AP-4 loss in CRISPR-edited zebrafish affects early embryo development

Abstract

Mutations in the heterotetrametric adaptor protein 4 (AP-4; ε/β4/μ4/σ4 subunits) membrane trafficking coat complex lead to complex neurological disorders characterized by spastic paraplegia, microcephaly, and intellectual disabilities. Understanding molecular mechanisms underlying these disorders continues to emerge with recent identification of an essential autophagy protein, ATG9A, as an AP-4 cargo. Significant progress has been made uncovering AP-4 function in cell culture and patient-derived cell lines, and ATG9A trafficking by AP-4 is considered a potential target for gene therapy approaches. In contrast, understanding how AP-4 trafficking affects development and function at the organismal level has long been hindered by loss of conserved AP-4 genes in key model systems (S. cerevisiae, C. elegans, D. melanogaster). However, zebrafish (Danio rerio) have retained AP-4 and can serve as an important model system for studying both the nervous system and overall development. We undertook gene editing in zebrafish using a CRISPR-ExoCas9 knockout system to determine how loss of single AP-4, or its accessory protein tepsin, genes affect embryo development 24 hours post-fertilization (hpf). Single gene-edited embryos display abnormal head morphology and neural necrosis. We further conducted the first exploration of how AP-4 single gene knockouts in zebrafish embryos affect expression levels and patterns of two autophagy genes, atg9a and map1lc3b. This work suggests zebrafish may be further adapted and developed as a tool to uncover AP-4 function in membrane trafficking and autophagy in the context of a model organism.

Introduction

Large multi-subunit coat protein complexes are essential for the movement of transmembrane protein cargoes and lipids to distinct membrane-bound organelles within cells. Disruption or loss of trafficking coat complexes is often severely detrimental to human health (reviewed Sanger et al., 2019). Mutations in the adaptor protein 4 (AP-4; ε/β4/μ4/σ4 subunits) heterotetrametric complex in humans result in a family of neurological disorders called AP-4-deficiency syndrome (Ebrahimi-Fakhari et al., 2020) characterized by spastic paraplegia, microcephaly, and intellectual disability (Abdollahpour et al., 2014; Abou Jamra et al., 2011; Bauer et al., 2012; Ebrahimi-Fakhari et al., 2018; Hardies et al., 2015; Jameel et al., 2014; Tessa et al., 2016).

AP-4 is recruited to the trans-Golgi network (TGN) by the small GTPase, Arf1 (Boehm et al., 2001), where it forms a coat and recognizes specific transmembrane protein cargoes (Burgos et al., 2010; Simmen et al., 2002; Yap et al., 2003). Molecular mechanisms underlying AP-4-deficiency syndrome in human patients have recently emerged with identification of new cargo proteins. Both ATG9A (Davies et al., 2018; Ivankovic et al., 2019; Mattera et al., 2017) and diacylglycerol lipase β, or DAGLB (Davies et al., 2022), are now established to be proteins sorted out of the TGN by AP-4. In addition, several accessory proteins have been identified in trafficking AP-4 cargoes (Davies et al., 2022, 2018; Mattera et al., 2020b). Accessory proteins in trafficking coats perform diverse functions regulating vesicle formation, scission, targeting, or uncoating (Merrifield and Kaksonen, 2014; Zouhar and Sauer, 2014). ATG9A export from the TGN and anterograde transport towards the cell periphery depends on RUSC1 and RUSC2 (Davies et al., 2018; Guardia et al., 2021). Regulation of ATG9A subcellular distribution is also partially regulated by Hook1 and Hook2, which coordinate retrograde trafficking as part of the FHF (FTS, Hook, and FHIP) complex (Mattera et al., 2020b). Tepsin was the first identified accessory protein in AP-4 coats (Borner et al., 2012), but its molecular role within the coat remains unclear. Tepsin is considered a member of the epsin family (Rosenthal et al., 1999) because it contains an Epsin N-Terminal Homology (ENTH) domain. However, multiple lines of evidence, including X-ray structural and biochemical data, suggest tepsin has functionally diverged from other epsin family members (Archuleta et al., 2017; Frazier et al., 2016; Mattera et al., 2015). While no pathogenic mutations have been identified, tepsin loss-of-function studies would greatly contribute to our understanding of AP-4 coat function and biology.

Emerging data on both AP-4 and its accessory proteins have significantly improved understanding of AP-4-deficiency syndrome at the cellular level, and loss of ATG9A export from the TGN can be used as a diagnostic marker of AP-4 related disorders in patient classification (Behne et al., 2020; Ebrahimi-Fakhari et al., 2021). ATG9A is an essential autophagy protein (Noda et al., 2000; Young et al., 2006) and has recently emerged as a lipid scramblase (Guardia et al., 2020; Maeda et al., 2020; Matoba et al., 2020) during early stages of autophagy (Orsi et al., 2012; Yamamoto et al., 2012). Experiments in mammalian cell culture suggest AP-4 impacts ATG9A sorting to the cell periphery (Davies et al., 2018; Mattera et al., 2020b), which links AP-4 trafficking indirectly to regulation of cellular homeostasis through the autophagy pathway (reviewed in Galluzzi et al., 2017). However, it remains unclear why the severe phenotypes associated with these disorders appear to be specific to the nervous system since AP-4 proteins are ubiquitously expressed.

Study of AP-4 trafficking pathways and deficiencies in organisms has been impacted by loss of conserved AP-4 genes across many common model organisms including budding and fission yeast (S. cerevisiae and S. pombe), worms (C. elegans), and fruit flies (D. melanogaster) (Hirst et al., 2013). Two knockout mouse models targeting the large subunits ε (gene: AP4E1) (de Pace et al., 2018) and β4 (gene: AP4B1) (Matsuda et al., 2008) have reported phenotypes linked to trafficking and autophagy defects, but mouse models have only partly recapitulated human disease. AP4B1 knockout mice present no gross motor or cognitive deficits. AP4E1 knockout mice display impaired performance in motor function tests and exhibit neuroanatomical abnormalities similar to those observed in human patients, though only partially model the severity of motor and cognitive phenotypes typical of AP-4-deficiency disorders. It remains very beneficial to establish a model system like zebrafish, because organisms having a shorter life span and well-established suite of tools will allow researchers to study a time-course of organismal development with a large sample size. One study has reported how a morpholino knockdown of ap4s1 gene (AP4 σ4 subunit) yields embryos with altered neural development including small head size, irregular eye development, and locomotor deficits (D’Amore et al., 2020).

In this study, we generated CRISPR gene-edited F0 knockout zebrafish containing single gene deletions of individual AP-4 (ap4e1, ap4b1, ap4m1, ap4s1) subunits or the accessory protein, tepsin (enthd2). We utilized an established CRISPR-ExoCas9 approach to produce larger deletions at an increased hit rate (Clements et al., 2017). We characterized phenotypes at 24 hours post-fertilization (hpf) and found all gene-edited embryos consistently display developmental defects characterized by abnormal head shape, enlarged yolk sacs, and neural necrosis. We monitored autophagy gene expression in knockout embryos using reverse transcription quantitative PCR (RT-qPCR) on two genes: atg9a, an essential scramblase, and map1lc3b, (LC3B; reviewed in Wesch et al., 2020), an autophagy marker used in cultured cells and organisms (Klionsky et al., 2021). CRISPR-edited tepsin knockout embryos showed a trend towards increased atg9a expression. Finally, in situ hybridization data suggest both genes display altered patterns early in zebrafish development. Together, these data suggest loss of single AP-4 or tepsin genes gives a consistent phenotype in zebrafish and suggests either AP-4 or tepsin loss may yield trafficking or autophagy defects.

Results

CRISPR-edited zebrafish lacking AP-4 or tepsin genes exhibit morphological defects

We investigated whether Danio rerio could be used as a model system for understanding and exploring AP-4 trafficking. We generated single gene knockout models using a modified CRISPR technology referred to as CRISPR-ExoCas9, in which fusion of an exonuclease (Exo1) to the Cas9 protein improves overall hit rate and produces larger deletions (summarized in Figure S1A) (Clements et al., 2017). This approach and overall ExoCas9 efficiency were confirmed by recapitulating pigment loss in the eye of tyrosinase CRISPR-edited knockout (KO) embryos (Figure S1B) in line with published data (Clements et al., 2017). This control also established injection trauma is not likely to explain phenotypes described below for single gene AP-4 or tepsin knockouts. We individually targeted genes corresponding to each AP-4 subunit (ap4b1, ap4e1, ap4m1, and ap4s1) and tepsin (enthd2). For clarity, we will describe knockout embryos using standard zebrafish gene nomenclature. Following injection of both ExoCas9 and small guide RNAs (sgRNAs), embryos were initially monitored over a period of three days post-fertilization (dpf), and recurring phenotypic variations from wild-type embryos of the same clutch were noted. A preliminary characterization revealed greater than 90% of injected embryos died (Table 1) by 3 dpf. Phenotypes were instead scored at 24 hours post-fertilization (hpf) in the F0 generation, and representative phenotypic classes for each gene knockout are shown in Figure S2).

Table 1: Preliminary characterization of AP-4 subunit and enthd2 single gene knockout phenotypes.

Total number of embryos injected with individual sgRNAs (N) and the number of surviving phenotypic embryos observed at 3 days post-fertilization (dpf).

Gene & sgRNA	Injected embryos (N)	Embryos surviving (3 dpf)
ap4b1 sgRNA-1	128	3
ap4b1 sgRNA-2	31	0
ap4e1 sgRNA-1	66	6
ap4m1 sgRNA-1	51	4
ap4s1 sgRNA-1	55	2
enthd2 sgRNA-1	28	2

A consensus phenotype emerged across ap4b1, ap4e1, ap4m1, ap4s1, and enthd2 single gene KO embryos. Each knockout model exhibited stunted head development, sometimes lacking a defined head region while still retaining a distinct tail region with somites (Figure 1; columns 1 & 2). At 24 hpf, single gene knockout embryos exhibited a global developmental delay with enlarged yolk sac and shorter length. These embryos contained fewer somites than expected at 24 hpf and looked more reminiscent of earlier developmental stages (e.g. 15-somite to 25-somite stages; (Kimmel et al., 1995). These phenotypes were not observed in uninjected wild-type embryos from the same clutch (Figure 1; column 3, “uninjected control”). Ap4b1, ap4e1, ap4m1, and ap4s1 KO embryos using two independent sgRNA targeting different sites produced similar phenotypic ratios (Table 2); approximately 50–60% of injected AP-4 gene knockout embryos displayed phenotypic traits characterized as intermediate or severe (Table 2). Three sgRNAs were utilized for enthd2 knockouts because one sgRNA (“sgRNA-2”, Table 2) demonstrated low hit rates and thus lower phenotypic ratios. However, enthd2 sgRNA-1 and sgRNA-3 independently gave hit rates near 30%, suggesting it may be more difficult to target tepsin than single AP-4 genes. The consistency across multiple guide RNAs and across all four AP-4 genes suggests these phenotypic effects arise from AP-4 loss during early development.

Fig. 1. AP-4 or tepsin gene loss causes irregular head size and neural necrosis in zebrafish.

Representative intermediate knockout phenotypes of each AP-4 subunit or tepsin gene at 24 h post-fertilization (hpf). Rows depict gene target, and columns depict injection condition, Knockout embryos display irregularly small heads with neural necrosis compared to the uninjected (wild-type) control. This phenotype is consistent across all guide RNAs for all five KO genes (N values; Table 1). Uninjected 24 hpf embryos are from the pair-matched clutch for each of the knockout embryos.

Table 2: Classification of AP-4 subunit and enthd2 single gene knockout embryos.

Percentages of embryos scored into each of three phenotypic scoring categories. Percentages calculated from total number of replicates (N) for each knockout type and rounded to the nearest whole percent

sgRNA	% None	% Intermediate	% Severe	Total embryos (N)
ap4b1 sgRNA 1	36	46	18	124
ap4b1 sgRNA 2	41	44	14	63
ap4e1 sgRNA 1	53	40	7	75
ap4e1 sgRNA 2	43	52	5	98
ap4m1 sgRNA 1	52	43	5	95
ap4m1 sgRNA 2	47	47	6	144
ap4s1 sgRNA 1	55	38	8	88
ap4s1 sgRNA 2	51	39	10	89
enthd2 sgRNA 1	72	24	4	82
enthd2 sgRNA 2	91	7	1	140
enthd2 sgRNA 3	68	31	2	124

We verified whether CRISPR-mediated gene editing was successful by sequencing embryos scored in the intermediate class (Figure 2; Figure S3). Representative reads for ap4b1 (Figure 2A, 2B), ap4e1 (Figure S3A, S3B), ap4m1 (Figure S3C, S3D), ap4s1 (Figure S3E, S3F), and enthd2 (Figure 2C, 2D) indicate CRISPR-edited embryos for each single gene exhibited nucleotide insertions, deletions, or mutations in regions surrounding the sgRNA target sites. Ap4b1 knockout occurred with the highest efficiency compared to other AP-4 genes (Figure 2A, percent reads), which is a notable result in the context of established patient mutations (see Discussion). Overall, sequencing data confirm successful CRISPR-mediated gene editing in zebrafish embryos for all genes, although tepsin (enthd2) knockouts exhibited lower efficiency.

Fig. 2. Sequencing validates successful ap4b1 and enthd2 single gene knockout zebrafish embryos.

Embryos displaying intermediate phenotypes (Fig. 1; Fig. S2) were analyzed by Amplicon EZ deep sequencing (Azenta). Percentage of reads containing insertions (green) or deletions (blue) in the sgRNA target site (gray box) are displayed for representative ap4b1 (A) and enthd2 (C) KO embryos. Sequencing data were obtained from a single representative 24 hpf embryo in the intermediate phenotype category. (B, D) Three different representative reads (R1, R2, R3) from sequencing of the ap4b1 (B) and enthd2 (D) KO embryos are shown together with the WT sequence (underlined target site highlighted in gray with nucleotide numbers marked). Both point mutations (yellow) and deletions (dashes) are observed.

Autophagy gene expression quantification in AP-4 zebrafish knockout models

We chose to follow up with two gene-edited knockout models to look at potential effects on autophagy genes. Published models of AP-4-deficiency syndrome in multiple cell lines focus on ATG9A export from the TGN to the cell periphery (Davies et al., 2018; Mattera et al., 2017). This suggests a link between AP-4 trafficking and autophagy, since ATG9A is both an essential autophagy protein (Noda et al., 2000; Young et al., 2006) and an AP-4 cargo. We selected ap4b1 KO embryos for follow-up, because these embryos consistently yielded the largest percentage demonstrating intermediate or severe phenotypes (Figure 1; Figure S2; Table 2). Human patients also contain mutations in the β4 subunit with higher frequency than other subunits (Ebrahimi-Fakhari et al., 2020; Gadbery et al., 2020). We also selected enthd2 KO embryos, because the function of tepsin within AP-4 coats remains elusive. We probed atg9a gene expression levels in both wild-type and knockout fish to establish whether ap4b1 or enthd2 loss affects atg9a expression. We also probed map1lc3b (LC3B). LC3B is a well-established autophagy marker in cultured cells and in model organisms from budding yeast to higher metazoans (reviewed in Klionsky et al., 2021) that influences autophagosome formation and expansion (Wesch et al., 2020; Xie et al., 2008).

We used reverse transcription quantitative PCR (RT-qPCR) to analyze atg9a gene expression in ap4b1 and enthd2 knockout embryos. We observed a trend towards elevated expression levels in tepsin (enthd2) knockout compared to wild-type embryos, but this result was not statistically significant likely due to variability across mosaic samples (Figure 3A). Atg9a expression in ap4b1 knockout embryos was no different from levels observed in wild-type embryos (Figure 3A). Map1lc3b expression levels in both ap4b1 and enthd2 knockout embryos were not statistically different from wild-type embryos (Figure 3B). In addition, we explored expression patterns of both autophagy genes using in situ hybridization. Wild-type atg9a (Figure S5A) or map1lc3b (Figure S5B) showed concentrated expression in the central nervous system, including the forebrain and notochord. There are no published reports of atg9a expression data in zebrafish embryos, but map1lc3b expression observed here is consistent with published data (B. Thisse & Thisse, 2004). Overall, we observed greater atg9a and map1lc3b staining in the tails of ap4b1 and enthd2 knockout embryos compared to wild-type embryos. While non-specific staining can be common in in situ hybridization experiments, we did not observe non-specific staining in wild-type embryos (Figure S5A, column 1). Additionally, knockout embryos show consistently darker staining for both atg9a and map1lc3b. This result partially supports the trend observed in RT-qPCR data indicating increased expression levels for atg9a. Overall, these data are worth further exploration, because they suggest both ATG9A and LC3B may either be slightly elevated or mis-localized during early development in zebrafish lacking AP-4 coat components.

Fig. 3. Autophagy gene expression levels in ap4b1 and enthd2 single gene knockout models.

Autophagy gene expression at 24 hpf was assayed by RT-qPCR of pooled wild-type, ap4b1 or enthd2 KO embryos. (A) Expression levels of atg9a were generally elevated in enthd2 knockout embryos though not statistically significant. (B) Map1lc3b expression levels were not significantly different from WT. Each data point represent a pool of approximately 40 embryos from 3 independent injections. Normalized expression values (determined in Bio-Rad Maestro) were calculated using a control gene (elfa). Data are presented as the mean ± SEM with p-value from one-way ANOVA with Dunnett’s multiple comparison post-hoc analysis.

Discussion

Summary.

Overall, this study suggests it is feasible to generate CRISPR-edited zebrafish lacking single genes associated with AP-4 coat components. Independently targeting single AP-4 subunit or tepsin (enthd2) genes yields embryos displaying similar gross morphological phenotypes at 24 hpf, including neural necrosis and stunted development. These models share some features with human patients, but they also exhibit limitations as tools for future study (discussed below).

Zebrafish models for studying AP-4 trafficking and deficiency.

This study aimed to establish whether zebrafish could serve as a suitable model system for studying AP-4 during development or trafficking. The observed phenotypes for zebrafish embryos lacking either single AP-4 or tepsin genes include neural necrosis, stunted head development, and a global developmental delay. These traits are somewhat reminiscent of the microcephaly and altered neuronal development that characterize AP-4-deficiency syndrome in human patients (Abdollahpour et al., 2014; Abou Jamra et al., 2011; Bauer et al., 2012; Ebrahimi-Fakhari et al., 2018; Hardies et al., 2015; Jameel et al., 2014; Tessa et al., 2016). The characterized phenotypes observed in gene-edited zebrafish are broadly consistent with published ap4s1 studies (D’Amore et al., 2020) in which the σ4 subunit was acutely depleted using morpholinos. Very few knockout embryos developed heads in this study, but in some cases, those with heads appeared to have irregular eye development (data not shown). Unfortunately, the very small sample size prevented assessment of eye development using the approach established for published ap4s1 knockdown embryos (D’Amore et al., 2020).

The similarities between phenotypes observed here (Figure 4A) when single AP-4 genes are genetically removed from zebrafish using CRISPR and human AP-4-deficiency syndrome may suggest both orthologs function broadly in similar trafficking and autophagy pathways (Figure 4B) established in mammalian cell culture (Davies et al., 2022, 2018; de Pace et al., 2018; Guardia et al., 2021; Ivankovic et al., 2019; Mattera et al., 2020b, 2017) and in human patients lacking AP-4 (Behne et al., 2020; Ebrahimi-Fakhari et al., 2021). Zebrafish are an established model organism for studying how proteins function in membrane trafficking (Cox et al., 2018; Edeling et al., 2009; Sarmah et al., 2010), as well as how protein function in the nervous system underlies the basis for complex disorders like hereditary spastic paraplegias (Naef et al., 2019). Since AP-4 genes are absent in other model organisms (Hirst et al., 2013), using zebrafish as a tool to study AP-4 effects on trafficking or autophagy may be beneficial in the context of uncovering AP-4 biology at the organismal level. However, substantial limitations exist. AP-4 single gene knockout embryos appear to have notably more severe phenotypes than those observed in human patients. It is possible that human embryos also exhibit high levels of lethality when AP-4 genes are lost or mutated early in human pregnancies, but we do not have a way to validate this. Few zebrafish embryos survive past 3 dpf, thus preventing observation at later developmental stages as well as generation of knockout lines. In addition, the mosaic nature inherent in F0 generation knockouts complicates a more thorough study of zebrafish physiology.

Fig. 4. Insight into AP-4 loss using knockout zebrafish models.

(A) AP-4 and enthd2 single gene knockout embryos display abnormal head development, enlarged yolk sacs, and shortened tails containing fewer somites. Autophagy gene expression patterns (represented in blue by atg9a) exhibit darker staining patterns in the zebrafish brain, notochord, and tail. (B) At the cellular level, AP-4 mediates ATG9A export from the TGN. ATG9A is required for generation and maintenance of autophagosomes, suggests links between AP-4 trafficking and autophagy. CRISPR-edited AP-4 single gene knockout zebrafish models suggest AP-4 loss may be linked to trafficking or autophagy defect early in development. (Created using Biorender).

AP-4 is considered an obligate heterotetramer, with all four subunits required for the protein complex to fold and carry out cellular function. However, data presented here indicate ap4b1 gene loss produced a higher percentage of severe phenotypes compared to other AP-4 genes. This trend is reminiscent of reported data highlighting genetic variation and pathogenicity observed in human populations and patients; published data report both predicted and observed pathogenic variants tend to cluster in the β4 and C-terminal μ4 subunits (Gadbery et al., 2020). This suggests the β4 subunit may be a genetic “hot spot” for viable mutations.

Unlike AP-4, tepsin loss has not been well-explored, particularly since there are no reported human patients with disease-causing mutations. The enthd2 knockout zebrafish presented here offers an early glimpse into the importance of tepsin in AP-4 coats. Overall, enthd2 knockout in this study resulted in less severe phenotypes than did AP-4 during early embryo development, but its loss nevertheless corresponded to a low frequency of intermediate and severe phenotypes displaying abnormal head development and neural necrosis. These data provide additional genetic evidence that tepsin and AP-4 occur in the same pathway, since tepsin loss in zebrafish phenocopies AP-4 loss, which is already well-established in cell culture (Borner et al., 2012; Davies et al., 2018; Frazier et al., 2016).

An emerging role for AP-4 and tepsin in autophagy?

Based on cell culture models, neurological symptoms observed in human patients are thought to arise partly from ATG9A mistrafficking (reviewed in Mattera, de Pace, et al., 2020). ATG9A protein expression is increased in patient-derived fibroblasts and other cell lines lacking AP-4 (Davies et al., 2018; Mattera et al., 2017). We explored whether expression differences occur at the transcriptional level in single gene knockout zebrafish models. Two independent pools of enthd2 knockout embryos exhibited higher atg9a expression levels, but variability in the third replicate resulted in a statistically insignificant result. This is worth further exploration, because these data were obtained from pools of enthd2 F0 generation embryos despite the enthd2 sgRNAs exhibiting relatively low hit rates. Low hit rates suggest these embryos are highly mosaic, which could confound accurate determination of atg9a expression levels. In spite of this complexity, enthd2 loss had a more pronounced effect on atg9a expression than did ap4b1 loss. The role of tepsin in AP-4 coats thus remains a particularly interesting area of investigation.

The RT-qPCR results (Figure 3) can be further considered alongside in situ hybridization data (Figure S5). Immunostaining for either autophagy gene (atg9a or map1lc3b) suggests altered expression patterns in ap4b1 or enthd2 single gene knockout embryos compared to wild-type embryos. Together these data indicate loss of AP-4 genes may globally dysregulate autophagy in zebrafish. Future work should explore whether these differences in expression patterns are explained by trafficking defects. This may be challenging to address directly in gene-edited fish. CRISPR-editing can give rise to a complete genetic loss, and surviving fish likely will find ways to compensate. Acute depletion strategies using knockdown approaches may be more appropriate for follow-up in some cases because embryos will survive and can be monitored at relevant developmental stages.

Materials and Methods

Zebrafish Care and Embryo Collection

Wild-type AB zebrafish were used for all experiments. Zebrafish were maintained on a 12h/12h light/dark cycle. At the end of the experiments, no longer than 3 dpf, embryos were euthanized. All procedures were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) at Vanderbilt University under protocol number M1800200–01 (PI: James Patton)

Reagents and Oligonucleotides

All chemicals were purchased from Sigma unless otherwise stated.

CRISPR injections

Zebrafish embryos were injected at the single cell stage (Gagnon et al., 2014) with modifications for design of single guide RNAs (sgRNAs) and use of ExoCas9 (Clements et al., 2017). The ExoCas9 construct (Clements et al., 2017) was prepared according to published protocol. Two sgRNA target sites were selected for each AP-4 gene, and three were selected for enthd2 (tepsin). All sgRNA target sequences are listed in Table 3. Oligonucleotides for sgRNAs were synthesized (Sigma). Final single guide RNAs were prepared according to published protocol (Clements et al., 2017).

Table 3.

sgRNA target site sequences.

sgRNA identifier	target sequence
ap4b1 sgRNA-1	GGTGAAAGCCAGTGCCACGG
ap4b1 sgRNA-2	GGACCCAGACCCAGTGGTGG
ap4e1 sgRNA-1	GG CATGGATGTAGCTGAAGG
ap4e1 sgRNA-2	GGCTGAGATGCCGCAGCTCG
ap4m1 sgRNA-1	GG GAAAGAATGAGATCTTTG
ap4m1 sgRNA-2	GAGGGGAATAACTGAAATGG
ap4s1 sgRNA-1	GGAGGAGGAATGGCCGCCCA
ap4s1 sgRNA-2	GGAAAGGGCGCATTGGGAAG
enthd2 sgRNA-1	GGTTGCGCCCTCTGCTGCTG
enthd2 sgRNA-2	GGCTGGGAGGAAACAGACAG
enthd2 sgRNA-3	GGCTGTCCTGGCTATCTCTTCG

Phenotype Identification and Scoring

Embryos were initially monitored twice daily following CRISPR injections. Embryos displaying divergent morphology were separated as “phenotypic” and surviving phenotypic embryos at 3 dpf were counted (Table 1). Subsequently, injected embryos were scored into three phenotypic classes under a dissecting microscope (Leica Zoom 2000) 24 hpf. Those with no apparent differences from wild-type embryos were scored into classification 1 (none). Those with abnormally small (or absent) heads and/or abnormally shaped eyes were scored into classification 2 (intermediate). Those with morphological abnormalities that prevented head and tail from being clearly distinguished were scored into classification 3 (severe). Representative embryos from each class are displayed in Figure S2.

For phenotypic characterization, live embryos were mounted 24 hpf in E3 medium (0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM Mg SO₄, 5% Methylene Blue) on a glass slide (Corning) with no coverslip and photographed under transmitted light on an EVOS FL inverted microscope (Thermo Fisher Scientific). Representative embryos were selected for imaging from total injected embryos (Table 2; N) Images were taken both in and out of the chorion. Knockout embryos were always imaged in conjunction with wild-type embryos from the same clutch.

Deep Sequencing

Representative embryos scored in the intermediate classification were chosen for each single gene knockout, and genomic DNA was isolated from individual embryos according to published protocol (Xing et al., 2014). PCR primers (Table 4) were designed to amplify the region surrounding each target site (Geneious primer design; product size <500 bp) in the genomic DNA template based on gene transcript data (Ensembl). Following amplification, samples were purified using a QIAquick PCR purification kit (Qiagen) and resuspended in 30 μL sterile water. DNA concentration was determined using a NanoDrop spectrophotometer (Thermo Fisher Scientific) and the samples were diluted to 20 ng/μL. Samples were sequenced using amplicon EZ sequencing (Azenta). INDEL graphs were generated using the Azenta bioinformatics pipeline.

Table 4.

Deep sequencing primers to amplify near sgRNA target sites.

Gene	5’ Forward Primer	3’ Reverse Primer
ap4b1	TCTTTGTACTGTGGTCTTCAGTGT	ACATGTTTCTCAGGGCCAGG
ap4e1	GCATGTGTCGACGCACTTAC	TGTTTTGGGAAGGCTCAGCA
ap4m1	AGCGTTTGGTCTTTGCATCT	GCACGGGGACAGTAATGGAA
ap4s1	GAAAACCACAACAGCCGCTC	AAAGCAACCGACGAGTTGTC
enthd2	CCACTGTGCCACCATCTTCA	GCCCACTGACTCATGGGAAA

RNA extraction

RNA extractions were performed on 40–45 embryos using TRIzol (Thermo Fisher Scientific) after the manufacturer protocol. The resulting pellets were resuspended in 100 μL of sterile water. Samples were visualized on an agarose gel, and RNA yield was quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific). Where indicated, complementary DNA (cDNA) was generated from RNA using M-MLV Reverse Transcriptase (Promega) according to manufacturer protocol.

In situ hybridization

RNA probes for in situ hybridization were designed and prepared (Table 5; Thisse and Thisse, 2007) utilizing the two-step amplification method (Hua et al., 2018) with cDNA from pooled embryos (described above). Published primers were ordered for the control gene, krt4 (Thisse et al., 2001). PCR products were purified using the QIAquick PCR purification kit (Qiagen).

Table 5.

In situ hybridization probe sequences.

Probe	Forward	Reverse	Forward T7	Reverse T7
map1lc3b	TCTGCCTACAACAAACGTGT	AGCTTCGTGTTTGGGTAGCC	AAGGGAGAGAAGCAACTGCC	GCGCATGCTAATACGACTCACTATAGGGACCAGCAGGAAGAAAGCCT
atg9a	CTTTCGGTTGCCTGTGTTGG	TGCATGAGCGACAGTTCTGT	AGGGAATCGCTTGGAGTTGG	GCGCATGCTATACGACTCACTAGGATGACCAGTTGGCGGACAT

In situ hybridization was modified after a published protocol (Thisse and Thisse, 2007). During the wash step (protocol step 40), embryos were washed twice for 30 minutes each in 0.2X Saline-Sodium Citrate (SSC) buffer (30 mM NaCl, 3 mM sodium citrate) and then washed for 5 minutes at room temperature with 1 mL 1X MAB buffer (100 mM maleic acid, 150 mM NaCl, pH adjusted to 7.5 with 10M NaOH) before being transferred directly to blocking buffer. Following this step, embryos were incubated overnight in 1:2000 anti-DIG Fab fragments (Sigma Aldrich 11093274910) in blocking buffer. Embryos were incubated at room temperature in the dark for 4–8 hours in 1 mL 1X MAB before moving directly to the alkaline buffer wash steps. Following staining, embryos were washed twice with PBST for 5 minutes each, followed by sterile water for 1 minute and stored for 5 days in a 50% benzylalcohol / 50% benzylbenzoate destaining solution. Representative embryos in the intermediate phenotypic category were mounted in glycerol on a bridge slide and imaged on a Nikon AZ 100M wide field microscope at 5X magnification.

Reverse transcription quantitative PCR

Reverse transcription quantitative PCR (RT-qPCR) was conducted using iTaq Universal SYBR Green Supermix (Bio-Rad) after manufacturer protocol using cDNA from pooled embryos (described above) and run on a CFX96 Thermocycler (Bio-Rad) in the Vanderbilt Cell and Developmental Biology Resource Core. Primers for atg9a and map1lc3b (Table 6) were validated by assessing the derivative of a 65–95°C melting curve (Figure S4A–C). Normalized expression values were calculated using Bio-Rad Maestro 1.1 software by the ΔΔCq method (defined by the software) with the published control gene, elfa (McCurley and Callard, 2008) as a benchmark for normalization (Figure S4D). Statistical significance was analyzed using one-way ANOVA with Dunnett’s multiple comparison post-hoc testing in GraphPad Prism 9 software.

Table 6.

qPCR primer sequences.

gene	Forward	Reverse
map1lc3b	AAGAGGTGCAGGCAAGGATC	CCAACACAGGCAACCGAAAG
atg9a	AAGGGAGAGAAGCAACTGCC	GACCAGCAGGAAGAAAGCCT