Identification of Toxoplasma gondii Genes Responsive to the Host Immune Response during In Vivo Infection

Toxoplasma gondii is an obligate intracellular protozoa parasite that causes the disease toxoplasmosis. It resides within host cells in a parasitophorous vacuole distinct from the host cell endocytic system. T. gondii was used as a model to investigate how obligate intracellular parasites alter their gene expression in response to the host immune response during infection compared to growth in host cells in vitro. While bacterial pathogens clearly alter gene expression to adapt to the host environment during infection, the degree to which the external environment affects gene expression by obligate intracellular pathogens sequestered within host cells is less clear. The global transcriptome of T. gondii was analyzed in vivo in the presence and absence of the IFN-γ-dependent host innate immune response. The parasites' in vivo transcriptome was also compared to its transcriptome in vitro in fibroblast cells. Our results indicate that the parasite transcriptome is significantly altered during in vivo infection in the presence, but not absence, of IFN–γ-dependent immunity compared with fibroblasts infected in vitro. Many of the parasite genes increased in vivo appear to be common to an early general stress response by the parasite; surprisingly putative oocyst stage specific genes were also disproportionately increased during infection.


Introduction
Most bacterial pathogens significantly alter their transcriptome within a host during infection compared to growth in vitro in culture. To do this pathogens must be able to sense changes in their microenvironment and translate those signals into alterations in gene expression important for pathogenesis including but not limited to protection against the host immune response. The degree to which obligate intracellular pathogens must sense and respond to their environment is unknown; as it is unclear how the environment in a host cell in vitro differs to that encountered in vivo during infection and how the immune response of the host effects this environment. Changes in the global transcriptome of a pathogen can be used as a tool to interrogate the specific specialized intracellular niche occupied by a pathogen during in vivo infection and how the environment in this niche is altered by a progressive immune response [1].
T. gondii is an obligate intracellular protozoal pathogen. It actively invades virtually any type of warm blooded vertebrate host cell [2], independent of phagocytosis, and replicates within a parasitophorous vacuole (PV) that is largely segregated from the endocytic system [3,4,5]. This is markedly distinct from the majority of intracellular pathogens that require uptake by phagocytosis and formation of a nascent phagosome prior to pathogen mediated vacuole modification or escape from the phagosome into the cytosol.
A number of elegant global gene expression studies have analyzed changes in gene expression during the parasite's lytic cycle in vitro in human fibroblasts (HFF cells) [6,7,8]. These have shown that T. gondii's transcriptome is highly cell cycle regulated with many genes transcribed maximally just prior to their use and then downregulated in a pattern described as ''just in time'' [7,8]. For example, genes important for parasite invasion of host cells are transcribed maximally just prior to parasite lysis from the host cell.
T. gondii also exhibits distinct developmental stages each with defined alterations in parasite gene expression. The asexual stages are the relatively quiescent bradyzoite stage contained within tissue cysts during chronic infection and the rapidly replicating tachyzoite stage that predominates during acute infection and reactivation of chronic infection. Feline hosts are the only known definitive host for the parasite where sexual reproduction occurs resulting in the fecal release of environmentally resistant oocysts that undergo sporulation in the environment to become infectious. Peroral infection with bradyzoite containing cysts or sporulated oocysts results in cyst rupture, invasion of the intestinal epithelium followed by differentiation to and systemic dissemination of the tachyzoite stage. A vigorous IFN-c-dependent cell mediated immune response eventually controls the tachyzoite stage of infection. However, some tachyzoites undergo a developmental switch to bradyzoites contained within tissue cysts and these persist long term largely in muscle tissue and the CNS. These bradyzoites can re-activate back to tachyzoites causing severe pathology in individuals with a suppressed IFN-c-dependent cell mediated immune response. Global gene expression studies in combination with previous EST date are beginning to elucidate transcriptomes unique to the tachyzoite, bradyzoite and oocyst stages of T. gondii development [9,10,11].
Although the T. gondii transcriptome is highly cell cycle regulated and the parasite differentiates between developmental stages, it is unclear how the parasite senses changes in its environment to trigger changes in its transcriptome. To date, twocomponent regulatory systems or other defined mechanisms for triggering transcriptional responses to extracellular environmental changes have not been described for T. gondii.
In this study we utilized global microarray expression analysis to identify the degree to which T. gondii alters its transcriptome in response to the in vivo microenvironment during the innate immune response (four days post-infection) relative to growth in human fibroblast (HFF) cells in vitro. To distinguish changes in the parasite transcriptome dependent or independent of the host immune response the parasite transcriptome was examined in IFN-c gene deleted mice in addition to wild type (WT) mice. IFNc is the key cytokine required for inhibition of T. gondii replication in vivo and in its absence parasite growth is largely unrestrained [12]. Our results indicate that the parasite transcriptome is significantly altered during in vivo infection in the presence, but not absence, of IFN-c-dependent immunity compared with fibroblasts infected in vitro. Therefore parasite transcriptional changes are largely induced in response to the IFN-c-dependent-innate immune response. This also suggests that in the absence of host immunity the in vitro environment in HFF cells versus in vivo in the peritoneum is relatively similar from the parasites' perspective.

Biological model used for parasite global transcriptome analysis
Whole-genome expression profiling was performed using the Affymetrix ToxoGeneChip microarray [13] to analyze the transcriptome response of T.gondii to the in vivo microenvironment during acute infection in the presence and absence of IFN-cdependent immunity. This array interrogates 8,058 predicted T. gondii genes (version 4.0 genome annotation) using 39 biased probes. The parasite transcriptome was compared between tachyzoites grown in vitro in HFF cells to parasites isolated from the peritoneum of infected WT or IFN-c KO mice on day four post-infection. Comparing the transcriptome of the parasite in both WT and IFN-c KO mice allowed identification of parasite transcriptome changes induced in response to the host immune response versus those dependent on the in vivo microenvironment but independent of IFN-c. The Type II genotype strain Prugniaud was utilized for these studies. The Type II genotype readily switches from tachyzoites to bradyzoites allowing analysis of tachyzoite to bradyzoite conversion during infection in the presence and absence of endogenous IFN-c. The Type II genotype is also the most common cause of human infections [14].
For T. gondii global expression analysis mice were infected with 2610 6 tachyzoites IP and on day four post infection parasites were isolated from the peritoneum for microarray analysis. The number of viable parasites in the peritoneum was determined on days one through four post-infection in WT mice and on day four postinfection in IFN-c KO mice by a modified plaque assay (see Materials and Methods). As shown in figure 1A numbers of parasites increased over the four days in WT mice although on day one post infection the number was decreased relative to input parasites. As expected parasite number on day four post infection was higher in IFN-c KO mice relative to WT mice consistent with the requirement for IFN-c in the control of T. gondii.
To determine the length of parasite exposure to endogenous IFN-c during the four day infection, the kinetics of IFNcproduction were evaluated by real time qRT-PCR. Sustained upregulation of IFN-c occurred within the first day post-infection in WT mice and was absent as expected in the IFN-c KO mice throughout infection (Fig. 1B). This indicated parasites in WT mice were exposed to increased levels of endogenous IFN-c for at least three days prior to collection for microarray analysis.
Array analysis compared the transcriptome of parasites isolated from WT mice or IFN-c KO mice relative to tachyzoites in HFF cells in vitro 48 hours post-invasion (approximately 85% intracellular). Parasites in HFF cells were chosen as a comparison for two reasons. First, HFF cells are the most common cell type used for studies on the parasite's lytic cycle as well as parasite transcriptome analysis allowing us to compare our in vivo results to the in vitro transcriptome studies in the field. Second, parasite replication in naïve murine macrophages appears largely unimpaired similar to that for HFF cells in vitro (data not shown).
Endogenous IFN-c exerts a pronounced effect on the T. gondii transcriptome during infection  data and distances between samples (Fig. 2). Both the in vitro parasites and the parasites isolated from IFN-c KO mice displayed similar variance and were highly correlated with one another. In contrast, parasites isolated from WT mice with an intact IFN-cdependent immune response did not correlate with parasites in vitro or in IFN-c KO mice and had significantly greater variance.
In figure 3, a Venn diagram shows the number of parasite genes increased or decreased in parasites isolated from WT or IFN-c KO mice compared to in vitro. The criteria for designation of differential expression between in vivo WT and in vitro samples was a greater than 2-fold change and determined as significantly changed by both a) unpaired t-test with benjamini hochberg correction (p,0.01) and by b) SAM analysis [FDR, False Discovery Rate (90th percentile): 0.04939%, Delta value: 1.603]. The stringency was relaxed to p,0.05 for changes between in vitro and IFN-cKO. For all analysis in the paper genes designated as increased or decreased in vivo reflect at least a 2 fold change in addition to the above criteria for significance. Parasites isolated from infected WT mice had 1151 genes whose transcription was upregulated and 1066 that were downregulated compared to parasites grown in vitro in HFF cells. In contrast, only 72 of these gene transcripts were significantly upregulated and 106 downregulated in IFN-cKO mice compared to in vitro. There was a small subset of parasite gene transcripts that were altered only in IFN-c KO mice and not in WT mice. Five of these were upregulated only in IFN-c KO mice and 27 were downregulated. All the genes with differential expression between the various groups are listed in Tables S1 and S2. The observation that parasite changes in gene expression generally required the presence of IFN-c to reach significance levels suggests that the in vivo environment the parasite encounters in the absence of IFN-c relative to in vitro growth in HFF cells is not substantially different from the perspective of the parasite.
Real time qRT-PCR validation of parasite genes representing select patterns of differential expression Microarray analysis identified four broad categories of transcriptional alterations in T. gondii in vivo compared to in vitro in HFF cells. The largest set of parasite transcripts induced in vivo, 1079, required the presence of IFN-c as they were only altered in WT mice and not in IFN-c KO mice. These were designated in vivo, IFN-c dependent. A smaller set of 72 transcripts was induced in parasites isolated from both WT and IFN-c KO mice. These were designated IFN-c independent-in vivo-induced transcripts. A very small number of genes (5) were upregulated in vivo but only in the absence of IFN-c. Finally, a large number of parasite genes were downregulated in WT mice (960), WT and IFN-c KO mice (106) or only IFN-c KO mice (27) relative to in vitro grown tachyzoites. A subset of genes was selected from the above expression patterns for evaluation by qRT-PCR to validate the microarray results. Table 1 shows the qRT-PCR and microarray expression fold change of ten predicted T. gondii genes in parasites isolated from WT or IFN-c KO mice relative to in vitro in HFF cells. The primers used for the PCR are listed in Table S3. Genes 44.m02656 (TGME49_031360), 59.m00026 (TGME49_069170), and 27.m00833 (TGME49_010950) were all preferentially upregulated in parasites isolated from WT mice relative to in vitro. TGME49_031360 is a hypothetical protein with a putative signal peptide and mass spectrometry data showing expression in both the conoid-enriched and conoid-depleted fractions in tachyzoites, after omitting proteins identified as contaminants derived from other subcellular organelles [15]. TGME49_069170 and TGME49_010950 are both hypothetical proteins with expression in partially sporulated oocysts based on EST and microarray data [10,16]. 76.m03412 (TGME49_084310) and 113.m01577 (TGME49_097280) were induced preferentially in IFN-c KO mice relative to in vitro. TGME49_084310 is a hypothetical protein and TGME49_097280 is a hypothetical protein with a putative prokaryotic membrane lipoprotein lipid attachment site. Both proteins have a putative signal peptide and show increased expression in bradyzoites and sporulated oocysts as opposed to tachyzoite stage [10,16]. Mass spectrometry data also shows expression of TGME49_097280 in the oocyst stage [16]. 72.m00003 (TGME49_080580), 42.m03338 (TGME49_024170), 49.m03400 (TGME49_044260), 37.m00739 (TGME49_017400) and 31.m00871 (TGME49_013050) were all decreased in vivo in both WT and IFN-c KO mice during infection compared to in vitro. TGME49_080580 is a classical bradyzoite differentiation antigen, P18 bradyzoite surface antigen/SAG4.2 [17]. TGME49_024170 is a putative SRS domain-containing protein common to a family of parasite GPI-anchored surface proteins some of which are developmentally expressed [18]. Microarray results for in vitro and in vivo generated bradyzoites imply enriched expression in bradyzoites for both genes [11]. The others are hypothetical proteins but proteomic data is available for  T gondii in vivo transcriptome does not correlate to a specific parasite cell cycle stage IFN-c is important for activating host cells to restrict growth of T. gondii and for host cell cytolytic activity against the parasite. In the absence of IFN-c parasite replication is relatively unrestrained. Approximately 40% of T. gondii genes have a significant cell cycle expression pattern in the tachyzoite stage [8]. To investigate the potential contribution of parasite cell cycle inhibition by innate immunity to the in vivo transcriptome, the parasite genes induced in vivo were compared to those previously shown to be cell cycle regulated [8]. Only 11.8% of the 1151 genes increased in vivo were designated as cell cycle regulated genes (137 total). These are highlighted in Table S2. 62 of these were classified as G1 and 75 as S/M phase. This suggests that the T. gondii in vivo transcriptome is not simply a consequence of parasite cell cycle arrest by IFN-cdependent innate immunity.
Transcripts for parasite stress response genes are induced early during acute infection in the presence of IFN-c-dependent innate immunity Although tachyzoite to bradyzoite conversion can be induced within a few days [11,19,20], the degree and kinetics of conversion in vivo during acute infection in the peritoneum and lymphoid tissue is less clear. To examine the extent to which bradyzoite specific gene transcription was induced four days post-infection in the peritoneum, the array expression values of a panel of known bradyzoite genes were examined from parasites isolated from WT mice compared to HFF cells in vitro. These included bradyzoite antigen-1 (BAG1), ENO1, bradyzoite rhoptry protein 1(BRP1), LDH2, SAG2Yand SRS 9 [11,21]. Transcripts of tachyzoitespecific genes, including ENO2, LDH1, SAG1, SAG2A and SRS2 were also examined [11,21]. Results are shown in Table 2. None of the canonical bradyzoite genes were significantly increased in vivo during infection based on our statistical cut offs. Expression of two parasite bradyzoite genes TGME49_080580 (P18/SAG 4.2) and TGME49_024170 (SRS-domain-containing protein) were further evaluated by qRT-PCR; both were downregulated in parasites from WT and IFN-c mice relative to in vitro (Table 1). Our microarray results were also compared with genes induced by compound 1, a trisubstituted pyrrole that was identified in a small molecule screen as a potent inducer of bradyzoite-specific antigens [22]. Out of the 94 parasite genes altered by compound 1 in vitro, only 13 were statistically increased in vivo relative to in vitro in the current study (Table S4). These results in total indicate that at least at day four post-infection tachyzoite to bradyzoite conversion is not strongly triggered in the peritoneum by the host innate immune response.
Although mature bradyzoite transcripts were not significantly induced during infection, it was possible that early stress response genes were induced during infection in the presence of IFN-cdependent innate immunity. To examine this possibility the current data set was compared with data sets for parasite genes induced by sodium nitroprusside (SNP), a nitric oxide donor, or by CO 2 starvation (ToxoDB.org; Bradyzoite Differentiation (Multiple 6-hr time points and Extended time series; Paul Davis, Florence Dzierszinski and David Roos). The Bradyzoite Differentiation Study identified 909 parasite transcripts induced greater than 2fold 24 hours after SNP treatment. Approximately 45% of these transcripts were also statistically increased in vivo in the presence of an intact immune response in the current study. These genes are listed in Table S4. In comparison, 211 transcripts were identified in the Bradyzoite Differentiation Study as upregulated greater than 2 fold in CO 2 depleted conditions for 30 hours. 33% of these transcripts were also statistically increased in vivo in parasites isolated from WT mice (Table S4). These results suggest that the in vivo microenvironment in the presence of an intact IFN-c dependent innate immune response induces alterations in parasite gene expression consistent, in part, with an early general stress response induced in vitro by SNP or CO 2 depletion.
Transcription of parasite genes coding for putative surface proteins and intracellular signaling cascades are disproportionately increased in vivo in the presence of IFN-c Genes statistically upregulated in vivo during infection relative to in vitro in HFF cells were classified into 11 categories based on known or implied function. The data are displayed as a pie chart with slices representing the proportion of genes in each category. Genes of unknown function are not included in the pie chart but comprised the majority of the genes. 68% of the genes up in parasites from WT mice and 51% of the genes down in parasites from WT mice were of unknown function. Figure 4A shows the functional categorization of T. gondii genes upregulated in vivo in WT mice compared to in vitro. This is compared to Figure 4B that shows categorization of genes downregulated in vivo in WT mice compared to in vitro. Two categories in particular appeared increased in vivo compared to those downregulated in vivo. Genes implicated in intracellular signaling were increased slightly preferentially in vivo. Putative surface proteins, the majority with putative prokaryotic membrane lipoprotein lipid attachment sites as evaluated by ToxoDB (ProSite), were also preferentially upregulated in vivo compared to the proportion downregulated in vivo. Prokaryotic membrane lipoprotein attachment site profiles for proteins are based on the presence of a putative defined signal peptide that is cleaved by a specific signal peptidase II upstream of a cysteine residue to which a glyceride-fatty acid lipid is attached. The putative surface proteins did not include transmembrane proteins characterized as putative transporters that were instead included in the category cellular transport.

Transcripts for oocyst genes are disproportionately increased in vivo during infection
To more broadly examine the effect of in vivo infection and the host innate immune response on parasite stage specific gene expression, genes with increased and decreased expression in vivo in WT mice were interrogated using data on ToxoDB (http:// ToxoDB.org) for stage specific ESTs or transcripts for either tachyzoites, bradyzoites or oocysts. For the comparison, stage specific genes were those who had EST data for only one stage of the parasite and/or at least a 10-fold preferential increase in the ToxoDB microarray data for a particular stage. Figure 5 shows the distribution of tachyzoite, bradyzoite and oocyst specific transcripts that were differentially expressed by T. gondii in vivo in WT mice versus in vitro based on these criteria. Out of the 171 stage specific genes increased in parasites from WT mice, 85 of the genes were tachyzoite specific, 25 bradyzoite specific and 61 oocyst specific (including partially and fully sporulated oocysts). The bradyzoite and oocyst specific genes increased in vivo in WT mice are listed in Table 3. qRT-PCR confirmed in vivo induction of two of the oocyst genes, TGME49_069170 and TGME49_010950 (Table 1). In the absence of IFN-c, only three oocyst specific  transcripts were increased in vivo suggesting that host immunity was important for the increased expression of the oocyst-associated genes.

Discussion
This study represents the first global transcriptional analysis of how Toxoplasma gondii, an obligate intracellular protozoa, alters its transcription when it transitions from an intracellular environment in vitro in host cells to an in vivo environment in host cells during infection in the presence and absence of IFN-c-dependent innate immunity. The study shows that the in vivo microenvironment in the absence of an IFN-c-dependent innate immune response results in little transcriptional adaptation on the part of the parasite. This is consistent with the relative lack of restraint on parasite replication in HFF cells and murine macrophages in vitro and in vivo in the absence of IFN-c. However, our results indicate the parasite clearly does respond to changes in the in vivo microenvironment due to the host IFN-c-dependent innate immune response even early in infection. This response shares similarities with transcriptional changes induced in vitro by exposure to sodium nitroprusside (SNP), a nitric oxide donor or CO 2 depletion.
It is clear from the present study that T. gondii senses its external environment from within its segregated compartment within a host cell and responds with alterations in gene expression in response to host immunity. It is well established that extracellular stresses can induce intracellular T. gondii to differentiate from tachyzoites to bradyzoites. However, this is the first study evaluating how, and the degree to which, the parasite responds transcriptionally to changes in the in vivo microenvironment during infection. The importance of these changes to parasite pathogenesis requires further investigation. How the parasite senses its environment is unknown. Our results suggest that the in vivo transcriptome is not simply a passive consequence on the part of the parasite due to an immune-induced parasite cell cycle block or inhibition of replication. To date, two-component regulatory systems used by bacteria to sense and respond to its environment have not been described for T. gondii. However, T. gondii could utilize proteins with redox-responsive residues such as iron-containing transcrip-tion factors to directly sense changes in redox potential to induce changes in gene expression similar to other pathogens [23,24]. Entamoeba histolytica has an endoplasmic reticulum stress-sensing mechanism activated by nitric oxide important for modulating parasite nuclear gene expression [25]. Generation of second messengers such as cAMP downstream of phosphodiesterases mediates transduction of extracellular stimuli, such as nutrient deprivation, to intracellular downstream singling pathways in pathogenic fungi [26,27]. In fact, Cyclic AMP and cGMP kinases have been implicated in T. gondii tachyzoite to bradyzoite conversion although the upstream signals remain unknown [28,29].
The IFN-c-dependent T. gondii in vivo transcriptome shares similarity with transcriptional changes induced in vitro in the first 24-30 hours of exposure to sodium nitroprusside (SNP) or low CO 2 . Both SNP and low CO 2 can induce bradyzoite conversion in vitro during prolonged exposure. SNP is a nitric oxide donor with additional toxicity particularly to the mitochondria due to the release of cyanide molecules. IFN-c can stimulate transcription of inducible nitric oxide synthase particularly in phagocytic cells resulting in nitric oxide that has cytostatic activity against T. gondii. Our results show that approximately 45% of the parasite genes induced in vivo overlapped with those previously identified as induced by SNP in vitro within 24 hours. This is likely due to the fact that nitric oxide is produced early during T. gondii infection of WT mice and may be mediating many of the in vivo IFN-cdependent transcriptional changes. However, approximately 30% of the genes induced in vivo overlapped with those induced in vivo following CO 2 depletion. These common genes suggest that the parasite in vivo IFN-c-dependent transcriptome response, in part, is indicative of a general early stress response in the parasite. This general stress transcriptional response may or may not be directly linked to eventual progression to mature bradyzoites.
In contrast to the overlap of genes of the IFN-c-dependent in vivo transcriptome with genes up-regulated early after SNP addition or CO2 depletion, canonical bradyzoite markers were not increased and were sometimes decreased in vivo during infection. The fact that we did not observe progression to mature bradyzoite transcripts in vivo may be due to a four day or less exposure of parasites to an effective innate immune response. We show that IFN-c is significantly induced within a day of parasite challenge. Although three days is sufficient to stimulate tachyzoite to bradyzoite conversion in terms of changes in gene expression in vitro with exogenous stress, it may not be long enough in vivo. However, it is also likely that tachyzoite to bradyzoite conversion occurs more readily in anatomical locations such as muscle and CNS rather than in primary lymphoid tissue where parasite clearance rather than bradyzoite conversion may be more likely.
Surprisingly a disproportionate number of the transcripts induced by the parasite during infection in an immune intact host are those traditionally considered as oocyst transcripts. Expressed sequence tags (EST) and global transcriptome analysis has largely been used to analyze gene expression for tachyzoites in vitro, bradyzoites or oocysts. This is the first study that evaluated the T. gondii tachyzoite transcriptome response to the in vivo microenvironment and innate immunity relative to in vitro culture in HFF cells. Consequently, some transcripts upregulated in oocysts may also be upregulated by in vivo factors during infection including host immunity.
A number of the transcripts increased during infection had low expression in tachyzoites grown in HFF in vitro. Transcripts with low expression can often be disproportionately represented in microarray analysis as a fold increase is more likely to be obtained for low expression transcripts. However, bradyzoite transcripts  Partially sporulated expected to be low or non-detectable in tachyzoites in vitro were not disproportionately increased in vivo. Therefore, the selective increase in a number of oocyst transcripts in vivo is unlikely to be explained strictly on the basis of their low expression in vitro.
Similarly it is likely that parasite genes without evidence of gene transcription in vitro or in oocysts or bradyzoites may still be expressed in vivo during infection. The current study highlights the importance of evaluating gene expression during infection in addition to particular stages of parasite development before concluding a gene transcript is not expressed. The majority of the parasite transcripts increased during infection was not annotated and lacked sufficient homology to other proteins to delineate potential functions. This is true for the parasite genome as a whole but is more pronounced with the genes whose transcripts are increased during infection. However, the increased expression of surface proteins with putative lipoprotein anchors in vivo relative to in vitro, suggests the possibility that upregulation of these surface proteins may play a protective role in shielding the parasite. The other subset of proteins increased in vivo during infection were those involved in intracellular signaling consistent with the parasite responding to changes in its environment induced by the innate immune response.
This study establishes that T. gondii's transcriptome is responsive to the in vivo microenvironment early during infection in the presence of the IFN-c-dependent innate immune response. In contrast, in the absence of IFN-c, the parasite's transcriptome is surprisingly similar to that during growth in vitro in HFF cells. The in vivo transcriptome response shares similarities with what appears to be a general early stress response on the part of the parasite in the absence of bradyzoite conversion at least within the first four days of infection in the peritoneum. The unknown questions are how the parasite senses its environment to induce alterations in gene expression and the importance of such transcriptional adaptation early during infection.
For in vitro microarray studies healthy confluent HFF cells were infected with T. gondii. A MOI was chosen to ensure the majority of HFF cells were infected and had one to two parasite-containing vacuoles (PVs). The cells were harvested 48 hours later as parasites were beginning to egress (10-30% extracellular). For in vivo transcriptome analysis C57BL/6 wild type (WT) or IFN gammagene deleted mice (IFNKO) were inoculated with 2610 6 parasites intraperitoneally (ip). Parasites were isolated from the peritoneal cavity of the mice four days post infection as previously described [30]. To ensure sufficient parasite RNA, parasites isolated from 3-5 mice were pooled together to serve as one biological replicate. Three to five biological replicates were analyzed per experimental group. To reduce host cell contamination, harvested samples were passed through a syringe and needle to rupture host cells and centrifuged to obtain a T. gondii pellet. Although this method reduced host cell contamination it was unable to effectively separate parasites from peritoneal cells resulting in significant host cell RNA in addition to parasite RNA. All animal experiments were conducted with the approval and oversight of the Institutional Animal Care and Use Committee at New York Medical College.

RNA isolation, labeling, and microarray hybridization
Total RNA was isolated using RNeasy mini elute columns (Qiagen, Valencia). RNase inhibitor (Ambion, Inc) was added to samples followed by storage at 280uC. Synthesis, fragmentation of cRNA probes and hybridization of the samples to the T. gondii custom chip [13] by Affymetrix was performed by the University of Pennsylvania Microarray Facility (U.Penn) as per GeneChip Expression Analysis Technical Manual (Affymetrix, Santa Clara, CA).

Preprocessing
Affymetrix Microarray suite 5.0 was used to quantify expression levels. Raw (.cel) files and the .cdf file were imported into RMADataConv in RMAExpress (http://rmaexpress.bmbolstad. com) [31,32,33]. A modified .cdf file was generated which excluded human and mouse genes present in the chip to eliminate any bias arising due to differences in the expression levels of host genes in the samples being compared during normalization. The output files containing only parasite genes were imported into RMAExpress. Gene expression values were quantified following normalization using Robust Multichip Average algorithm [31,32,33].

Statistical analysis
Normalized expression data was analyzed using GeneSpring GX version 9 GX9 (Agilent Technologies, Inc. software. CA). Statistically significant differences in gene expression between experimental groups were determined using unpaired t-test with Benjamini Hochberg FDR correction to minimize false discovery. Genes showing $2 fold difference and p values ,0.01 for comparisons between in vitro and in vivo (wt) and p,0.05 when comparing in vitro and in vivo (IFNKO) samples were selected.
Normalized data was also analyzed using two class unpaired SAM (significance analysis of microarrays) [34] as provided in MeV, multi experiment viewer version 4.3 [35,36]. Fold differences $2 fold and a FDR, False Discovery Rate (90th percentile): 0.04939%, Delta value: 1.603 for comparisons between in vitro and in vivo (WT) and a FDR False Discovery Rate (90th percentile): 1.91679%, Delta value: 1.847 when comparing in vitro and in vivo (IFNKO) samples were set as parameters to select for differentially expressed genes. For the purpose of the present study, gene expression was only considered changed if a gene was changed .2 fold and considered significantly altered when analyzed by both SAM analysis and GeneSpring GX9. Tox-oDB.org version 4.3 was initially used as a reference for identification of putative genes and data from later versions of ToxoDB (http://ToxoDB.org) up until version 7.1 was used for characterization of genes and proteins based on bioinformatics, information for functional characterization, ESTs and mass spectrometry and for many of the comparisons to published and pre publication microarray data.

Real Time Quantitative RT-PCR
RNA samples were reverse transcribed with Superscript III (Invitrogen). cDNA was amplified using the FastStart Universal SYBR Green Master (ROX) (Roche Diagnostics) for regular assays and using FastStart Universal Probe Master (Roche Diagnostics) when using TaqManH Gene Expression Assays (Applied Biosystems) on an ABI-7900 PCR system following the manufacturer's instructions (Applied Biosystems). Reactions containing cDNA generated without reverse transcriptase or containing primers alone were included as controls. Primers and probes used to amplify the genes are listed in Table S3. One to three biological replicates were analyzed in duplicate in this study. The cycle threshold (Ct) value was determined from a representative experiment unless specified otherwise. A relative quantitation method Delta Delta Ct [37] was used to determine the relative expression of a gene after normalization to housekeeping gene T. gondii actin. All of the probes utilized were tested against cDNA from uninfected mice to confirm specificity for parasites.
IFN-c levels were determined using TaqmanH Gene Expression Assay: Mm00801778_m1 after normalization to mouse glyceraldehyde-3-phosphate dehydrogenase with TaqmanH Gene Expression Assay: Mm99999915_g1. The cycle threshold (Ct) value was determined from three biological replicates analyzed in triplicate. An unpaired Student's t-test was used to determine significant differences and p-values less than 0.05 were considered significant.
Parasites were also isolated from the peritoneal cavity of infected mice on days one through four post-infection to assess parasite viability by a modified plaque assay. Parasites were serial diluted and allowed to invade HFF cells in 96-well plates overnight. Cells were fixed with 2.5% formaldehyde buffered in PBS and stained with a rabbit polyclonal anti-Toxoplasma antibody (Fitzgerald Laboratories) followed by an Alexa488-conjugated goat antirabbit secondary antibody (Invitrogen). The number of PVs that contained more than 2 parasites per PV was counted in each well and that number was multiplied by the total dilution factor to determine the number of viable parasites in the peritoneal cavity of infected mice. Since we were unable to completely separate lymphoid cells from intracellular parasites, this method proved more accurate for quantification of the number of parasites in an infected cell than a traditional plaque assay.

Microarray data accession number
The complete data set obtained in this study has been deposited in the ArrayExpress database E-MEXP-3579 ''In vivo transcriptome analysis of Toxoplasma gondii''.

Supporting Information
Table S1 Genes whose expression is downregulated in vivo in wild type C57/BL6 and IFNKO mice compared to in vitro samples.

(XLS)
Table S2 Genes whose expression is increased in vivo in wild type mice and IFNKO mice compared to in vitro samples. Highlighted rows show genes that are cell cycle regulated. Sheet 1, lists genes regulated in vivo in wild type mice only. Sheet 2, lists genes regulated in vivo in both wild type and IFNKO mice. Sheet 3, lists genes regulated in vivo in IFNKO mice only. Column descriptions: columns in vitro 1-4 show the normalized log expression values for genes in the individual in vitro samples. Columns in vivo 1-4 and IFNKO 1-3 show the normalized log expression values for genes in the individual in vivo samples from wild type mice and IFNKO mice respectively. (XLS) Table S3 Primers and probes used to amplify genes.

(XLS)
Table S4 Genes whose expression is increased in vivo in wild type mice vs in vitro samples also induced by Compound 1, sodium nitroprusside (SNP) and reduced CO 2 .