Background

Toxoplasma gondii is an obligate intracellular parasite that can infect any nucleated cell of warm-blooded animals. The parasite has both sexual and asexual cycles where the sexual cycle takes place in the intestinal cells of the definitive feline host and the asexual cycle occurs in all warm-blooded animals [1]. The asexual stages of T. gondii consist of the rapidly replicating tachyzoite and the slow growing encysted bradyzoite. In the host, the tachyzoite is the prominent stage during initial acute infection [2]. Once the tachyzoite is subjected to stress from the host immune response, it differentiates to the bradyzoite form and eventually establishes a chronic infection [3]. The bradyzoite persists for the lifetime of the host as intracellular cysts present in striated muscle and the central nervous system [4].

T. gondii is one of the most prominent parasites in humans with prevalence rates between 10 and 80 percent worldwide, depending on the country [5]. Complications such as hydrocephaly, retinochoroiditis, mental retardation and even death can occur in developing fetuses [6, 7]. In immunocompetent humans, infection with T. gondii is generally asymptomatic presenting flu-like symptoms in approximately 10 percent of individuals [8]. Patients with compromised immune systems, such as those infected with HIV, are at great risk of developing severe symptoms such as Toxoplasmic Encephalitis (TE) and ocular infection that may result in blindness [9, 10]. Disease is largely associated with sporadic reactivation of the latent bradyzoite back to the rapidly replicating tachyzoite [11]. Currently there are no drugs that can combat the bradyzoite form of the parasite or effective vaccines to protect against infection. These issues highlight the critical need to understand the cellular triggers that control development between the tachyzoite and bradyzoite stages.

Previous work to characterize the transcriptome of tachyzoite and bradyzoites from T. gondii has primarily used microarray technology from samples prepared in tissue culture [12–14]. Transcriptomic studies have also used a combination of in vitro and in vivo samples with in vitro tachyzoites, in vivo bradyzoites, and oocysts collected from infected felines as well as tissue culture tachyzoites and bradyzoites, developing oocysts, and bradyzoites purified from mouse brains 21 days post-infection [12, 15] While these studies have provided valuable insight into this developmental process, the information that can be extracted is limited because tissue culture conditions for tachyzoite and bradyzoite development do not precisely model animal infections. Other in vivo microarray studies have compared peritoneal-derived tachyzoites from different strain types of T. gondii from wild type and interferon-γ (IFN-γ) deleted mice, and transcriptional changes in the brain of mice eight days after T. gondii infection [16–18]. These studies have highlighted important aspects of the developmental process, but the dynamic range of microarrays is restricted and sample preparations are unable to be simultaneously processed and analyzed for both host and pathogen. One way to overcome these limitations is RNA sequencing (RNA-seq), a breakthrough molecular tool that can provide the transcript profile (transcriptome) of total cellular RNA with a large dynamic range and improved sensitivity [19]. RNA-seq has detected novel T. gondii tachyzoite transcripts and alternative splicing between strains [20]. RNA-seq of tissue culture-derived bradyzoites has shown dysregulation of bradyzoite genes in the deletion mutant of a mucin domain containing cyst wall protein CST1 [21]. Using RNA-seq, transcriptome analysis has been performed to compare mouse brains that were uninfected or infected with T. gondii for 32 days [22]. These data highlight the sensitivity and depth of knowledge that can be obtained from RNA-seq studies; however, a time course of T. gondii infection and simultaneous analysis of the parasite transcriptome has not been performed.

To provide a more comprehensive analysis of T. gondii and the host during both acute and chronic infection, we collected RNA-seq data from three experimental groups of mice: uninfected, 10 and 28 days post-infection. Because T. gondii preferentially establishes cysts in the brains of mice and reactivation of cysts is the main cause of TE, we chose to analyze the brains of mice. A novel aspect of this dataset is that parasites were not purified from the brain tissue but instead, samples were rapidly processed so that RNA-seq reads represent the "interactome" between host and pathogen during the peak of acute and chronic infection. We report that many genes involved in T. gondii transcription, translation and metabolism are highly expressed during chronic infection. For the host, we find that more genes are increased in abundance during chronic versus (vs) acute infection, attesting to a continuously active host response even at 28 days post-infection.

Results

Sequencing and mapping the T. gondii/host interactome

To study parasite-host gene expression dynamics we used RNA-seq on samples collected from forebrains of mice infected with type II strains of T. gondii (schematized in Figure 1). Type II strains of T. gondii have been detected in the mouse brain as early as 4 days post-infection and numbers continue to increase until 10 days post-infection [23, 24], the peak of acute infection [25]. Cyst structures are present in the brains of mice at 21 days post-infection [26], which is generally considered to be the beginning of chronic infection. By 28 days post-infection, cysts have stably formed in the brain while parasite numbers have decreased elsewhere in the body [23, 27]. When brains were sectioned and analyzed for parasite distribution, high numbers of T. gondii were observed in the frontal lobe at 32 days post-infection [22]. We examined mouse brains at 10 and 28 days post-infection using an In Vivo Imaging System (IVIS), which confirmed that parasites were primarily localized in the forebrains of mice (Figure 1). Therefore to maximize parasite transcripts as well as to compare the same host tissue during acute and chronic infection, we chose to collect mouse forebrain samples at 10 and 28 days post-infection as well as uninfected mice. Nine mouse forebrains were sequenced individually: three uninfected, three infected for ten days and three infected for 28 days. Nearly one billion 100 base-pair (bp) paired-end RNA sequences were generated. Between 81,000,000 and 114,000,000 reads were obtained from each forebrain sample (Table 1). Between 69-76% of the reads aligned to the M. musculus reference genome while approximately 0.1% aligned to the T. gondii TGME49 reference. Since T. gondii was not purified from the forebrains as a means to rapidly process the samples and preserve the interactome, uninfected mouse forebrain samples were mapped to the T. gondii TGME49 reference to determine the extent of false positive reads. A small number of uninfected mouse reads aligned to the T. gondii genomic reference file (Table 1). The reads aligned to T. gondii ribosomal associated RNA or small (~200-300 bp) hypothetical proteins, none of which were considered differentially expressed and were treated as background.

Figure 1
figure 1

Schematic of T. gondii /host dataset generation. (A) Nine mice were divided into three experimental groups: uninfected, 10 days post-infection, and 28 days post-infection. Infected mice were given 10,000 type II ME49 parasites and sacrificed on the corresponding days. The forebrains were removed and homogenized in TRIzol, and RNA was extracted and purified. A cDNA library was generated from the RNA prior to IlluminaHiSeq2000 sequencing. Raw reads were aligned to either the T. gondii or M. musculus genomes, normalized and analyzed for differential gene expression. (B) To examine T. gondii in the brains of mice at the designated time points, mice were infected with 10,000 parasites of a bioluminescent T. gondii[62]. Shown are representative brains for 10 and 28 day post-infection mice, after the brains were soaked for 5 minutes in luciferin prior to imaging.

Full size image
Table 1 Mapped paired-end reads of individual mouse forebrain samples
Full size table

Transcript abundance of T. gondiiduring acute and chronic infection in mice

Abundance estimates for each T. gondii gene were calculated and the three biological replicates for each time point were averaged before differential expression analysis was performed using Cuffdiff, a Cufflinks program. Cuffdiff calculates the fold change to determine which genes are differentially regulated between time points. Cuffdiff also calculates a p-value and q-value to determine if the fold change is significant. Each forebrain was treated as a biological replicate in Cuffdiff and therefore variation between replicates was considered when assigning a p-value. A p-value and q-value <0.05 were considered significant. It is important to note that parasites were not purified from mouse forebrains prior to RNA extraction. As a result, the concentration of T. gondii RNA could not be normalized between time points. To address the potential difference in parasite numbers between experimental time points quantitative PCR was performed using genomic DNA extracted from the forebrains at the time of RNA extraction. Using a standard curve generated from serial dilutions of genomic DNA extracted from a known number of parasites, we found approximately 500 parasites per 350 ng of genomic DNA in 10 day post-infection samples and approximately 1500 parasites per 350 ng of genomic DNA in the forebrains of mice infected for 28 days. These numbers are in agreement with the average number of reads generated from each time point with approximately twice as many reads aligning to the T. gondii genome in the 28 day samples than the 10 day samples (Table 1). To further examine differences in parasite numbers between acute and chronic infection samples, the FPKM fold change of housekeeping genes α-tubulin, actin, glyceraldehyde 3-phosphate dehydrogenase 1 and 2 (GAPDH 1 and 2), and hexokinase were examined between acute and chronic time points (Table 2). The fold change of these housekeeping genes between chronic and acute time points were 0.6-1.8, suggesting that global parasite transcript levels in our acute and chronic infection samples do not dramatically change.

Table 2 Fold change between chronic and acute infection for previously characterized T. gondii genes
Full size table

When comparing the fold change of previously characterized tachyzoite-specific genes, surface antigen 1 (SAG1) and microneme protein (MIC) 1, and bradyzoite specific genes bradyzoite antigen 1 (BAG1) and enolase 1 (ENO1), we saw large differential expression between acute and chronic infected mice (Table 2). These results along with qPCR performed on genomic DNA and quantitation of the reads that map to the T. gondii genome (Table 1) suggest that the significant differential expression of genes between time points is not due to an overabundance of transcripts in chronically infected mice compared to acutely infected mice. This data is supported in previous work that determined there is minimal increase in parasite numbers in the brain between 10 and 20 days [28]. To account for minimal variability of parasite numbers at 28 days post-infection differentially expressed T. gondii genes with a fold-change of >5 were considered for further analysis. The low abundance of SAG-1 and high abundance of BAG-1 between 28 day post-infection and 10 day post infection time points also suggests that tachyzoites are the primary stage present at 10 days post-infection, while the majority of parasites at 28 days post-infection are in the bradyzoite stage.

To explore the similarities in expression of T. gondii genes during acute and chronic infection, the 100 most highly expressed genes from acute and chronic time points were compared. Forty-two of the most highly expressed T. gondii genes during the acute stage were also among the 100 most highly expressed transcripts during the chronic stage (Figure 2A). To help interpret the biological functions of these genes, statistically over-represented GO terms were compiled (Figure 2B) using the Blast2GO program, a GO term analysis program for non-model organisms [29]. The GO term categories for the genes with the greatest abundance in both acute and chronic samples were transcription, translation, macromolecule biosynthesis and cellular metabolism.

Figure 2
figure 2

Top 100 T. gondii genes from acute and chronic infection. (A) Venn diagram compares the 100 T. gondii genes with the greatest FPKM value from acute and chronic time points. Of the top 100 genes for each stage, 42 genes were similarly abundant (green), and 58 were only abundant in acute (blue) or chronic (yellow) infection. (B) GO terms were assigned to the top 100 T. gondii genes from acute and chronic infection. Genes were grouped based on whether they are similar or different between time points prior to GO term analysis.

Full size image

Differential expression of T. gondiibetween acute and chronic infection in mice

To further investigate the transcriptome of T. gondii during acute and chronic infection, differential expression analysis was performed. Fold change and significance values were calculated for each of the ~8900 annotated T. gondii genes (Additional file 1). From this analysis, we found 547 significantly differentially expressed genes (DEGs, with p-value and q-value <0.05) between acute and chronic infection. Of these, the DEGs with a fold change of >5 are presented in Tables 3 and 4.

Table 3 T. gondii DEGs that were more abundant >5-fold in acute vs chronic infection
Full size table
Table 4 T. gondii DEGs that were more abundant >5 fold in chronic vs acute infection
Full size table

Sixty-three T. gondii genes were >5 fold more abundant in acute compared to chronic infection (Table 3). SAG-related sequence (SRS) are a family of GPI-anchored surface antigens related to the first characterized T. gondii surface antigen, SAG1 [30]. SRS2/SRS29C was the most differently expressed gene in acute compared to chronic infection, 305-fold. Five additional SRS genes were >5 fold more abundant: SRS20A, SAG1/SRS29B, SAG2/SRS34A, SRS54 and SRS52A. Four genes for rhoptry proteins (ROP) were >5 fold more abundant during acute infection: ROP9, ROP16, ROP39, and ROP40. ROP16 is involved in decreased synthesis of cytokines in mouse bone marrow-derived macrophages [31]. ROP9 is a tachyzoite-specific protein with no known function [32]. ROP39 and ROP40 have homology to ROP2, but the functions have yet to be elucidated. Because of the extensive study of tachyzoites, most of the DEGs highly abundant in acute compared to chronic infection (Table 3) have been previously identified as tachyzoite-specific markers. Twenty-six of the 63 acute infection DEGs encode hypothetical proteins, with no homology to any annotated protein in the BLAST database. Characterizing these hypothetical proteins could be vital to understanding parasites during acute infection.

Fifty-one of the 107 T. gondii DEGs associated with chronic infection with a >5 fold change are annotated as hypothetical (Table 4). Among the chronic infection DEGs, four microneme proteins were identified MIC12, MIC13, MIC17A and MIC17C. MIC12 and MIC13 were previously shown to be bradyzoite specific [12], but our data revealed MIC17A and MIC17C abundant specifically during chronic infection. Another interesting group of chronic infection DEGs are those involved in glycolysis: glucose-6-phosphate isomerase, pyruvate kinase, lactate dehydrogenase 2, and glucosephosphate mutase. Previous data showed that tachyzoites and bradyzoites use the glycolytic pathway differently with lactate dehydrogenase 2 and pyruvate kinase being up-regulated during the bradyzoite stage [33, 34]. Our data shows that in addition to these two previously described bradyzoite-specific glycolytic enzymes, glucose-6-phosphate isomerase and glucosephosphate mutase are also more abundant in chronic vs acute infection. These results strengthen the idea that bradyzoites do not have an active TCA cycle because the transcripts of the key enzymes in the TCA cycle were less abundant in chronic infection compared to acute. Four SRS genes were also identified as more abundant in chronic compared to acute infection. SRS35A, also known as SAG4 and P18, is a long-known bradyzoite-specific marker [35]. SAG2C/SRS49D and SRS9/SRS16B are previously identified bradyzoite-specific genes that are important for persistence of infection [36–39]. SRS13 was identified as up-regulated in a microarray analysis comparing tissue culture tachyzoites to mouse-derived bradyzoites, but the function has yet to be determined [12]. Several novel chronic infection DEGs were found including DnAK-TPR, heat-shock protein 21, calcium dependent protein kinase CDPK5 (Table 4). The high number of novel and hypothetical DEGs highlights the fact that much is still unknown about T. gondii during animal infection. These hypothetical proteins have no known homology to proteins in the host and therefore could be essential and specific to parasite function during infection. Identification of novel DEGs that could play critical roles in T. gondii infection is the first step in elucidating potential targets for both vaccine and drug development.

To highlight the accuracy and sensitivity of RNA-seq, q-PCR was performed on a family of CCCH zinc fingers in T. gondii that are more abundant in our data set during chronic infection. TGME49_224630, TGME49_262970, and TGME49_311100 contain CCCH zinc finger domains and had a fold change of 86, 26, and 4.7, respectively (Table 4 excluding TGME49_311100 which did not meet the >5 fold cut-off). These genes were chosen based on the possible similarity in function and the range of transcript differential expression between chronic and acute infection. The increase in abundance of these transcripts between 10 day and 28 day post-infection was observed using qPCR when normalized to the house keeping gene tub1a. The fold change between chronic and acute infection for TGME49_224630, TGME49_262970, and TGME49_311100 were 53, 79, and 6.1 respectively. Not only does this data demonstrate the range and accuracy of RNA-seq, but also confirms the validity of the differential expression analysis.

Differential expression of host genes during T. gondiiinfection

To understand the transcriptional changes of the host during stages of T. gondii infection, differential expression analysis was conducted. Differential expression was determined between acute vs uninfected, chronic vs uninfected, and acute vs chronic time points. Genes were considered differentially expressed if the p-value and q-value was <0.05 and the fold change between time points was >2-fold. The host underwent extensive transcriptional changes during T. gondii infection (Figure 3A). When comparing acute T. gondii infected mice with uninfected, 1004 mouse genes were more abundant during acute infection and 143 were less abundant (Figure 3A). Over twice as many mouse genes, 2510, were more abundant in mice with a chronic T. gondii infection compared to uninfected while only 132 genes were less abundant. Finally, 1872 mouse genes were more abundant and 190 were less abundant in chronically vs. acutely infected mice. This increase of differentially regulated host genes during chronic infection is illustrated by Venn diagram analysis of DEGs >2-fold (Figure 3B). To understand the similarities and differences between DEGs more abundant during acute vs uninfected and chronic vs uninfected time points, we identified genes that had increased abundance at both time points as well as those found only in acute or chronic (Additional files 2, 3 and 4). Out of the 1004 DEGs more abundant in acute vs uninfected, 902 were also abundant in the chronic vs uninfected group. More mouse genes, 1608, were increased in abundance in the chronic vs uninfected time point that were not considered differentially expressed in acute. This data suggests genes activated during the peak of acute infection are maintained during chronic infection and that an entirely new subset of transcripts are expressed during chronic infection. The increase in differentially expressed genes during chronic infection could also be due to recruitment of cells to the site of infection.

Figure 3
figure 3

More host genes have increased abundance during chronic infection. (A) DEGs in the mouse with a fold change >2 were grouped based on increased abundance (grey) and decreased abundance (black) between acute vs uninfected, chronic vs uninfected, and chronic vs acute time points. (B) A Venn diagram was created to compare DEGs with increased abundance in acute vs uninfected (purple) and chronic vs uninfected time points (red). Of the 1004 more abundant DEGs in acute vs uninfected, 902 were also more abundant in the chronic vs uninfected group (magenta). (C) To explore the function of DEGs analyzed in the Venn diagram, a GO term enrichment analysis was performed.

Full size image

GO term enrichment analysis of DEGs in mice during T. gondiiinfection

To understand the functions of DEGs in mice during acute and chronic T. gondii infection and to characterize the overlap between DEGs at these time points (Figure 3B) we performed GO term enrichment analyses. Among the host genes more abundant during both acute and chronic infection stages, overrepresented GO terms were related to stress and immune responses (Figure 3C). This suggests infection with T. gondii stimulates the immune response and/or immune cell recruitment into the brain even when the majority of parasites are in the encysted stage, as seen by the increased abundance of BAG1 and decreased abundance of SAG1 (Table 2). It also suggests a specific subset of host genes are responsible for immune stimulation during chronic infection that are distinct from acute, although many acute infection associated genes are still activated.

Only a small number of host DEGs were less abundant during both acute and chronic infection (Figure 4A). Go term analysis of DEGs with decreased abundance showed little functional overlap between the different stages of infection. The few commonalities in GO categories included secondary metabolism, membrane organization and phagocytosis (Figure 4B). The proposed functions of the DEGs that were less abundant specifically during acute infection were different from those less abundant during chronic infection. GO terms for genes with decreased abundance specifically during acute infection were proteolysis, protein metabolic process, and anatomical structure, while GO terms specific to chronic infection were cell-cell communication and primary metabolic function. The enrichment of metabolism-associated processes among less abundant DEGs suggests a link between T. gondii infection and host metabolism, possibly as a means of restricting parasite growth.

Figure 4
figure 4

Analysis of host genes with decreased abundance during T. gondii infection. (A) A Venn diagram was created to compare DEGs >2-fold decreased abundance in acute vs uninfected (purple) and chronic vs uninfected (red) mice. Forty host genes were less abundant during both acute and chronic infection (magenta). (B) To explore the function of DEGs analyzed in the Venn diagram, a GO term enrichment analysis was performed. Functionally enriched GO terms display little similarity between time points.

Full size image

Analysis of the mouse genes with increased abundance during T. gondiiinfection

Given that more host transcripts had >2-fold increased abundance during chronic infection (Figure 3B), further examination of the most abundant mouse genes during acute and chronic time points was performed to enhance the understanding of the host response. DEGs with a FPKM fold change >20 were compared between acute vs uninfected and chronic vs uninfected mice (Figure 5A). 155 genes met this cut-off in the acute vs uninfected group while 540 genes had 20-fold or higher FPKM values in chronic vs uninfected time points. Of these more abundant DEGs, 146 were shared between acute vs uninfected and chronic vs uninfected groups. Only 9 more abundant DEGs had a fold change >20 in acute vs uninfected that were not highly abundant in chronic vs uninfected time points. Several of these acute-infection specific host genes belong to the family of guanylate-binding proteins, which are GTPases that are induced by interferon-γ (Table 5). Conversely, 394 DEGs had a fold change of >20 in chronic vs uninfected that did not meet this cut-off in acute vs uninfected. The majority of genes increased in abundance during acute infection are maintained into chronic infection. These results show that few host genes are specifically increased during acute T. gondii infection. It also suggests a unique set of host genes are differentially expressed during chronic infection.

Figure 5
figure 5

Few highly expressed DEGs are specific to acute infection. (A) To better characterize abundant DEGs, a Venn diagram of the more abundant DEGs in the mouse with a fold change >20 between acute vs uninfected (purple) and chronic vs uninfected (red) time points. Of these more abundant DEGs, 146 were shared between acute and chronic infection (magenta). Only 9 DEGs had a fold change >20 in acute vs uninfected that were also not highly abundant in chronic vs uninfected time points. (B) To analyze the function of the host more abundant DEGs in chronic vs acute time points, GO term enrichment analysis was performed. Many genes differentially expressed in chronic infection are associated with immune regulation and stress response.

Full size image
Table 5 Mouse DEGs more abundant in acute, but not chronic infection
Full size table

Elevation of immune response genes during chronic infection

To assess the function of the mouse genes increased in abundance specifically during chronic infection, GO term enrichment was performed on genes with a fold change >20 in chronic vs acute infection. GO terms for genes highly increased in abundance during chronic infection were related to stress and immune responses (Figure 5B). Table 6 shows the top 50 DEGs more abundant during chronic infection. Many of these genes are immunoglobulin heavy chain variable regions, which share sequence similarity and can be difficult to differentiate between. The variable regions of immunoglobulins are responsible for specificity of antibodies to antigens, suggesting a different subset of antibodies being produced in response to T. gondii antigens that may not be abundant during acute infection. Another highly differentially expressed set of genes are H2-EB2 and H2-M2, both of which are involved in antigen presentation. H2-EB2 is an MHC class I membrane-associated protein while H2-M2 is an MHC class II membrane-associated protein. These antigen-presenting proteins are key players in the continued stimulation of the immune system at later time points. This increased in abundance of immune genes during chronic infection indicates a unique set of DEGs may be involved in the chronic infection immune response and/or a novel population of immune cells are recruited into the brain after acute infection.

Table 6 The top 50 mouse DEGs more abundant in chronic vs. acute infection
Full size table

Discussion

Infection with T. gondii is often asymptomatic in immune competent individuals, but presents serious health risks if acquired congenitally or if a person becomes immune-compromised. Rising evidence is showing complications, such as psychiatric disorders and increased rates of suicide, occur in people with healthy immune systems who have a chronic T. gondii infection [40, 41]. Currently, drug treatment is only effective against the acute stage of infection and there are no therapeutics interventions available to target the encysted form during chronic infection. There are also no vaccines against T. gondii approved for use in humans. This lack of therapeutic intervention highlights the need to better understand the biological differences between the acute rapidly replicating form of the parasite and the chronic associated cyst stage.

Our data provides a list of candidate genes that could be targeted for novel therapeutics or gene deletion to create a non-persistent vaccine strain (Tables 3 and 4). Many of these genes highly differentially regulated are hypothetical proteins with no known orthologs in other organisms. Hypothetical T. gondii genes not found in the mammalian host could be excellent pathogen specific drug targets. The fact that multiple T. gondii microneme proteins are more abundant in chronic infection stages raises the question of whether parasites during chronic infection are actively invading cells. If parasites are invading at this time, this could explain the continued stimulation of the host response so late in infection. Although microarray analysis of human fibroblasts showed that tissue culture derived bradyzoites stimulated a weaker immune response than tachyzoites after two days [42]. An additional T. gondii gene that is more abundant in chronic infection is cAMP-dependent protein kinase. cAMP-dependent protein kinase is crucial for growth of tachyzoites and is proposed to be critical in tachyzoite to bradyzoite stage conversion[43, 44], but the mechanism is unknown. Another potentially interesting T. gondii gene that is more abundant in chronic infection is calcium dependent protein kinase 5, CDPK5. The T. gondii paralog, TgCDPK1, was shown to be necessary for tachyzoite motility, invasion and egress [45]. Studying the role of CDPK5 in bradyzoite development could be pivotal in understanding the biology of this stage.

Our study also provides crucial insight into host response to the parasite during both acute and chronic infection. It shows that, at least in the beginning stages of chronic infection, the host immune system is still actively combating infection. Chronic infection of T. gondii is typically thought of as a period in which the parasite transitions to the encysted, less immune stimulatory form resulting in a dampening of the immune response. Another study assessing the mouse transcriptome showed many immune associated genes are still expressed at 32 days post T. gondii infection [22]. To assess the overarching function of genes increased in abundance between acute and chronic infection Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed (Additional file 5). The KEGG database is a bioinformatics tool that assembles large-scale molecular datasets, such as gene lists, into biological pathway maps (example Additional file 6). Analysis of our dataset suggests active NK cells are recruited to the brain during chronic infection by the increased abundance of perforin, granzymes A and B, and IL-10 (additional files 3 and 4). NK cells have long been known to be essential for the control of acute T. gondii infection [46], but their role in chronic infection maintenance has yet to be elucidated. NK cells are a significant source of IFN-γ during acute T. gondii infection [47, 48] and IFN-γ is also necessary to maintain chronic infection [49], but producers of IFN-γ during chronic infection have not been determined. In mice with an established chronic T. gondii infection, NK cells are a major source of IFN-γ essential to combat infection with H5N1 influenza virus [50]. Chronic T. gondii infection was equally effective to protect against lethal influenza virus whether the mice had been infected with T. gondii for 1 month or 4 months, suggesting that NK cells are active in late stages of chronic infection. Similarly, NK are elicited in peritoneal exudate 6 months after T. gondii infection [51]. Our data suggests that NK cells play a role in chronic T. gondii infection maintenance, which will be the focus of future studies in understanding infection persistence.

Production of nitric oxide (NO) is crucial for control of T. gondii growth [52, 53], and triggers differentiation of tachyzoites into bradyzoites in tissue culture [54]. Mice deficient in the inducible nitric oxide synthetase gene succumb to non-lethal doses of T. gondii, but only during chronic infection [52, 53]. One host gene differentially expressed during chronic infection is arginase-1 (ARG-1), which had a fold change of nearly 70 in chronic vs. acute samples (Table 6). Arginine is not only a substrate for NO production, but it is an essential amino acid for T. gondii[55]. ARG-1 depletes host cell arginine, possibly as means to starve the parasite, but in type I strains, T. gondii initiates expression of ARG-1 via ROP16, potentially to preserve infected tissue [31]. While NO is detrimental for parasite growth, it also results in inflammation and subsequent destruction of host tissue. In the brain, microglial cells are the main producers of NO and have the potential to cause neuronal degradation. T. gondii-infected astrocytes secrete factors that decrease NO production by microglial cells, thus preserving both host and pathogen during latent infection [56]. It is unclear whether the increase in ARG-1 transcripts is induced by T. gondii or the host as ROP16 is more abundant in T. gondii during acute infection (Table 3), and ARG-1 is more abundant in the host during chronic infection. Furthermore, ROP16 is polymorphic and in type II strains, such as ME49 used in this study, does not maintain STAT3/6 activation and may not initiate ARG-1 expression [57]. An alternative mechanism may be responsible for the increased abundance of ARG-1 during chronic infection. Another highly differentiated host genes during chronic infection is kallikrein-6 peptidase. Kallikrein peptidases have been implicated in infection through involvement in vasodilation and permeability [58], but in the context of bacterial infections, kallikreins are also involved in the generation of NO [58, 59]. Specifically, kallikrein-6 has been shown to be up-regulated in the CNS during inflammation, possibly as a means to promote lymphocyte survival [60, 61]. Our data suggest kallikrein proteases could be involved in the parasite’s transition from the rapidly replicating form to the encysted form in the host. Together these data suggest that regulation of NO production during chronic infection is of vital importance for both the host and the parasites, and will be a future avenue of research.

Conclusions

The depth of RNA-seq coverage allowed, for the first time, simultaneous sampling of both host and microbe during acute and chronic stages of animal infection. In this study, we show the majority of highly expressed T. gondii genes common to both acute and chronic infection are involved in transcription and translation, underscoring that parasites in both stages are actively synthesizing proteins. Similarly, most of the T. gondii genes highly expressed during chronic infection are involved in metabolism, highlighting the metabolic activity of the cyst at 28 days post-infection. For the host, analysis of transcripts at 10 and 28 days post-infection compared to uninfected mice showed that more immunity associated host genes are increased in abundance at 28 days post-infection vs 10 days post-infection. The increase in abundance of T. gondii genes during chronic infection is in conjunction with the heightened host response; indicative of the constant battle for survival between the host and the parasite. Discussed here are only a few examples of hypotheses that can be generated from this transcriptome data set. This dataset is novel because information from the host and pathogen is provided at multiple time points, allowing for the interplay between both to be studied. Many platforms, such as KEGG pathways and DAVID, are available for the research community to further investigate these data and cater to their scientific interests. This data provides the potential to elucidate mechanisms required for Apicomplexan parasites to maintain a relationship with their hosts, which will lead to better therapeutics, vaccines and diagnostic methods.

Methods

Ethics statement

Animals were housed under conventional, specific-pathogen-free conditions and were treated in compliance with guidelines set by the Institutional Animal Care and Use Committee of the University of Wisconsin School of Medicine and Public Health (IACUC), according to IACUC approved protocol number M01545. This protocol adheres to the regulations and guidelines set by the National Research Council. The University of Wisconsin is accredited by the International Association for Assessment and Accreditation of Laboratory Animal Care.

IVIS detection of T. gondiiin the mouse forebrain at 10 days and 28 days post-infection

The ME49 strain of T. gondii with a deletion of the gene HPT and an insertion of the coding region for firefly luciferase, as previously described [62], was used for these experiments. 6–8 week old BALB/C mice (National Cancer Institute, Charles River Laboratories, Frederick, MD) received an intraperitoneal (i.p.) inoculation of 104 freshly lysed tachyzoites. Mice were imaged using IVIS (PerkinElmer) at 10 days and 28 days post-infection. Mice were anesthetized with isoflurane and intravenous (i.v.) injected with 3 mg of luciferin, the substrate for luciferase, and imaged ventrally, dorsally, and then sacrificed. The brains of the mice were removed and soaked in luciferin for 5 minutes prior to imaging.

Generation of mRNA and RNA-seq

A ME49 strain of T. gondii that was recently passaged through the sexual cycle was used to inoculate mice for RNA-seq analysis. ME49 was maintained as tachyzoites in monolayers of Human Foreskin Fibroblasts in Dulbecco’s Modified Eagle’s Medium supplemented with 10% FBS, 2 mM L-glutamine, and 1% penicillin-streptomycin. 6–8 week old CBA/J mice (National Cancer Institute, Charles River Laboratories, Frederick, MD) were either left uninfected or i.p. injected with 104 parasites and were sacrificed at 10 days and 28 days post-infection. Uninfected mice were sacrificed along with the 28 day post-infection group. We selected T. gondii infected mice that were healthy and not displaying any signs of disease, so samples would not contain host transcripts involved with inappetence, dehydration or general malaise to confound our analyses. To minimize changes to the transcriptome, the forebrains were rapidly and precisely sectioned at the intersection of the optic nerves using a mouse brain matrix (Zivic Instruments) with less than one minute between animal sacrifice to forebrain homogenization in 3 mL of TRIzol. Total RNA was isolated according to manufacturer’s protocol. RNA was purified using Promega SV total RNA isolation system according to manufacturer’s protocol. RNA was submitted to the University of Wisconsin Biotechnology Center for purity analysis using the Agilent 2100 Bioanalyzer and sequencing using the Illumina HiSeq2000. Sequencing was performed on each individual mouse and samples were not pooled. Infection was quantified in the un-used hindbrains collected at 28 days post-infection and stained with fluorescein labeled Dolichos biflorus agglutinin (Vector Laboratories) for cyst detection. All 28 day post-infection hindbrains contained a minimum of 10,000 cysts.

Determination of T. gondiiparasite numbers in mouse forebrain samples

Genomic DNA was extracted from each mouse forebrain at the time of RNA extraction using TRIzol according to manufactures instructions. DNA was purified by phenol/chloroform extraction followed by ethanol precipitation. Genomic DNA was used as the template for quantitative PCR using T. gondii primers for the housekeeping gene alpha-tubulin (TUB1A). Tub1A Forward primer 5’-GACGACGCCTTCAACACCTTCTTT-3’, Tub1A Rev 5’-AGTTGTTCGCAGCATCCTCTTTCC-3’. Primer efficiency for TUB1A was 2.002 with an R2 value of .99 using T. gondii genomic DNA. To determine parasite burden in the mouse forebrain samples, a standard curve was generated using a genomic DNA preparation of known parasite numbers. Quantitative PCR was performed on serial dilutions of parasite genomic DNA, using TUB1A primers, ranging from 10 to 1×106 parasites. A best-fit logarithmic line was generated with an R2 of 0.999. The equation of the line along with Ct values obtained from qPCR of TUB1A on genomic DNA from each forebrain sample was used to extrapolate parasite numbers. qPCR was performed on each sample in duplicate using BIO-RAD iTaq Universal SYBR Green Supermix product number 172–5121.

Quantitative PCR of T. gondiiCCCH zinc fingers

Sequences for TGME49_224630, TGME49_269270, and TGME49_311100 were obtained from ToxoDB.org. Sequences were run through BLASTp to confirm presence of CCCH zinc finger motifs. cDNA was generated from the same RNA samples used for RNA sequencing with Invitrogen Superscript III Reverse Transcriptase cDNA synthesis kit. All CCCH zinc fingers were normalized to the T. gondii house keeping gene tub1A. Efficiencies were determined using in vitro bradyzoite cDNA. RNA was extracted from 5 day bradyzoites grown under low CO2 and high pH conditions using TRIzol. cDNA was generated using the Invitrogen Superscript III Reverse Transcriptase cDNA synthesis kit. Efficiencies were calculated using the slopes of a 1:10 dilution series (neat through 104) and the formula E = 10[-1/slope]. Efficiencies for tub1A, TGME49_224630, TGME49_269270, and TGME49_311100 were 1.96, 1.89, 2.14, and 2.04 (Between 95-107% efficient). Quantitative PCR was performed using Bio-Rad iTaq Universal SYBR Green Supermix on an Applied Biosystems StepOnePlus Real-Time PCR system. Primers were used at a 300 nM concentration and an extension temperature of 60°C for 60 seconds for all primer sets except TGME49_224630 which was run at 56.5°C for 60 seconds. Relative quantification was calculated using Pfaffl’s method [63]. The three biological replicates were used and conducted in duplicate. Wells with only one melt curve and temperature were used, and duplicate Ct values were all at or below a 0.25 difference in cycle threshold value. Primers: Tub1A, same as previously described above

TGME49_224630 Forward 5’-GCGAGGATGAGTGTGGG-3’

Reverse 5’- AGGCGTCACCGTTTGG-3’

TGME49_269270 Forward 5’- GCTTACCGAGGATGACCTGCT-3’

Reverse 5’-CCGTACACTGGTGGCGATCAT-3’

TGME49_311100 Forward 5’- TTTGCCCACACAGCCGAAGAA-3’

Reverse 5’- GCCACAGATGCCTTCCGTAAC-3’.

RNA-seq, mapping and differential expression analysis

Approximately 950,000,000 paired end 100 bp reads were generated from Illumina HiSeq2000 sequencing. Aligning RNA-seq data for eukaryotic organisms becomes difficult when mapping to a genomic reference because of the presence of introns in the reference and polyadenylated transcripts in the data. If an RNA-seq read spans an exon-exon junction or a polyadenylated region of a transcript it will be "unmappable" to the reference genome and is discarded. Bioinformatics software, such as TopHat, has been created to consider exon-exon boundaries during the mapping process [64]. Raw reads were uploaded onto the Galaxy platform [65–67]. Reads were filtered by Sanger quality score using FASTQ Groomer v. 1.0.4 and paired end reads were aligned against the genomes of T. gondii (TGME49 version 9.0; ToxoDB.org) and M. musculus (GRCm38 version 74.38; ensemble.org/Mus_musculus) references uploaded into Galaxy using TopHat2 [68]. Parameters for TopHat2: Max edit distance of 2, final read mismatch of 2, anchor length of 8, minimum intron length of 70, maximum intron length of 500000, max insertion and deletion length of 3, number of mismatches allowed of 2, and a minimum length of read segments of 25. Reads were not treated as strand specific as they were paired end reads. The total numbers of reads were as followed: Uninfected mouse 1 was 112075860, Uninfected mouse 2 was 112948998, Uninfected mouse 3 was 103209252, 10 day post-infection mouse 1 was 102581171, 10 day post-infection mouse 2 was 125546828, 10 day post-infection mouse 3 was 81630704, 28 day post-infection mouse 1 was 103765423, 28 day post-infection mouse 2 was 113094439, and 28 day post-infection mouse 3 was 114177191. The number of reads that mapped to the T. gondii genomic reference file and M. musculus genomic reference files are listed in Table 1.

The program Cufflinks [64] was used to convert aligned reads of BAM files, generated from Tophat2, into relative expression values for each gene represented as FPKM (fragments per kilobase of exon per million mapped reads). Cuffdiff was used to detect significant changes in differential expression between the experimental groups. When running Cuffdiff, a GTF file obtained from ToxoDB.org and ensemble.org/Mus_musculus was used as a guide and the TopHat2 aligned BAM files from each biological replicate were used as the source of comparison between experimental groups. A geometric library normalization and a pooled cross-replicate dispersion estimation method was used when comparing differential expression between each experimental group. Genes with a p-value, for statistical significance, and q-value, to detect the false discovery rate, of <0.05 were considered differentially expressed. The "gene differential testing" output file from Cuffdiff was used to identify differentially expressed genes.

GO term analysis

FASTA sequences of the most abundant T. gondii genes from different time points during infection were loaded into the program Blast2Go [29]. T. gondii is not an available organism on many GO term analysis programs, making Blast2Go ideal for uncommon models. Sequences were run with the blastx program against the nr database. Aligned sequences were mapped and assigned GO term annotations. Combined graphs were generated representing the most enriched GO terms in the provided gene list. A score was assigned to determine significance of enrichment. For analysis of M. musculus, a gene list was generated from each experimental time point based on gene names in column C of Additional files 2,3, and 4. Ensembl gene IDs were obtained from these gene names, exported from BioMart and uploaded to the functional annotation tool. A functional annotation chart of the enriched GO terms was generated using GO terms associated with biological process. GO terms were assigned a p-value to indicate significance of enrichment [69]. Only GO terms with a p-value <0.05 were used to represent functional enrichment. For further analysis of M. musculus, the online Database for Annotation, Visualization, and Integrated Discovery (DAVID) was used (david.abcc.ncifcrf.gov).

Availability

Raw short read RNA-seq data has been submitted to ToxoDB and to NCBI with SRA: SRS550800, http://www.ncbi.nlm.nih.gov/biosample/2615816.