sensitivity were observed between H n/n A ϩ/h and H n/n A h/h , these data provide further support that positive selection, at least using hygromycin resistance as a selection, cannot be used to isolate homozygous null mutants from heterozygous populations. This result contrasts with analogous experiments performed on the murine ES cells in which cells containing one or two copies of neo r can be distinguished on the basis of G418 sensitivity (18).
Targeted gene replacement has become a powerful tool to evaluate gene function in Leishmania and related parasites (1,4,7,11,(15)(16)(17). In diploid organisms with a sexual cycle, the general strategy for obtaining null mutants is to target the first allele with a specific construct to generate the heterozygote, followed by sexual crossing to generate the homozygote. As Leishmania spp. are ostensibly asexual in nature, at least under laboratory conditions, and are presumed to be diploid at most loci, elimination of a wild type diploid locus has required two independent drug resistance markers, e.g. neo r and hyg r . Moreover, a third independent marker is theoretically required to verify gene function by complementation of an observed phenotype. The ability to create null mutants, therefore, with only a single targeting construct greatly facilitates genetic investigations into complex metabolic pathways, such as that for purine salvage, because the number of drug resistance markers available for gene targeting in Leishmania is limited (6,15,38,39). As a consequence, other drug resistance markers are preserved for subsequent genetic manipulations, such as the creation of strains with multiple homozygous mutations, e.g. H n/n A h/h . Indeed, the H n/n A h/h cell line is the first Leishmania strain constructed by targeted gene replacement to contain more than one homozygous mutation. Given the availability of preexisting L. donovani mutants genetically deficient in either APRT (9) or adenosine kinase (40), one can envision a stratagem to generate all the requisite strains with multiple muta- tions in purine genes to thoroughly dissect the multifarious and divergent purine acquisition pathway in Leishmania. This purine salvage pathway is of particular interest, because of the purine auxotrophy established in all protozoan parasites studied to date (41). Thus, the pathway may provide selective targets for antiparasitic therapeutic strategies. It should also be noted that the ability to obliterate both wild type alleles of a genetic locus with only a single targeting construct also significantly expedites the construction of mutants in this genus of protozoan parasite. As vector construction and strain generation and characterization are laborious and time-consuming endeavors, the ability to generate mutant lines of L. donovani after one round of transfection substantially conserves both time and financial expenditures.