Integrative analysis of the Trypanosoma brucei gene expression cascade predicts differential regulation of mRNA processing and unusual control of ribosomal protein expression

Abstract

Background: Trypanosoma brucei is a unicellular parasite which multiplies in mammals (bloodstream form) and Tsetse flies (procyclic form). Trypanosome RNA polymerase II transcription is polycistronic, individual mRNAs being excised by trans splicing and polyadenylation. We previously made detailed measurements of mRNA half-lives in bloodstream and procyclic forms, and developed a mathematical model of gene expression for bloodstream forms. At the whole transcriptome level, many bloodstream-form mRNAs were less abundant than was predicted by the model.

Results: We refined the published mathematical model and extended it to the procyclic form. We used the model, together with known mRNA half-lives, to predict the abundances of individual mRNAs, assuming rapid, unregulated mRNA processing; then we compared the results with measured mRNA abundances. Remarkably, the abundances of most mRNAs in procyclic forms are predicted quite well by the model, being largely explained by variations in mRNA decay rates and length. In bloodstream forms substantially more mRNAs are less abundant than predicted. We list mRNAs that are likely to show particularly slow or inefficient processing, either in both forms or with developmental regulation. We also measured ribosome occupancies of all mRNAs in trypanosomes grown in the same conditions as were used to measure mRNA turnover. In procyclic forms there was a weak positive correlation between ribosome density and mRNA half-life, suggesting cross-talk between translation and mRNA decay; ribosome density was related to the proportion of the mRNA on polysomes, indicating control of translation initiation. Ribosomal protein mRNAs in procyclics appeared to be exceptionally rapidly processed but poorly translated.

Conclusions: Levels of mRNAs in procyclic form trypanosomes are determined mainly by length and mRNA decay, with some control of precursor processing. In bloodstream forms variations in nuclear events play a larger role in transcriptome regulation, suggesting aquisition of new control mechanisms during adaptation to mammalian parasitism.

Fig. 1

Gene expression from DNA to mRNA in Trypanosoma brucei. The upper panel is a schematic time-lapse image of a polymerase II complex progressing along a chromosome. Coding regions are in dark colours and 3′ and 5′ untranslated regions are in lighter colours. The capped spliced leader is in orange. Kinetic constants for the different processes are indicated and the formulae that comprise the model are shown below the figure 5′ trans splicing (rate constant k₁) and 3′ polyadenylation (rate constant k₂) compete with nuclear degradation of the precursor (rate constant k₃) and the 5′ spliced intermediate (rate constant k₄). Assuming that the two processes are coupled, the rate constant for 3′ polyadenylation (k₂) of mRNA A is expected to equal the rate constant for 5′ trans splicing (k₁) of mRNA B, and the same applies for mRNA and mRNA C (k₂ of mRNA B = k₁ for mRNA C). Based on the observation that that long mRNAs had unexpectedly low abundances, we added a factor (α) that incorporates length-dependent precursor degradation into the model [16]. In the analysis described in this paper, we tested alternative versions of this and used a different factor for PC trypanosomes (see text and Table 1). Finally, there is degradation of the mature mRNA (rate constant k₅). Values for k₅ are based on transcriptome-wide decay measurements. In addition, the growth rate of the cells is included via the specific growth rate μ. The growth rate affects the abundance of every RNA species, as during growth the pre-existing RNA species get diluted

Fig. 2

Correlation between predicted and measured mRNA amounts. The numbers of mRNAs per cell were predicted for the subset of mRNAs with reliable measurements of gene copy number, half-life, and abundance. The predictions were then compared with measured values of mRNA abundance for (a) procyclic forms (model PC-A) and (b) bloodstream forms (model BS-D). Each diamond represents a different open reading frame. Correlation coefficients are for log₂-transformed values. The pink dotted lines are for y = 2x, y = x and 2y = x. Results for ribosomal protein genes are in cyan

Fig. 3

Relationship between prediction and measurement for bloodstream and procyclic forms. For each unique gene considered, the amount of mRNA that was measured was divided by the predicted amount, and the log₂ of the values for both stages plotted. BS - bloodstream form, model BS-D; PC - procyclic form, model PC-A. A value of 0 indicates that the prediction was perfect; this could however also be true if transcription was faster, and processing slower, than the values that were set in the model. Measures above 0 indicate either that a processing half-time is shorter than 1.7 min, or that the steady-state transcription rate is greater than the value that was used for the model. values below 0 indicate that the splicing or polyadenylation half-time is longer than 1.7 min. Spots above the line “PC = 2xBS” are mRNAs for which the measured/prediction results was at least 2-fold higher in procyclic than in bloodstream forms; those below the “BS = 2xPC” line have better processing in bloodstream forms than in procyclics

Fig. 4

Ribosome profiling results. a Ribosomes per coding sequence (CDS) per cell, procyclic forms (PC). Our previously published results (labelled “Jensen”) [20] are compared with the set reported here (labelled “Silicotryp”). b As in (a) but for bloodstream forms (BS). c Developmental regulation: the numbers of ribosomes per kilobase are compared from bloodstream and procyclic forms. The grey-shaded area indicates impossible densities of less than 30 nt per ribosome. d The relationship between mRNA half-life and ribosome density for procyclic forms. Only mRNAs with reliable measurements of abundance and half-life were considered, and impossible ribosome densities were excluded

Fig. 5

Effect of methods on the measurement of mRNA abundance. a Poly(A) + mRNA from bloodstream forms, fragmented to 30–70 nt before sequencing; comparison of Silicotryp results vs Jensen results [20]. b Poly(A) + RNA from bloodstream forms, 30–70 nt fragments compared with published results for rRNA-depleted (ribo-minus) RNA [16], prepared from bloodstream forms cultured under identical conditions

Fig. 6

Ribosome density measurements reflect variable loading of mRNA on polysomes. The number of ribosomes per kb of total mRNA (note log scale) is plotted against the proportion of the same mRNA that is in the polysomal fraction. Results for specific functional classes are highlighted in different colours