Engineering Toxoplasma gondii secretion systems for intracellular delivery of multiple large therapeutic proteins to neurons

Abstract

Delivering macromolecules across biological barriers such as the blood-brain barrier limits their application in vivo. Previous work has demonstrated that Toxoplasma gondii, a parasite that naturally travels from the human gut to the central nervous system (CNS), can deliver proteins to host cells. Here we engineered T. gondii's endogenous secretion systems, the rhoptries and dense granules, to deliver multiple large (>100 kDa) therapeutic proteins into neurons via translational fusions to toxofilin and GRA16. We demonstrate delivery in cultured cells, brain organoids and in vivo, and probe protein activity using imaging, pull-down assays, scRNA-seq and fluorescent reporters. We demonstrate robust delivery after intraperitoneal administration in mice and characterize 3D distribution throughout the brain. As proof of concept, we demonstrate GRA16-mediated brain delivery of the MeCP2 protein, a putative therapeutic target for Rett syndrome. By characterizing the potential and current limitations of the system, we aim to guide future improvements that will be required for broader application.

Fig. 1. Targeting of therapeutic proteins to the rhoptry secretion organelles by fusion to toxofilin.

a, Illustration of a T. gondii cell. IMC, inner membrane complex (parasite outline). b, Scheme of the genetic constructs used. c, Intracellular RH T. gondii immunostained with the rhoptry marker anti-ROP2/4, in HFF cells. d, Intracellular RH T. gondii stably expressing different toxofilin-fused therapeutic proteins associated with human neurological diseases, in HFF. e, Intracellular T. gondii expressing different variations of toxofilin-fused zinc finger nucleases and Cas9. The scheme of each genetic construct is displayed above the image of the T. gondii expressing it. The images shown are representative of the protein localization over several independent transfections (numbers of repeats for each are provided in Extended Data Table 2). Scale bars, 10 μm.

Fig. 2. Targeting therapeutic proteins for intracellular delivery by T. gondii’s dense granules using fusion to GRA16.

a, Illustration of an intracellular parasitophorous vacuole (PV) containing four T. gondii parasites secreting a dense granule protein (yellow) targeted to the host cell nucleus (HCN). b, Scheme of the genetic constructs used. c,d, Intracellular RH T. gondii stably expressing HA-tagged GRA16 alone (c) or GRA16-fused proteins associated with human neurological diseases (d), in HFF cells. Images represent localization of the proteins to the parasitophorous vacuole and host cell nucleus (MECP2opt, SMN1, TFEBopt), to the parasitophorous vacuole alone (ASPA, ASPAopt, GALC), or the most frequently observed localizations in the pool of stably transfected parasites. Rightmost image of each panel shows a close-up view of the parasitophorous vacuole. The numbers of repeated independent transfections for each are provided in Extended Data Table 2. e, Fluorescence quantification of the anti-HA signal in the nuclei of infected host cells. Data represent mean ± s.d. N = cells per condition, left to right: 32, 10, 30, 11, 39, 23, 6, 17, 20, 47. a.f.u., arbitrary fluorescence units. Significance represents the difference between the fusion protein and the parental strain (‘no construct’), calculated using one-way ANOVA with multiple comparisons (Dunnett test); ****P < 0.0001, *P < 0.0332. Scale bars, 10 μm. Source data

Fig. 3. Dual protein secretion, kinetics of dense granule protein delivery and protein delivery to neurons.

a, Representative images of Synaptophysin-TdTomato neurons infected with T. gondii delivering GRA16-MeCP2 alone, toxofilin-Cre (TCre) alone or both. White arrowheads, nuclear GRA16-MeCP2; grey arrows, Synaptophysin-TdTomato punctae from Cre recombination. Scale bars, 20 µM. b, Mean fluorescent intensity (MFI) of anti-HA in the nuclei of infected neurons. N = 50 images per strain per biological replicate. Black circles, biological replicates; grey circles, technical replicates. c, Quantification of Synaptophysin-TdTomato punctae in infected neurons. Symbols, individual FOVs. N = 9 FOVs per experiment, 2 independent experiments. ND, not detected. d, Quantification of the percent of Synaptophysin⁺HA⁺ neurons of the total Synaptophysin⁺ neurons per strain. N = 100 neurons per strain per replicate. Symbols, biological replicates. b–d, Bars denote mean ± s.e.m., unpaired t-test. e, Overview of the automated pipeline used for imaging and statistical analysis of T. gondii-infected cells. f–l, Quantitative characterization in infected HFF cells. For all graphs, data represent mean ± s.d. f, Infection rate. g, PV per infected cell. h, Nuclear protein delivery rate (% of infected cells with HA-positive nuclei). i, Mean nuclear fluorescence intensity in cells with a single PV. j, Increase in PV size demonstrates a doubling time of 7 h for all lines, in agreement with previous reports. k,l, Manual quantification of protein delivery rate (k) and nuclear fluorescence intensity (l) for validation of the automated pipeline. N = images per condition (N for each condition provided in Methods and in Source data). m, Representative images of infected LUHMES-derived neurons (3 independent repeats). Yellow dashed lines mark the host nucleus. n, Infection rate for GRA16-HA T. gondii in LUHMES neurons. Data represent mean ± s.d. N = images per condition (N for each condition provided in Methods and in Source data). o, Uninfected and infected WT and MECP2-KO neurons, 12 hpi. Magenta dashed lines mark neuronal nuclei. Yellow arrowheads mark infected neurons. Scale bars, 10 μm. p, Fluorescence quantification of the anti-MeCP2 signal in the nuclei of infected neurons. Data represent mean ± s.d. N = cells per condition, left to right: 427, 50, 157, 391, 510, 95, 202, 391. Significance represents the difference between each condition and the untreated MECP2-KO control, two-way ANOVA with multiple comparisons (Dunnett test); ****P < 0.0001. Source data

Fig. 4. Probing the functionality of T. gondii-delivered MeCP2 via heterochromatin binding, pull-down assay and single-cell sequencing in human cortical organoids.

a, Mouse primary neurons inoculated with GRA16-MeCP2. Top images show a close-up view of the soma and T. gondii (2 independent repeats). Blue arrowheads mark co-localization of GRA16-MeCP2 with foci of heterochromatic DNA. Scale bars, 10 μm. b, Mouse neuroblastoma cells inoculated with GRA16-MeCP2 and GRA16-HA T. gondii (2 independent repeats). Scale bars, 10 μm. c, Pull-down assay of protein lysates from HFF cells infected with GRA16-MeCP2 or WT ME49 T. gondii. L, protein ladder; In, Input lysate; Me(+), protein pull down with methylated DNA probes; Me(−), protein pull down with non-methylated DNA probes; primary antibody, anti-MeCP2. Representative blot from 3 independent repeats. d, UMAP of single cells based on human+T. gondii gene quantification, coloured by percentage of Toxoplasma and human transcript counts. e, Same UMAP as in d but coloured by exogenous constructs transcript counts. f, UMAP of single cells based on human gene expression. Main clusters were identified and coloured by cell subtypes. CyclingProg, cycling progenitors; vRG_oRG, vRG_oRG2, ventricular and outer radial glia; N_IP, neuronal intermediate progenitors; N1, N2, N3, neuronal clusters; N_UPR, N_UPR2, N_UPR3, neurons with signature of unfolded protein response; N_Met, neurons with signature of metabolic regulation; N_Proj, neurons with signature of axonal regulation. g, Heat map showing the distribution of differentially expressed genes between neurons of organoids infected with GRA16-MeCP2 vs uninfected (left),GRA16-HA vs uninfected (middle) and GRA16-MeCP2 vs GRA16-HA organoids (right). h, Violin plot showing mean expression of genes belonging to the ‘Reactome transcriptional regulation by MECP2’ pathway (10.3180/R-HSA-8986944.1); norm. enrich. = 3.19, P = 0.001417 for the GRA16-MeCP2 vs GRA16-HA comparison. i, UMAP of single infected neurons coloured by expression of CREB1 and MEF2C. j, Violin plots showing the counts distribution of CREB1 and MEF2C between neurons from organoids infected with GRA16-HA and GRA16-MeCP2. All scRNA-seq data include 3 biological replicates (single-cell suspensions dissociated from 3 different organoids) for each condition. Source data

Fig. 5. T. gondii delivers MeCP2 to the CNS following intraperitoneal administration in mice.

Mice were injected intraperitoneally with saline, the parental ‘Pru’ (reduced virulence) strain, GRA16-HA or GRA16-MeCP2 Pru T. gondii as labelled. At 18 dpi (a–d) or 1 and 3 mpi (e–m), brains were collected and processed for analysis. a, Representative maximal projection image of a cyst. Scale bar, 20 µM. b, Quantification of cyst numbers per 7 sections per mouse at 18 dpi. N = 12 mice per group. Blue symbols represent mice used to quantify punctate HA staining in c and d. Data represent mean ± s.d. c, Representative confocal images from brain sections stained with anti-HA (yellow) and anti-NeuN (magenta) from 18 dpi mice. Arrows point to cells that show co-localization of anti-HA and anti-NeuN staining; arrowheads point to anti-HA staining that did not co-localize with anti-NeuN staining. HA staining for GRA16-HA-infected tissue was typically diffuse while GRA16-MeCP2 was punctate. Scale bar, 10 µM. d, Quantification of the HA⁺NeuN⁺ punctate staining as a percentage of all HA⁺NeuN⁺ staining; 150 FOVs per mouse, N = 3 mice per group. e, Parasite burden in indicated organs by qPCR of the parasite-specific B1, normalized to GAPDH. All results normalized to the average brain burden for parental T. gondii at 1 mpi. f,g, Quantification of cysts per 9 sections per mouse, N = 5 mice per group at 1 (f) and 3 (g) mpi. h, Representative image of brain section stained with anti-Iba1 antibodies (microglia and macrophages). Scale bar, 110 µM. i,j, Number of Iba1⁺ cells per brain section per mouse at 1 (i) and 3 (j) mpi. N = 3 sections per mouse. k, Representative image of brain section stained with anti-CD3ε antibody (T cells). Scale bar, 110 µM. l,m, Number of CD3⁺ cells at 1 (l) and 3 (m) mpi. N = 5 mice per group for the infected mice (1 and 3 mpi); N = 3 mice per group for saline. The saline group was combined from the 1 mpi (black diamonds) and 3 mpi (grey diamond) cohorts. For all graphs, data represent mean ± s.e.m.; one-way ANOVA with subsequent two-tailed t-tests were used to calculate significance.

Fig. 6. 3D distribution of rhoptry- and dense granule-mediated protein delivery by T. gondii in cleared brains, following intraperitoneal administration in reporter mice.

Ai6 reporter mice were injected intraperitoneally with toxofilin-Cre or GRA16-Cre Pru T. gondii or saline. At 21 dpi, brains were collected, processed, cleared using the CRYSTAL technique and imaged. a, Distribution of ZsGreen+ cells in different brain regions, averaged over all the samples in each group. Top: toxofilin-Cre (magenta); middle: GRA16-Cre (cyan); bottom: saline (yellow). b, 3D visualization of group-average distribution of ZsGreen+ cells within the whole brain. c, Total ZsGreen+ cell volume per brain region, obtained from the averaged data per group. Data represent mean ± s.d. Biologically independent repeats (mice) per group: N = 8 (toxofilin-Cre), 7 (GRA16-Cre) and 2 (saline). Scale bars, 2,000 μm.

Extended Data Fig. 1. Testing truncated GRA16 for protein delivery and profiling the intrinsic disorder of the tested fusions.

(A) Intracellular RH T. gondii stably expressing different fragments of GRA16, fused to either a HA tag alone or to the HA-tagged murine MeCP2 (codon optimized). All tagged proteins localized to the PV. However, only the GRA16 truncates without MeCP2 were detected in the host cell nucleus (HCN), as well as the full-length GRA16 with or without MeCP2. Rightmost image of each set shows a close-up of the parasite vacuole, and the text above it indicates the localization of the tagged protein in the image. Yellow dashed lines mark the host cell nucleus. Left and top: illustration of the full-length GRA16, tested C-terminally truncated GRA16, and the fused HA or HA-MECP2 sequences used in the respective constructs. NLS-Nuclear localization signal. (B) Fluorescence quantification of the anti-HA signal in the nuclei of infected host cells. Error bars represent mean ± SD. Take note that the y-axis of the leftmost graph is divided into two segments, to account for the high values of GRA16[505]-HA. AFU-arbitrary fluorescence units. Mean relative nuclear localization is normalized to the full-length GRA16[505]-HA. N=cells/condition, left to right: 32,10,32,19,13 (GRA16-HA truncates), 32,17,12,15,18 (GRA16-MeCP2 truncates). Significance represents the difference between the expressed protein and the parental strain (‘no construct’), calculated by two-way ANOVA for the effect of the GRA16 variant and the fused sequence (HA or HA-MECP2) with multiple comparisons (Dunnett test). * P = 0.0332;** P = 0.0021;**** P < 0.0001. Scale bar = 10μm. (C) Prediction of protein intrinsic disorder for all tested GRA16 fusion proteins, using IUPred2 (disordered protein regions, blue line) and ANCHOR2 (disordered binding regions, red line). Dashed line marks the start of the heterologous protein, and the grayed-out region before it is GRA16. Disorder score = average (IUPred2 score, ANCHOR2 score) of the fused heterologous protein. (D) Correlation between host nuclear localization and the average intrinsic disorder score of the fused heterologous protein. Linear regression line drawn in gray. Source data

Extended Data Fig. 2. Delivery of proteins to human brain organoids.

Representative immunostaining for (A) anti-SOX2 (green), anti-MAP2B (pink), anti-HA (red), DAPI (blue). (B) anti-TUJ1 (white), anti-Toxo (red), DAPI (blue). (C) anti-NeuN (green), anti-MeCP2 (white), anti-HA (red) and DAPI (blue). Scale bars = 20 μm. (D) enlarged sections containing infected and uninfected cells from C. Scale bars = 10 μm. N = 3 replicates (organoids) for each strain and combination of antibodies.

Extended Data Fig. 3. Extended single-cell sequencing analyses including quality checks and normalization, identification of bona fide infected cells, clustering, gene markers and the MECP2 reactome pathway.

(A) Violin plots showing the distribution of the relevant metrics for quality checks across all the replicates of each infection condition, and the droplets identified as doublets and as empty droplets. All scRNA-seq data includes three biological replicates (single cell suspensions dissociated from 3 different organoids) for each condition. (B) UMAP plots of the single cells colored by experimental conditions, barcoded replicates, and bona fide infected cells in the UMAP embeddings including human + T. gondii genes quantification; (C) Violin plots showing the distribution of human vs T. gondii transcripts across all replicates, after having identified the infected cells; (D) UMAP plots of the single cells colored by experimental conditions, barcoded replicates, and bona fide infected cells in the UMAP embeddings including only human genes quantification; (E) UMAP and diffusion map plots of the single-cell colored by Leiden cluster; UMAP plots colored by paradigmatic marker of neurodevelopmental cell types.

Extended Data Fig. 4. MeCP2 does not alter peripheral inflammation, and brain distribution following intravenous administration.

(A–I) Mice were inoculated with 20,000 freshly lysed Pru (reduced virulence) T. gondii tachyzoites of the identified strains. At labelled time points, organs were harvested, fixed, sectioned, stained, and analyzed. N = 5 mice/group/time point for the infected conditions and N = 2 mice/group for the uninfected controls (independent biological repeats). The uninfected control mice are combined from the two time cohorts (1 from 1 mpi and 1 from 3 mpi). (A–C) Representative images of formalin-fixed, H&E-stained tissue sections from labelled organs at 1 mpi. Scale bar=210 µM. (D–F) Representative images of lung sections that were given inflammation scores of 0, 1, or 2, respectively. Inflammation scoring system: Tissue with no inflammation = 0, some inflammation = 1, and notable inflammation = 2. Scale bar = 210 µM. (G–H) Graph of average inflammation score for labelled organs at 1 mpi (G) and 3 mpi (H). Inflammation score generated from 8 FOV/organ, 1 section/organ/mouse. Bars, mean ± SEM. ND = not detected. (I) Percent weight change of infected and saline-inoculated mice during the first 20 days of infection. Mice were weighed before infection or saline-inoculation and throughout infection as indicated. Symbols, mean ± SD. (J) Male mice were injected in the tail vein with 50,000 tachyzoites of ME49 GRA16-MeCP2 T. gondii. At 21 days post-injection, the tissues were harvested and DNA was extracted. T. gondii genomes per mg tissue was measured by qPCR of the T. gondii B1 gene. Genome number was calculated from the Ct values based on a standard curve generated from a serial dilution of a B1 plasmid, assuming 35 B1 copies per genome. At 21 dpi, all peripheral tissues have significantly lower amounts of T. gondii than the brain (P < 0.001, one-way ANOVA), N = 5 mice/group. Bars show mean ± std, and each color corresponds to a different mouse.

Extended Data Fig. 5

Per-mouse distribution of the fluorescent signal in mice treated with saline, GRA16-HA and GRA16-MeCP2 across 7 representative brain slices.

Extended Data Fig. 6. Per-mouse brain distribution quantification and extended methods for the CLARITY data processing, cell segmentation and alignment to the Allen brain atlas.

(A) Distribution of rhoptry- and dense granule-mediated protein delivery by Pru T. gondii in different brain regions, cell volume per brain region, per animal in each group (mouse ID numbers are denoted in the legend). (B) Raw images of Purkinje neuron autofluorescence in the cerebellum of saline-injected control mice. The depicted slices are taken from sample 4843. The number of biologically independent samples in the saline group is N = 2. (C) Representative output from the 3D cell detection pipeline in whole-brains. All operations are three-dimensional, however for simplicity, we show here the output only at a slice of sample 4864. (a) Raw data. (b) Image restoration after intensity normalization and denoising. (c) Overlay of the perimeter of detected cells (magenta edge) and raw data. Numbers 1–4 correspond to the four boxes shown in (a). (D) Representative output of alignment to the Allen Brain Atlas. Coronal view of the aligned raw data (grayscale), onto the Atlas (magenta contour lines). Tiles correspond to slices of the coronal view at different depths of the sample. The output corresponds to sample 4864. (E) Validation of segmentation, with respect to manual annotations by domain experts. (a) Convergence of true positive (TP) and true negative (TN) pixels, with respect to the number of z-slices included in the computation. (b,c,d) overlay of segmentations and raw data, shown in slices along the (b) xy, (c) yz, and (d) xz plane. Color code: yellow: correctly detected cell core, cyan: non-detected boundary region around cell cores, magenta: false negatives. (e) 3D view of segmented data colored as described above. Images (b)-(e) are taken within the software ITK-SNAP (Yushkevich et al 2006).

Extended Data Fig. 7. Mutant double-knockout of LDH1 and LDH2, which are reported to have inhibited growth in vivo, are viable and grow as well as WT T. gondii in vitro.

(A) Scheme of the genetic constructs used in the sequential CRISPR-Cas9 mediated gene knock-out of LDH2 and LDH1 by Cas9-mediated insertional replacement with the selection markers DHFR (for LDH2) or mNeonGreen (for LDH2). We note the location of the different amplicons targeted by the diagnostic PCRs. (B) Gel electrophoresis results showing the amplicons from PCRs, spanning homology regions of insertion cassettes in ME49 T. gondii with the genetic knockouts ΔLDH2ΔLDH1 (I) and ΔLDH2 (II). The WT ME49 (C) parental strain is included as a control. M = MW marker (1 Kb Plus DNA Ladder). The numbers 1–8 above each gel correspond to the targeted PCR amplicons shown in (A).(C, D) Fold change in gene of interest mRNA expression relative to housekeeping gene Actin1 in ME49 and derived knockout mutants ME49ΔLDH2 (C) and ME49ΔLDH2ΔLDH1 (D). Comparative statistical analysis was performed using independent two tailed t-tests. Welch’s correction was used in the statistical analysis of ME49 vs. ME49ΔLDH2ΔLDH1. Error bars are Mean ± SD representative of three biological replicates. (E) Plaque assays comparing the growth of ME49 and derived knock-out mutants. 500 parasites were seeded into 6-well plates with DMEM supplemented with 10% or 1% FBS, and cultured for 7 days. (F) Relative plaque area for parasites grown in DMEM supplemented with 10% FBS (N = 130 plaques) or 1% FBS (N = 183 plaques). Relative plaque area analysis was done using Kruskal Wallis One-way ANOVA with Dun’s post hoc. Error bars are Mean ± SD representative of three biological replicates. (G) Number of plaques in a 307.88mm2 area for parasites grown in DMEM supplemented with either 10% FBS and 1% FBS. Plaque count analysis was done using One-way ANOVA with Tukey’s post hoc. Error bars are Mean ± SD, representative of three biological replicates. (H) Red-green invasion assays between ME49 and derived knock-out strains. Parasites were allowed to invade an HFF monolayer for 30 minutes prior to fixation and staining. Analysis was done via One-way ANOVA with Tukey’s post hoc. Error bars are Mean ± SD, representative of three biological replicates, consisting of two technical replicates each, N ≥ 100. (I) Parasite replication assays between ME49 and derived knock-out strains under different media culturing conditions. The coloring indicates the number of tachyzoites observed in each PV. Analysis done using One-way ANOVA with Tukey’s post hoc. Error bars are Mean ± SD, representative of three biological replicates, N = 150 vacuoles/condition. Source data