Analysis of the spatial and temporal arrangement of transcripts over intergenic regions in the human malarial parasite Plasmodium falciparum

Plasmodium falciparum, the aetiological agent of the most severe form of human malaria, imposes a significant health and socioeconomic impact on those regions of the world where this parasite is endemic [1]. This malarial parasite has a lifecycle that alternates between a human host and mosquito vector, requiring multiple morphological and biological adaptations to successfully invade, colonise and divide within diverse cellular environments. Progression of parasites through this complex life cycle and the manifestation of virulence within the human host are both tightly linked to the temporal and spatial control of gene expression [2–9]. Over recent years we have garnered a greater appreciation of the interplay between the molecular mechanisms operating at the genetic and epigenetic levels in regulating developmentally-linked gene expression [4–6, 8]. These insights have been provided by global analyses of the temporal programme of steady-state transcript accumulation [10–12], mRNA stability and RNA polymerase II complex activity [13–16]. Yet despite these advances, and with access to a fully-annotated genome [17], we know relatively little regarding the fundamental organisation of the transcriptional unit in this important pathogen. This bottleneck arises from the extreme AT nucleotide bias in the intergenic regions (IGR). Here AT content typically exceeds 80-90%, imposing significant challenges for amplifying, cloning and sequencing of these regions as well as the application of bioinformatics tools (e.g. the unambiguous mapping of sequence reads from massive parallel sequencing of cDNA). Thus, we understand very little regarding the nature of the transcriptional unit outside of the open reading frame (ORF).

Determining the coordinates of the transcriptional start and stop sites is important. Sequences adjacent to transcriptional start sites likely comprise the cis-acting elements to which the regulatory and basal components of the RNA polymerase II complex bind. Moreover, these coordinates identify sequences in the 5′ and 3′ untranslated regions (UTR) of the transcript. These UTR similarly contain cis-acting sequences that direct translational efficiency, mRNA capping and stability. Knowing the number and position of transcription start sites in P. falciparum is potentially important as it may provide key clues to the different molecular mechanisms employed in the control of transcription. For example, is there a generally relaxed transcriptional activation process that relies on molecular mechanisms downstream to regulate temporal patterns of steady-state transcript accumulation? This model is certainly supported by recent reports of a global programme of temporal mRNA stability during intraerythrocytic development [14]. Or, does the parasite utilise a single predominant transcription start site that employs specific cis–trans interactions over a core promoter to drive temporal expression? This was not previously a favoured model given the apparent paucity of specific transcription factors in the parasite’s genome [18–20], but it has recently regained support following the identification and characterisation of an expanded family of novel specific transcription factors (ApiAP2) in apicomplexan parasites [21–26]. A combination of both models is likely at play – but resolving the issue of where these key transcriptional coordinates are located is essential.

Studies on the size and organization of IGR in fungal species, which share a similarly compact genome as P. falciparum, suggest that transcriptional and RNA processing cis-acting regulatory sequences leave a “footprint” on the IGR [27, 28]. IGR that contain divergent transcripts, i.e. the flanking open reading frames (ORF) are orientated in a head-to-head fashion (see Figure 1A), are larger than those IGR with convergent transcriptional units where the flanking ORF are organised tail-to-tail. These studies indicate that gene spacing is not random, but is instead organised to facilitate the spatial arrangement of transcriptional units, and also that 5′ UTR are larger than 3′ UTR. A provisional analysis of IGR spaces from the incomplete chromosome 3 of P. falciparum indicates the same gene spacing patterning is present [29]. However, to date, no studies have addressed the spatial and temporal organisation of transcripts over these IGR.

As indicated above, there is a critical lack of data concerning the P. falciparum transcriptional unit outside of the ORF. Expressed sequence tag (EST) data from 3′ rapid amplification of cDNA ends (3′ RACE) and RNA ligase mediated RACE (RLM-RACE) provide some coverage. For example, RLM-RACE provides transcription start data for 1465 ORF (c. 27% of total) and is available through the Full-Malaria database (http://fullmal.hgc.jp) [30, 31]. These data indicate that P. falciparum transcriptional start sites are generally located at multiple loci, often spread over several hundred basepairs, some 150-450 bp upstream of the ORF. In addition to these genomic approaches, there are also a number of single-gene studies that provide transcript size data from Northern blots (see Additional file 1 and Additional file 2). Whilst many of these studies do not report the physical mapping of transcriptional start and stop sites, they do generally indicate two features of the P. falciparum transcript that seem at odds with the available EST data. First, transcripts are typically much larger than the ORF, suggesting a significant fraction of a transcript is untranslated. Second, one or two major transcripts are most often observed, which would suggest either that only one or two major transcription start sites exist, or that if many transcription start sites are utilised then these are either very close together or else only one or two give rise to a major stable transcript. Assays of promoter structure that are complemented with physical mapping of the transcription start site suggest that transcripts initiate at one, or at two closely located, transcription start sites and that these extend between 400-1900 bp upstream of the ORF [5, 32–37]. Despite what appears to be a disparity between the size of UTR predicted from EST and Northern blot studies, no systematic comparison of these data has been carried out to date to explore this difference.

We describe here a study that explores the size and organisation of IGR in P. falciparum and correlates this with UTR data available from Northern blots and EST databases. Our findings suggest that P. falciparum transcripts have a large UTR which appears preferentially apportioned to the 5′ end of the ORF. As this would suggest that significant amounts of the IGR that flank ORF are included in transcripts, we explore how transcriptional units are spatially and temporally organised over these IGR. Further, by showing a similar IGR arrangement in other apicomplexan parasites important for human and animal health, we suggest that our findings may impact more widely in understanding the molecular control of transcription across this phylum.

This study set out to address a fundamental gap in our understanding of the P. falciparum transcriptional unit outside of the ORF. Specifically, we examined the size and apportionment of the UTR as well as the spatial and temporal organization of the transcriptional units within the IGR that flank these ORF. In terms of the size and apportionment of UTR, our data would indicate; (i) that UTR are long, typically some 800–1800 bases, (ii) that the size of the UTR is independent of the size of the coding sequence and (iii) that 70-80% of the UTR is preferentially apportioned 5′ of the ORF. This would indicate that transcriptional start and stop sites lay between 600-1350 bp and 200-450 bp either side of the ORF. Apart from lengthening our current understanding of the extent of the transcriptional landscape in P. falciparum, these more distal transcriptional coordinates have implications for our search and validation of regulatory cis-acting regions. In silico searches for sequence motifs enriched in the flanking regions of functionally related and/or cotranscribed genes typically use 1kbp of flanking sequence [64, 65]. Whilst this would seem suitable for searching downstream of an ORF, it is perhaps not sufficient to identify all potential 5′ positioned regulatory elements. That said, a ScanACE analysis of at least 2kbp of flanking sequence has provided an extensive catalogue of putative ApiAP2 transcription factor binding sites [22]. Testing of these putative sites will require functional analyses of promoter activity. Our data regarding the extent of UTR coverage, as well as the significant chance of transcript overlap, provides insights that may help guide selection of sites more likely to be trans-acting factor binding sites to be tested in these studies.

Of note was the discrepancy between the sizes of UTR predicted from Northern blot and EST data; with those predicted from EST data invariably being shorter. This discrepancy is unlikely to result from a selection bias in the cohort of 105 genes used in this study as the mean size of all 5′ UTR from the EST data for these genes (305 ± 182 bp) is very similar to that published for 1465 genes for which 5′ EST data is available (303 ± 155 bp) [31]. More likely, bias introduced into the EST data by; (i) reduced processivity of reverse transcriptase over AT rich sequences, (ii) partial RNAseH activity in early generation enzymes and (iii) the use of oligo(dT) for first strand cDNA synthesis in some EST datasets, are all at play. Northern blot data are similarly prone to systematic error as often these are “guestimates” based on the use of a limited set of size standards during electrophoretic size fractionation. We also recognise the limitations arising from analysis of 105 genes by Northern blot analysis (c. 2% of all genes). This study does, however, represent the most complete meta-analysis of Northern blot data in P. falciparum to date.

Assuming a range of UTR between 800 and 1800 bases would indicate that 40-90% of all IGR space in the relatively compact genome of P. falciparum is included in at least one transcript. Since it would appear likely that there is significant transcriptional unit overlap, the actual extent of this transcriptional landscape over the genome would be reduced, although our data would suggest it is still considerably more than previously predicted from the available RNAseq and EST coverage. Why these UTR are so large in P. falciparum is intriguing. The size of the UTR, in part, would require that it is long enough to contain the cis-regulatory elements necessary for RNA metabolism. Whilst we know relatively little about these, the high level of selective constraint throughout intergenic regions in P. falciparum provides evidence of an evolutionary “footprint” for these non-coding elements [66, 67]. Selective constraint is slightly, although not significantly, higher in proximal intergenic regions [66], i.e. regions more likely encoded in the UTR. In itself, however, the presence of these cis-regulatory elements doesn’t provide an explanation for the length of the UTR. The extreme AT bias of these IGR, however, may provide some explanation for this phenomenon. Like P. falciparum, transcripts in D. discoidium have long UTR with a median length of 724 bp for the 14124 5′ UTR sequences deposited in Dictybase. Both organisms share a highly biased AT-rich genome, effectively resulting in a binary nucleotide code within the IGR. This reduction in information content may necessarily lead to an expansion of sequences necessary to encode/utilise regulatory information, although this is perhaps an oversimplified interpretation of the observation. Critically, the genomes of both organisms show evidence of extensive overrepresentation of homopolymeric poly(dA).poly(dT) tracks [68, 69], and these tracts are more highly overrepresented within the IGR (own unpublished data). Thus, a requirement to maintain non-coding cis-regulatory elements embedded within flexible poly(dA).(dT) tracts that are prone to expansion could account for the increased length of UTR in P. falciparum. This proposal would suggest that some regions within the UTR are less essential than others - an observation borne out by our own (Hasenkamp S, Russell K, Ullah I, Horrocks P: Functional analysis of the 5′ untranslated region of the phosphoglutamase 2 transcript in Plasmodium falciparum, in press) and other studies that have determined the effect on reporter gene expression following deletion of UTR sequences [70–73]. Deletions of several hundred bases of the proximal 5′UTR appear to have a minimal effect on the absolute and temporal expression of the reporter gene, suggesting some plasticity in the size of the P. falciparum transcript.

Our analysis of IGR organisation in P. falciparum would indicate; (i) that the observed 1:1.8:1 relationship for IGR types A, B and C, respectively, is close to the predicted 1:2:1 ratio expected of independently-organised monocistronic transcriptional units and (ii) that IGR size directly correlates with the nature of the transcriptional activity that occur over them with a ratio of 2.86:2.05:1. Szafranski et al., using partial genome sequence from S. cerevisiae, D. discoidium, A. thaliana and P. falciparum, reported a provisional investigation of features of AT-rich organisms that may assist in genome annotation [29]. In doing so, they predicted that relatively compact genomes would share a 3:2:1 gene spacing rule for IGR types A, B and C. Their study couldn’t correlate this 3:2:1 rule to AT content due to the limited diversity of organisms investigated. Here we have extended this analysis of IGR to encompass the entire genomes of 13 organisms, exhibiting a range of AT content and genome density, albeit with a focus on other apicomplexan parasites. In this larger study, we confirm that IGR size does not correlate with AT content, whereas we do find, perhaps not unexpectedly, that IGR size does correlate with the overall genome density, with a close linear relationship (R² between 0.84-0.98) for genome densities between 2.3-4.6 Kb/ORF. This correlation, although weaker does extend out to the 9.1 Kb/ORF gene density found in T. gondii, although here the 3:2:1 gene spacing rule apparently collapses to an approximate 1.5:1.5:1 ratio. A novel finding in this study, however, was the differing spatial arrangement of IGR size within different chromosomal compartments in P. falciparum, where IGR lengths, irrespective of their type, are longer in subtelomeric regions. Multigene families that encode proteins likely to mediate interactions with the host environment are preferentially located in this compartment and are best exemplified by the var family that encodes the P. falciparum erythrocyte membrane protein (PfEMP1) [9, 41, 46]. PfEMP1 are exposed on the surface of infected erythrocytes where they mediate adhesion to host cell surface ligands and, through clonal variation of the PfEMP1 expressed, help to establish a chronic infection in the face of a human immune response mounted against infected erythrocytes. We would speculate that this immune response may act a balancing selection pressure to that operating in the chromosomal internal compartment to reduce gene density through reduction in IGR size [74]. Repetitive sequence elements within the longer IGR in subtelomeric regions may assist in the organisation of chromosome ends at the nuclear periphery, a necessary factor in the epigenetic regulation of clonal expression, or may promote recombination to drive the generation of antigenic diversity in these multigene families.

Taken together, our data provides a theoretical framework for the spatial and temporal organisation of transcripts over the IGR, data that are not available from current microarray, EST and RNAseq analyses. With the potential for the next generation of directional RNAseq data to extend cDNA coverage into the IGR, we propose here a series of testable hypotheses that result from our theoretical framework. Specifically, we would predict; (i) UTR are typically between 800 and 1800 bp in size, (ii) 70-80% of UTR are preferentially organised to the 5’ of the transcript, (iii) 40-90% of the IGR sequences are transcribed, resulting in 70-80% of the entire genome organised within a transcript, (iv) that whilst UTR do not temporally overlap, a significant proportion will spatially overlap and (v) that a small number (up to 200) of bidirectional promoters exist. In addition, our findings suggest that how we think about the transcriptional landscape across the P. falciparum genome should be revised to a view that is more dynamic in terms of direction, timing and extent of coverage of transcription over the genomic template. These insights should impact on how we design studies to define and characterise functional elements that govern processes such as developmentally-linked gene expression and monoallelic expression of virulence-linked multigene families. Finally, since we show the organisation of IGR in related apicomplexans appears to follow the same spatial rules, aspects of this work may translate more widely across this group of parasites important to human and veterinary health.