Genome organization is a major component of gene expression control in response to stress and during the cell division cycle in trypanosomes

The trypanosome genome is characterized by RNA polymerase II-driven polycistronic transcription of protein-coding genes. Ten to hundreds of genes are co-transcribed from a single promoter; thus, selective regulation of individual genes via initiation is impossible. However, selective responses to external stimuli occur and post-transcriptional mechanisms are thought to account for all temporal gene expression patterns. We show that genes encoding mRNAs that are differentially regulated during the heat-shock response are selectively positioned in polycistronic transcription units; downregulated genes are close to transcription initiation sites and upregulated genes are distant. We demonstrate that the position of a reporter gene within a transcription unit is sufficient to reproduce this effect. Analysis of gene ontology annotations reveals that positional bias is not restricted to stress–response genes and that there is a genome-wide organization based on proximity to transcription initiation sites. Furthermore, we show that the relative abundance of mRNAs at different time points in the cell division cycle is dependent on the location of the corresponding genes to transcription initiation sites. This work provides evidence that the genome in trypanosomes is organized to facilitate co-coordinated temporal control of gene expression in the absence of selective promoters.


Summary
The trypanosome genome is characterized by RNA polymerase II-driven polycistronic transcription of protein-coding genes. Ten to hundreds of genes are co-transcribed from a single promoter; thus, selective regulation of individual genes via initiation is impossible. However, selective responses to external stimuli occur and post-transcriptional mechanisms are thought to account for all temporal gene expression patterns. We show that genes encoding mRNAs that are differentially regulated during the heat-shock response are selectively positioned in polycistronic transcription units; downregulated genes are close to transcription initiation sites and upregulated genes are distant. We demonstrate that the position of a reporter gene within a transcription unit is sufficient to reproduce this effect. Analysis of gene ontology annotations reveals that positional bias is not restricted to stress -response genes and that there is a genome-wide organization based on proximity to transcription initiation sites. Furthermore, we show that the relative abundance of mRNAs at different time points in the cell division cycle is dependent on the location of the corresponding genes to transcription initiation sites. This work provides evidence that the genome in trypanosomes is organized to facilitate co-coordinated temporal control of gene expression in the absence of selective promoters.
causing globally important parasitic diseases of humans and livestock. In trypanosomatids, synthesis of mRNA by RNA polymerase II (RNAP II) occurs via polycistronic transcription [5]. Co-transcriptional processing to individual monocistronic mRNAs is mediated by trans-splicing of a 39-nucleotidecapped exon to the 5 0 end of all protein-coding genes. Linked endonucleolytic cleavage and polyadenylation of the upstream mRNA complete the maturation process. This mechanism of gene expression is reflected in the structure and organization of the genome, where protein-coding genes are densely packed in polycistronically transcribed tandem arrays containing tens to hundreds of genes with greater than 50 per cent of the nucleotide sequence of the array present in mature mRNAs [6][7][8]. The RNAP II promoters driving transcription of these polycistronic arrays have yet to be characterized mechanistically, and selective initiation of transcription by RNAP II of particular gene arrays has not been demonstrated. Hence, it is thought that transcription is constitutive and thus the majority of gene expression regulation is mediated posttranscriptionally. Polycistronic transcription and trans-splicing are not unique to trypanosomatids, and are fundamental to the biology of many branches of the eukaryotic tree of life. In addition to trypanosomatids, many diverse eukaryotes (including appendicularia, ascidians, cnidarians, dinoflagellates, nematodes, platyhelminthes and rotifers) partially or entirely rely on this form of transcription for expression of their protein-coding genes [9][10][11][12][13].
The RNAP II promoters for protein-coding genes in trypanosomes are not fully characterized, but the location of the RNAP II transcription initiation sites for the procyclic developmental form of Trypanosoma brucei have been determined by precise transcriptional mapping [8]. A total of 191 RNAP II initiation sites were identified for protein-coding gene arrays, 129 were found at the 5 0 end of the polycistronic gene arrays and 62 occurred within tandem gene arrays, indicating a more complicated pattern of transcription initiation than is apparent from analysis of the genome sequence alone. Despite a superficial similarity to bacterial operons, the identities of genes within transcription units in trypanosomes appear to lack functional clustering. There are a few exceptions, the most notable being the Trypanosoma brucei tubulin gene array on chromosome 1 that contains multiple repeats of the a-and b-tubulin genes [14,15]. However, it is unclear how this organization provides function to the cell, and in other trypanosomatids the a-and b-tubulin genes are in separate loci [14]. Some evidence has been provided that polycistrons can contain differentially expressed gene clusters [16]; however, caution should be exercised when analysing gene expression data from multi-locus high-copynumber gene families as the identity of the source locus cannot be unambiguously resolved. Despite a lack of obvious functional clustering, trypanosomatid genomes are highly syntenic [17]. On average, 70 per cent of the set of genes comprising each trypanosomatid genome share the same genomic context with other trypanosomatids [17,18]. This high degree of gene order conservation is astonishing given that the ancestors of trypanosomatids diverged hundreds of millions of years ago. A rationale for the strong conservation of gene order has yet to be identified.
In trypanosomatids, the absence of gene-specific promoters and the dependence on polycistronic transcription impacts on the ability of the cell to modulate gene expression in response to external stimuli. For example, on cellular stress in yeast and metazoa, there is an immediate response that operates through post-transcriptional mechanisms followed by a gene-specific transcriptional response. In the specific case of the heat-shock response, there is a rapid and selective inhibition of splicing [19,20]. A set of mRNAs, including those encoding heat-shock proteins (HSPs), escape the inhibition of splicing, and continue to be synthesized and exported. Other polyadenylated mRNAs are retained within the nucleus [21,22]. In addition, the half-life of some mRNAs, including HSP70, increases dramatically [23]. The subsequent transcriptional response is mediated by competition for HSP90 binding between heat-shock transcription factors and thermally sensitive proteins [24]. In trypanosomatids, the initial response to heat shock is similar. There is inhibition of splicing followed by a rapid decrease in levels of mRNA owing to increased turnover [25][26][27]. As in yeast and metazoa, a set of mRNAs, including those encoding HSPs, is excluded from this process and increases in relative abundance over the first hour of heat shock [27,28]. The lack of individual gene promoters means that a subsequent selective transcriptional response does not appear to be available and it has remained unclear how global patterns of gene expression are regulated in response to heat shock.
Here, an investigation of mechanisms compensating for the lack of a selective transcriptional response has led to the finding that there is a genome-wide functional organization of heatshock-responsive genes. Rapidly downregulated genes tend to be located proximal to the transcription initiation site and upregulated genes tend to be distal. Furthermore, we demonstrate that spatial positioning of a reporter gene within a transcription unit is sufficient to alter temporal regulation of the corresponding mRNA's behaviour during heat shock. Extension of this observation to all genes with ascribed annotations reveals that selective positioning of groups of genes is not limited to heat-shock-responsive genes but is a general phenomenon of genome organization in Trypanosoma brucei.

Heat-shock expression data and genome position analysis
The GenBank file for the Trypanosoma brucei genome (Tbrucei_-TriTrypDB-1.0.gff ) was downloaded from TriTrypDB [29]. The locations of defined transcription initiation sites were retrieved from Kolev et al. [8]. The position for each gene relative to its nearest transcription initiation site in the correct direction was calculated. In cases where there was no identified transcription initiation site in the correct direction between a given gene and the end of the available sequence data, this sequence end was assumed to be the location of a putative transcription initiation site. Distances in nucleotides to transcription initiation sites were calculated based on the midpoint of each open reading frame. Pseudogenes, variant surface glycoproteins, expression-site-associated genes and genes with multiple genomic locations such as 'retrotransposon hot spot protein' were removed from all calculations. Heat-shockinduced changes in mRNA abundance were obtained from a previous study [27]. Fold change in mRNA abundance following heat shock was calculated from these microarray data. Only genes that were twofold or more differentially regulated in response to heat shock were selected for further analysis.
rsob.royalsocietypublishing.org Open Biol 2: 120033 All calculations, genome distance measurements and statistical tests were performed using Perl scripts.

Cells and reagents
Trypanosoma brucei Lister 427 procyclic forms were grown in SDM-79. All genetic manipulations used standard techniques. Cells were grown without antibiotic selection and below a density of 1 Â 10 7 cells ml 21 for two passages before any experiment. Measurements of mRNA half-lives were performed on cultures with cell densities between 4 and 7 Â 10 6 cells ml 21 . For heat shock, 20 ml aliquots of cultures were placed in pre-warmed 30 ml glass centrifuge tubes in a 418C water bath; the cultures took the first 4-5 min of the time course to reach 418C. At selected time points, the tubes were removed from the water bath and cells pelleted by centrifugation at 3000g for 60 s. The pellet was resuspended in 1 ml serum-free medium; the cells were recovered by centrifugation in a microfuge for 15 s and resuspended in 50 ml of residual supernatant, and immediately frozen in an ethanol dry ice bath. In total, it took 3-4 min from water bath to freezing. RNA preparation and northern blotting were performed as previously described [27].

Gene ontology category analysis
To determine if particular groups of genes had biased locations within transcription units, the following analysis was performed. The complete list of genes with gene ontology (GO) term annotations was downloaded from TriTrypDB [29]. For each GO term category, the mean distance for the constituent group of genes to their respective transcription initiation sites was calculated. To determine whether these mean distances were higher or lower than expected if there was a random distribution of genes in the genome, a Monte Carlo resampling test was performed. For each GO term category containing 10 or more genes, the mean distance for the constituent genes to their nearest transcription initiation site was compared with the mean distance of randomly composed groups of genes of the same size. This process was repeated 10 000 times and the proportion of randomly selected groups that achieved a mean distance to transcription initiation sites of less than the GO term group was recorded.
To calculate an expected distribution for randomly composed GO term groups, the following procedure was performed. Each of the GO term categories containing 10 or more genes was randomly re-constituted from the set of genes with GO annotations. In cases where individual genes belonged to multiple GO categories, this relationship structure between GO categories was maintained by assigning the same randomly selected gene to all shared categories. The Monte Carlo distance test (described above) was then performed on this randomly resampled GO category dataset. An expected distribution was then calculated from 100 replicates of this randomization procedure.

Cell division cycle transcriptome analysis
The previously published cell-cycle-dependent transcriptome for early G1, late G1, S and G2/M phases of the cell division cycle of procyclic form Trypanosoma brucei were downloaded from Archer et al. [30]. The data were extracted and analysed in context of characterized transcription initiation sites, as above. A sliding window approach was taken to analyse these data. The window size was set to 20 kbp and was moved in 1 kbp steps in the direction of transcription away from the characterized transcription initiation sites. For each step, for each cell cycle stage, the mean mRNA abundance of all genes occurring with the 20 kb window (across all transcription units) was calculated. For each window position, the mean of all cell cycle stages was calculated and the log 2 ratio of the individual cell cycle stage expression level relative to the mean was then taken.

Genes differentially regulated in response to heat
shock are not randomly distributed within polycistronic transcription units A previous analysis of the heat-shock response in procyclic form Trypanosoma brucei identified 1058 mRNAs whose abundance changed in response to heat shock [27]. In the analysis presented here, the location of the heat-shock-responsive genes on each chromosome was determined. All 1058 mRNAs showing differential abundance in response to heat shock in the microarray experiment were selected. Three criteria were applied to the filter list. First, all mRNAs with a less than twofold response were removed to reduce the number of false positives arising from inaccuracies in the microarray data. Second, all mRNAs likely to be transcribed by RNA polymerase I (i.e. variant surface glycoprotein and expression site-associated genes) were removed. Third, mRNAs arising from dispersed multi-copy genes (i.e. GRESAG4 and 'retrotransposon hotspot protein') for which the microarray data cannot unambiguously distinguish the originating genes were also removed. The final list contained 211 mRNAs whose relative abundance decreased and 566 mRNAs whose relative abundance increased after heat shock (electronic supplementary material, file S1). Visual inspection of the distribution of the genes in this list relative to defined RNAPII transcription initiation sites [8] suggested that the genes corresponding to mRNAs whose relative abundance increased in response to heat shock were located further away from transcription initiation sites than those that decreased (figure 1; electronic supplementary material, file S1).
To quantify and test this observation, the fold change in abundance of individual mRNAs was plotted against the distance of the cognate genes from their transcription initiation sites (figure 2a). This revealed that there was a significant positive correlation between these values, with Spearman's rank correlation coefficient r ¼ 0.476, the probability of this being p , 0.0001 (Pearson product -moment correlation coefficient ¼ 0.501, p , 0.0001). To determine whether this correlation reflects a bias in the relative position of heatshock responsive genes, the location of these genes was interrogated in the context of the underlying distribution of all genes in the genome. More than 36 per cent of the genes whose mRNA abundance decreased following heat shock occur within 20 kbp of a transcription initiation site (figure 2b); this is more than twice the value expected if downregulated genes were distributed randomly in the rsob.royalsocietypublishing.org Open Biol 2: 120033 genome (17.8%, figure 2b). In contrast to this, genes whose mRNA abundance increased in response to heat shock are under-represented near transcription initiation sites, with only 3 per cent of responsive genes occurring in the same interval (figure 2c). Moreover, genes encoding mRNAs that increased after heat shock are over-represented at distances greater than 120 kb from the nearest transcription initiation site (figure 2c). The mean distances for each group of genes are significantly different (all p , 0.0001, determined by Monte Carlo resampling of the data). While the full complement of genes that are necessary to mediate the heat-shock response is not yet defined, the two verified heatshock-responsive HSPs [27] are both located near the ends of transcription units: both the tandem array of 10 HSP83 genes and the HSP70 (Tb11.01.3110) gene are located at larger-thanaverage distances from transcription start sites. This analysis provides evidence that there is a genome-wide functional positioning of genes within transcription units that contributes to the differential temporal response of mRNAs to stress, genes proximal to the initiation sites are downregulated following heat shock and genes distant to initiation sites are upregulated.

4.2.
Genome positioning is sufficient to mediate differential regulation in response to heat shock The analysis above suggested that the distance of a gene from a transcription initiation site is sufficient to mediate a change in relative abundance of its corresponding mRNA during heat shock. To test this directly, a reporter transgene (encoding the neomycin resistance gene) was inserted at one or other of two positions in a transcription unit on chromosome 4 ( figures 1 and 3a). The transcription unit on chromosome 4 from Tb927.4.2110 to Tb927.4.3190 was selected as it is long (approx. 294 kbp), it is clearly defined at each end by two inflection points in strand coding potential and it had no evidence for additional internal transcription initiation sites. The reporter constructs were designed to replace the sequence between two consecutive open reading frames (inter-ORF) with an a-to b-tubulin inter-ORF followed by the neomycinresistant gene followed by a b-to a-tubulin inter-ORF (figure 3a). Tubulin inter-ORFs were chosen as the abundance of tubulin mRNAs are representative of the behaviour of an average mRNA following heat shock [27]. The transgene was inserted either between Tb927.4.3140 and Tb927.4.3150 (the midpoint between these two ORFs is approximately 284 kbp from the initiation of transcription) or between Tb927.4.2120 and Tb927.4.2130 (in this case, the midpoint is approximately 8 kbp from transcription initiation site). In either location, the transgene results in the expression of an identical mRNA with a b-tubulin 5 0 UTR followed by the neomycin phosphotransferase ORF followed by a b-tubulin 3 0 UTR. Independent cloned cell lines were isolated after growth in G418; none had an obvious growth defect, and no deleterious effect of transgene expression was observed (data not shown). For both the distal and proximal position, the response of the transgene mRNA to heat shock in three independent clones was determined by northern blotting (figure 3b). At each of four time points, the mRNA level was quantified using a phosphorimager and normalized against total RNA (figure 3c). The rate of decrease was different for the distal and proximal transgene mRNAs. At all time points following heat shock, the level of the distal transgene mRNA was higher than that of the proximal (figure 3c). This was particularly apparent at 30 min, when the distal transgene mRNA had reduced to 82 per cent compared with 64 per cent from the proximal transgene. As a control, the steady-state level of the endogenous actin mRNA was determined in parallel for all time points. The decrease in abundance in all six cell lines was similar; there was a small difference between distal (14%) and proximal (18%) cell lines at 60 min, but this probably resulted from variations in estimating the low levels of actin mRNA at this time point. This analysis shows that the position of a gene from a transcription initiation site is sufficient to alter the relative abundance of the corresponding mRNA to heat shock. This position-dependent differential response provides a mechanism that contributes towards a differential abundance in the absence of a selective transcriptional response. mRNAs corresponding to genes distal to transcription initiation sites persist for longer following heat shock. Thus, as mRNAs corresponding to the proximal genes decrease more rapidly, this leads to a corresponding increase in the relative abundance of mRNAs corresponding to the distal genes.

Multiple categories of genes exhibit transcription unit positioning bias
The above analyses showed that the position of a gene relative to a transcription initiation site is important for rsob.royalsocietypublishing.org Open Biol 2: 120033 controlling the differential abundance of its mRNA in response to heat-shock-induced stress. To test whether other factors influence the location of genes within polycistronic units, two approaches were taken. First, other functional categories of genes were investigated to determine whether they showed positioning biases. GO annotations were used to group genes into categories. For each GO term category containing 10 or more genes, the mean and median distance of the genes to the nearest transcription initiation site was calculated (electronic supplementary material, file S2). The mean distance of this group was then compared with the mean distance of 10 000 randomly composed groups containing the same number of genes. The proportion of randomly selected groups that achieved a mean distance to transcription initiation sites of more than the GO term group was recorded (figure 4a; electronic supplementary material, file S2). For example, if 71 per cent of the randomly generated groups were further away from transcription start sites than the GO term group, then the GO term group would obtain a score of 0.71. This analysis showed that there are some GO term categories whose constituent genes are on average closer to transcription initiation sites than expected if genes were distributed randomly in the genome (figure 4b and  table 1). Similarly, there are some GO term categories whose constituent genes are on average more distant from transcription start sites than expected if genes were distributed randomly in the genome (figure 4b and table 1).
To control for discrepancies in gene density within transcription units and for differential transcription unit length, a further analysis was performed where a set of false GO term categories was reconstituted from randomly selected genes, each containing the same number of genes as the real GO term categories (see §3). The same distance-based analysis was completed as before and the procedure was repeated 100 times (figure 4b, red-shaded box). This shows that if genes were randomly distributed in the genome, an even distribution of GO term groups across transcription units would be observed. For the 16 GO term categories that achieved a score of greater than or equal to 0.99 (i.e. 1% chance that randomly selected genes would obtain a mean distance less or equal to that observed for the GO category), four categories are concerned with translation (table 1). Interestingly, this set does not include genes encoding proteins involved in translation elongation and initiation, but rather the structural components of the ribosome (electronic supplementary material, file S2). In addition to the translation components, genes encoding components of the cytoskeleton and flagellum are also highly enriched proximal to transcription initiation sites (table 1).
Of the three GO term categories that achieved a score of 0.01 or less (i.e. 1% chance that randomly selected genes would obtain a mean distance greater or equal to that observed), two are concerned with transcription (table 2). Interestingly, within the 0.95 limit, all of the GO categories concerned with transcription were present. Hence, proteins of the transcription machinery are enriched at large distances from transcription initiation sites. The mean behaviour of these groups of genes in response to heat shock displays a highly significant positive correlation (Pearson r ¼ 0.312, p , 0.00001, Spearman r ¼ 0.314, p , 0.00001; electronic supplementary material, file S2), such that those mRNAs corresponding genes more distant from transcription initiation sites increase in abundance while those close to transcription initiation sites decrease. While the correlation is highly significant, it is low; therefore the requirements to respond to stress does not account for all the positioning biases of genes within transcription units observed here. Hence, this analysis suggests that differential positioning  Figure 2. (a) Plot of fold change in mRNA abundance after heat-shock against distance of corresponding gene from nearest correct orientation transcription initiation site for mRNAs that were greater than twofold differentially regulated. (b) Histogram of proportion of genes at different distances from transcription initiation sites. Black bars indicate distribution of all genes in the genome. Grey bars indicate the distribution of genes whose mRNA abundance decreases by twofold or more on heat shock. (c) Histogram of proportion of genes at different distances relative to transcription initiation sites. Black bars indicate distribution of all genes in the genome. Grey bars indicate the distribution of genes whose mRNA abundance increases by twofold or more on heat shock. rsob.royalsocietypublishing.org Open Biol 2: 120033 with respect to transcription initiation sites is also driven by other factors.

mRNA abundance during the cell division cycle is dependent on position relative to transcription initiation sites
The second approach was to analyse other expression datasets to determine whether mRNA abundance was related to gene position relative to transcription initiation site. The most marked effects were obtained from an analysis of mRNA abundance during the cell division cycle from a previous study [30]. Analysis of this data showed that mRNA abundance is dependent on position of the corresponding gene relative to the transcription initiation site at several points in the cell division cycle (figure 4c). In early G1, transcripts corresponding to genes positioned between 25 and 60 kbp from transcription initiation sites are relatively more abundant than during the remainder of the cell cycle (figure 4c). In late G1, relative mRNA abundance has a clear linear relationship with distance of the corresponding genes from transcription initiation sites ( figure 4c). Here, the more distal a gene is from a transcription initiation site, the more abundant its mRNA will be in late G1 relative to other phases. In S-phase, mRNA abundance decreases with distance of the corresponding gene from the transcription initiation site. In G2/M, there is an increase in distal gene mRNA abundance (figure 4c). Taken together, this analysis shows that gene position has a pronounced effect on the relative abundance of mRNAs at different time points in the cell division cycle.

Discussion
The main findings in this paper are as follows. (i) Genes encoding mRNAs that increase during heat shock are not randomly distributed within transcription units, but tend to be located distal to the transcription initiation site. (ii) Genes encoding mRNAs that are downregulated rapidly on heat shock tend to be located close to a transcription initiation site. (iii) The location of a gene within a transcription unit is sufficient to modify the behaviour of the corresponding mRNA in response to heat shock. (iv) Positional bias is not limited to the heat-shock-responsive genes but rather multiple categories of genes display positional bias relative to transcription initiation sites. (v) Relative mRNA abundance in the different phases of the cell division cycle is related to the distance of the corresponding gene to transcription initiation sites. Taken together, these observations provide the first demonstration of multiple competing rules for gene location within rsob.royalsocietypublishing.org Open Biol 2: 120033 polycistronic transcription units. They also show that spatial positioning contributes significantly to the temporal expression of genes and thus provide the first evidence for functional organization of the genome of Trypanosoma brucei. The heat-shock response was initially investigated as the mRNA dynamics have been well documented. On heat shock, RNAP II transcription initiation is reduced and the half-life of many mRNAs is reduced [25,27]. In addition, there is a selective inhibition of the maturation of many mRNAs, but not HSP70 or HSP83 [28,31]. The effect of these changes is to reduce the total mRNA pool by 50 per cent in 1 h [27]. In this context, the data presented here are consistent with a model where, during heat shock, the initiation of transcription is reduced or stops, but elongation continues. There are no direct measurements of RNAP II transcription rates in trypanosomes but it is unlikely to differ greatly from the 4.3 kb min 21 determined in mammalian cells [32]. Therefore, as a cell enters heat shock, the reduction in initiation would affect genes proximal to the transcription initiation sites within a few minutes. However, genes distal to the initiation site would continue to be transcribed for longer as RNAP II complexes that had initiated before heat shock complete their transcription cycle. In the case of a gene approximately 250 kbp from a transcription initiation site, the journey time for RNAP II would be approximately 60 min and thus transcription of genes at these distances would continue for the first hour of heat shock. Any increased pausing of RNAP II would extend this further, thus providing a spatial mechanism to achieve specific temporal regulation.
By extending the analysis of position and function to include all genes with annotated GO terms, we have shown that there are specific biases in the location of genes involved in multiple different processes. Genes involved in translation, the cytoskeleton and the cell cycle are located proximal to transcription initiation sites, whereas genes involved in transcription and RNA processing are located distal to transcription initiation sites. We propose that the distal positioning of these genes may play a role in global regulation of transcription.
We have shown that relative transcript abundance during the cell division cycle is related to the position of the corresponding gene relative to transcription initiation sites. Hence, correct spatial positioning is critical for correct temporal expression in the cell division cycle. As genome position is not the sole factor controlling gene expression level, there are a number of possible reasons for the deviations from the linear responses observed in this analysis. One possibility is that they represent cell-cycle-stage-specific transcription initiation sites that have yet to be described. It will be interesting to see whether mapping the transcription start sites in a cell-cycle-dependent manner will facilitate the discovery of additional position-dependent effects. It is likely that the spatial effects we have observed here will play an important role in the regulation of gene expression in other groups of eukaryotes that use polycistronic transcription of tandemarranged genes. It will also be interesting to see whether changes in the composition of the active polymerase complex [33] play a contributing role in these distance-dependent effects.
Is there evidence for a similar functional organization in the genomes of related species? A direct comparison with the genome of Leishmania major was not possible as the transcription initiation sites have not been systematically mapped. In T. brucei, 32 per cent (62/191) of transcription calculate mean distance of group to initiation sites generate random group of same size as GO group calculate mean distance of randomly generated group to initiation sites compare mean distances between the GO group and the randomly generated group count how many times the randomly generated group achieved a mean distance larger than that of the GO group rsob.royalsocietypublishing.org Open Biol 2: 120033 initiation sites occur without an inversion of stands with protein-coding potential, and this invalidates an analysis based on locating transcription initiation sites at strand inversion points. It will be interesting to see whether the functional organization is conserved once the transcription initiation sites are mapped in other kinetoplastids. Moreover, it will be interesting to see whether the same categories of genes display the same patterns of location.
In eukaryotes, temporal patterns of gene expression usually result from regulated initiation of transcription. In trypanosomes, this mechanism is absent and post-transcriptional mechanisms are thought to account for the majority of the regulation of gene expression. The genome-wide spatial organization of genes described here uncovers a new layer of expression-level control and facilitates an alternative mechanism to achieve temporal regulation of Table 1. Gene ontology (GO) groups whose constituent genes are significantly enriched proximal to transcription initiation sites. Trials column shows the proportion of 10 000 trials that achieved a mean distance greater than the GO group. Grey shading indicates those groups for which less than 1 per cent of randomly selected groups achieved a smaller mean distance.
gene ontolody ID GO term number of members mean distance (bp) trials expression in the absence of specific initiation. Moreover, it provides specific evidence that a temporal programme of gene expression regulation has been hard-wired into the genome organization. Hence, a major component of temporal gene expression regulation in trypanosomes is achieved through spatial organization.