Quantitative insertion-site sequencing (QIseq) for high throughput phenotyping of transposon mutants

Abstract

Genetic screening using random transposon insertions has been a powerful tool for uncovering biology in prokaryotes, where whole-genome saturating screens have been performed in multiple organisms. In eukaryotes, such screens have proven more problematic, in part because of the lack of a sensitive and robust system for identifying transposon insertion sites. We here describe quantitative insertion-site sequencing, or QIseq, which uses custom library preparation and Illumina sequencing technology and is able to identify insertion sites from both the 5' and 3' ends of the transposon, providing an inbuilt level of validation. The approach was developed using piggyBac mutants in the human malaria parasite Plasmodium falciparum but should be applicable to many other eukaryotic genomes. QIseq proved accurate, confirming known sites in >100 mutants, and sensitive, identifying and monitoring sites over a >10,000-fold dynamic range of sequence counts. Applying QIseq to uncloned parasites shortly after transfections revealed multiple insertions in mixed populations and suggests that >4000 independent mutants could be generated from relatively modest scales of transfection, providing a clear pathway to genome-scale screens in P. falciparum QIseq was also used to monitor the growth of pools of previously cloned mutants and reproducibly differentiated between deleterious and neutral mutations in competitive growth. Among the mutants with fitness defects was a mutant with a piggyBac insertion immediately upstream of the kelch protein K13 gene associated with artemisinin resistance, implying mutants in this gene may have competitive fitness costs. QIseq has the potential to enable the scale-up of piggyBac-mediated genetics across multiple eukaryotic systems.

Figure 1.

Outline of QIseq process. (A) Library generation. (1) Randomly sheared genomic DNA containing piggyBac transposon (Tn) insertions is ligated to a custom Splinkerette (Spl) adapter (light blue). (2) Separate 5′- and 3′-Tn-end-specific libraries are generated by linear amplification from the transposon (orange) using 5′- or 3′-specific primers, followed by priming from the adapter. (3) A nested PCR is performed with primers specific to the Tn end and adapter end, containing a P5 (green) and P7 (red) sequence, respectively, to allow the product to bind to the Illumina flow cell for sequencing. (B) Sequencing protocol. After binding to the flow cell at the P7 end, a Tn-specific sequencing primer initiates base incorporation. Because all insertion sites will start with the same bases at the end of the Tn, no imaging takes place during the first cycles, with normal imaging initiated only at the genomic DNA sequence. After denaturation, Tn tag sequences are then read using the same sequencing primer, before index tags, which allows sequencing of multiple libraries on one flow cell, and are read using a Splinkerette (Spl)-specific sequencing primer. Finally, a reverse read is generated using a paired-end turnaround, hybridizing the P5 end to the flow cell and sequencing using a Spl-specific reverse oligo.

Figure 2.

Reproducibility and sensitivity of QIseq. Insertion sites identified during sequencing of multiple single and pooled piggyBac lines were compared across biological replicates and time points. (A) Correlation plot of insertion sites identified by 5′ and 3′ libraries in small pool of 41 piggyBac insertions. (B) Correlation plot of the same insertion sites identified in biological replicates. (C) QIseq sensitivity analysis of the gDNA of PB-3 mutant line was diluted as 1:10, 1:100, 1:1000, 1:10,000 and mixed with three different mutant lines as biological reps. QIseq identifying and monitoring sites over a >10,000-fold dynamic range of sequence counts. (D) Percentage of mutants identified in small pool growth cycles, over a >10,000-fold dynamic range of sequence counts in 18 cycles.

Figure 3.

Distribution and frequency of identified piggyBac insertion sites in the unknown mixed populations (MPs). Distribution of the 254 QIseq insertion sites of the 19 MPs (Supplemental Table S3) shown along the 14 chromosomes; different colors represent different pools. (Inset) Normalized QIseq insertion site: Pools MP1–MP5 were based on 31,000, MP6–MP10 on 62,000, MP11–MP15 on 124,000, and MP15–MP19 248,000 parasites, respectively. The bar plot shows the normalized amount of insertion sites per 31,000 parasites for each unknown MP.

Figure 4.

Using QIseq to measure growth of 41 piggyBac mutants within a single pool. As described in the Methods, 41 individual piggyBac mutants were pooled and grown over 24 cycles, with the relative abundance of each clone measured by QIseq counts every three cycles. (A) Growth of each piggyBac mutant growth measured by QIseq reads. (B) Relative abundance of each piggyBac mutant during each measured growth cycle, represented by percentage of reads. (C) Identifying neutral or advantageous mutations and deleterious mutations within the pool. Bars represent fold change in read number for each mutant over time; the top quartile (N = 10) represents the neutral mutation phenotypes (red) and the bottom quartile represents the deleterious or ‘loser’ mutation phenotypes (blue) in competitive growth conditions.

Figure 5.

Using QIseq to measure growth of 128 piggyBac mutants in a single pool. (A) Growth of each mutant within two biological replicates (rep1, rep2) of the same pool, grown over 12 cycles. (B) Permutation-based statistical test of reproducibility between replicates and between different screens. (C) The same mutants in different replications always had a significantly higher correlation compared with randomly reshuffled mutant pairs (Wilcoxon test P < 1 × 10⁻⁵). (D) Estimates of population growth by individual piggyBac mutants within the pool of clones grown together over 24 cycles. (E) Estimates of QIseq sampling of piggyBac mutant populations. While all piggyBac mutants continue to proliferate, the relative abundance of most slower growing mutants is predicted to decrease (“bottom quartile”) relative to the few faster growing parasites (“top quartile”). (F) Classification of phenotypes from the large pool of 128 mutants identified neutral or advantageous mutations (red; top quartile of growth, N = 32) and deleterious mutations (blue; bottom quartile).

Figure 6.

Mutant-associated functions in competitive growth assay for the top and bottom quartiles. The gene enrichment is performed with GO term representation by a hypergeometric test; P-value adjusted with the Benjamini-Hochberg method. The solid circle sizes represent the number of mutants in the group.