Adult zebrafish husbandry was approved by The University of Manchester Animal Welfare and Ethical Review Board, and all experiments were performed in accordance with U.K Home Office regulations (PPL: P132EB6D7). The zebrafish used in this study were raised and maintained at The University of Manchester Biological Services Unit, under standard conditions, with adults housed in mixed sex tanks with a recirculating water supply maintained at 28°C under a 14/10 hour light/dark cycle, as previously described (19). Fertilized eggs were collected following natural spawning, and incubated at 28°C in fresh E3 medium. The embryos were staged according to standard guidelines (20). After termination of the experiment, all embryos were killed prior to protected status [5 days post fertilization (dpf)] using a lethal dose of Tricaine Methanesulfonate (MS222) and freezing at -20°C.

A guide RNA (gRNA) with no predicted off-target sequences was designed for exon 4 of samhd1 using CHOPCHOP (21). gRNA incorporating this target sequence was generated from a polymerase chain reaction (PCR) amplification (Forward primer: 5’- TAATACGACTCACTATAGGCGTCACATTAAGCAGCTCGGGTTTTAGAGC-3’; Reverse primer: 5’-AAAAGCACCGACTCGGTGCCACTTTTTCAAG-3’) including the remaining sequence of S. pyogenes chimeric single gRNA through in vitro transcription using a HiScribe T7 Quick kit (New England Biolabs). Precipitation of the gRNA and synthesis of Cas9 RNA was performed as previously described (22). The resulting RNAs were mixed with 0.05% phenol red, 120 mM KCl and 20mM HEPES, ph7.0, and ~1 nl of the mix was injected into the yolk of wild type (WT) AB fertilized eggs at the one cell stage to produce F0 crispants. To identify founders, F0 adults were individually crossed with wild type (WT) AB animals and the resulting embryos were assessed for indels using Hyp188I restriction analysis. This process identified a founder animal harboring a 23bp deletion in exon 4 (samhd1 ^Δ23) that was predicted to lead to a frameshift. Following confirmation of germline transmission, the F0 samhd1 ^Δ23 was crossed again with WT AB fish to produce an F1 generation. Following F1 heterozygous mutant incrosses, we ascertained that homozygotes (samhd1 ^Δ23/Δ23) were viable and capable of breeding. Subsequent generations were genotyped following PCR and Sanger sequencing of fin clip genomic DNA, using primers that flanked the 23bp deletion (forward primer 5’-GTGTTTAATGACCCCATCCA-3’; reverse primer: 5’-CCTATGGAGTGCTCAAATCG-3’).

To generate the amino acid sequence, the zebrafish WT samhd1 transcript was obtained from Ensembl (23) and inputted into ExPASy for translation (24). For the samhd1^Δ23/Δ23 sequence, the 23 bp deletion determined by Sanger sequencing was removed from the WT transcript before translation. For this study, all experiments were performed on an F4/F5 generation and comparisons were made between age matched WT and homozygous mutants.

Total RNA was extracted from pooled groups of whole larvae (n=20) or dissected larval heads (n=30) at 5 days post-fertilization (dpf) using a standard TRIzol (Invitrogen) method. Complementary DNA (cDNA) was synthesized from 800ng RNA as previously described (18). Quantitative PCR (qPCR) was performed on a StepOne Plus Real Time PCR machine (Applied Biosystems). To assess samhd1 expression in whole larvae, cDNA was combined with power SYBR green mastermix (Applied Biosystems) and primers (Eurofins Genomics) that flanked the 23bp deletion (Forward: 5’-AGAACATCATCTGCCGCCGG-3’; Reverse: 5’-CCAGTTCCTTCGCCCAGTCC-3’). Primers targeting the housekeeping gene hprt1 were used to normalize expression (Forward: 5’-GGACTTCATCCTCAAGAG-3’; Reverse: 5’-GTTCTAGCAGCGTCTTCATCG-3’). To assess expression of ISGs (isg15 (18), isg12 (Dr03140917_g1) and stat1b (Dr03151121_m1)) and the cholesterol biosynthesis gene hmgcrb (Dr03128326_m1) in dissected larval heads, a Taqman (Life Technologies) protocol was performed, as previously described (18).

Anaesthetized WT and samhd1 ^Δ23/Δ23 embryos were fixed overnight in 4% paraformaldehyde (PFA; Alfa Aesar) at 2 dpf before washing in 1% PBS-Tritonx. To image the samples, 80% glycerol was added to the embryos which were placed into a nunc glass base dish 12 mm. The samples were individually imaged in a ventral orientation using a Leica M165FC light stereo microscope with DFC7000T camera, and processed using LAS-X v3.3.3.16958 software (Leica). To assess for microcephaly, the distance between the eyes was measured alongside full length of the embryo (μm). The head/body ratio was determined from these two values, and normalized to the average WT head/body ratio, to produce a microcephaly index (ratio). Measurements were obtained using Image J (version 1.52a). n=8 embryos per group, for each biological replicate, repeated 3 independent times.

At 2 dpf, PFA fixed WT and samhd1 ^Δ23/Δ23 embryos were stained using the Click- iT Plus TUNEL assay kit (Invitrogen) following the manufacturer guidelines. For imaging, samples were mounted in a lateral position in 80% glycerol on a nunc glass base dish 12 mm (Thermo Fisher) and imaged on a Leica M205 FA Stereo fluorescence microscope using a 5x/0.50 PlanAPO LWD objective, captured using a DFC 365FX camera and processed using LAS AF v3.1.0.8587 software (Leica). Images were analyzed using the manual cell counting software on Image J (version 1.52a). Analyses were performed on n= 6-8 embryos from 3 independent replicates.

Swimming was measured from 3 – 5 dpf in WT and samhd1 ^Δ23/Δ23 larvae using DanioVision camera chamber and ethovision XT software (Noldus, version 11), as previously described (25). Cumulative duration of movement was measured in 3 independent replicates, with n=22-24 larvae per replicate.

To assess for brain hemorrhages, WT and samhd1 ^Δ23/Δ23 embryos (n=20 per group) were treated with atorvastatin (ATV; Merck) at 32 hours post-fertilization (hpf), as described previously (25). To visualize bleeds, ATV treated embryos were stained from 54 hpf using an o-dianisidine (Sigma) protocol (26). The percentage of embryos with hemorrhages within each treatment/genotype group was determined from 3 independent replicates.

Whole blood from AGS patients with confirmed mutations in TREX1, RNASEH2A, RNASEH2B, RNASEH2C, SAMHD1, ADAR1 or IFIH1 was taken to allow for whole genome RNA sequencing, as described previously (27, 28). In addition, samples from non-AGS patient age-matched controls were obtained. Briefly, AGS patients were identified through either direct clinical contact with the appropriate physicians, or through a referral process. The use of patient materials has been approved from the Leeds (East) research ethics committee (reference number: 07/Q1206/7) and South Centre – Hampshire A research ethics committee (reference number 17/SC/0026), whilst consent was also obtained from parents of the affected patients.

The Paxgene (PreAnalytix) whole blood RNA extraction method for RNA sequencing has been described previously, and the AGS patient RNA sequencing data-sets have been published elsewhere (27–29). RNA concentration was assessed with a spectrophotometer (FLUOstar Omega, Labtech), and 1 μg of mRNA from each sample was diluted into 20 μl nuclease-free water. Subsequently, the quality and integrity of the RNA samples were assessed using a 2200 TapeStation (Agilent Technologies). A poly-A enrichment library was then generated using the TruSeq^® Stranded mRNA assay (Illumina), according to the manufacturer’s protocol, and 76 bp (+ indices) paired-end sequencing carried out on an Illumina HiSeq 4000. Raw sequencing output was demultiplexed (allowing one mismatch) and the Binary Base Call (BCL) sequence file format was converted to the text-based sequencing data file format (FASTQ) using Illumina’s bcl2fastq software, version 2.17.1.14. Low quality bases and adaptor sequences were trimmed using Trimmomatic, and reads aligned to the Genome Reference Consortium Human Build 37 (GRCh37) genome using the two-pass mode of the Spliced Transcripts Alignment to a Reference (STAR) aligner (v2.5.3a), as well as to the transcriptome according to the GENCODE v19 human gene annotation (downloaded from https://www.gencodegenes.org/human/release_19.html). For each patient sample, the RNA-Seq by Expectation Maximization (RSEM) software package (v1.3.0) was used to calculate gene expression values, in transcripts per million (TPM). Subsequently, fold change was determined from TPM values, and then Log transformed (Log2).

All data were analyzed using GraphPad Prism 8.1.2 (GraphPad software Inc). All zebrafish experiments, except from ATV treatment were analyzed using either an unpaired t test, or Mann Whitney test, depending on normality, established by the Shapiro Wilks normality test. ATV treatment was analyzed using a two-way ANOVA with sidak’s multiple comparisons test. AGS patient RNA seq data was analyzed using a one-way ANOVA with Dunnetts multiple comparisons test, comparing each AGS gene with the control group.

Following on from the samhd1 MO model, which successfully recapitulated aspects of the AGS5 phenotype, a stable mutant zebrafish line was subsequently generated to facilitate long term studies (18). To generate a stable samhd1 loss of function model, a gRNA was designed using the CHOPCHOP software, targeting exon 4 of the samhd1 gene ( Figure 1A ). A 23 bp deletion was confirmed by Sanger sequencing and PCR in the F1 generation ( Figures 1B, C ). To determine whether this mutation resulted in loss of transcript, RNA was harvested from WT and samhd1 ^Δ23/Δ23 larvae at 5 dpf for qPCR analysis of the samhd1 gene, where the mutants exhibited almost a complete loss in samhd1 expression compared to WT larvae (P<0.05) ( Figure 1D ). Translation analysis of genomic sequences highlighted that the samhd1 ^Δ23 mutation introduces a premature stop codon prior to the HD domain which is predicted to produce a truncated amino acid sequence ( Supplementary Figure 1 ).

One of the primary hematological hallmarks of AGS is the excessive upregulation of IFN-I and the subsequent ISG signature, which is often used as a clinical diagnostic tool (7, 27, 30). Therefore, in an attempt to phenocopy this observation, we measured ISG expression in larval heads ( Figure 2A ). Variable expression of isg12, isg15 and stat1b was observed in samhd1 ^Δ23/Δ23 larvae compared to the WTs. However, although variable, the samhd1 ^Δ23/Δ23 larvae appeared to display increased expression of each ISG, although this did not reach statistical significance. Expression analysis of other ISGs displayed similar variability (data not shown).

Severe neurological manifestations are another important clinical feature of AGS, with patients presenting with a range of CNS symptoms (7, 31). To characterize the initiation of neurological phenotypes in the zebrafish model, we first measured head size, as microcephaly is a common clinical manifestation of AGS (32–34). The distance between the eyes after ventral imaging of fixed embryos at 2 dpf was used to measure microcephaly ( Figure 2B ). Full body length was also measured to determine a head to body length ratio, also referred to as a microcephaly index, normalized to the mean of the WT group ( Figure 2C ). The microcephaly index was significantly reduced by 15% in samhd1 ^Δ23/Δ23 embryos (P<0.0001), thus indicating that loss of samhd1 is associated with smaller head size in zebrafish ( Figure 2D ).

Previous studies have suggested that microcephaly can be attributed to an increase in neuronal cell death (35). In an attempt to assess this, TUNEL staining was used to identify apoptotic cells in the head region. Manual counting of apoptotic cells in the indicated region ( Figure 2E ) revealed a significant increase in brain cell death in the samhd1 ^Δ23/Δ23 embryos (P<0.001) ( Figure 2F ).

Subsequently, we aimed to establish whether the microcephaly and increased brain cell death in samhd1 ^Δ23/Δ23 embryos also contributed to functional defects. From a clinical perspective, a wide range of motor disabilities are observed across all human AGS subtypes (7). Swimming behavior can be readily assessed in zebrafish larvae as a readout for motor dysfunction, as observed in other zebrafish models of disease (25). We tracked swimming behavior in the larvae between 3 - 5 dpf. No difference in movement was observed at either 3 dpf or 5 dpf (data not shown), however, a ~50% reduction in cumulative duration of movement was observed in the larvae at 4dpf (P<0.01), implying a transient physical deficit exists in the samhd1 ^Δ23/Δ23 model ( Figures 2G, H ).

We next tested for AGS5-related cerebrovasculopathy, as the samhd1 MO model developed spontaneous cerebral hemorrhages (18). At baseline conditions, only a small proportion (5%) of samhd1 ^Δ23/Δ23 embryos developed cerebral hemorrhages, whilst no hemorrhages were observed in the WT embryos.

Pharmacological inhibition of the 3-hydroxy-3-methylglutaryl-CoA reductase (hmgcr) pathway, using statins, has previously been shown to induce brain specific hemorrhages in zebrafish (25, 26). The hemorrhages arise due to a reduction in de novo cholesterol biosynthesis, resulting in lowered cholesterol essential for neurovascular integrity (26). As only a small percentage of embryos spontaneously hemorrhaged in the samhd1 ^Δ23/Δ23 model at baseline, we tested whether the mutants may be more susceptible to brain hemorrhage following exposure to low concentrations of ATV. Following water bath incubation at ~30 hpf, ATV-induced hemorrhages were consistently identified from ~52 hpf. The incidence of hemorrhages increased in a dose-dependent manner for both mutant and WT groups, however, there was a significant increase in cerebral hemorrhage frequencies in the samhd1 ^Δ23/Δ23 embryos at the lower doses of 0.25 μM (P<0.05) and 0.5 μM (P<0.05) ATV: 17% versus 38% and 63% versus 87% respectively ( Figures 3A, B ). We postulated that increased susceptibility to ATV-induced cerebral hemorrhages may indicate a deficiency in hmgcr expression in the homozygotes. Therefore, we measured hmgcr transcript by qPCR in larval heads and identified a ~50% reduction in samhd1 ^Δ23/Δ23 larvae compared to WT ( Figure 3C ). Together, these data suggest that the samhd1 ^Δ23/Δ23 embryos possess a genetically-induced susceptibility to cerebral hemorrhages that is associated with dysregulated cholesterol biosynthesis.

AGS5 patients frequently present with cerebrovascular disease (13–16). Furthermore, reductions in cholesterol have been associated with increased risk of hemorrhagic stroke in the general population (36–40). As such, we hypothesized that a reduction in cholesterol biosynthesis gene expression could be specifically associated with AGS5 and not observed across other AGS subtypes in which cerebrovasculopathy is not reported. To test this hypothesis and to validate the translational relevance of the samhd1 ^Δ23/Δ23 zebrafish observation, we assessed the expression of several genes encoding enzymes involved in the multiple arms of the cholesterol biosynthesis pathway ( Supplementary Figure 2 ) in existing RNA sequencing data obtained from AGS1-7 patient whole blood samples. This analysis revealed an overall increase in the expression of 9 of the 14 cholesterol biosynthesis genes measured (ACAT2, MVD, CYP51A, DHCR24, MSMO1, NSDHL, HSD17B, EBP, SC5D), compared to an age matched control group ( Figure 4 ). This pattern was found across all AGS subtypes, excluding AGS1 patients, where no differences were observed. Although directionally inverse to the zebrafish model, these observations implicate defective cholesterol biosynthesis in the pathophysiology of AGS.