CRISPR gRNA phenotypic screening in zebrafish reveals pro-regenerative genes in spinal cord injury

Abstract

Zebrafish exhibit robust regeneration following spinal cord injury, promoted by macrophages that control post-injury inflammation. However, the mechanistic basis of how macrophages regulate regeneration is poorly understood. To address this gap in understanding, we conducted a rapid in vivo phenotypic screen for macrophage-related genes that promote regeneration after spinal injury. We used acute injection of synthetic RNA Oligo CRISPR guide RNAs (sCrRNAs) that were pre-screened for high activity in vivo. Pre-screening of over 350 sCrRNAs allowed us to rapidly identify highly active sCrRNAs (up to half, abbreviated as haCRs) and to effectively target 30 potentially macrophage-related genes. Disruption of 10 of these genes impaired axonal regeneration following spinal cord injury. We selected 5 genes for further analysis and generated stable mutants using haCRs. Four of these mutants (tgfb1a, tgfb3, tnfa, sparc) retained the acute haCR phenotype, validating the approach. Mechanistically, tgfb1a haCR-injected and stable mutant zebrafish fail to resolve post-injury inflammation, indicated by prolonged presence of neutrophils and increased levels of il1b expression. Inhibition of Il-1β rescues the impaired axon regeneration in the tgfb1a mutant. Hence, our rapid and scalable screening approach has identified functional regulators of spinal cord regeneration, but can be applied to any biological function of interest.

Fig 1. In vivo pre-screening identifies highly active sCrRNAs.

A: Example gels to assess sCrRNA in vivo mutation rate by resistance to restriction enzyme digest (RFLP) are shown. These indicate > 90% mutation rate (top), medium mutation rate (middle) and no detectable mutation rate (bottom). Each lane is derived from one animal. The chart shows the distribution of > 350 sCrRNAs by in vivo mutation rate. B-D: Activities of individual sCrRNAs show weak correlation with predicted in silico efficiency using different prediction rules (see results). E: sCrRNA activities are not correlated with their relative position to the start codon in vivo.

Fig 2. Phenotypic screening reveals modifiers of spinal cord regeneration.

A: A schematic representation of the spinal cord regeneration assay. Percentages indicate expected proportions of larvae with an injury site bridged by axons in controls. B: Example images of unlesioned, non-bridged (star indicates gap of neuronal labeling) and bridged spinal cord (white arrow) are shown (lateral views). Scale bar = 50 μm. C: Results of spinal cord regeneration screen for all screened genes at 48 hpl are shown. Significant reductions in bridging, normalised to control lesioned animals, are observed for cst7 (p < 0.0001), sparc (p = 0.04), tgfb1a (p = 0.03), tgfb3 (p = 0.005), tnfa (p < 0.0001), ifngr1 (p = 0.0013,) hspd1 (p = 0.011), tbrg1 (p = 0.0494, serpinb1 (p = 0.0279), and mertk (p = 0.0195); * indicates significance at 48 hpl; # indicates significance at 24 hpl (see S2 Fig); number of larvae per experiment are indicated at the bottom of each bar. For dpm3 no viable larvae could be raised. A single sCrRNA targeting a key functional domain was used to target ctsd, abca7, sparc, clip3, abca1b, tnfa, tgfb1a and tgfb3. Two sCrRNAs were used to target all remaining genes. D: Mutant analysis confirms axonal phenotypes for sparc (p = 0.0189), tgfb1a (p = 0.019), tgfb3 (p = 0.043) and tnfa (p = 0.024), but not for cst7 (p = 0.079) at 48 hpl. The table compares the magnitude of effects between acute injection and in mutants. Fischer’s exact test was used for all analyses.

Fig 3. Loss of tgfb1a leads to prolonged inflammation.

A: Lateral views of lesion sites in larval zebrafish are shown with the indicated markers and experimental conditions at 48 hpl. B-C: Quantifications show that numbers of macrophages were not altered by injecting any of the indicated haCRs (B; one-way ANOVA with Bonferroni’s multiple comparison test; Theoretical power = 0.85 to see a similar increase as for neutrophils in C), neutrophils were increased in number in tgf1b haCRs injected animals (one-way ANOVA with Bonferroni’s multiple comparison test, P = 0.0006), but not in tgfb3 haCRs injected animals (p = 0.32). D: Animals injected with tgfb1a haCRs, but not those injected with tgfb3 haCRs (P = 0.36), displayed marked increases in il1b expression levels in the lesion site compared to lesioned controls at 48 hpl (one-way ANOVA with Bonferroni’s multiple comparison test, P = 0.0211). All transcript levels were normalized to uninjected, unlesioned controls. E-F: tgfb1a heterozygous (one-way ANOVA with Tukey’s multiple comparison test, P = 0.0497) and homozygous mutant animals (P = 0.039) show increased numbers of neutrophils, comparable to haCR-injected animals. G: Inhibition of Il-1β with YVAD rescued axonal bridging compared to the DMSO-treated control group in animals injected with tgfb1a haCRs (Fisher’s exact test * p <0.05, ** p<0.01). Numbers in B, C, F indicate numbers of animals; in D numbers of independent experiments. Error bars represent standard error of the mean (SEM). Scale bars = 50 μm.