High-throughput phenotyping using parallel sequencing of RNA interference targets in the African trypanosome

Abstract

African trypanosomes are major pathogens of humans and livestock and represent a model for studies of unusual protozoal biology. We describe a high-throughput phenotyping approach termed RNA interference (RNAi) target sequencing, or RIT-seq that, using Illumina sequencing, maps fitness-costs associated with RNAi. We scored the abundance of >90,000 integrated RNAi targets recovered from trypanosome libraries before and after induction of RNAi. Data are presented for 7435 protein coding sequences, >99% of a non-redundant set in the Trypanosoma brucei genome. Analysis of bloodstream and insect life-cycle stages and differentiated libraries revealed genome-scale knockdown profiles of growth and development, linking thousands of previously uncharacterized and "hypothetical" genes to essential functions. Genes underlying prominent features of trypanosome biology are highlighted, including the constitutive emphasis on post-transcriptional gene expression control, the importance of flagellar motility and glycolysis in the bloodstream, and of carboxylic acid metabolism and phosphorylation during differentiation from the bloodstream to the insect stage. The current data set also provides much needed genetic validation to identify new drug targets. RIT-seq represents a versatile new tool for genome-scale functional analyses and for the exploitation of genome sequence data.

Figure 1.

The RNAi library expression system. (A) The Sce* strain expresses the tetracycline repressor (TetR) and T7 phage RNA polymerase (RNAP) for the control of dsRNA expression, while inducible homing endonuclease (I-SceI) cleavage facilitates site-specific RNAi library integration at a locus that has been validated for reproducible and robust inducible expression. (B) RNAi plasmid library constructs replace the I-SceI gene and cleavage site. The RNAi vector consists of opposing T7 promoters regulated by Tet-operators. The RNAi target fragments serve as a template for the production of dsRNA and also provide unique sequence identifiers for each clonal population. (C) Tetracycline induction of dsRNA synthesis. These long (av. ∼600 bp) dsRNAs generate a pool of heterogeneous siRNAs that mediate sequence-specific destruction of the cognate mRNA.

Figure 2.

RIT-seq. (A) Schematic of the RNAi library and the growth conditions analyzed. The RNAi target fragment provides a unique identifier for each cell and its progeny, while dsRNA production is induced by tetracycline (Tet) addition (Fig. 1). Five different possible outcomes are illustrated. (B) Schematic of amplification, sequencing, and mapping of the RNAi target fragments. Only sequences containing a terminal RNAi-vector junction sequence, GCCTCGCGA, were mapped. The box shows a sample region with sequence mapping frequency viewed in Artemis (Carver et al. 2008); each peak represents a unique RNAi target fragment.

Figure 3.

Example plots in Artemis of the four different loss-of-fitness patterns illustrated in Figure 2; mRNA ablation is associated with a defect in all four induced conditions (blue), a more pronounced defect in bloodstream form cells (red) or procyclic cells (green), or a differentiation defect (purple). Cells carrying RNAi target fragments that negatively impact fitness through dsRNA expression and RNAi-mediated ablation are relatively depleted as the population expands and these changes are reported by the depth of sequence coverage relative to the uninduced control (no Tet). Each peak represents a unique RNAi target fragment. The genes shown encode a dynein heavy chain (Tb927.4.870), an intraflagellar transport protein (Tb927.10.14470), dihydrolipoamide acetyltransferase (Tb927.10.7570), and an uncharacterized zinc ion binding protein (Tb09.160.265).

Figure 4.

RIT-seq digital data analysis and validation. (A) The scatter-plots illustrate groups of genes that display different loss-of-fitness profiles in the BF^D3 (upper plot) and PF experiments (lower plot). Trend lines and slope scores are shown. (IFT) Intraflagellar transport (all genes identified to date); (PDH) pyruvate dehydrogenase (four-subunit complex); (TCP-1) T-complex protein 1 (eight-subunit chaperone ring complex); (n) number of genes (see Supplemental File 1B for details). (B) The distribution of Z-scores for 7435 genes and for each of the four experiments based on DEG-seq analysis; a Z-score of >3.3 represents a significant loss-of-fitness. (C) The plot shows mean Z-scores, based on DEG-seq analysis, for genes with the GO-term annotation “translation” and groups of characterized genes for each of the four experiments. Genes were selected based on a combination of TriTrypDB and PubMed searches (See Supplemental File 1B for details). (D) The four-set Venn diagram shows the distribution of genes based on DEG-seq analysis. The binary codes score the output from each of the four experiments (1–4) in the order indicated, with “0” representing a significant loss-of-fitness. For example “0-0-1-0” genes are associated with a loss-of-fitness in all three experiments involving RNAi induction during growth as bloodstream-form cells. The DIF experiment involved RNAi induction throughout growth as bloodstream forms, differentiation and growth as procyclic forms, so any gene required for growth in either life-cycle stage or for differentiation should register a loss-of-fitness in this experiment. The four cohorts illustrated in Figure 2A and Figure 3 are highlighted (bold colored text).

Figure 5.

Genetic profile for genes associated with loss-of-fitness in all four RNAi-induced experiments. (A) Genes in the 0-0-0-0 group are presented as a Z-score heat-map above the associated GO-term profile. All highly significant (P < 0.01) associations from the GO-slim set are shown. See Supplemental File 2 for full GO-term analysis. The term shown in blue is analyzed in more detail in B and discussed in the text. (B) Distribution of cohorts of genes potentially associated with post-transcriptional control of gene expression. GO terms were used to extract GeneID lists from TritrypDB. See Supplemental File 1B for ZC3H GeneIDs. Binary coding is as described in the legend to Figure 4D.

Figure 6.

Developmentally regulated genetic profile for genes associated with loss-of-fitness in all three experiments involving the bloodstream form (BF) and compared to the equivalent procyclic form (PF) profile. (A) Genes in the BF (0-0-1-0, red bars) and PF (1-1-0-0, gray bars) groups are presented as a Z-score heat-map above the associated GO-term profile. All highly significant (P < 0.01) associations from the GO-slim set are shown. See Supplemental File 2 for full GO-term analysis. Every term shown is notably under-represented in the PF group. Terms shown in blue are analyzed in more detail in B and discussed in the text. (B) Mean Z-scores for cohorts of genes associated with flagellar motility and the proteasome complex. (IFT) Intraflagellar transport. (C) Mean Z-scores for cohorts of genes associated with energy metabolism. (PDH) Pyruvate dehydrogenase. (B–C) See Supplemental File 1B for details of cohorts of genes analyzed.

Figure 7.

Genetic profile for genes associated with loss-of-fitness only in the differentiation experiment. (A) Genes in the 1-1-1-0 group are presented as a Z-score heat-map above the associated GO-term profile. All significant (P < 0.03) associations from the GO-slim set are shown. See Supplemental File 2 for full GO-term analysis. Terms shown in blue are analyzed in more detail in B or discussed in the text. (B) Examples of cohorts of genes displaying loss-of-fitness (upper panels) or gain-of-fitness (lower panel) associated with differentiation. The upper panels show the genes in the 1-1-1-0 group that are associated with the GO terms “carboxylic acid metabolic process” or “carboxylic acid transport.” (AAT) Amino acid transporter. The lower panel shows the genes with a low Z-score in the DIF experiment (increased reads) but otherwise neutral Z-scores.