Bioinformatics

To identify potential apicomplexan polyadenylation complex subunits, TBLASTN was used to search the genomes of P. falciparum, T. gondii, and S. neurona. TBLASTN was chosen to allow for the identification of possible protein-coding regions that have not yet been annotated by the respective communities. The three apicomplexan genomes were downloaded from Plasmodb and Toxodb; unmasked genome sequences were used for this study. The Arabidopsis genome was included in this analysis to provide examples of identification of potential subunits in a genome from a phylum apart from metazoans; the genome sequence was downloaded from the TAIR website. The human polyadenylation complex subunits used as queries in this study are listed in Table 1. To present the results, the “e-value” of the best match for each subunit was transformed by taking the -log(10) value for each hit; these transformed values were then plotted.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Human poly(A) complex subunits used as queries for TBLASTN.

https://doi.org/10.1371/journal.pone.0203317.t001

Virtual northern blots were generated using a S. neurona RNASeq dataset (accession SRR1784218); these reads were generated from RNA isolated from S. neurona SO SN1 merozoites [32] and were used without modification. Mapping was performed using the Large Gap Read Mapping tool in the CLC Genomics Workbench package with the following parameters: References = the genomic regions surrounding the respective genes, Maximum number of hits for a segment = 10, Maximum distance from seed = 50,000, Multi match mode = ignore, Mismatch cost = 2, Insertion cost = 3, Deletion cost = 3, Similarity = 0.9, Length fraction = 0.5, Override default distances = No, Create Report = Yes, Create non mapped list = No.

Cell culturing and isolation of DNA and RNA

Sarcocystis neurona (strain SN3.E1), T. gondii (strain RH), and N. caninum (strain NC-1) were propagated by serial passage in monolayers of BT cells, as described [33]. Upon lysis of the infected monolayers, extracellular zoites were harvested by passing through 23 and 25-gauge needles and filter-purified to remove host cell debris. Zoites were pelleted and stored at -80°C until used for DNA or RNA isolation.

For analysis of intracellular parasites, freshly egressed merozoites were inoculated onto BT cell monolayers and allowed to invade for 1 hr. Extracellular zoites were removed from the cultures by washing 2X with PBS and then incubated for 48 hrs to allow schizont development. The infected monolayer was harvested from the flask using a cell scraper, pelleted, and stored at– 80°C until used.

For isolation of genomic DNA from S. neurona, harvested merozoites were incubated overnight in 10 mM Tris-HCL (pH 8.0), 100 mM EDTA, 1% sarkosyl, 2 mg/ml proteinase K, followed by extraction with phenol-chloroform and precipitation with 2 volumes 95% ethanol. The resulting nucleic acid pellet was resuspended in TE containing RNase A and stored at 4°C.

RNA was isolated from parasites or infected host cells by adding 1mL Trizol to cell pellets, incubating for 5 minutes at room temperature, then adding 200 μL chloroform and incubating for 3 minutes at room temperature. Samples were centrifuged at 12,000 RPM for 15 minutes, and the top layer was transferred to a fresh tube. 500 μL of isopropanol was added and mixed, and allowed to sit overnight at -20°C. The tubes were then spun at 12,000 RMP for 10 minutes and the pellet was washed with 1 mL 75% EtOH. The pellet was air dried for 5–10 minutes and resuspended in RNase free water.

Poly(A) site profiling

Libraries for genome-wide poly(A) site profiling was performed following the procedures described in Ma et al. [34] and Pati et al. [35], using the oligonucleotide primers described in S1 Table. So-called poly(A) tag (PAT-Seq) libraries were sequenced on an Illumina MiSeq instrument; the sequencing data are available under Bioproject ID PRJNA436572. Sequencing reads were trimmed, demultiplexed, and mapped to the respective genomes using CLC Genomics Workbench (latest version used was 10.1); the results of these analyses are summarized in S1 File. Subsequent analyses were performed following the pipelines described in Bell et al. [36] and de Lorenzo et al. [37]. It is important to note that, prior to defining poly(A) sites and quantifying poly(A) site usage, mapped reads were reduced to single nucleotide coordinates that correspond to the mRNA-poly(A) junction defined by the reads. To perform the nucleotide composition study, custom genome annotations for T. gondii, S. neurona, and N. caninum were produced by extending annotated genes by some 500, 500, and 2000 bp, respectively, and then extracting the respective 3’-UTRs; poly(A) sites that mapped to the resulting extended 3’-UTRs were used for this study. (The lengths of these extensions were arrived at empirically, by systematically mapping reads to annotations with differing lengths of extensions, and choosing the lengths that yielded the greatest number of mapped reads.) A separate revised annotation of the S. neurona genome was generated using a list of poly(A) site clusters (PACs) produced by grouping individual sites that lie within 24 nts of each other (this parameter is the same as that used in refs. 34 and 35); in this annotation, PACs that mapped to annotated 3’-UTRs were identified and, where appropriate, used to extend to annotations. The extended genes were used for the gene expression analyses.

To evaluate global shifts in mRNA lengths, a weighted normalized mRNA length for genes impacted by APA was calculated for each condition (merozoites and schizonts). For this, the fractional usage of each site was multiplied by the fractional length of the corresponding mRNA, with 100% usage at a single site being set as 1.0, and the length of the longest mRNA being set as 1.0. Fractional mRNA lengths were calculated using the corrected S. neurona annotation described in the preceding paragraph. The fractional usages of all PACs for each gene were then summed, and the differences between these values in merozoites and schizonts calculated; positive values denote increased usage of distal poly(A) sites in merozoites. The results were binned into increments of 0.05, the fraction of all affected genes whose values fell within the different increments were tabulated and the results plotted as shown. More details may be found in S2 File.

Gene expression analysis

Gene expression was estimated by mapping individual PAT-Seq reads to the revised S. neurona genome annotation indicated in the preceding subsection using Bedtools as described elsewhere [37, 38]; the utility of 3’-end tags for quantifying gene expression is discussed in Lohman et al. [39]. The resulting mappings of raw reads were ported into CLC Genomics Workbench and the gene expression tools in the package used to quantify relative expression. The process and results are explained in detail in S3 File.

A bioinformatics study of the apicomplexan polyadenylation complex

While the polyadenylation complexes of animals and yeast have been extensively characterized, less is known about analogous complexes in most other eukaryotes. However, a conclusion drawn by the authors of a study of the genome of G. lamblia is provocative [40]; these authors noted that the genome of this organism seems to possess genes encoding only three of the subunits (CPSF73, CPSF30, and PAP) seen in the mammalian complex.

With this curious result in mind, a TBLASTN analysis of three apicomplexan genomes–P. falciparum, S. neurona, and T. gondii–was conducted, using the suite of human polyadenylation complex subunits as queries. For comparative purposes, the Arabidopsis genome was also searched. TBLASTN was chosen for this study to allow for the identification of possible orthologs in situations in which the community annotations have yet to identify the respective proteins. The only clear matches in any of the apicomplexan genomes were to CPSF73 and PAP (Fig 1A and 1B; panels A and B in S1 Fig). There were also suggestive, although less than definitive, matches for WDR33, PABN1, and CFIm25. The alignments of the closest apicomplexan matches to the latter three human proteins are shown in S1 Fig (panels C-E). To provide context, the sequences of respective orthologs from Arabidopsis, rice, and Physcomitrella patens (moss) were included in these alignments. As can be seen, the bulk of the sequence similarities between the apicomplexan and human proteins involve commonly-occurring sequence motifs (the WDR repeat in the case of WDR33, the RRM motif in the case of PABN1, and the NUDIX motif in the case of CFIm25).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. TBLASTN analysis of apicomplexan poly(A) complex subunits.

Results of a TBLASTN search was conducted using the collection of human polyadenylation complex subunits as queries against three apicomplexans genomes. A. thaliana was included to illustrate results obtained when orthologs exist. The y-axis on this plot is the negative log(10) of the e-value of the strongest match obtained in the BLAST search. The plot on the left shows the results obtained with CPSF subunits, and that on the right those obtained with other subunits. In these plots, values greater than 20 may be taken to denote possible orthologs. CPSF subunits are shown in panel A, and other subunits in panel B.

https://doi.org/10.1371/journal.pone.0203317.g001

Whereas clear orthologs of one of the two mammalian subunits that recognize the AAUAAA motif (WDR33) are not apparent in the three apicomplexan genomes studied here (panel C of S1 Fig), the situation is less clear for possible counterparts of the other protein, CPSF30 (or CPSF4) that also contacts AAUAAA. There is a protein in the three apicomplexan species that shows limited similarity to the human CPSF30 (panel F of S1 Fig). This similarity reflects the presence, in the apicomplexan and human proteins, of an array of three CCCH zinc finger motifs (the respective human domain is boxed in red in panel F of S1 Fig). The apicomplexan proteins lack two of the CCCH motifs seen in the human protein (as well as its yeast counterpart; these motifs are highlighted by the green boxes in panel F of S1 Fig). However, when selected plant CPSF30 proteins are included in the alignments, a more pronounced similarity is seen. This reflects several novel features of the plant proteins–the presence of the three (and not five) CCH motifs and also of a domain that bears resemblance to the so-called YTH domain (highlighted by the blue boxes in panel F of S1 Fig). The latter domain is associated with proteins that bind to RNAs in which particular adenosine bases have been modified by methylation at the N6 position. The CPSF30-YTH arrangement is a feature seen in all known examples of higher plant CPSF30 proteins [41]. The three apicomplexan proteins show significant similarity in the CCCH motifs, and also in the YTH domain (panel F of S1 Fig). This arrangement is confirmed by mapping of RNA-Seq reads S. neurona (panel G of S1 Fig). These results show that the parts of the putative S. neurona CPSF30 gene that encode the zinc finger arrays and the YTH domains may be connected by individual rRNA-Seq reads, providing support for the hypothesis that, as is the case in plants, these two domains are present in a single polypeptide.

There are two poly(A) complex subunits that can be identified in the apicomplexan genomes, CPSF73 and PAP. In animals, plants, and yeast, CPSF73 is highly conserved in both sequence and size. The apicomplexan orthologs are predicted to be considerably larger, with central region that is similar to the human CPSF73 and a N- and C- terminal domains that is absent in other eukaryotic counterparts (Fig 2A, panel A in S1 Fig). To confirm that the annotations of the S. neurona CPSF73 gene are accurate, S. neurona RNA-Seq reads were mapped to the region of the S. neurona genome that includes the predicted CPSF73 gene. The results confirm the accuracy of the annotation, supporting the existence of a single ca. 8000 nt mRNA (Fig 2B). No indications of splicing or other modifications that might interrupt the protein-coding region were apparent, again supporting the conclusion that the S. neurona CPSF73 protein is much larger than its other eukaryotic counterparts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Expression of the S. neurona CPSF73 and PAP genes.

A. Alignment of the S. neurona CPSF73 amino acid sequence (SN3_01500330_CPSF73) with the CPSF73 sequences from Arabidopsis (A. thaliana CPSF73) and humans (HsCPSF73). The extent pf conservation among the three sequences is depicted on the “Conservation” line, with red denoting high sequence conservation and dark blue low conservation. Analyses and depictions were done using CLC Genomics Workbench. B. Characterization of mRNAs encoding the S. neurona CPSF73 protein. RNA-Seq reads were mapped to the region of the S. neurona genome that includes the predicted CPSF73 gene using CLC Genomics Workbench. Mapping results were presented as genome browser tracks. Blue tags denote authentic paired-end reads, whereas green and red tags denote broken reads. C. Alignment of the predicted S. neurona PAP amino acid sequence (Sn SN3 PAP) with the PAP sequences from Arabidopsis (A. thaliana PAP) and humans (HsPAP), and with the corrected S. neurona polypeptide discussed in the text (Sn SN3 PAP revised). The extent pf conservation among the four sequences is depicted on the “Conservation” line, with red denoting high sequence conservation and dark blue low conservation. Analyses and depictions were done using CLC Genomics Workbench. The location of possible PAP-encoding sequences whose annotation is supported by RNA-Seq reads is shown by the arrows below the alignment. D. Characterization of mRNAs encoded by the region that includes the predicted S. neurona PAP gene. RNA-Seq reads were mapped to the region of the S. neurona genome that includes the predicted PAP gene using CLC Genomics Workbench. Mapping results were presented as genome browser tracks. A possible discontinuity in transcription in this region is highlighted with a red oval. The location of possible PAP-encoding sequences whose annotation is supported by RNA-Seq reads is shown by the arrows above the annotation.

https://doi.org/10.1371/journal.pone.0203317.g002

Apicomplexan genes encoding poly(A) polymerase (PAP) are also apparent in the BLAST analysis shown in Fig 1. As is the case with CPSF73, the apicomplexans PAP is predicted to be much larger than its plant and human counterparts (Fig 2C, panel B in S1 Fig). To confirm the accuracy of this annotation, S. neurona RNA-Seq reads were mapped to the region of the S. neurona genome that includes the predicted PAP gene (Fig 2D). In contrast to what is seen with the S. neurona CPSF73 gene, the transcriptomics analysis of the S. neurona PAP gene annotation suggest that the annotation may be inaccurate, but that the S. neurona PAP is indeed much larger than its other eukaryotic counterparts. Specifically, the annotation corresponding to the 5’ end of the putative PAP-encoding mRNA is not well supported by the RNA-Seq reads, in that there is little support for the splicing pattern in this annotation. RNA-Seq support for the structure of the remaining portion of the PAP-encoding gene is stronger; this portion includes all of the sequences that bear similarity to other PAPs (Sn SN3 PAP revised in Fig 2C). It is difficult to reconcile the two portions of the predicted gene with a single polypeptide, as the unspliced 5’ portion of the predicted mRNA would result in a very long 5’-untranslated region with numerous upstream AUG codons preceding any possible start codon for the PAP coding region. A possible discontinuity in the RNA-Seq reads (highlighted with the red oval in Fig 2D) raises the possibility that this locus may give rise to two mRNAs; the sequences whose processing pattern is supported by RNA-Seq reads (see the arrow over the annotation in Fig 2D) would encode a polypeptide with all of the PAP-related sequences. However, even given this correction of the S. neurona PAP annotation, the predicted protein would still be considerably larger than the Arabidopsis and human orthologs. Therefore, as is the case with CPSF73, the S. neurona PAP would seem to be a larger protein than its eukaryotic counterparts.

The apicomplexan species possess an A-rich polyadenylation signal

In mammals, two subunits of the poly(A) complex–WDR33 and CPSF30 –bind to the AAUAAA motif in the pre- mRNA [10, 11, 42]. As indicated above (Fig 1), no clear orthologs of WDR33 were identified in the three apicomplexan genomes. This observation raises questions regarding the nature of the poly(A) signal in apicomplexans. To examine this, a global study of poly(A) sites in three apicomplexans—S. neurona, N. caninum, and T. gondii—was conducted, with the purpose of characterizing sequences surrounding poly(A) sites in these organisms. For this, RNA was isolated from the three organisms at two distinct stages of growth—from extracellular parasites and from parasite-infected host cells. The isolated RNA was used to prepare so-called poly(A) tag libraries and these libraries were sequenced on the Illumina platform. The sequence data were analyzed using the methods described previously [36, 37] so as to identify polyadenylation sites. In these analyses, the data from extracellular and intracellular parasites were pooled to provide an overall view of possible poly(A) signals. The analysis focused on sites that mapped to annotated 3’-UTRs, or to reconstructed 3’-UTRs in the case of annotated genes lacking this feature (see Methods). Subsequently, the nucleotide composition surrounding these sites was determined.

For these three organisms, a distinctive and common nucleotide base composition surrounding poly(A) sites was observed (Fig 3). There was a decided region of A-richness around 20 nts 5’ (upstream) from the poly(A) site and a possible (G,C)A dinucleotide at the processing site itself. A motif analysis failed to identify highly-abundant motifs (such as AAUAAA) in this region, or elsewhere in the vicinity of these collections of poly(A) sites (not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Nucleotide composition surrounding poly(A) sites in three apicomplexan parasites.

Poly(A) sites were identified using high throughput PAT-Seq data and the nucleotide compositions determined as described [36, 43, 44]. Datapoints represent the relative fractional composition of each nucleotide at each position extending from 50 nts upstream to 50 nts downstream of the experimentally-determined or mock sites. The organism under study is indicated beneath each plot. Red arrows denote the prominent A-rich region situated upstream from the poly(A) site.

https://doi.org/10.1371/journal.pone.0203317.g003

An additional feature, a bias for A immediately following the poly(A) site, was also seen in all three organisms (Fig 3). This latter feature raises the possibility of internal priming by the oligo-dT primer used for the library preparation. However, two facts argue against this. The first is that tracts of 6 or more A were masked in the mapping process, thus eliminating from the study sequences possibly derived from internal priming at modest tracts of A. In addition, similar computational analyses of the regions surrounding oligo-A tracts in 3’-UTRs failed to show the distinctive A-rich upstream region or other base composition trends in (S2 Fig). This argues against extensive internal priming by RT, since this would be expected to yield results in Fig 3 similar to those shown in S2 Fig.

Alternative polyadenylation in S. neurona

Intracellular development of T. gondii and N. caninum occurs by endodyogeny, a relatively rapid process (6–8 hours) that produces two daughter zoites from a mother cell. In contrast, S. neurona propagates by endopolygeny, which progresses over approximately 72 hours and produces 64 daughter merozoites from a single mother cell schizont. This lengthy developmental process permits preparation of RNA samples from parasites that are distinctly extracellular (invasive form) or intracellular (propagative form). Accordingly, the poly(A) site profiles of S. neurona in the extracellular merozoite (M) and intracellular schizont (S) stages were determined and compared to identify poly(A) sites and associated genes whose usage varies significantly during asexual development. For this analysis, individual poly(A) sites situated near other sites were grouped into clusters as described elsewhere [36]. This process yielded 26,560 poly(A) site clusters (or PACs; S4 File). Of these, 24348 could be associated with at least one annotated gene; 23831 could be linked with one annotation, while 516 were associated with two or more adjacent annotations (S4 File). These PACS defined poly(A) sites in 4594 genes, or about 65% of all genes in the current S. neurona genome annotation.

At least two PACs were observed in 79% of all S. neurona genes (S3 Fig), revealing a possibility of extensive APA. Accordingly, an analysis of possible APA in the merozoites and schizonts in S. neurona was conducted. The results showed that 306 PACs exhibited highly-significant differences in usage in the two stages (S5 File). These 306 sites were associated with 204 gene annotations in the S. neurona genome. The number of sites showing less-significant but still detectable APA was 783; these sites affect 454 genes, which is more than 6% of annotated S. neurona genes [32]. This analysis suggests a significant extent of APA in S. neurona during asexual development. Interestingly, the set of 454 genes impacted by APA in S. neurona includes a significant number that encode membrane transporters and ribosomal proteins (Table 2, S5 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Genes encoding transporters and ribosomal proteins that are affected by APA.

https://doi.org/10.1371/journal.pone.0203317.t002

One important mechanism by which APA contributes to the regulation of gene expression in animals is via the alteration of the lengths mRNAs, usually involving differential use of sites situated within 3’-UTRs. To test whether this mechanism may apply to apicomplexans, an analysis of the lengths of mRNAs derived from genes impacted by APA in S. neurona was conducted. For this, a normalized aggregate mRNA length for each gene was calculated; this metric was arrived at by determining the contributions that each mRNA isoform makes to the total numbers of species encoded by the respective gene, and then calculating a normalized aggregate length with the longest isoform being set as 1. When the normalized aggregate lengths for the genes listed in S5 File were compared in the M and S stages, the majority of these genes showed a trend towards longer mRNAs in the M stage (Fig 4A). Instances wherein dramatic or complete switching between two isoforms were rare; rather, the changes reflected quantitative shifts in poly(A) site usage (examples are shown in Fig 4B), usually resulting in changes in the lengths of the 3’-UTR.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Changes in mRNA lengths for genes affected by APA.

A. Analysis of poly(A) site shifts for S. neurona gene that possess poly(A) sites whose usage changes significantly during development. For each gene, a weighted and normalized mRNA length was calculated as described in Methods, and the differences in this value between the merozoite and schizont stages calculated. The values were binned into discreet groups and the numbers of genes with changes falling into each bin plotted as shown. Positive values denote genes whose mRNA lengths are longer in merozoites than in schizonts. B. Browser tracks showing typical APA patterns for mRNAs encoding a membrane transporter (SN3_00700860) and a ribosomal protein (SN3_00900035). In both cases, the 5’-3’ orientations of these genes are right to left.

https://doi.org/10.1371/journal.pone.0203317.g004

Developmental changes in gene expression in S. neurona

The poly(A) tag data can provide an assessment of overall gene expression levels. Accordingly, changes in gene expression between the two stages of S. neurona were determined by quantifying and analyzing the total numbers of poly(A) tag reads that map to S. neurona genes. This analysis yielded a set of 576 genes whose expression differed significantly in merozoites or schizonts (S3 File). Included in this list were genes that are expected to reflect the different biological needs of the intracellular propagating stage relative to the extracellular invasive stage of S. neurona. Thus, genes whose expression was at least 5-fold higher in merozoites than schizonts included those that encode proteins associated with gene regulation (transcription factors, protein kinases, and second messenger binding or production; S3 File). Metabolic pathways involving phosphoinositide and sterol metabolism were also disproportionately up-regulated in the extracellular stage. Genes whose expression was at least 5-fold higher in the intracellular stage included those whose products are involved in DNA replication, glycolysis, gluconeogenesis, fermentation, and sphingolipid biosynthesis (S3 File). There was no appreciable overlap between the sets of genes showing differential expression and those affected by APA (Fig 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Comparison of genes affected by APA and developmentally-regulated expression.

Venn diagram showing the minimal overlap between genes affected by APA, those up-regulated in merozoites, and those up-regulated in schizonts. Gene lists are from S3 and S5 Files.

https://doi.org/10.1371/journal.pone.0203317.g005

The nature of the apicomplexan polyadenylation complex

The results of the bioinformatics study presented here raise interesting questions regarding the apicomplexan polyadenylation complex. Only two of the mammalian poly(A) complex subunits are clearly identifiable; these are the CPSF73 (or CPSF3) subunit and poly(A) polymerase. These two proteins arguably sit at the core of the polyadenylation reaction. CPSF73 is the endonuclease that processes the pre-mRNA [45, 46] and thus presents a substrate for PAP, the enzyme that adds the characteristic poly(A) tract to the 3’ end of the processed mRNA [2]. It is thus not surprising that these were easily identified in this analysis. In mammals and yeast, PAP is coordinated with the cleaved pre-mRNA by the CPSF subcomplex, and several protein-protein interactions between CPSF subunits and PAP have been reported; however, these do not include direct contacts between CPSF73 and PAP. Among other polyadenylation complex subunits that do engage in direct interactions with PAP are Fip1 [47–51], CPSF160 [52], CPSF100 [53, 54], and CFIm25 [49, 55]. Curiously, clear-cut orthologs of these proteins were not found in the three apicomplexan genomes studied here. PAP has been reported to interact with CPSF30 in some organisms [49]; the apicomplexan genomes studied here encode a possible CPSF30 ortholog, albeit one with somewhat limited sequence similarity with the human protein (panel F in S1 Fig). These considerations collectively raise interesting questions as to how the activities of the endonuclease (CPSF73) and PAP are coordinated in time and space. There may be functional counterparts to other PAP-interacting proteins that similarity searches cannot definitively identify. Alternatively, the CPSF-PAP complex in apicomplexans may be highly-reduced, such that only three proteins (CPSF73, PAP, and CPSF30) suffice for poly(A) signal recognition, pre-mRNA processing, and polyadenylation. These questions cannot be answered at the moment, but the disparate models represented have important implications for the functioning of the poly(A) complex, in the apicomplexans as well as other eukaryotes.

The matter of poly(A) signal recognition is of particular interest. The apicomplexan genomes studied in this report lack clear orthologs of one of the two subunits (WDR33) that associate with the motif AAUAAA in mammals, and they possess a possible ortholog of the other such subunit (CPSF30), albeit one with properties not seen in the mammalian counterpart. There does seem to be, in apicomplexans, a signal analogous to AAUAAA, and it reasonable to expect that the underlying mechanisms for recognition should be somewhat conserved. These considerations raise additional interesting questions. The apicomplexan genomes studied here possess genes that encode numerous WD repeat-containing proteins, any of which may be orthologs of WDR33. Alternatively, recognition of the A-rich PAS may not require a WDR33 ortholog, but instead could be performed by the CPSF30 ortholog. Plants are able to survive and thrive when CPSF30 is removed by mutation, providing precedent for the proposition that both WDR33 and CPSF30 are not required for mRNA 3’ end formation. This precedent makes somewhat more feasible the possibility that WDR33 orthologs may in fact be absent in apicomplexans. This possibility is further reinforced by the observation that clear orthologs of CPSF160, the scaffolding subunit upon which WDR33 and CPSF30 assemble in recognizing AAUAAA [42], are also not apparent in the apicomplexans; the possible absence of this subunit removes the means by which WDR33 and CPSF30 cooperate in the mammalian complex.

Perhaps the most provocative result of the bioinformatics study is the absence of identifiable orthologs of subunits of the Cleavage Stimulatory Factor (CstF) in the apicomplexans. This complex recognizes RNA sequences 3’ of the processing site [5], and plays important roles in processing and in alternative poly(A) site choice [7, 56, 57]. As is the case for WDR33, each of the three CstF subunits (CstF77, CstF64, and CstF50) possess motifs that are commonly found in proteins families. Also like WDR33, CstF50 possesses an array of WD repeats, and more than 30 WD repeat-containing proteins are found in Plasmodium, Toxoplasma, and Sarcocystis genome annotations. CstF64 possesses a distinctive RRM-type RNA-binding domain, and this domain is found in more than 40 proteins in different apicomplexan genome annotations. While the matches for any of these WDR and RRM proteins to CstF50 and CstF64 are rather tenuous, it remains possible that one or more may in fact serve analogous functions in apicomplexans. Surprisingly, however, even tenuous matches to CstF77 cannot be found in apicomplexan annotations, or when genomes are searched using TBLASTN. Given that CstF77 is the core scaffold upon which the other two subunits assemble to form a functional complex, this raises the intriguing possibility that mRNA 3’ end formation in apicomplexan species occurs without the participation of a CstF-like subcomplex.

Stage-specific APA and gene regulation

As is the case in most eukaryotes, the majority of S. neurona genes possess more than one poly(A) site (S3 Fig); the median number of sites per gene is 3, and average is slightly greater than 5 (S4 File). Thus, the propensity for APA is great in this organism. It is thus not surprising that more than 200 genes are impacted by APA when comparing extracellular merozoites with intracellular schizonts. Interestingly, the analyses suggested that APA has a disproportionate impact on genes that encode ribosomal proteins and membrane transporters (S5 File, Table 2). For the most part, these impacts are such that the respective mRNAs have somewhat longer 3’-UTRs in merozoites (such as shown in Fig 4B), with no appreciable difference in overall transcript levels. This suggests the differing properties of the 3’-UTRs that dominate in the two developmental stages may play arole in posttranscriptional or translational control in the expression of these genes. Certainly, higher levels of ribosomal proteins and membrane transporters are consistent with the need for enhanced protein translation and nutrient acquisition, respectively, during intracellular development by schizonts. While these findings are preliminary and speculative, a role for APA in controlling protein levels during development of apicomplexan parasites seems plausible and warrants future investigation.