Genetic analysis of tachyzoite to bradyzoite differentiation mutants in Toxoplasma gondii reveals a hierarchy of gene induction

Authors


Summary

Developmental switching in Toxoplasma gondii, from the virulent tachyzoite to the relatively quiescent bradyzoite stage, is responsible for disease propagation and reactivation. We have generated tachyzoite to bradyzoite differentiation (Tbd) mutants in T. gondii and used these in combination with a cDNA microarray to identify developmental pathways in bradyzoite formation. Four independently generated Tbd mutants were analysed and had defects in bradyzoite development in response to multiple bradyzoite-inducing conditions, a stable phenotype after in vivo passages and a markedly reduced brain cyst burden in a murine model of chronic infection. Transcriptional profiles of mutant and wild-type parasites, growing under bradyzoite conditions, revealed a hierarchy of developmentally regulated genes, including many bradyzoite-induced genes whose transcripts were reduced in all mutants. A set of non-developmentally regulated genes whose transcripts were less abundant in Tbd mutants were also identified. These may represent genes that mediate downstream effects and/or whose expression is dependent on the same transcription factors as the bradyzoite-induced set. Using these data, we have generated a model of transcription regulation during bradyzoite development in T. gondii. Our approach shows the utility of this system as a model to study developmental biology in single-celled eukaryotes including protozoa and fungi.

Introduction

Numerous eukaryotic pathogens, including parasitic protozoa and yeast, have developmental stages in their life cycle that are essential for disease propagation and causation. We use the highly complex unicellular parasite, Toxoplasma gondii, as a model for studying such development. Information gained in this manner can be applied as a general developmental model to other members of the Apicomplexan family (including Plasmodium, Cryptosporidium and Eimeria), as well as to other eukaryotic systems. T. gondii is a pathogen responsible for severe systemic disease in immunocompromised individuals, congenital disease with hydrocephalus and mental retardation in children and blindness in immunocompetent individuals (Luft and Remington, 1992; Wong and Remington, 1993; 1994). With the advent of the AIDS epidemic, T. gondii came to attention as one of the leading causes of encephalitis in AIDS patients and, today, an increasing number of immunocompromised individuals (e.g. as a result of AIDS, cancer chemotherapy, transplants, etc.) are susceptible to this parasite.

Toxoplasma gondii exists in two stages in humans and all intermediate hosts: tachyzoites, which replicate rapidly and are susceptible to the host’s immune system and drug treatment, and bradyzoites, which replicate slowly and evade both the host’s immune response and currently available drug therapies (Weiss and Kim, 2000). In most cases, clinical disease in adults results from recrudescence of the dormant bradyzoite form to the tachyzoite stage. The interconversion between the two stages, at the heart of the parasite’s survival and pathogenicity, is poorly understood at a genetic level.

Tachyzoite to bradyzoite transformation has been characterized both ultrastructurally and, to some degree, mechanistically (Soete et al., 1993; Bohne et al., 1996; Weiss et al., 1996; Boothroyd et al., 1997; Sasono and Smith, 1998). For example, only bradyzoites are surrounded by a cyst wall and contain large amounts of amylopectin granules (Sasono and Smith, 1998; Weiss and Kim, 2000). One of the key signals in vivo that controls parasite replication, and presumably bradyzoite formation, is the host production of γ-interferon. In vitro models of tachyzoite to bradyzoite differentiation have been established, some of which mimic the ‘stress’ of the host’s immune response. These conditions include treatment with γ-interferon, mitochondrial inhibitors (atovaquone or antimycin A), high pH (pH 8.1) and high temperature (43°C) (Bohne et al., 1994; Soete et al., 1994; Tomavo and Boothroyd, 1995; Weiss et al., 1995; Silva et al., 1998). The molecular mechanisms that the parasite uses to sense the environmental signals have not been identified, but roles for cAMP and cGMP signalling pathways in bradyzoite formation have been reported (Kirkman et al., 2001). Additionally, multiple bradyzoite-specific antigens (SAG4, BSR4, MAG1, CST1, SAG2C/2D, p21, BAG1/5) and stage-specific enzymatic paralogues (LDH2, ENO1) have been identified providing further evidence that the two stages are structurally and biochemically distinct entities (Tomavo et al., 1991; Bohne et al., 1995; 1996; Parmley et al., 1995; Yang and Parmley, 1995; 1997; Denton et al., 1996; Yahiaoui et al., 1999; Toursel et al., 2000; Holpert et al., 2001). Control of bradyzoite-specific gene expression has been demonstrated at both transcriptional and post-transcriptional levels (Yang and Parmley, 1997; Knoll and Boothroyd, 1998), but little is known about the mechanisms and genetic pathways that regulate these events.

Using bradyzoite-specific expression of green fluorescent protein (GFP) as a selection strategy, we have isolated mutant parasites that do not express GFP and do not convert to bradyzoites under bradyzoite-inducing conditions. The characterization of the mutants and, in combination with a bradyzoite cDNA microarray, their use in identifying genetic pathways in T. gondii development are presented.

Results

Generation of a Toxoplasma clone that expresses GFP in a bradyzoite-specific manner

To obtain bradyzoite-specific expression of GFP, we engineered a plasmid (pLDH2-GFP) in which the GFP coding region is flanked by the upstream and down-stream regions of the bradyzoite-specific LDH2 gene. Endogenous LDH2 message is undetectable in tachyzoites but is induced 1–2 days after a high pH switch in vitro (Yang and Parmley, 1997). After stable integration of this construct into tachyzoites, three rounds of switching to bradyzoite conditions and fluorescence-activated cell sorting (FACS) for GFP-expressing parasites, we isolated a clone (BSG-4) (bradyzoite-specific GFP expressor-4) that specifically expresses GFP under bradyzoite conditions. GFP expression in this clone is apparent 72 h after a high pH in vitro switch and correlates with cyst wall biosynthesis, as determined by staining with the Dolichos biflorus agglutinin of the CST1 antigen (Figs 1A and 2A). Although weaker, both markers are also clearly evident by 48 h after switching (data not shown). Note that the CST1 staining can sometimes appear to precede GFP expression [as seen for one of the cysts in Fig. 2B, WT (BSG-4)], although we believe that this represents detection sensitivity (as CST1 is detected with a lectin versus autofluorescence of GFP) and that the relative temporal expression pattern of CST1 and GFP correlate closely.

Figure 1.

FACS analysis showing bradyzoite-specific expression of GFP in WT (BSG-4) but not TBD-1 parasites. Parasites were subjected to high-pH in vitro switch conditions for 72 h, and GFP fluorescence was analysed by FACS. The GFP fluorescence is on the x-axis and the cell number is on the y-axis. The GFP fluorescence of the tachyzoite population and bradyzoite population are labelled.

A. WT (BSG-4) parasites.

B. TBD-1 mutant.

Figure 2.

TBD-1 parasites have reduced GFP fluorescence and cyst wall biosynthesis under multiple bradyzoite-inducing conditions. Parasites were subjected to bradyzoite-inducing conditions, and IFA was performed for the cyst wall with Dolichos biflorus lectin. GFP detection was by autofluorescence.

A. pH 8.1 for 72 h.

B. Treatment with atovaquone for 6–7 days.

As seen by FACS analysis, a portion of the parasites do not express GFP under bradyzoite-inducing conditions (Fig. 1A). These parasites may express GFP at low levels, may be non-viable under bradyzoite conditions and/or may not have differentiated into bradyzoites [the variability of in vitro bradyzoite induction has been described previously (Soete et al., 1993)]. Similarly, using standard tachyzoite culture conditions, up to 10% of the BSG-4 parasites are sometimes observed to express GFP (the ‘tail’ in the tachyzoite sample of Fig. 1A) and low levels of the CST1 protein (as assessed by staining with the Dolichos biflorus agglutinin; data not shown). This phenomenon of ‘spontaneous switching’ in tachyzoite cultures has been described previously with type II strains of T. gondii (Bohne et al., 1993).

Southern blot analysis of BSG-4 reveals that it contains three GFP-containing plasmids, integrated at a single site in the Toxoplasma genome, in a tandem orientation (data not shown). The site of integration and which exact repeat(s) is responsible for the GFP expression are not known but, in every respect tested, these parasites behave as wild type in their differentiation properties including induction of the known bradyzoite-specific markers (CST1, BSR4, BAG1, LDH2, SAG4A and MAG1) and their ability to make morphologically normal cysts during infection in a murine model (data not shown). This strain will henceforth be referred to as ‘wild type’ (WT).

Generation of differentiation mutants by sequential rounds of switching and sorting

After chemical mutagenesis of WT parasites, sequential rounds of switching to bradyzoites and sorting for Gfp parasites, we identified populations that were substantially reduced in their ability to express GFP under bradyzoite conditions. From each of four independently mutagenized and selected populations, we cloned one ‘tachyzoite to bradyzoite differentiation’ (Tbd) mutant. Figure 1B shows the FACS analysis of TBD-1 under high pH bradyzoite conditions: in contrast to WT, TBD-1 parasites show only rare instances of GFP expression (note the slightly longer ‘tail’ on the FACS analysis, indicating a small percentage of vacuoles with detectable fluorescence). TBD-2, -3 and -4 have indistinguishable FACS profiles compared with TBD-1 (data not shown). The Tbd mutants were also analysed for their ability to express GFP at later stages in bradyzoite induction and were again found to show only rare instances of GFP expression even 4–7 days after bradyzoite induction (data not shown).

In vitro analysis of mutant parasites

The Tbd parasites also show a defect in the levels of the bradyzoite cyst wall protein (CST1) synthesized under high pH conditions (Fig. 2A) or in the presence of atovaquone (Fig. 2B). Indistinguishable results were obtained for all four mutants (data not shown). For each mutant, a small subset (<10%) does express CST1 (Fig. 2B), as was seen with GFP (Fig. 1B). The quality of the CST1 staining on the Tbd mutants varies and can consist of a thin rim of cyst wall staining with Dolichos biflorus agglutinin as shown in Fig. 2A (lower right vacuole) or some brighter, albeit incomplete, cyst wall staining as shown in Fig. 2B. Note that there is a background level of Dolichos biflorus agglutinin staining of all parasite vacuoles (as seen in Fig. 2A, TBD-1, upper left vacuole), which is distinct from the Dolichos staining of the cyst wall specific to bradyzoites (Fig. 2A and B, WT). The rare CST1+/Gfp+ vacuoles in the Tbd mutants are also positive for other bradyzoite markers and appear to progress normally along the previously described pathway, expressing early (BSR4, BAG1), middle (LDH2, CST1, SAG4A) and late (p21) bradyzoite markers in a normal temporal manner, as determined by immunofluorescence assays (IFAs; data not shown). To confirm the results of the IFAs, we used Western blot analysis and saw substantially decreased expression of GFP, SAG4A, BSR4 and BAG1 in the Tbd mutants (data not shown). Growth and replication rates of the WT and Tbd parasites under both tachyzoite and bradyzoite conditions showed no detectable differences, as measured by uracil incorporation assays (data not shown).

In vivo analysis of mutant parasites

To characterize the in vivo phenotype of the Tbd mutant parasites, we tested WT, TBD-1 and TBD-2 in a murine model of chronic infection. Five independent experiments were performed, and the brain cyst load was measured at various time points of ‘early’ (days 17–18 after injection) and ‘late’ (days 23–26 and days 49–51) chronic infection. Two representative experiments are shown (Fig. 3). Brain cyst burden, as assessed by GFP-positive cysts, was severely attenuated in both mutants at ‘early’ and ‘late’ time points tested. As the Tbd parasites were selected on the basis of decreased GFP expression under bradyzoite conditions, we also measured brain cyst burdens using Dolichos biflorus agglutinin staining for CST1 and observed closely similar decreases relative to WT. Additionally, we quantified pepsin-resistant cysts in the brain material as a functional assay of bradyzoite formation and, again, saw a parallel decrease compared with WT (12% of WT levels for TBD-1 and 8.9% of WT levels for TBD-2; data not shown). The overall architecture of the in vivo WT and rare Tbd mutant cysts was grossly similar; in all cases, the cysts showed no staining with SAG1 (tachyzoite-specific), but similarly positive staining with bradyzoite-specific SAG4A, CST1 and p21 (data not shown). There was no difference in the clinical course of the mice during either acute or chronic infection with either WT or Tbd mutant parasites.

Figure 3.

TBD-1 and TBD-2 have a markedly attenuated cyst burden in outbred mice. Equal number of tachyzoites were injected i.p. into female outbred Swiss Webster mice using groups of five mice for each strain. Mice were sacrificed at days 23–26, and brain cyst burden was assessed in 100 μl of brain material. Data points are shown for two independent trials (A) and (B). The y-axis is a log scale for (A) and a linear scale for (B). The mean cyst burden is labelled with (—) for each parasite strain. All mice that had no cysts were serotested, and only data for seropositive mice are shown.

Interestingly, there was variability in cyst burden, most notably in WT and TBD-2 parasite strains. In two of the mice infected with TBD-2 (e.g. one mouse in Fig. 3B), there were large numbers of brain cysts, even though the mice exhibited no clinical symptoms of a severe infection. Bradyzoites from brain cysts were harvested from mice infected with WT, TBD-1 and TBD-2 (including the mouse with the high cyst burden shown in Fig. 3B), by pepsin digestion, a method that should allow only the CST1+ parasites to survive. Once maintained in tissue culture, these parasites were checked for their ability to differentiate in vitro and, in all cases, there was no change in phenotype compared with the starting parasite cultures. Thus, passage through a mouse did not attenuate or enhance the phenotype of the mutants and, importantly for TBD-2, the large cyst number did not represent a revertant phenotype.

Transcriptional profile of Tbd mutants

To assess which genes are key to differentiation and to develop a detailed genetic phenotype of each mutant, we compared the transcriptional profiles of WT and Tbd parasites under bradyzoite conditions using cDNA microarray analysis. A T. gondii bradyzoite cDNA microarray has been generated from the bradyzoite cDNA library (Manger et al., 1998) containing ≈ 4000 expressed sequence tags (ESTs) and representing a total of at least 613 contigs (Cleary et al., 2002 ). Using a time course, an expression profile of parasites undergoing in vitro differentiation to bradyzoites has been generated and compared with tachyzoites (Cleary et al., 2002). Transcripts corresponding to 31 genes were identified as being ‘induced’ in bradyzoites (i.e. ≥1.95-fold more abundant in bradyzoites than in tachyzoites), 17 genes were identified as being ‘repressed’ or less abundant in bradyzoites (≤0.66-fold the tachyzoite levels), and 182 contigs were identified as having constitutive expression profiles (Cleary et al., 2002). The remaining contigs were classified as ‘intermediate’ (increased expression in bradyzoites but did not meet our induced criteria) or ‘indeterminate’ (too variable between experiments to call with confidence).

We have used this cDNA microarray to analyse Tbd mutants and identified subsets of genes that are expressed in the mutants at significantly lower levels compared with WT parasites (Tables 1 and 2 and Web Fig. 2; see Supplementary material). Overall, TBD-1 had 44 genes, TBD-2 had 40 genes, TBD-3 had 64 genes, and TBD-4 had 57 genes that showed significantly lower transcript levels compared with WT bradyzoites 3 days after transfer to high pH medium. A substantial set of the genes with reduced expression in the Tbd mutants was previously identified as being bradyzoite induced in WT parasites. For TBD-1, -2, -3 and -4, 21/44 (48%), 19/40 (47%), 21/64 (33%), and 23/57 (40%), respectively, of genes with reduced expression were genes previously classified as being bradyzoite induced. The remainder of the genes with decreased expression in each Tbd mutant fell into the ‘intermediate’, ‘constitutive’ or ‘indeterminate’ categories. For each Tbd mutant, we also analysed the number of genes that were significantly higher in the mutant compared with WT parasites (e.g. were induced in the mutant parasites under bradyzoite conditions compared with WT parasites under similar conditions). In this category, TBD-1 had 19 genes, TBD-2 had 20 genes, TBD-3 had nine genes and TBD-4 had six genes.

Table 1. Expression patterns of bradyzoite-induced genes in the Tbd mutants.
B/T day3ContigGene nameTBD-1TBD-2TBD-3TBD-4
  1. The 31 T. gondii bradyzoite-induced genes and their relative expression patterns in the Tbd mutants under bradyzoite conditions compared with WT parasites under similar conditions. The induction level in WT bradyzoites at day 3 versus tachyzoites (B/T day3), contig number, gene name (based on BLAST search) and relative expression level in the four Tbd mutants versus WT bradyzoites at day 3 are shown. The contigs, other than those that are reduced in all mutants, which are included in our ‘model’ (Fig. 4) are labelled with an asterisk (*) in column 2. The genes are shown in five categories (category 1–5): (1) decreased expression in all mutants tested; (2) decreased expression in three/four mutants; (3) decreased expression in two/three mutants; (4) decreased expression in one/four mutants; and (5) those with WT expression levels. The values that meet our ‘reduced expression’ criteria (relative expression + SEM <0.75) are in bold; values that do not meet our criteria are in plain font or plain font with an asterisk (for genes that have mean relative expression <0.75 but in which the SEM is high and therefore the relative expression + SEM is not <0.75). Genes for which data are not available for a given mutant are listed as ‘NA’. B/T day3-values below 1.95 are from genes found to be ‘induced’ only in day 4 bradyzoites. Ctoxoqual_3500 is ‘induced’ only in day 2 bradyzoites.

Category 1
4.814135 SAG2C/D 0.09 0.070.220.14
2.873897Mucin domain 0.07 0.100.200.18
2.783956 + 4040  0.10 0.090.220.21
3.694227 NA0.050.220.38
4.943985 + 4432  0.29 0.320.140.15
2.443993  0.08 0.180.380.30
3.094054 + 4243 LDH2 0.23 0.270.360.31
2.024080Met aminopeptidase 0.22 0.250.370.48
2.784130 SRS9 0.33 0.410.290.32
3.454192 + 4416  0.37 0.410.480.37
2.343973  0.33 0.590.460.27
2.461284Oligopeptidase 0.41 0.430.640.60
2.174062  0.48 NA0.550.54
2.433906 + 4342 BAG1 0.39 0.480.680.61
1.914131 + 4245VEE repeat protein 0.47 0.650.600.58
Category 2
1.633915*  0.47 0.45 0.45 0.83
2.514068*  0.08 0.83 0.07 0.35
4.464196* SAG4A 0.08 0.71* 0.16 0.12
1.554217*  0.55 1.60 0.57 0.52
2.544309 0.45* 0.22 0.69 0.55
Category 3
3.424101*  0.18 0.110.74*1.00
2.714343* 1.220.96 0.13 0.33
Category 4
2.933908 ENO1 0.790.800.62* 0.56
2.333919 0.65*0.72*0.68* 0.62
2.433969 + 4095*  0.07 0.57*0.62*0.49*
2.054074 0.61*0.940.86 0.66
2.384210* 0.65* 0.43 0.900.80
1.974303  0.63 0.760.71*0.94
1.854436 DRPA 0.65* 0.60 0.790.85
Category 5
2.033500 NA1.141.15NA
1.473900 NANA1.151.20
Table 2. Expression patterns of nonbradyzoite-induced genes in the Tbd mutants.
B/T Day3ContigGene nameTBD-1TBD-2TBD-3TBD-4
  1. Toxoplasma gondii contigs that are not bradyzoite induced but have reduced expression in at least two out of the four Tbd mutants relative to their expression patterns in WT parasites under similar conditions. The expression level in WT bradyzoites at day 3 versus tachyzoites (B/T day3), contig number, gene name (based on BLAST search) and relative expression level in the four Tbd mutants versus WT bradyzoites are shown. The genes are shown in three categories (categories 1–3): (1) decreased expression in all Tbd mutants tested; (2) decreased expression in three/four mutants; and (3) decreased expression in two/three mutants. Bold font, NA and asterisk are as in the footnote to Table 1.

Category 1
0.98369014-3-3 0.33 0.48 0.61 NA
NA4179Vesicle transport proteinNA 0.08 0.17 0.32
1.844321 NA 0.27 0.29 0.64
Category 2
1.331000  0.61 0.64 0.63* 0.74
1.722161 MAG1/p65 0.41 0.38 1.02 0.44
2.403008  0.73 0.67 0.29 0.45*
1.384018 0.81 0.64 0.57 0.58
1.364061  0.51 0.74 0.76 0.63
1.744149  0.65 0.47 0.70* 0.43
Category 3
1.422047  0.74 0.961.32 0.66
1.092071  0.60 0.811.31 0.67
1.093972PITSLRE kinase1.19NA 0.72 0.70
1.363980Vacuolar ATPase 0.75 NA 0.36 1.32
1.763981 0.990.92 0.62 0.55
1.114 0.89NA 0.60 0.56
1.204126  0.71 0.66 1.101.08
1.364184  0.52 1.15 0.58 1.13
1.394231  0.49 0.61 1.191.47
NA4233 1.230.99 0.73 0.59
1.364239 0.73*0.70* 0.64 0.57
0.914244 1.031.73 0.46 0.35
1.164333  0.71 0.811.19 0.50
1.06942  0.70 0.73 1.111.90

Of the 31 bradyzoite-induced genes identified previously using WT bradyzoites and the cDNA microarray (Cleary et al., 2002), 15/31 (48%) genes had reduced expression in all Tbd mutants tested, an additional 5/31 (16%) were decreased in 75% of mutants tested, and another 2/31 (6%) were decreased in 50% of mutants tested. A total of 7/31 (22%) bradyzoite-induced genes were decreased in only one Tbd mutant, and 2/31 (6%) had WT expression levels in all Tbd mutants tested (Table 1). Genes with decreased expression in all Tbd mutants included the bradyzoite-induced surface antigens [SAG2C/D, SRS9 (predicted)], enzymes (LDH2, methionine aminopeptidase, oligopeptidase), heat shock proteins (BAG1), a gene with homology to the mucin domain-containing homologue of a Cryptosporidium parvum surface protein (Barnes et al., 1998) and a gene with conserved VEE repeats that are found in the D260, RESA-H3 and LSA-3 proteins of P. falciparum (al-Yaman et al., 1997).

The expression profile of the Tbd mutants for non-bradyzoite-induced genes was characterized (Table 2), and three genes were identified that had decreased expression in all the Tbd mutants tested; six additional genes were decreased in 75% of the mutants, and another 14 genes were decreased in 50% of the Tbd mutants. As expected, for all but two of the genes that are normally ‘repressed’ in WT bradyzoites at day 3 of bradyzoite development, the Tbd mutants had higher expression levels (data not shown). The two exceptions were Ctoxoqual_4801, a gene for which we have expression data in only one mutant and with only one spot that met our filter criteria, and Ctoxoqual_3007, which had lower expression in one Tbd mutant but higher than WT levels in the other Tbd mutant tested. Out of the 182 genes identified as having constitutive expression in the WT bradyzoites versus tachyzoites (Cleary et al., 2002), 17 had significantly decreased expression in at least one Tbd mutant, three had reduced expression in two Tbd mutants, and the remainder (14 genes) had reduced expression in only one mutant. Genes with reduced expression in multiple Tbd mutants included Ctoxoqual_3972 (with homology to PITSLRE kinase), Ctoxoqual_4 and Ctoxoqual_4244. The remainder of the genes with identifiable homologies that were decreased in one Tbd included histone H2A variant(Ctoxoqual_260), cyclophilin (Ctoxoqual_4538), ubi-quitin-like protein (Ctoxoqual_4143) and a variety of ribosomal genes. All data for the microarray portion of the paper are available on www.stanford.edu/~blader/Usinghetalwebfig1.xls (complete microarray data set for all ESTs) and www.stanford.edu/~blader/Usinghetalwebfig2.xls (data set assembled by contigs highlighting contigs with decreased expression in the Tbd mutants compared with WT parasites).

Discussion

Using a GFP-based selection, we have identified four tachyzoite to bradyzoite differentiation mutants and analysed the transcriptional profile of these via a cDNA microarray. Using this approach, we have identified a hierarchy of genes associated with bradyzoite formation and generated a model for the cascade that leads to changes in their respective transcript levels (Fig. 4). We have also identified a class of genes that, although not developmentally regulated, have decreased transcript levels in most of the mutants tested and may thus be dependent on similar transcription factors and/or serve as the ‘gatekeepers’ for the induction of bradyzoite formation. This work lays the foundation for understanding the developmental pathways from tachyzoite to bradyzoite differentiation.

Figure 4.

A proposed model of hierarchical gene expression in T. gondii. Multiple stimuli (pH, atovaquone, γ-interferon, high temperatures) all lead into a pathway that results in the induction of bradyzoite-induced genes. A cascade reflecting a hierarchy of bradyzoite-induced gene expression in the Tbd mutants is shown.

Interestingly, all four Tbd mutants have a leaky phenotype, in that ≈ 10% of the parasite population still differentiates under bradyzoite conditions. This could represent a limitation of our selection, i.e. tight mutants might die in the high pH, bradyzoite-inducing conditions of the selection. Alternatively, it might be that mutations that completely disrupt differentiation are not viable even as tachyzoites, although this seems less likely. Of note, Tbd mutants generated in a different strain (RH) by a different method (insertional mutagenesis) and selection [using hypoxanthine–xanthine–guanine phosphoribosyltransferase (HXGPRT); please refer to accompanying paper by Matrajt et al., 2002) also have a leaky phenotype. The fact that those Tbd parasites that do develop into bradyzoites apparently have WT levels of all markers (as measured in in vivo cysts) suggests that the defect in the mutants is in the core pathway of differentiation and that this step is only partially blocked or somehow circumvented at low frequency. Likewise, the commitment to the developmental pathway in WT parasites is not absolute: only ≈ 85–90% of parasites switch to bradyzoites in vitro using even the most efficient stimulus (high pH) (Soete et al., 1993). Whether the non-differentiating parasites are not viable or not responsive to the differentiation signals is not clear. In vivo, WT tissue cysts can be observed with occasional tachyzoites (defined by the expression of key tachyzoite-specific antigens) among the many bradyzoites, although whether these represent parasites that failed to differentiate into bradyzoites or ones that are switching back to tachyzoites cannot be determined (Ferguson et al., 1994).

The Tbd mutants, although selected exclusively under high-pH conditions, were found to be resistant to another switch stimulus – treatment with sublethal doses of the mitochondrial inhibitor atovaquone. The mechanisms by which different stresses induce bradyzoite formation are not known, although the temporal appearance of bradyzoite-specific genes has previously been found to be closely similar using different methodologies (Tomavo and Boothroyd, 1995; Soete and Dubremetz, 1996). Our data further corroborate the notion that multiple bradyzoite-inducing conditions lead into a common pathway of bradyzoite-specific gene development.

The leakiness of the in vivo phenotype of the Tbd mutants was similar to the in vitro phenotype, indicating that an in vitro selection can be used to generate parasites with the desired in vivo phenotype. The decreased cyst burden in mice infected with TBD-1 was consistent, in contrast to the occasional ‘breakthrough’ phenotype of TBD-2. These rare instances of apparent breakthrough could be caused by a spurious result during intraperitoneal injection (e.g. into a particularly sensitive organ), an effect of host genetics (outbred mice were used) or an ability of this particular mutant to ‘circumvent’ the blocked regulator of differentiation.

The generation of these Tbd mutants with an attenuated cyst burden and a robust serological response to tachyzoites indicates that these parasites might serve as a useful basis for the development of a vaccine strain for livestock. Identification of the mutated gene by complementation is hindered in T. gondii by two main facts: the leaky phenotype of our mutants and significant difficulties, to date, in complementing complex phenotypes. The 10% leaky phenotype, both in vitro and in vivo, makes selections for revertants very inefficient. Additionally, the techniques currently available in T. gondii, although successful for clean auxotrophic defects (Donald and Roos, 1995; Black and Boothroyd, 1998), have not yet been used successfully to complement more subtle phenotypes.

Using a cDNA microarray and the Tbd mutants, we have generated a transcriptional profile for Tbd mutants under bradyzoite conditions. Using this approach, we could hope to identify three categories of genes: (i) the gene(s) that is(are) mutated in the Tbd mutants; (ii) genes that are along the differentiation pathway and involved in bradyzoite formation or maintenance; and (iii) genes that are co-ordinately controlled by the same transcription factor that controls bradyzoite-specific gene induction. We used chemical mutagenesis to increase the chance of obtaining mutants in essential genes (insertion into which will be lethal) and to obtain saturation of the genome (the transfection efficiency in the Pru strain is much lower than in RH). However, as a result, identifying the mutated genes (category ‘i’ above) is more difficult, and so we have focused on categories ‘ii’ and ‘iii’.

The Tbd mutants appear to have a global defect in differentiation; 20/31 (65%) of the genes identified as bradyzoite induced in WT parasites have decreased expression in at least 75% of the mutants tested. Additionally, 83% of the genes that have decreased expression in all Tbd mutants are bradyzoite induced. There are three genes that have decreased expression in all Tbd mutants tested but are not bradyzoite induced by our criteria. One gene (Ctoxoqual_4321) has a 1.83 induction in WT bradyzoites at day 3 compared with tachyzoites, which, although it does not meet our strict criteria, may be functionally bradyzoite induced. For another gene, Cotxoqual_4179, we do not have data for its expression profile during differentiation. The bradyzoite-induced genes that have reduced expression in all Tbd mutants are probably involved in bradyzoite maintenance. One of the genes that falls into this category, BAG1, has been successfully disrupted in T. gondii by two groups of investigators. One group (Zhang et al., 1999) observed a 78–93% decrease in cyst formation in vivo, whereas the other (Bohne et al., 1998), using a different starting strain, saw no clear phenotype in vitro or in vivo. There are no data on the essential nature of the other genes identified here, although attempts to disrupt the LDH2 gene have not been successful (U. Singh and J. C. Boothroyd, unpublished results; S. Parmley, personal communication). (Note that the technology for a regulated back-up copy of a gene has not yet been developed for T. gondii, and so essential genes cannot be deleted unless a fully active copy is also present.)

Only two of the bradyzoite-induced genes were expressed in all mutants at WT levels at the time point (day 3) tested, and a further seven were at WT levels in all but one Tbd mutant. These genes may represent a response to ‘stress’ (but not related specifically to differentiation) or, alternatively, genes that may be very early in the differentiation pathway, upstream of the genes that are disrupted in our mutants. Ctoxoqual_3500 was found to be induced only in day 2 bradyzoites, suggesting that this gene may act early in the differentiation pathway, and Ctoxoqual_ 3900 was induced only in day 4 bradyzoites, suggesting that the time point used in this study may fail to detect differences in expression for this gene. Whether this implies a ‘linear’ pathway to differentiation or indicates that our current selection strategy will only identify severely defective parasites is unclear. It should also be noted that we chose to identify and focus on mutants that were significantly altered in their GFP and CST1 expression. We did not pursue mutants in which the efficiency of switching was attenuated by only 50%. Future work on such parasites will aid in clarifying this issue. Using the expression profiles of the four Tbd mutants for the bradyzoite-induced genes, we have generated a working model reflecting a hierarchy in bradyzoite-induced gene development (Fig. 4).

We have also identified a class of non-bradyzoite-induced genes that have decreased expression in most of the Tbd mutants analysed. These may encode proteins that have a regulatory effect on bradyzoite formation or are co-dependent on common transcription factors. Among these genes, three are of special interest: a 14-3-3 homologue, an apparent PITSLRE kinase and a probable vacuolar ATPase. The 14-3-3 gene’s transcript is expressed in T. gondii in the tachyzoite (Ajioka, 1998), bradyzoite (Manger et al., 1998) and enteroepithelial stages as well as in the sporozoites of the parasites (Koyama et al., 2001). Expressed in all eukaryotic cells, the 14-3-3 proteins function by binding to a variety of signalling proteins including kinases, phosphatases and transmembrane receptors as regulatory molecules. Owing to the diverse interactions, 14-3-3 proteins affect a multitude of processes including cell cycle control, apoptosis and regulation of mitosis (Fu et al., 2000). The contig Ctoxoqual_3972 has homology with the PITSLRE protein kinases (P-value = 5.8 × 10−4). This family of protein kinases has been shown to function as a negative regulator in checkpoint control during mitosis, and overexpression in CHO cells decreases DNA replication (Bunnell et al., 1990). Bradyzoite development in T. gondii has been linked to decreased growth rates (Bohne et al., 1994). Thus, a downregulation in the PITSLRE protein kinase might increase parasite growth rates, thereby overcoming the signal for differentiation. We were not able reproducibly to identify a difference in the growth kinetics of our Tbd mutants. In contrast, Matrajt et al. (2002) report a markedly increased growth rate under bradyzoite conditions for their mutants, which are on the RH background (Matrajt et al. 2002). Whether this strain difference or some other factor is responsible for this heightened replication rate is not yet clear.

Ctoxoqual_3980 has clear homology with Plasmodium falciparum vacuolar ATPase A subunit (P-value = 1.1 × 10−42). These enzymes regulate cytoplasmic pH in T. gondii tachyzoites (Moreno et al., 1998), and bradyzoite-specific expression of a P-type ATPase has been reported (Holpert et al., 2001), indicating that tachyzoites and bradyzoites may have independent means of regulating their intracellular pH. Multiple roles relating to growth and differentiation have been attributed to this enzyme, including regulation of cell proliferation (Manabe et al., 1993), phenotypic modulation of myofibroblasts (Otani et al., 2000), mouse post-implantation and early embryonic development (Inoue et al., 1999; Sun-Wada et al., 2000) and plant growth and development (Schumacher et al., 1999).

Using the combined genetic and genomic approaches, we have generated a hierarchical genetic pathway for bradyzoite formation in T. gondii, providing an initial template for understanding the developmental biology of this parasite. A similar genetic approach to the analysis of development in Toxoplasma has been performed by Matrajt et al. (2002). They too found a hierarchy of genes based on how they are affected in the mutants. Through such approaches, we can now focus on genes that appear to occupy key positions in the differentiation cascade and functionally test genes identified as important in bradyzoite formation using targeted gene knock-out. A similar reverse genetics approach was undertaken to identify genes required for meiosis and spore formation in yeast (Rabitsch et al., 2001). Disruption of genes identified as late in the cascade should affect bradyzoite maintenance but not initial bradyzoite formation. In contrast, disruption of genes that appear to be early in the cascade should result in parasites that are unable to initiate differentiation. Functionally characterizing genes in the bradyzoite differentiation pathway of T. gondii is a key step in fully understanding the developmental biology of this important and complex parasite.

Experimental procedures

Parasite culture

All tissue culture was done with the Pru strain of Toxoplasma, which has been engineered to lack HXGPRT: Pru/hxgprtΔ (a gift from D. Soldati; Donald and Roos, 1998). Parasites were grown in human foreskin fibroblasts (HFFs) under standard Toxoplasma culture conditions, transfected and cloned by limiting dilution as described previously (Knoll and Boothroyd, 1998). Parasite growth rates under tachyzoite and bradyzoite conditions were measured using a radiolabelled uracil incorporation assay (McFadden et al., 1997).

In vitro differentiation

In vitro bradyzoite induction procedures were performed with the high pH (8.1) method as described previously (Soete et al., 1993; Weiss et al., 1995; Silva et al., 1998) or by atovaquone (Tomavo and Boothroyd, 1995) with the following modifications. Approximately 2 × 105 parasites were allowed to invade cells in a 24-well plate for 4 h. The medium was then replaced with tachyzoite medium supplemented with 800–1000 μM atovaquone for 6–8 days. Bradyzoite induction under all methods was assessed and followed by inverted fluorescence microscopy and GFP expression or cyst wall detection using the Dolichos biflorus lectin (see below).

Plasmids

Replacing the SAG1 coding region in the plasmid pLDH2-SAG (Seeber et al., 1998) with the coding region for GFP generated the plasmid pLDH2-GFP. We used a variant of GFP that has been optimized for fluorescence in T. gondii (Kim et al., 2001). Briefly, pLDH2-SAG was digested with NsiI and PacI to remove the SAG1 coding region and ligated with the GFP coding region (purified from pGRA-GFP by digestion with NsiI and PacI). The plasmid pLDH-GFP-TUB-CAT was constructed by putting the LDH2 5′-GFP-LDH2-3′ cassette in a 5′-to-5′ orientation in the pTUB-CAT vector (Kim et al., 1993): pTUB-CAT was digested with HindIII and KpnI, blunted, and the LDH2 5′-GFP-LDH2-3′ insert (digested with XhoI, SacI, XmnI and blunted) was ligated into the vector.

Transfection and selection

Electroporation was carried out as described previously (Kim et al., 1993; Soldati and Boothroyd, 1993; Black et al., 1995). pLDH-GFP-TUB-CAT (50 μg) was linearized at the XmnI site and integrated into Toxoplasma using restriction enzyme-mediated integration (REMI) with NotI. Parasites were selected with 20 μM chloramphenicol, and stable populations were generated. This population was switched to bradyzoites using a high-pH induction and then passed through a fluorescence-activated cell sorter (FACS) to select parasites with high fluorescence. The parasites were then expanded in tissue culture as tachyzoites and subjected to three more rounds of bradyzoite switching and Gfp+ sorting before being cloned by limiting dilution. Clones were analysed by FACS, and those that were Gfp under tachyzoite conditions and Gfp+ under bradyzoite conditions were selected for further analysis.

Southern blot analysis

DNA from the selected clones was probed with the GFP coding region and the 5′ region of the LDH2 promoter to assess the number of plasmid inserts and their orientation. A clone characterized in this manner, bradyzoite-specific GFP expresser (BSG-4), was subsequently used for mutagenesis and mutant selection.

Mutagenesis

Chemical mutagenesis was done using the standard protocol with ethylnitrosylurea (ENU; 300 μg ml−1) at 37°C for 1 h (Radke et al., 2000).

FACS analysis and sorting

For live cell sorting of bradyzoites, bradyzoite induction medium was removed, cells were washed once with PBS, the monolayer was scraped and passed twice through a 27-gauge needle, followed by once through a 30-gauge needle to release parasites from the host cells. These were then centrifuged at 1800 r.p.m. for 10 min at room temperature and resuspended in sterile PBS or tachyzoite medium (without phenol red) for FACS analysis. Tachyzoite cultures were processed in a similar manner except that release from the HFFs used only a 27-gauge needle. FACS sorting was done on a FACSTAR cell sorter with sorted parasites being put back into tachyzoite tissue culture conditions and grown until confluent. FACS analysis was done on a SCANFORD machine using parasites treated with 3% formaldehyde for 20 min at room temperature.

Immunofluorescence assay (IFA)

Toxoplasma was grown in HFFs on glass coverslips in 24-well plates for 3 days under high-pH bradyzoite conditions or for 6–8 days in 800–1000 μM atovaquone. Cells were fixed with 3% formaldehyde (20 min at room temperature), quenched with 100 mm glycine (2 min at room temperature), then permeabilized with 0.2% Triton X-100 in PBS (20 min at room temperature) and blocked with 3% bovine serum albumin (BSA) in PBS (1 h at room temperature or overnight at 4°C). All primary and secondary antibody incubations were for 1 h at room temperature. The Dolichos biflorus agglutinin (Vector) was used at 1:1000, and a monoclonal antibody against BSR4 (p36, Tg4A12; a kind gift from J. F. Dubremetz) was used at 1:300. Secondary antibodies were used at 1:1000 [streptavidin–Texas red; (Gibco BRL) and Texas red anti-mouse (Vector Laboratories)]. For immunofluorescence on fixed brain material from infected mice, primary antibodies were used at 1:300 dilution, and secondary antibodies were used at 1:1000 dilution. All samples were visualized on an Olympus BX60 microscope with a 35 mm camera.

In vivo phenotype of Tbd mutants

Outbred female Swiss Webster mice (8–9 weeks old) were injected intraperitoneally (i.p.) with 500 tachyzoites harvested from healthy, non-lysing cultures, as described for FACS sorting above, and resuspended in sterile PBS. Input numbers of parasites were checked by plaque assays performed in quadruplicate for each of the three strains. Mice in the chronic phase of infection (days 17–18, 23–26 or 49–51) were euthanized, and serum and brains were harvested for analysis. Brain material was ground gently with a mortar and pestle, 1 ml of PBS was added, and the material was put through an 18-gauge needle three times. Five 20 μl samples of this material were loaded on slides, covered with a coverslip and analysed for cyst number using an Olympus BX60 microscope. A total of 100 μl of brain material was counted for each mouse at each time point. An equal portion of each mouse brain was also digested in 170 mm NaCl–pepsin (0.1 mg ml−1)–60 mm HCl for 1 min at 37°C and then neutralized with 94 mm Na2CO3 (Manger et al., 1998). These parasites were placed in 24-well plates containing HFFs, and plaque numbers were counted at 5–7 days after inoculation. A sample of the digested parasites was put on a fresh HFF monolayer and maintained in tissue culture under standard conditions for subsequent testing of the efficiency of their differentiation. The remainder of the brain material was prepared for IFA as described above. The brain cyst burden in all mice was assessed by (i) GFP expression in cysts; (ii) CST1 staining for the cyst wall; (iii) pepsin digestion and resistance. Additionally, the brain material from a subset of mice was stained with antibodies to tachyzoite (SAG1) and bradyzoite (SAG4A, p21) antigens. These markers did not identify any additional parasite material in the brain tissue. Successful infection of the mice was confirmed by Western blot analysis using serum from the infected mice and whole Toxoplasma tachyzoite lysate (Dzierszinski et al., 2000).

RNA extraction

Parasites grown for 72 h in bradyzoite-inducing conditions were harvested by removing the media, scraping the monolayer and performing sequential rounds of syringing (a 27-gauge needle twice and a 30-gauge needle once). Parasites were pelleted by spinning at 2600 r.p.m. for 15 min at room temperature, and RNA was harvested using Trizol reagent (Gibco BRL). RNA was stained with SYBR green (Molecular Probes), and the fluorescence intensity of the large-subunit ribosomal RNA of HFFs and T. gondii was measured to estimate the percentage of human and Toxoplasma RNA in each preparation (human rRNA is substantially larger than that of Toxoplasma).

Microarray experiments

The cDNA microarray was generated from the in vivo bradyzoite EST library (Manger et al., 1998). Details of the genes represented, PCR conditions, hybridization and other controls were as described previously (Cleary et al., 2002). Briefly, 3 μg of total T. gondii bradyzoite RNA was used in a first-strand cDNA synthesis reaction (Superscript II). The cDNA was resuspended in 30 μl of TE and subjected to a labelling step with random nanomers, Klenow and a Cy5 fluorophore (Amersham Pharmacia Biotech). A reference sample (500 ng of PCR product using the T3 and T7 primers on the Bluescript-II SK vector) was labelled concurrently with the Cy3 fluorophore. The labelled products were mixed and hybridized on a microarray at 65°C overnight. The microarrays were washed, scanned with a GenePix 4000 microarray scanner (Axon Instruments) and gridded using the SCANALYZE program (written by M. Eisen and available on the Web at http://rana.lbl.gov/EisenSoftware.htm). A minimum of two in-dependent bradyzoite inductions and microarray experiments were performed with each of the WT and mutant strains. All data analysis was performed using the Stanford microarray software database (Sherlock et al., 2001) and normalizing the Cy5 signal intensity for each spot by the Cy3 intensity for that same spot. The reproducibility of the microarray experiments was assessed by calculating the correlation coefficient values from two independent, replicate experiments for each of the strains. The particular arrays used for these analyses lacked sufficient representation of the highly abundant tachyzoite ESTs, and so data for those will not be presented or discussed.

The algorithm for data analysis has been detailed previously (Cleary et al., 2002). Briefly, data were collected for all ESTs that met a spot quality filter. For each parasite strain, the experimental variability of each EST was evaluated in the two arrays by calculating the magnitude of the mean deviation relative to the mean ratio, and all ESTs for which the value (avg dev/avg) was >0.40 were removed from the analysis. Data for all remaining ESTs were averaged for each of the mutant strains and WT parasites, and the relative expression value of each EST was calculated as: avg Tbd mutant expression/avg WT expression (avg Tbd/avg WT). The data were compiled into the previously assigned gene assemblies, and an average relative expression level and standard error of the mean (SEM) were determined for each contig. A ratio <0.75, including SEM, was defined as significantly below WT (i.e. a ratio of 0.60 ± 0.10 was considered significant but 0.70 ± 0.10 was not). We chose a 0.75 cut-off to allow for both the 10%‘leaky’ differentiating phenotype and experimental variability. The initial identification of bradyzoite-induced, bradyzoite-repressed or constitutive genes has been validated by both Northern blot and SAM analysis (Cleary et al., 2002). The SAM technique identifies whether differences in mean expression levels, regardless of their magnitude, are statistically significant based on repeat measurements of a given spot (Tusher et al., 2001). This approach has the greatest statistical significance with more than two independent measurements and therefore was not applied to our Tbd mutant analysis.

Acknowledgements

We gratefully acknowledge J. F. Dubremetz for the Tg4A12 antibody, Dominique Soldati for the Pru/hxgprtΔ strain, Kami Kim for the GFP-containing plasmid, and Mark Gilbert and Tim Knaak for assistance with FACS analysis. We thank Sandeep Jaggi for analytical support, Kimberly Chong for microarray preparation, Gavin Sherlock and the Stanford Microarray Database for bio-informatics support and guidance, Michael Cleary, Ira Blader and all members of the Boothroyd laboratory for scientific discussions and helpful suggestions. This work was supported by the National Institutes of Health (grants AI41014 to J.C.B. and K08 AI01453 to U.S.). U.S. is funded by a Burroughs Wellcome Fund Career Devel-opment Award.

Ancillary