T. gondii, like Eimeria tenella and Cryptosporidium parvum, uses glucose as its main source of energy in its rapidly dividing tachyzoite stage 16. Phylogenetic analyses have shown that two of the glycolytic genes in T. gondii, enolase and glucose-6-phosphate isomerase, are closely related to their corresponding homologs in plants, suggesting that they were acquired and potentially suitable as drug targets due to their distinct evolutionary origin 17. Glycolysis has also been actively studied with respect to stage differentiation in T. gondii. Three key glycolytic enzymes - glucose-6-phosphate isomerase [ToxoDB:76.m00001], lactate dehydrogenase (LDH) and enolase (ENO) [ToxoDB:59.m03410] - exhibit developmentally regulated expression 18. Stage-specific cDNAs have been isolated that encode distinct isoforms of LDH: LDH1 (tachyzoite) and LDH2 (bradyzoite) 19. Experimental evidence based on the detection of their respective mRNA and protein products indicates that LDH1 is post-translationally repressed while LDH2 is transcriptionally induced in bradyzoites 19. Similarly, stage-specific cDNAs have also been isolated for distinct forms of ENO: ENO1 (bradyzoite) and ENO2 (tachyzoite) 20. Stage-specific expression of the two enolases is brought about by the presence of specific cis-regulatory elements in the promoter regions of these genes 10. The regulation of the genes involved in glycolysis presents an intriguing case study from developmental, evolutionary and regulatory perspectives.

We analyzed the upstream sequences of 11 genes involved in tachyzoite glycolysis to identify conserved, over-represented sequence motifs (Table 1). We report the analysis of two candidate motifs here: motif GLYCA, also found upstream of six orthologs in E. tenella, and motif GLYCB, found exclusively in T. gondii. These motifs were not reported in the aforementioned studies on stage-specific regulation of the enolase gene 18. Motif GLYCA, represented by the consensus 5'GCTKCMTY (Figure 1a) is an 8 bp well-conserved sequence occurring at least once per sequence on the forward strand (Figure 1b). It does not show significant positional conservation, but motifs found upstream of orthologs in E. tenella are found to be 100% conserved in sequence to their counterpart in T. gondii. Motif GLYCA is not found in the upstream regions of the bradyzoite isoforms of the stage-specific glycolytic genes (ENO2 and LDH1). Motif GLYCB is also an 8 bp motif represented by the consensus sequence 5'TGCASTNT (Figure 1a), with 6 of 8 bases conserved in more than 90% of the occurrences. This motif is present once per sequence and can occur on either strand (Figure 1b). Motif GLYCB was also found in the upstream regions of the bradyzoite-specific copies of enolase and LDH (data not shown).

Mutagenesis of GLYCA to the sequence 5'AACAAACA in the ENO2 promoter resulted in a small increase in promoter activity. Mutagenesis of GLYCB to the sequence 5'CAACACAC within the ENO2 promoter resulted in a small decrease in promoter activity (Figure 1c, d). However, when both motifs were mutagenized, a larger decrease in promoter activity was seen. These results are complex in comparison to patterns seen with motifs for other groups of genes (see below). It must be noted that the changes in expression levels caused by mutagenizing each individual sequence in the ENO2 promoter are of small magnitude, but statistically significant. It is possible that the effects of mutagenizing each motif are not very severe in their effect, while the double mutant shows a large decrease in reporter expression, indicating a definite role for both of these motifs, in concert, to affect downstream gene expression. An alternative scenario to explain this result is one in which mutagenesis of GLYCA gives rise to a chimeric motif that enhances downstream gene-expression only in the presence of wild-type (WT) GLYCB. The strong evolutionary conservation of motif GLYCA in E. tenella and the significant decrease in reporter activity in the double mutant lend support to their role in regulating gene expression. Further experiments are needed to fully resolve these intriguing results.

Purines and pyrimidines are the building blocks of nucleic acids in living cells. All protozoan parasites examined thus far are unable to synthesize purines de novo and depend upon salvage enzymes to obtain purines from the host 21. Most protists, however, possess a full set of de novo pyrimidine biosynthesis enzymes, with one exception, C. parvum, which has lost the de novo pathway and evolved to also salvage pyrimidines from the host cell 22. Enzymes involved in nucleotide metabolism in protozoan parasites can serve as promising drug targets because they are essential to the parasite's survival and are also evolutionarily distinct from host enzymes in some cases 22. In T. gondii, it was found that de novo pyrimidine biosynthesis is essential for the virulence of the parasite 23. We examined eight genes encoding enzymes involved in nucleotide biosynthesis and salvage in T. gondii and selected two conserved motifs found in their upstream regions as candidates for experimental validation. Motif NTBA is an A-rich 9 bp motif represented by the consensus 5'GCAAAMGRA (Figure 2a). It is very well conserved in four orthologs in E. tenella. Motif NTBA is present only once upstream of each gene and is always found on the positive strand. It is primarily located at 1,000-1,500 bp upstream of the translation start (Figure 2b). Motif NTBB is an 8 bp long T-rich motif and is exclusive to T. gondii. It is represented by the consensus sequence 5'TTTYTCGC (Figure 2a) and is also found only once upstream of each gene on the forward strand. The two motifs are typically present within 300-400 bp of each other (Figure 2b).

To establish the biological significance of these motifs, we mutagenized NTBA to the sequence 5'AAGCGCAAG and NTBB to the sequence 5'GTGTGTG (Figure 2c). Mutagenesis of either of these motifs individually in the promoter of the gene encoding uracil phosphoribosyl transferase (UPRT) [ToxoDB:583.m00018] showed no significant change in promoter activity. Mutagenesis of both motifs within the UPRT promoter resulted in a seven-fold increase in reporter gene-expression, indicating that the two motifs function in repressing gene-expression and possibly possess redundancy in function (Figure 2d).

Micronemes are secretory organelles found in apicomplexan parasites and serve as compartments for the storage and trafficking of micronemal proteins, a family of proteins that function as ligand for host-cell receptors 24. These proteins play a very important role in the active process of host-cell adhesion and invasion during the parasite life cycle. We analyzed the upstream sequences of 12 microneme protein-encoding genes in T. gondii and corresponding upstream sequences of four orthologs in E. tenella. We identified two well-conserved sequence motifs in this data set that we subsequently selected for further experimental characterization. Motif MICA is an 8 bp motif represented by the consensus sequence 5'GCGTCDCW (Figure 3a). It is found at least twice in the majority of the upstream regions occurring on either strand and does not show conservation of position relative to the translational start site (Figure 3b). This motif was also found upstream of E. tenella micronemal protein genes. In the reverse orientation, this motif closely resembles the 5'WGAGACG motif that has been identified in previous studies to function as a regulatory element in several promoters of T. gondii 8. Motif MICB is an 8 bp motif with the very well conserved sequence 5'SMTGCAGY (Figure 3a); the core 'TGCA' nucleotides are conserved in 100% of occurrences. This motif occurs once upstream in all 11 micronemal protein genes in T. gondii, but was not found in the corresponding orthologs in E. tenella. It does not show conservation of position relative to the translational start site, and is always found on the forward strand (Figure 3b).

To characterize the functional significance of these conserved motifs, each was mutagenized to an 8 bp polyA sequence (5'AAAAAAAA; Figure 3c). The mutagenesis of motif MICA in the Mic8 (Micronemal protein 8) [ToxoDB: 50.m00002] promoter led to a tenfold reduction in reporter activity, and the mutagenesis of motif MICB led to a threefold reduction in reporter expression. When both MICA and MICB were mutagenized in the same promoter, it had a dramatic effect on promoter activity (the raw value of firefly expression levels (440 units) was comparable to that of non-transfected cells (386 units) (Figure 3d)). From these data, we infer that both MICA and MICB act positively to enhance gene expression from the Mic8 promoter, and together exert an additive effect on downstream gene-expression, as is indicated by the loss of expression when both MICA and MICB are mutagenized (Figure 3d).

Examination of stage-specific expressed sequence tag libraries in E. tenella and T. gondii indicates that the coccidia regulate de novo ribosome biosynthesis at the transcriptional level 25. In a recent study 26 the authors examined a large set of cytoplasmic ribosomal proteins in T. gondii (79 genes in all) and describe the presence of two well-conserved motifs, TRP-1 (motif RPA; 5'CGGCTTATATTCG) and TRP-2 (motif RPB; 5'YGCATGCR) (Figure 4a) identified by MEME in all promoters. The sequence of TRP-2 (RPB) is similar to the 8 bp element 5'TGCATGCA reported to be overrepresented in the non-coding regions of the apicomplexans C. parvum, T. gondii and E. tenella 27. This sequence is also similar to one of the binding sites of the AP2-domain containing transcription factors as inferred from protein-based microarray studies conducted in P. falciparum 28. In a study of the promoter strengths of eight of the ribosomal protein genes, no correlation could be found between multiple occurrences of one or both motifs and promoter strength in the eight promoters 29. However, the biological function of these motifs was not reported. We conducted analyses on a subset of these genes (eight promoters) and also recovered the motifs TRP-1 (RPA) and TRP-2 (RPB) as described by van Poppel et al. 29 (Figure 4b). We mutagenized these motifs in our analyses to ascertain if they functioned in a sequence-specific manner to affect promoter activity.

Motif TRP-1 (RPA) in the RPL9 (Ribosomal protein L9) promoter [ToxoDB:76.m00009] was mutagenized to the sequence 5'CGAAGTATGCGAG (retaining the WT sequence at 3 of the 13 nucleotide positions due to mutagenesis challenges presented by the length of this motif) and motif TRP-2 (RPB), which occurs twice in the RPL9 promoter, was mutagenized at both sites (singly and jointly) to the sequence 5'TAAATAAA (Figure 4c). TRP-1 (RPA) did not affect reporter expression when mutagenized individually or in combination with TRP-2 (RPB). This observation may be attributed to the fact that not all of the bases in this motif were mutagenized, indicating that the three WT positions might be crucial and sufficient for the function of this motif or that this motif may serve a function during a different stage of development or not serve a function related to gene expression. These results warrant further examination. Mutagenesis of one of the copies of motif RPB resulted in a 50% reduction in promoter activity, while mutagenesis of both the copies of RPB caused a 75% reduction in gene expression relative to the WT promoter (Figure 4d). These data indicate that TRP-2 (RPB) enhances gene expression from the RPL9 promoter; the presence of additional copies of this motif likely confers additional strength to the promoter.

We examined the occurrences of each of the motifs to determine if there was over-representation within upstream regions relative to coding regions. Table 2 lists the genome-wide occurrences of each of the candidate motifs within the upstream and the coding regions of the genome, respectively, as computed by MAST (Motif Analysis and Search Tool) 15. In order to normalize for the different sizes of the two data sets, the motif count is represented as number of motifs per 10 kbp (motif density). Of the eight candidate motifs selected in this study, the RPB (TRP-2) motif (5'YGCATGCR) has the highest occurrence within upstream regions, 4,030 occurrences upstream of 1,311 genes. When normalized to the total size of each database (upstream or coding), the candidate motifs (except GLYCA and MICB) were found to be significantly (two- to four-fold) over-represented (p < 0.001) in the upstream regions relative to the coding regions (Table 2, Figure 5).

We calculated the expected frequency of motifs within the upstream and coding regions based on the motif length, degeneracy and the composition and size of the database (Materials and methods). The expected occurrences of most of the motifs are almost equal in both databases (upstream and coding) because of the similarity in size and nucleotide composition of the two databases. The motifs are not found to occur at a significantly greater frequency than expected, exceptions being NTBA, which is found at a higher frequency than expected (p < 0.05) within the upstream and coding regions, and motifs NTBB and RPA, which are found at frequencies higher than expected in the coding regions only (Table 3 in Additional data file 1).

Thus, while most of the regulatory motifs are present at a slightly higher frequency in the upstream regions when compared to the coding regions, they do not occur at a higher frequency than expected in either upstream or coding regions. These analyses highlight the limitations of approaches that use statistical overrepresentation of motifs as a reliable and sufficient property to identify biologically relevant motifs. It is possible that a functional regulatory motif may not be detectable by sequence alone. The surrounding sequence context and other still elusive signals may be involved in enabling it to function as a regulatory motif.

To examine enrichment of specific Gene Ontology (GO) categories among all genes containing any of the eight candidate upstream motifs, we retrieved first-level GO annotations for all of the motif-containing genes (Table 2 in Additional data file 1) for each of the three main GO categories: 'cellular component', 'molecular function' and 'biological process'. We also included lower level GO annotation IDs for the specific pathways/functional groups included in this study (Materials and methods). Table 4 in Additional data file 1 lists the GO categories that were significantly enriched within the motif-containing gene sets. Some of the motif-containing gene sets are also enriched in GO terms related to the corresponding function/pathway used to initially identify the motif, indicating that the regulatory motif may indeed be a subset-specific or pathway-specific motif. On the other hand, some motif-containing gene sets do not show enrichment for a particular GO category, but rather to a more general, functional classification. For example, genes containing the motifs discovered in the analysis of ribosomal protein-coding genes (RPA and RPB) are enriched in annotated higher-level GO categories such as organelle and regulation of biological process. This indicates that a large number of genes that contain the RPA (TRP-1) and RPB (TRP-2) motifs can be assigned to ribosome or translational-specific functions, indicating a broad subset specificity for this motif. Genes that contain the MICA or MICB motifs do not show any GO category enrichment, indicating a more general role for these upstream motifs. When deeper-level GO annotations for particular processes (such as 'ribosome' [GO:0005840]) are enumerated among the motif-containing genes, we find that the genome-wide lists of genes that contain RPA and RPB motifs are also enriched in corresponding GO categories ('ribosome' and 'translation'), indicating an even stronger specific association of these motifs with the corresponding processes (Table 3).

Promoter organization in T. gondii has been studied in a few genes thus far 781011. In these studies, it has been observed that a gene-proximal region is necessary for minimal gene expression and additional upstream sequence helps to enhance expression from the same promoter. However, very little is known about the mechanism of gene regulation and the prevalence and type of transcriptional signals and regulatory apparatus in this organism. Analyses of genome sequences and individual gene-specific experiments point out two deviations from what has been observed in other model eukaryotes. First, canonical eukaryotic promoter elements such as the TATA box have not been found in T. gondii promoter regions 8, although a highly divergent TATA binding protein has been reported 9. Furthermore, there is a stark paucity of known specialized transcription factors encoded in the genome 9. A similar scenario is seen in two other apicomplexan parasites, P. falciparum and C. parvum 3031. This paradox can be explained in two ways: these organisms do not employ a specialized transcriptional apparatus to regulate their genes; or a specialized transcriptional machinery exists but is so divergent from known eukaryotic counterparts that its components cannot be detected by simple similarity-based searches. Recent studies have shown that the T. gondii genome encodes a rich repertoire of histone-modifying enzymes, and epigenetic regulation has been purported to be responsible for stage-switching in the parasite 3233. More recently, chromatin immunoprecipitation (ChIP)-on-chip experiments conducted on 1% of the T. gondii genome reveal a strong association between specific histone modification marks and active promoter regions 34. It is likely that histone-mediated regulation is responsible for regulation of genes to a sizeable extent in T. gondii. Serial analysis of gene expression (SAGE) studies of genes expressed during key life-cycle stages 13 have shown that the mRNA pool of T. gondii is highly dynamic and gene expression is controlled in a time- and stage-dependent manner. These studies have also shown that co-expressed genes in T. gondii do not cluster in the genome with respect to chromosomal location. Searches of the Plasmodium genome sequence for transcription factors using secondary structure similarity have revealed the presence of putative transcription factors that were missed in simple sequence-based searches 35. A divergent, putative, specialized transcription factor ApiAP2 has also been reported in the apicomplexa 36. A large percentage of proteins in T. gondii are 'hypothetical proteins' with no known function and might possibly encode parasite-specific functions, including transcriptional regulatory proteins. It is plausible that such highly divergent regulatory proteins utilize very different cis-elements for their recruitment, which would explain the absence of canonical cis-elements in the promoters studied thus far.

We have exploited the availability of genome sequence for T. gondii to identify conserved upstream motifs in diverse groups of functionally related genes. We identified over-represented motifs by de novo pattern finding and tested their function in vitro, in the parasite, by specifically mutagenizing them in their native promoter context and measuring reporter activity. For each group, two candidate motifs were selected and characterized for their function in their endogenous promoter. We find that seven out of eight motifs identified by de novo pattern finding show a statistically significant role in promoter activity. We have shown that conserved over-represented motifs play a definite role in gene-expression, and can affect promoter activity either positively or negatively (Figures 1, 2, 3, 4). It is exciting to note that some of the motifs that affect gene expression also exhibit cross-species conservation, as shown by their presence upstream of orthologs in E. tenella (Table 1).

Our studies have shown that in spite of the lack of a priori knowledge concerning the nature of regulatory sequences and/or expression profiles, it is possible to identify putative cis-regulatory elements. We have shown that elements identified purely on the basis of computational techniques can be functionally relevant for gene expression. We previously characterized putative cis-regulatory elements in the genome of C. parvum in a similar fashion where we established a correlation between over-represented elements and co-ordinate gene expression 37. In a comparison of the genes common to both studies (genes of the glycolytic and nucleotide metabolism pathways), we do not detect identical motifs in T. gondii and C. parvum. Given the evolutionary divergence and difference in genome organization and content between these two parasites, it is not unexpected that they do not share some specialized components of the regulatory machinery, or these components may have evolved so rapidly as to be unrecognizable as orthologs.

The use of asynchronous populations of parasites, as is the case here, is expected to dilute out expression effects to some extent, and this is reflected in the occasional small, but significant, change in promoter activity upon mutagenesis of a motif (Figure 1d, motif GLYCA; p < 0.05). In spite of these limitations, our study has successfully identified six novel cis-regulatory elements and established the functional significance of one previously reported conserved upstream element. The study of stage-specific gene regulation in T. gondii has been an active area of investigation, and small-scale microarray studies have reported genes that are preferentially expressed in either developmental stage 3839. Recently, three T. gondii studies have reported the presence of novel cis-regulatory elements in promoters of genes regulated in a stage-specific manner 101112. However, gene regulation within the tachyzoite stage has not been well studied. The lack of a synchronized population of tachyzoites (with respect to their cell cycle) in in vitro culture makes it difficult to address gene regulatory questions in the actively multiplying tachyzoite, as any given population of cells in culture consists of parasites at different points of their cell cycle.

Our study reports the presence of different cis-regulatory elements controlling gene-expression within the tachyzoite stage of the parasite and is among the first to show evidence for the presence of modular organization of promoters in T. gondii. Using site-specific mutagenesis of conserved upstream motifs and subsequent reporter assays, we have shown that the identified cis-elements can exhibit co-operative, additive or redundant effects within a promoter and operate in a sequence-specific manner to control gene expression (Figures 1, 2, 3, 4). The putative transcription factors or repressors recruited by these elements in order to facilitate gene regulation remain to be determined. One of the limitations of investigating gene expression within the tachyzoite stage is the lack of a parasite population enhanced in the production of a specific transcription factor that could be recruited by these cis-regulatory elements. Consequently, the detection of such putative transcription factors or repressor proteins from a mixed population of parasites by experiments such as electrophoretic mobility shift assays has proven to be challenging and yield inconsistent results (data not shown).

Whole genome sequence (v.3.3) and gene-predictions for T. gondii were obtained from ToxoDB (release 4.0) 40. Scripts were written in PERL to extract the upstream sequence (2 kb or until the previous upstream gene is encountered, whichever sequence was smaller) for every predicted gene to create an 'upstream sequence database'. This database was screened for possible missed protein coding regions by performing a BLASTX against all protein sequences in the non-redundant database at NCBI. Sequences that contained greater than 80% similarity to coding sequences were pruned. Groups of genes (hereafter referred to as seed genes) used to identify common over-represented upstream motifs were selected based on the hypothesis that genes belonging to the same biochemical or functional pathway should be co-regulated and, hence, possess common upstream regulatory elements. Members from each of the four groups considered in this study (glycolysis, nucleotide metabolism, micronemal proteins and ribosomal proteins) were identified based on existing annotation from ToxoDB. Additionally, BLAST analyses using sequence information of orthologs from P. falciparum and C. parvum were employed to identify the corresponding genes in T. gondii when necessary. Given the limited annotation at the time of this study, it is possible that the gene lists under each functional group are not exhaustive. The pattern-finding algorithm MEME 15 was used to identify over-represented conserved motifs in the upstream regions of genes within each functional group. MEME was run using the parameters minw = 8, maxw = 20, in all three modes (tcm, oops and zoops) and the results were manually examined to pick candidate motifs. Upstream sequences of corresponding orthologs from E. tenella (a related coccidian parasite) were also used whenever possible to identify evolutionarily conserved motifs. In order to narrow down the results of MEME, a rule-based approach was adopted to pick candidate motifs for subsequent experimental validation. Candidate motifs included those that were over-represented in the upstream regions in comparison to the background set (entire upstream regions database), showed considerable conservation in sequence and/or position, and were present in all the sequences within each group.

The program MAST 15 was used to search the most recently annotated version of the T. gondii genome (release 4.3) for the presence of each candidate motif in all 7,817 genes in the genome divided into coding regions (16,862,741 bp) and upstream regions (11,685,162 bp). To normalize for the different sizes of the coding regions database and the upstream regions database, the motif density was computed by calculating the number of motifs per 10,000 bp. Chi square analysis was performed to examine whether there was a significant difference in occurrence of each motif between upstream and coding regions.

The expected frequency of motifs within each set (the upstream regions database and the coding regions database) was calculated as follows. The upstream regions database has a base composition of 48% AT (A = 23%, T = 25%, G = 25% and C = 26%) and the coding regions database has a base composition of 42% AT (A = 22%, T = 20%, G = 30%, C = 28%). Given these statistics, the 8 bp motif GLYCA (GCTKCMTY) will be expected to occur once every 1,780 bases in the upstream regions database and once every 2,406 bases in the coding regions database (after accounting for positional degeneracy). Taking the database sizes into account, the expected motif density per 10,000 bases was calculated for each motif in each database. Chi square analysis was used to examine the statistical significance of the differences in the observed and expected frequencies of motifs in the upstream and coding regions.

For each group of functionally related genes considered in this study, a promoter that contained a single occurrence of the candidate motif was chosen to study the role of the motif in driving gene expression. Promoter sequences were PCR-amplified from parasite genomic DNA and a two-step overlap-extension PCR technique was employed to carry out site-directed mutagenesis 44 to alter the candidate motif sequence (see Table 1 in Additional data file 1 for primer sequences). All or a majority of the bases in each motif were substituted by base-specific transversions, thus destroying the original sequence of the candidate motif but maintaining the spacing within the promoter (Figures 1c, 2c, 3c and 4c). Successful mutagenesis was confirmed by sequencing or by restriction digest analysis. The Gateway™ cloning system was used to clone the WT and mutagenized promoters individually upstream of a firefly luciferase-expressing vector (test-firefly). As an internal control, a constitutive promoter (T. gondii α-tubulin promoter)-driven renilla luciferase-expressing construct (α-tub-renilla) was co-transfected along with the experimental construct (Additional data file 2).

T. gondii RH tachyzoites were cultured in human foreskin fibroblasts (hTERT cells, BJ Biomedicals) as previously described 8. Transient transfection was performed via electroporation, using freshly lysed parasites, needle-passaged and filtered through a 3 micron filter and resuspended in cytomix 845. Immediately prior to use, freshly prepared 2 mM ATP and 5 mM glutathione were added to the cytomix and sterile-filtered. For each co-transfection, 2 × 10⁷ parasites were transfected via electroporation with a mixture of sterile circular plasmid DNA of α-tub-renilla (control) and test-firefly mixed in a ratio of 2.5:1. (40 μg test + 16 μg control). Electroporation was performed in a 2 mm gap cuvette using a BTX electroporator: 1.8 kv, 100 Ω, 25 μF. Post-electroporation, the parasites were allowed to rest for 15 minutes in the cuvette and then transferred to T25 tissue culture flasks. Then, 18-24 hours post-electroporation, the cells were scraped and lysed using passive lysis buffer (Promega, Madison, WI, USA) and a dual luciferase assay was performed with the extract using the Promega Dual Luciferase kit. Briefly, the different substrate requirements for each enzyme, firefly luciferase and renilla luciferase allowed us to assay reporter expression for each construct sequentially within the same extract. Reporter activity from the WT or mutagenized promoter was measured relative to the internal control, eliminating errors due to variation in parasite populations and individual transfections. Enzyme activity was measured using a dual luciferase-ready luminometer. Each electroporation experiment was performed in triplicate and luciferase assays were performed in duplicate for expression measurements. The unpaired Students t-test was used to calculate the statistically significant difference in expression levels between WT and mutagenized promoter activity; p < 0.05 was considered statistically significant.