Zebrafish models of human-duplicated SRGAP2 reveal novel functions in microglia and visual system development

Abstract

The expansion of the human SRGAP2 family, resulting in a human-specific paralog SRGAP2C, likely contributed to altered evolutionary brain features. The introduction of SRGAP2C in mouse models is associated with changes in cortical neuronal migration, axon guidance, synaptogenesis, and sensory-task performance. Truncated SRGAP2C heterodimerizes with the full-length ancestral gene product SRGAP2A and antagonizes its functions. However, the significance of SRGAP2 duplication beyond neocortex development has not been elucidated due to the embryonic lethality of complete Srgap2 knockout in mice. Using zebrafish, we show that srgap2 knockout results in viable offspring and that these larvae phenocopy "humanized" SRGAP2C larvae, including altered morphometric features (i.e., reduced body length and inter-eye distance) and differential expression of synapse-, axonogenesis-, and vision-related genes. Through single-cell transcriptome analysis, we demonstrate a skewed balance of excitatory and inhibitory neurons that likely contribute to increased susceptibility to seizures displayed by Srgap2 mutant larvae, a phenotype resembling SRGAP2 loss-of-function in a child with early infantile epileptic encephalopathy. Single-cell data also shows strong endogenous expression of srgap2 in microglia with mutants exhibiting altered membrane dynamics and likely delayed maturation of microglial cells. Microglia cells expressing srgap2 were also detected in the developing eye together with altered expression of genes related to axonogenesis in mutant retinal cells. Consistent with the perturbed gene expression in the retina, we found that SRGAP2 mutant larvae exhibited increased sensitivity to broad and fine visual cues. Finally, comparing the transcriptomes of relevant cell types between human (+SRGAP2C) and non-human primates (-SRGAP2C) revealed significant overlaps of gene alterations with mutant cells in our zebrafish models; this suggests that SRGAP2C plays a similar role altering microglia and the visual system in modern humans. Together, our functional characterization of conserved ortholog Srgap2 and human SRGAP2C in zebrafish uncovered novel gene functions and highlights the strength of cross-species analysis in understanding the development of human-specific features.

Keywords: Danio rerio; brain development; eye development; gene duplication; human evolution; microglia; zebrafish.

Figure 1.. Functional analysis of srgap2 in the developing zebrafish.

(A) Top left, phylogenetic tree of human, mice, and zebrafish SRGAP proteins based on their full length amino acid sequence using the Unweighted Pair Group method with Arithmetic Mean method. Top right, schematic of inferred SRGAP2 gene family evolutionary history across human chromosome 1 . Bottom, cartoon summarizing the results of previous studies, showing that SRGAP2 functions after homodimerization in concert with F-actin (brown oval) to dictate cell membrane dynamics (bottom left) or heterodimerize with SRGAP2C producing no functional product (bottom right). (B) Co-immunoprecipitation of human-specific SRGAP2C and zebrafish Srgap2 in HEK293T cells showed interaction between these proteins. (C) Temporal expression of srgap2 in the developing embryo according to publicly available RNA-seq data (black line represents the best fit line with the standard error in dark gray) and normalized quantitative RT-PCR data from whole-embryo RNA collected at 6, 10, 24, 72, and 120 hpf (blue boxes, each dot represents a biological replicate). The light-gray box represents a critical stage in zebrafish neurogenesis between 6 and 24 hpf . (D) srgap2 expression in embryonic (24 hpf) and adult (>12 months old) tissues from a published RNA-seq dataset . (E) Spatial endogenous expression of srgap2 at 24 hpf and 3 dpf via in situ hybridization shown in blue. Scale bar 100 μm. (F) Illustration of the approaches to creating knockout srgap2 zebrafish. Top, a stable knockout line was generated by injecting SpCas9 coupled with one gRNA targeting exon 4. Middle, G₀ knockouts were generated by co-injecting SpCas9 coupled with four gRNAs targeting early exons. Bottom, humanized larvae were created by injecting in vitro transcribed SRGAP2C mRNA at the one-cell stage.

Figure 2.. Developmental and cellular phenotypes of diverse zebrafish models of SRGAP2.

(A) Measurements of central line distance (ANOVA: F_{(4, 321)}= 12.84, genotype effects p-value= 1.04×10⁻⁹, FDR-adjusted p-values Het= 4.40×10⁻⁷, Hom= 6.29×10⁻⁷, Pooled= 0.015, SRGAP2C= 1.36×10⁻⁴), Euclidean distance between the eyes (ANOVA: F_(4,321)= 23.49, genotype effects p-value= 4.72×10⁻¹⁷, Dunnett’s test FDR-adjusted p-values: Het= 6.77×10⁻¹¹, Hom=4.69×10⁻¹⁰, Pooled= 0.05, SRGAP2C= 2.19×10⁻⁹), and head angle (ANOVA: F_(4,315)= 0.49, genotype effects p-value= 0.746) in 5 dpf larvae from stable srgap2 knockout (Het n= 43, Hom n= 86), G₀ knockouts (n= 34), SRGAP2C-injected (n= 44), and control larvae (n= 124). Dots represent an imaged larva with the color indicating the imaging plate (a co-variable included in the statistical analyses). The red dotted line corresponds to the mean value for the control group. Representative images of each measurement are included on the top of each plot. (B) Correlation of the fold change (FC) between srgap2 G₀-knockouts and SRGAP2C-injected larvae at 5 dpf, with common DEGs highlighted (red= upregulated (FC > 2), blue= downregulated (FC < −2)). Top representative GO terms enriched in common DEGs between srgap2 G₀-knockouts and SRGAP2C-injected larvae (complete results in Table S5). Color of the bar represents the direction of the genes (red= commonly upregulated, blue= commonly downregulated). (C) Correlation of the FC between srgap2 G₀-knockouts and SRGAP2C-injected larvae across development using data from 24, 48, and 72 hpf larvae, with common DEGs highlighted, complete results can be found in Tables S7, S8. (D) Clustering of the 28,687 profiled cells colored as 24 cell types based on the expression of gene markers. Expression of srgap2 across cell types (left side, shaded in gray), with the size of the circle representing the percentage of cells in that cluster expressing srgap2 and the color of the circle representing the average scaled expression in the cluster. Enrichment test for the overlap between marker genes for each cell type and the differentially expressed genes at 3 dpf from bulk RNA-seq data (right side), with the size of the circle representing the odds ratio for the enrichment and the color of the circle the -log(BH-adjusted p-value) of the Fisher’s exact test. Asterisks indicate an FDR-adjusted p-value < 0.05.

Figure 3.. Neuronal alterations in SRGAP2 mutants.

(A) Neuronal clusters (hypothalamus, thalamus, optic tectum, hindbrain, Purkinje cells, and neurons rich in glutamate receptors) selected to perform a differential gene expression test was performed to DEGs in the SRGAP2 mutants compared to the control group. Bar plot represents the top GO terms overrepresented in the 14 commonly upregulated genes (complete results in Table S14). (B) Ratio of cells classified as excitatory (vglut2⁺) to inhibitory (gad1b⁺) between the srgap2 G₀-knockouts, SRGAP2C-injected, and controls (srgap2 G₀ knockouts: 0.78±0.15, p-value= 0.031; SRGAP2C-injected: 0.82±0.09, p-value= 0.017, controls= 0.57±0.13; t-tests versus controls). (C) Ratio of excitatory (vglut2:DsRed+) to inhibitory (dlx6:GFP) cell area quantified from images of 3 dpf srgap2 G₀-knockout, SRGAP2C-injected, SpCas9 control injected, and uninjected wild type larvae (G₀ knockout: Exc:Inh ratio=1.21±0.07, p-value=3.0×10⁻⁴, SRGAP2C: Exc:Inh ratio= 1.16±0.05, p-value= 7.0×10⁻⁴, SpCas9-injected controls Exc:Inh ratio= 0.98±0.03, p-value= 0.959; Mann-Whitney U-tests p-values vs wild-type controls). Images include representative samples per group, scale bars 100 μm. (D) High-speed events (HSE, >28 mm/s) identified in 15 min recordings of 4 dpf larvae (srgap2 knockouts (stable Hom_parent and G₀), SRGAP2C-injected, and SpCas9-injected controls, n= 36 larvae per group) with and without PTZ. Frequency of HSE per min were compared to controls (0 mM PTZ: ANOVA p-value for genotypic effect= 0.415, average HSE/min= 0.006±0.02, no significant differences between groups; 2.5 mM PTZ: ANOVA genotype effect p-value= 1.1×10⁻⁶, Hom_parent= 0.010, G₀-knockouts= 2.2×10⁻⁶, SRGAP2C-injected= 3.90×10⁻⁵). (E) Local field potential (LFP) recordings in the optic tectum of 4 dpf larvae (G₀-knockouts, SRGAP2C-injected, and SpCas9-injected controls, n=21–30 per group) were obtained and scored by two independent researchers. Representative traces per group are shown. Asterisks in graphs represent a p-value below 0.05 for the comparison against the control group. ns= not significant.

Figure 4.. Cross-species conservation of SRGAP2 as a microglial gene.

(A) Top GO terms with significant overrepresentation in genes upregulated (red) or downregulated (blue) in microglial cells from SRGAP2 mutants from Figure 2D. (B) Sphericity values for individual microglial cells (mpeg1.1⁺) at 3 and 7 dpf in srgap2 knockouts, SRGAP2C-injected, and scrambled gRNA-injected controls. Each dot represents a single microglial cell (average of 4–5 cells per larvae from 3–4 larvae per genotype per timepoint were obtained). Representative images for the median sphericity value of larvae at 3 and 7 dpf for each genotype are included below the graph (scale bars: top images= 100μm, bottom images= 5 μm). Asterisks denote a Tukey post-hoc p-value < 0.05. 3dpf: srgap2 G₀ knockouts: 0.70±0.09, p-value= 0.0085; SRGAP2C-injected: 0.73±0.09, p-value= 0.0021, controls: 0.58±0.12; 7dpf: srgap2 G₀ knockouts: 0.74±0.11, p-value < 2.2×10⁻¹⁶; SRGAP2C-injected: 0.78±0.08, p-value < 2.2×10⁻¹⁶, controls: 0.46±0.13. (C) Evaluation of 610,596 prefrontal cortex cells from human, chimpanzee, macaque, and marmoset (human: 171,997, chimpanzee: 158,099, macaque: 131,032, marmoset: 149,468) showing the levels of SRGAP2 and SRGAP2C expression across species, highlighting the microglial cluster with a dotted square. Micro: microglia. Expression of SRGAP2 and SRGAP2C in microglial subtypes across species with subtypes ordered from highest expression left to right. huMicro: human-specific microglia, hoMicro: Hominidae-specific microglia. (D) Microglial cells from human, chimpanzee, macaque, and marmoset (human: 8,819 cells, chimpanzee: 6,000 cells, macaque: 9,000 cells, marmoset: 7,099 cells) from the prefrontal cortex and middle temporal gyrus were used to identify common DEGs between human and non-human primates, finding 340 common upregulated and 323 common downregulated genes. Top GO terms with significant overrepresentation in common DEGs are included.

Figure 5.. SRGAP2 impacts the retina.

(A) Section of a 3 dpf NHGRI-1 larva staining srgap2 expression via in situ hybridization, labeling predominantly the optic nerve (ON), retinal pigmented epithelium (RPE), and the ganglion cell layer (GCL). D: dorsal, V: ventral. (B) Retinal ganglion cells (RGCs) were selected and a differential gene expression performed between SRGAP2-mutants (srgap2 knockouts and SRGAP2C-injected) versus controls, identifying 60 upregulated genes and 84 downregulated genes, with their top overrepresented GO terms included in bar plots. (C) Human and macaque cells from retinal organoids (43,857 human and 19,894 macaque) were integrated to identify genes with increased expression in either species, with their top overrepresented GO terms included in bar plots (complete results in Tables S22 and S23). (D) Motion response to changes in light were assessed in 4 dpf srgap2 knockouts (Hom_parent and G₀-knockouts), SRGAP2C-injected, and SpCas9-scrambled gRNA-coupled control larvae using a 10 min acclimation period followed by an abrupt light change. Plot includes trend lines for change in distance moved observed in each evaluated group (n= 24 per group, standard error for each line included as a shaded gray), which were different between all groups compared to controls (Kolmogorov-Smirnov tests p-values: Hom_parent= 9.16×10⁻¹¹, G₀-knockouts= 5.93×10⁻⁸, SRGAP2C-injected= 1.11×10⁻¹²). (E) Optomotor responses were evaluated in 4 dpf larvae using an optimized protocol that quantifies the percentage of larvae relative to moving stripes. Boxplot includes the percentage of OMR-positive larvae (aligned to the visual stimulus) in srgap2 knockouts (Hom_parent and G₀-knockouts) and SRGAP2C-injected, which was higher compared to controls (Dunn’s Benjamini-Hochberg adjusted p-values: Hom_parent= 0.0113, G₀-knockouts= 0.0040, SRGAP2C-injected= 0.0040). Asterisks denote a p-value below 0.05.