Mosaic pik3ca hotspot transgenes cause vascular malformations in zebrafish embryos

A) The structure of human PIK3CA and the predicted structure of the zebrafish Pik3ca orthologue is remarkably conserved, with one additional amino acid in zebrafish Pik3ca shifting the kinase domain amino acids by 1. The position of E542, E545 and H1047 residues found mutated in PROS (and the orthologous E542, E545 and H1048) are labelled in magenta and are predicted to occur in very similar structural regions of the protein in both human and fish. B) Schematic representation of the transgenic pik3caPROS constructs generated for this study, containing β-actin:pik3caPROS-mScarlet and the cardiac reporter cmlc2:GFP flanked by Tol2 response elements, and is co-injected with Tol2 mRNA into 1-cell zebrafish embryos. Embryos were screened for successful integration of the transgene from 48 hpf onwards, indicated by mosaic cardiac cmlc2:GFP, and mScarlet expression. At 72 hpf, some PROS mosaic embryos had tail vascular malformations. C) Maximum Z-projection of a 72 hpf Tg(fli1:eGFP);casper tail region, showing typical endothelial vessel structures marked by fli1:eGFP fluorescence (green). Scale bar = 50 μm D) Representative diagrams and fluorescent microscopy images of the spectrum of vascular phenotypes seen in Tg(fli1:eGFP) embryos injected with pik3caH1048R-mScarlet. These ranged from unaffected and mild, to moderate and severe. Fli1:eGFP expression in unaffected embryos is as uninjected. Embryos scored ‘mild’ displayed expansion and mispatterning of isolated inter somitic vessels that was only visible when viewing eGFP fluorescence. Moderate and severe scores were applied to embryos where blood pooling was evident under brightfield, particularly obvious in the PCV. Yellow arrows indicate mis-patterned vessels in ‘mild’ embryos, and white arrows indicate overlap of fli1:eGFP (green) and mScarlet (magenta) expression in moderate and severe examples. Quantification of this and numbers screened can be seen in Figure 3. Scale bars = 100 μm. E) Cross sections of trunk blood vessels at 72 hpf, showing the dramatic expansion of the PCV. F) Quantification of these vascular malformation phenotypes for each construct, scored based on severity, as a percentage of the total numbers injected. Numbers indicate fli1:eGFP; cmcl2:GFP+ embryos screened for vessel malformations at 72 hpf, representative of three biological repeats per construct. Error bars represent S.E.M. G) Representative images of juvenile Tg(β-actin:pik3caE545K-mScarlet; fli1:eGFP); casper (top) and Tg(β-actin:pik3caH1048R-mScarlet; fli1:eGFP fish that harboured no detectable vascular anomalies at 72 hpf, but were found to have vascular anomalies when imaged two weeks later (E545K-injected – 6/16, H1048R-injected – 3/9). Dotted lines encompass vessel malformations in PROS animals, and equivalent regions in control

Evidence for cell-autonomous and non-cell-autonomous mechanisms of pik3caPROS vascular overgrowth

A) Representative maximum projection confocal images of 72 hpf Tg(fli1:eGFP);casper embryos expressing either wildtype pik3ca-mScarlet (no overgrowth) or a pik3caPROS hotspot variant and exhibiting a mild, moderate or severe vascular phenotype. For each, separated mScarlet fluorescence (magenta), fli1:eGFP channels (green) and merged channels are shown, with a zoomed area to the right. Blue arrows indicate mScarlet cells in close proximity to deformed vessels. Yellow arrows indicate white cells co-expressing fli1:eGFP and mScarlet. Scale bars = 50 μm B) Quantification of mScarlet cells by volume within GFP+ endothelial cells. Imaris was used to visualize confocal Z stacks of vessel overgrowth at 72 hpf. 3D structures of each channel, plus a co-localization channel were converted into surfaces, enabling volume calculations for vessels (green), mScarlet+ cells outside vessels (purple) or areas of colocalization (white). Volumes of each were expressed as percentage of total cell volume in 20x magnification view, dots indicate individual images (8 lesions were analysed in total).

Early mesodermal pik3caPROStransgene expression causes endothelial mis-patterning

A) Plasmid design of the transgenic construct tbxta:pik3caE545K-mScarlet. B) Simplified diagram of a tbxta progenitor cell which may differentiate into the lateral plate mesoderm, the notochord, or presomitic mesoderm which gives rise to endothelia, the notochord or muscle respectively. Mosaic tbxta:mScarlet is visible at 5 hpf (representative images) before onset of fli1:eGFP reporter expression. By 18 hpf, tbxta is most highly expressed in the notochord. C) Representative maximum projection confocal images of embryos from a range of stages between somitogenesis (∼18 hpf) and tailbud (∼28 hpf) to visualise the start of fli1:eGFP expression and endothelial development. In uninjected embryos, lateral stripes of fli1:eGFP are present from 16hpf, with sprouting of ISVs occurring from anterior to posterior from around 19 hpf onwards. In tbxta:pik3caE545K-mScarlet injected embryos, ectopic fli1:eGFP cells can be seen dorsal to the initial lateral plate mesoderm stripe. Scale bars = 50 μm D) Quantification of ectopic fli1:eGFP+ cells in injected tbxta:pik3caE545K-mScarlet, or injected with a similar construct lacking the pik3ca-ires element (unattached to the lateral mesoderm stripes of fli1:eGFP). 3 biological replicates shown with each dot representing 1 embryo. Significance was calculated with a two-tailed Mann Whitney test. ****=p<0.001.

Mosaic mesodermal pik3caPROS expression induces pan-lineage dysregulation

A) Schematic of experimental workflow for scRNA-seq. Tg(fli1:eGFP) embryos were injected with tbxta:mScarlet (control) or tbxta:pik3caE545K-mScarlet (PROS) and grown until 19 hpf, dissociated, and all live cells sorted by FACS before processing for 10x sequencing. B) UMAP showing 40 cell clusters found the aggregated dataset, split by experimental condition. C) UMAP of aggregated dataset coloured by mesodermal or other germ layer origin D) FeaturePlots split by experimental condition, showing expression levels of mScarlet (red scale), eGFP (yellow scale) and pik3ca (blue scale) E) The numbers of cells per cluster was expressed as a percentage of total cells for each experimental condition, and then the difference between percentages in PROS and control clusters visualised in a UMAP heat plot. Blue indicates a higher percentage of control cells relative to PROS, and red indicates a higher proportion of PROS relative to control cells. F) Deseq2 DE analysis and gene enrichment analyses using GO (G:), KEGG (K:) and Reactome (R:) databases were performed to compare PROS mosaic vs control cells in various cluster groups, or in mScarlet+ cells across the dataset.

pik3caPROSpromotes widespread transcriptional changes in ligand-receptor expression patterns

A) CellChat was used to identify potential interactions between ligand and receptor pairs in clusters. The top 10% strongest pathway interactions of ‘control’ cells relative to ‘PROS’ and vice versa shown, normalised to cell number. Blue arrows indicating whether a cluster is a sender or a receiver of the signal, with arrow thickness proportional to connection strength. Abbreviations: LPM=lateral plate mesoderm, PSM, presomitic mesoderm, Noto = notochord, Endoth=endothelial, KV = kuppfer’s vesicle, NC=neural crest, Epi=epiderm, Epi/Phar epidermal/pharyngeal, Otic Plcd=otic placode, Telen=telencephalon, SC=spinal cord, Diffn. Neuron = differentiating neurons, B) UMAP showing merged data to illustrate the methodology of NICHES, whereby a subset of clusters from Figure 4 is reclustered (Top UMAP). Ligand-receptor interactions between each group were quantified using NICHES, converting it into a signaling object of cell:cell interactions (Bottom UMAPs, grouped by either sending cluster type (left) or receiving cluster type (right). Colours in interaction maps were retained from the original clusters in the top UMAP. Epi/Ph = epiderm/pharyngeal mix, PSM = presomitic mesoderm, Myo=myotome, LPM/Endothelial = grouped lateral plate mesoderm and endothelial clusters. C) Heatmap depicting log2 fold change in expression of the top differentially expressed “sending” ligand (left) and “receiving” receptor (right) genes in PROS relative to control embryos, clustered by row. Ligand-receptor mechanisms were aggregated by common ligand or common receptor (unaggregated data are shown in Supplemental Figure 4C). Only differences for which p.adj<0.05 by Wilcoxon test between PROS and WT are shown. D) Examples of pathways showing dysregulated expression of ligand-receptor (LR) pairs. The expression values of LR pairs for each pathway were averaged as a module score. FeaturePlots (left) show intensity of this module for each cell:cell interaction, split by experimental condition. Examples of circus plots for each pathway are shown on the right. Signal direction between clusters (nodes) is shown by arrows, with darker and thicker arrows corresponding to greater connectivity strength. Connectivity scale intensity is matched between control and PROS plots for each LR pair.

Hotspot pik3caPROS transgenes cause vascular malformations and defects in muscle, cartilage and skeletal patterning in larval fish

A) Representative images of severe vascular malformations found in 72 hpf Tg(fli1:eGFP) embryos previously injected with either β-actin:pik3caH1048R, H1048L, E542K or E545K. Uninjected embryos, and those injected with WT pik3ca showed no phenotype at this timepoint. In pik3caPROS embryos, colour panels on the left show blood pooling in malformed vessels, with matched merged fluorescence images of the same area on the right showing enlarged fli1:eGFP+ vessels (green), and variable numbers of mosaic pik3caPROS-mScarlet cells (magenta). Scale bar = 100 μm. B)Tg(fli1:eGFP);casper fish injected with β-actin:pik3caE545K-mScarlet but negative for vascular malformations at 72 hpf developed a ‘rippled’ appearance by two weeks of age (SL4.6) caused by protrusions of pik3caPROS-mScarlet cells in the embryo flank (10/25). Scale bar = 100 μm. C) Patterning defects in the developing spine of staged matched Tg(β-actin:pik3caE545K-mScarlet; fli1:eGFP); casper injected embryos, showing irregular spacing of cartilage centra during early spine development (3/9 fish in which pattern had developed, prior to bone deposition), premature deposition of bone in patches posterior to stage-matched controls (1/3 fish at a stage where bone deposition is just starting in anterior and so therefore premature deposition is noticeable), and thickened vertebra structure (1/6 fish in which vertebrae were formed) D) Comparison of stage-matched uninjected and Tg(β-actin:pik3caE545K-mScarlet; fli1:eGFP); casper injected fish during tail fin development with patterning defects, revealed by alcian blue staining for cartilage and alizarin red co-staining for bone. 3/9 cartilage hypural patterning defects, 5/7 tail shape defects.

Ubiquitous and mesodermal expression of pik3caPROS causes vascular malformations

A) Quantification of vascular phenotypes in 72 hpf Tg(fli1:eGFP) embryos injected with pik3caE545K-mScarlet constructs driven by β-actin or tbxta, or an injected no-pik3ca control (tbxta:mScarlet). Represents 3 biological replicates. Error bars are S.E.M. p-values are from chi:squared tests on raw counts, * p<0.05, ***p<0.001, **** p<0.0001.

scRNA seq analysis of pik3caPROS mosaic embryos

A) UMAP FeaturePlots showing expression intensity (purple) of transgenesis or cell lineage markers. B) Mapping of 18hpf (top) and 19 hpf (bottom) published dataset identities onto these data to call clusters. C) UMAP of re-clustered cluster 0, with identities from the zebrahub 19 hpf dataset mapped as in B. D) Single cell velocity analysis was performed on PROS and control data, and the vector field overlaid on a UMAP projection, with arrow indicating velocity dynamics and strength across clusters. The greatest variance in arrow length and density between PROS and control is found in the cluster groups outlined (pink dashed). The velocity length of this subset is displayed on a heatmap, with cells showing the highest abundances of unspliced mRNA in red. E) As Figure 4E, but to show the extent of the statistically significant changes in the proportion of cells within cluster groups. Two-sided Pearson’s chi-squared proportion test. ** p<0.01, ***p<0.001, ****p<0.0001

Mosaic pik3caPROS causes pan-lineage transcriptional changes, including to ligand-receptor signalling genes.

A) Volcano plots showing significantly differentially expressed genes after DE analysis of cells within cluster 0. Left – comparison of all cluster 0 PROS mosaics vs control, Right – mScarlet+ vs mScarlet-cells in PROS mosaics only. B) Heatmap showing fold change enrichment (Seurat wilcox test p<0.05, p.adj<0.1) of Hallmark (H:) and Progeny (P:) pathways in PROS mosaic cluster 0 mSc+ vs mSc-cells. Violin plots confirm significant upregulation of selected Hallmark signatures in mScarlet+ cells (wilcox test, **** = p<0.001) C) Heatmap showing fold change enrichment (Seurat wilcox test p<0.05, p.adj<0.1) of Hallmark (H:) and Progeny (P:) pathways in PROS mosaics relative to control N.S = non-significant, and Cluster 0 UMAP subset showing expression intensity of Hallmark terms, split by experimental condition. D) As in Figure 4F), but comparing only mScarlet- (ie, WT) cells between experimental conditions. E) Heatmaps depicting log2 fold change in expression of the top differentially expressed “sending” (left) and “receiving” (right) ligand-receptor gene pairs in PROS relative to control embryos, clustered by row. Only differences for which p.adj<0.05 by Wilcoxon test between PROS and WT are shown.