Pervasive sequence-level variation in the transcriptome of Plasmodium falciparum

Abstract

Single-nucleotide variations (SNVs) in RNA, arising from co- and post-transcriptional phenomena including transcription errors and RNA-editing, are well studied in a range of organisms. In the malaria parasite Plasmodium falciparum, stage-specific and non-specific gene-expression variations accompany the parasite's array of developmental and morphological phenotypes over the course of its complex life cycle. However, the extent, rate and effect of sequence-level variation in the parasite's transcriptome are unknown. Here, we report the presence of pervasive, non-specific SNVs in the P. falciparum transcriptome. SNV rates for a gene were correlated to gene length (r[Formula: see text]0.65-0.7) but not to the AT-content of that gene. Global SNV rates for the P. falciparum lines we used, and for publicly available P. vivax and P. falciparum clinical isolate datasets, were of the order of 10^-3 per base, ∼10× higher than rates we calculated for bacterial datasets. These variations may reflect an intrinsic transcriptional error rate in the parasite, and RNA editing may be responsible for a subset of them. This seemingly characteristic property of the parasite may have implications for clinical outcomes and the basic biology and evolution of P. falciparum and parasite biology more broadly. We anticipate that our study will prompt further investigations into the exact sources, consequences and possible adaptive roles of these SNVs.

Figure 1.

A global view of single-nucleotide variations in Plasmodium falciparum (data shown for a representative sample, MRA_1236 replicate 1). Plot showing the pervasive nature of SNVs. From outward: Track 1: SNV positions represented as vertical lines; Track 2: Ideogram of chromosomes, proportional to chromosome lengths; Track 3: Heatmap of RNAseq coverage at recorded positions, converted to log₁₀ scale for visualization; Track 4: Area plot of frequency of variation at recorded positions. Histogram showing distribution of SNVs on an averaged gene body.

Figure 2.

SNV types and effects. (A) Base changes (percentage of total). (B) Base shift patterns (%X > denotes the proportion of SNVs where base X changed to another base; %>X denotes the proportion of SNVs where the reference base was X). The distribution of predicted functional effects: (D) representative heatmap of the spectrum of amino acid changes. (E) Representative heatmap of the abundance of the most common dinucleotide patterns flanking each focal (original/reference) base. (A–C) Error bars represent 95% confidence intervals; number of replicates = 2 (3D7_CTRL, 3D7_ART, 3D7_TEMP, MRA_1241), 3 (MRA_1236, MRA_1240). (D) and (E) show data from sample MRA_1236 replicate 1.

Figure 3.

SNV rates in MRA and 3D7 Plasmodium falciparum lines. *** denotes significance at P < 0.001 (two-tailed) and *denotes P < 0.05 as calculated using a Dunett's test with PF3D7_Ctrl as the control group. Error bars represent 95% confidence intervals; number of replicates = 2 (3D7_CTRL, 3D7_ART, 3D7_TEMP, MRA_1241), 3 (MRA_1236, MRA_1240).

Figure 4.

SNV rates of MR4 and 3D7 Plasmodium falciparum lines as compared with those for other samples and species. Error bars represent 95% confidence intervals; number of replicates = 2 (3D7_CTRL, 3D7_ART, 3D7_TEMP, MRA_1241, P. vivax Mixed Culture, P. vivax Hypnozoites), 3 (MRA_1236, MRA_1240, PF_Mali Isolate, E. coli, B. subtilis), 4 (P. vivax Blood Stages).