A CRISPR platform for targeted in vivo screens identifies Toxoplasma gondii virulence factors in mice

Abstract

Genome-wide CRISPR screening is a powerful tool to identify genes required under selective conditions. However, the inherent scale of genome-wide libraries can limit their application in experimental settings where cell numbers are restricted, such as in vivo infections or single cell analysis. The use of small scale CRISPR libraries targeting gene subsets circumvents this problem. Here we develop a method for rapid generation of custom guide RNA (gRNA) libraries using arrayed single-stranded oligonucleotides for reproducible pooled cloning of CRISPR/Cas9 libraries. We use this system to generate mutant pools of different sizes in the protozoan parasite Toxoplasma gondi and describe optimised analysis methods for small scale libraries. An in vivo genetic screen in the murine host identifies novel and known virulence factors and we confirm results using cloned knock-out parasites. Our study also reveals a potential trans-rescue of individual knock-out parasites in pools of mutants compared to homogenous knock-out lines of the key virulence factor MYR1.

Fig. 1

Single-step CRISPR knockout pools from arrayed single-stranded oligos. a Schematic of tailored CRISPR screening method. gRNA encoding oligonucleotides are arrayed in multiwell plates and the gRNA of interest are selected for pooled Gibson cloning into pCAS9-T2A-HXGPRT. The vector library is transfected into parasites to generate a pool of mutants for testing in vitro and in vivo. b Schematic of pCAS9-T2A-HXGPRT. Regions of homology used for Gibson cloning of oligonucleotides are shown in purple. c Gene disruption using integrated pCas9-T2A-HXGPRT. RHΔhxgprt and PruΔhxgprt were transfected with vectors targeting UPRT or a control gene, GRA29. Transfectants were grown under M/X selection for 6 days before performing a plaque assay under 5-fluorodeoxyuridine (FUDR) selection to test for UPRT disruption. The percentage of UPRT knockout (KO) was calculated by comparing plaques numbers in the presence and absence of FUDR. The mean is shown with SD error bars. n = 2 biologically independent experiments. d Parasite survival is reduced after transfection of plasmid pools. RHΔhxgprt and PruΔhxgprt were transfected with 10 μg pCAS9-T2A-HXGPRT targeting UPRT or a mixed pool of 1000 gRNA. Parasites were grown in the presence of M/X and the resultant plaques counted after 7 days. The percentage of parasites surviving compared to parasites seeded was calculated in a plaque assay. The mean is shown with SD error bars. n = 3 biologically independent experiments. See also Supplementary Fig. 2. Source data are provided as a Source Data file

Fig. 2

Variable pool sizes produce reproducible CRISPR selection results. PruΔhxgprt were transfected with pCAS9-T2A-HXGPRT carrying 200, 800 or 3200 gRNA in triplicate transfections. The log 2 fold change (lfc) of the gRNA presence after 6 days growth in M/X compared to the DNA input is shown. Normalisation of sequencing counts using either a DESeq2 or b total read count for the 200, 800 and 3200 libraries and the published Sidik et al. dataset. Density plots show improved reproducibility between replicates for small libraries using DESeq2. c Overlay of 200, 800 and 3200 library gene lfcs with (solid line) and without (dotted line) centering of the 3200 library. d Ranked-order plots showing gene lfcs for 200, 800 and 3200 gRNA libraries. Essential and non-essential control genes are marked in red and blue, respectively. Error bars represent SEM. n = 3 technical replicates. e, f Correlation plots with Pearson’s correlation coefficient, r, of gene lfcs show that phenotypes are reproducible e across library sizes and f correlate well with published phenotype scores. Essential and non-essential control genes are marked in red and blue, respectively. See also Supplementary Fig. 3 and Source data in Supplementary Data 2

Fig. 3

In vivo CRISPR screen identifies known and novel virulence factors. a Schematic of experimental design. gRNA were picked in triplicate, combined and used in triplicate Gibson cloning reactions. Vector pools were combined and transfected into PruΔhxgprt in five replicates. After 6 days under M/X selection, subsets were taken for sequencing and remaining parasites were combined to generate the mouse inoculum. Eight mice were infected intraperitoneally (i.p.) with 200,000 parasites and the parasites retrieved from the peritoneum after 5 days. Orange stars indicate samples analysed by sequencing. b, c Rank-ordered plot of gene lfcs in vitro (b) and in vivo (c). Essential and non-essential control genes for in vitro and in vivo growth marked in red and blue, respectively. Translocon components MYR1 and MYR3 marked in orange. Error bars represent SEM. n = 5 technical replicates. d Discordance plot showing gene lfcs in vitro and in vivo. Colour indicates strength of discordance score. Error bars show 95% confidence range. Examples of known and newly identified virulence factors labelled with red and black text, respectively. See also Supplementary Fig. 4 and Source data in Supplementary Data 3

Fig. 4

Confirmation of virulence defects identified in the CRISPR pool in vivo. a In vitro and in vivo phenotype scores for genes targeted for knockout (KO) and mouse infection. b C57BL/6 mice were infected intraperitoneally (i.p.) with 50,000 tachyzoites of the parental line PruΔku80 (two experiments; n = 10 mice) and ΔGRA12 (two experiments; n = 10 mice), ΔTGME49_289150 (one experiment; n = 5 mice), ΔTGME49_313400 (two experiments; n = 10 mice) and control strains ΔGRA15 (one experiment; n = 5 mice), ΔROP18 (one experiment; n = 5 mice) and ΔMYR1 (one experiment; n = 5 mice). ΔROP18, ΔMYR1, ΔTGME49_289150 and ΔGRA12 were significantly different from PruΔku80 with P values of 0.0075, <0.0001, 0.0004 and <0.0001, respectively, using the Mantel–Cox test. c Brains were isolated from surviving mice and cysts enumerated at 24 or 28 days post infection. ΔTGME49_313400 n = 2, ΔGRA12 n = 10, ΔTGME49_289150 n = 5. P values were <0.0001 as calculated by one-way analysis of variance (ANOVA) with Tukey’s multiple comparison test. d Serum was isolated from mice surviving acute infection and tested for antibodies against Toxoplasma-soluble antigens in an ELISA. P values were ΔGRA12 P = 0.006, ΔTGME49_289150 P = 0.0058 and ΔTGME49_313400 P = 0.0009 compared to the naive sample as calculated by one-way ANOVA with Tukey’s multiple comparison test. ** refers to P < 0.01, *** to P < 0.001 and **** to P < 0.0001. See also Supplementary Figs. 5 and 6. Source data are provided as a Source Data file