Foundations of an Ovine Model of Fragile X Syndrome

Abstract

Background: Fragile X Syndrome (FXS) is an X-linked neurodevelopmental disorder characterised by intellectual disability, developmental delays, anxiety, and social and behavioural challenges. Currently, no effective treatments exist to address the root cause of FXS. Mouse models are the most widely used for studying molecular pathogenesis and conducting preclinical treatment testing. However, therapeutic interventions that show promise in rodent models have yet to succeed in clinical trials. After evaluating the current models, we have developed an ovine model to address this clinical translation gap. We expect this model to more accurately reflect the human condition in brain size, structure, and neurodevelopmental trajectory. We aim to establish this model as a valuable preclinical platform for testing therapies for FXS. Methods: To generate the sheep model, we used CRISPR-Cas9 dual-guide editing to knock out the Fragile X Messenger Ribonucleoprotein 1 (FMR1) gene in ovine embryos. Results: Two founder animals were created, one ram (male) and one ewe (female), both of which carried FMR1 gene knockouts. The ewe carries inactivating mutations on both alleles, with the edits in both animals resulting in no detectable Fragile X Messenger Ribonucleoprotein (FMRP) as expected. Both founders have undergone molecular characterisation and basic health checks, with the female founder showing increased joint flexibility, a characteristic of FXS. The ram has been used for breeding, with the successful transmission of the edited allele to his offspring. Importantly, specific lamb cohorts for postnatal treatment testing can be produced efficiently utilising accelerated breeding methods and preimplantation selection.

1. Introduction

Fragile X Syndrome (FXS) is a neurodevelopmental condition characterised by intellectual disability, developmental delay, anxiety, and social and behavioural challenges. Autism Spectrum Disorder (ASD) is also very common in FXS, with FMR1 being the leading monogenic cause of ASD [1,2,3]. FXS presents with considerable variation in the symptoms and degree of severity. Some individuals are mildly affected, while others can be profoundly affected and require full-time, lifelong care. Most FXS cases are caused by an expanded CGG repeat (>200) in the 5′ UTR of the FMR1 gene, located on the X chromosome [4,5]. An expansion of the repeat to over 200 units can result in the repeat and promoter becoming hypermethylated, resulting in reduced or completely inactivated transcription and a consequential lack of FMRP production. There are cases of individuals with FXS without a CGG repeat expansion (approximately 1% of cases) [6] carrying FMR1 loss of function mutations. These cases prove that FXS can be functionally modelled by FMR1 gene inactivation.

Accurate animal models can play critical roles in biomedical research, allowing the exploration of pathology and disease mechanisms and for use in preclinical drug testing. Despite the extensive use of rodent models of FXS for treatment testing, current therapies for FXS primarily focus on symptomatic management, with no treatments approved that target the underlying causal pathology.

This paper reviews the most commonly used FXS mammalian models and the rationale for a new large animal model. It describes the initial development of a sheep model of FXS, including the molecular characterisation of the two founder animals and successful transmission of the CRISPR-induced FMR1 knockout allele.

1.1. Mouse Models of FXS

Due to their small size, low maintenance costs, and established protocols for genetic manipulation, mice are often the first species used to model genetic conditions. The human and mouse FMR1 genes have a 95% DNA sequence identity, and the FMRP has a 97% protein identity [7]. The murine FMR1 gene typically contains 9–11 CGG repeats at the site of the human expansion, far fewer repeats than seen in the population average for unaffected humans (~29–30 repeats) [8].

The optimal FXS model would replicate the exact molecular cause observed in the majority of cases: the expansion of the CGG repeat, resulting in the hypermethylation and transcriptional silencing of FMR1.

Various KI models of the FMR1 repeat have been made [9,10,11] including the recent CRISPR/Cas9-mediated model carrying 341 repeats [11]. However, an analysis of the animals also revealed no evidence of hypermethylation-based FMR1 transcriptional silencing. Unfortunately, the KI mouse models do not show the FXS phenotypes because of the lack of silencing. Despite this, these models are valuable for investigating germline and somatic repeat instability and the effects of the premutation length repeat. From these results, it is apparent that modelling FXS requires a reduction in or total loss of FMRP.

The most commonly used FXS models are simple FMR1 gene knockouts. The first line reported in 1994 was made by The Dutch-Belgium Fragile X Consortium [12]. The model was engineered by inserting a cassette into exon five of the murine FMR1 gene, resulting in an FMR1 KO allele expressing no full-length FMR1 mRNA and no FMRP detected in the testes, kidneys, liver, and brain [12]. However, because residual levels of mutant mRNA were expressed in this KO mouse, another KO line, known as KO2, was generated. This line was engineered by deleting both the FMR1 promoter and exon one [13]. The KO2 model produces 0.09% of WT levels of FMR1 mRNA and no detectable FMRP in the brain lysate. Both KO models have been widely used for behavioural and therapeutic testing. However, the original KO model is still the most commonly used [14].

The phenotypic characterisation of the FMR1 KO mice has been extensive due to their widespread use and commercial availability. The FMR1 KO mice recapitulate many FXS phenotypes, including physical, molecular, cognitive, and behavioural [12,15,16,17,18,19,20,21,22,23,24] Some phenotypes were consistently observed across studies, regardless of genetic background. These phenotypes include macro-orchidism, hyperactivity, dendritic spine abnormalities, and increased seizure susceptibility (Table 1). The KO mouse model also replicated molecular phenotypes seen in individuals with FXS, including increased basal protein synthesis in multiple brain regions, increased long-term synaptic depression (LTD), and dendritic spine abnormalities [15,16,17].

However, there are phenotypic inconsistencies between studies utilising mouse KO models [23]. These are predominantly variable in behavioural or cognitive phenotypes. Table 1 summarises a selection of results from mouse FMR1 gene knockout studies, demonstrating both between-study and strain similarities and differences in some crucial FXS phenotypes. For example, some studies have shown that KO mice exhibit impaired cognitive function, learning deficits, and social and communication anomalies, all of which are symptoms of FXS. However, other studies show no difference in these important characteristics (Table 1).

Some of the between-study inconsistencies can be explained by background or strain genetics [25]. The FMR1 KO mice were initially made on the C57BL/6 strain; subsequently, KO mice have been crossbred with different strains to investigate the effects of genetic background. Spencer et al. examined five genetic backgrounds crossed with the original FXS B6 mice and showed that multiple FXS behaviours were dependent on certain background strains. Others have supported these findings; for example, Dobkin et al. showed a strain-specific difference in spatial memory between the FVB and C57BL/6 lines [19]. Pietropaolo et al. also showed a difference in the pre-pulse inhibition (PPI) response, a measure of sensory gating in C57BL/6 and FVB, and the presence of repetitive actions in C57BL/6 [21]. Rather concerningly, contradictory results have also been reported between studies ostensibly using the same strain of mice for a small number of phenotypes. An example of this is PPI, which has been reported to be enhanced, impaired, or unchanged across studies, all of which used the C57BL/6J strain (Table 1).

A meta-analysis of FXS mouse studies attributed some of the variability between studies to the age of the mice used, housing, and the testing methodology [14]. Due to the large number of experimental variables in cognitive testing, the differences across studies may be due to the difficulty of controlling the testing methodologies. The inconsistencies in behavioural outcomes in mice, compared to more robust molecular and physical phenotypes such as altered LTD, increased basal protein synthesis, and macro-orchidism, suggest that non-behavioural biomarkers of treatment may be more robust measures of the treatment outcome.

The need for non-behavioural outcomes was also exemplified by the KO rat model. These rats showed macro-orchidism, attention deficits, and repetitive behaviours [26,27]. Consistent with the FXS KO mouse model, KO rats showed an altered LTD and increased protein synthesis [28]. However, unlike the FMR1 KO mouse model and people with FXS, the rats do not have a hyperactive phenotype [27]. Although hyperactivity is a common phenotype in humans, it may confound rodent behavioural tests as many cognitive and behavioural tests for rodents involve the animal’s movement between regions or chambers [29].

Collectively, the results from rodent studies have been instrumental in advancing our understanding of FXS at a mechanistic level and have revealed potential new therapeutic targets. For example, clinical trials of the mGluR antagonists, such as Fenobam, were based on the mechanism described by the mGluR theory [30]. Unfortunately, therapeutic success in rodent trials has not always translated into success in humans. More than 70 FXS preclinical animal model drug trials reported between 2005 and 2017 indicated promise in rescuing FXS phenotypes. Of these, 63 (~90%) of the studies led to a clinical trial; however, no treatment for FXS has been approved [31]. These disappointing results suggest that an alternative or complementary animal model may be useful for preclinical testing. This failure of animal model success to clinical translation is not restricted to FXS but is common for many neurological disorders.

Table 1.

Knockout mouse model phenotypes. The table shows nine studies characterising The Dutch–Belgium Fragile X consortium (1994) FMR1 KO mouse. Phenotypes that showed no difference between the KO mouse and the WT littermates are highlighted in purple (=). Tick (✓) indicates the presence of the named phenotype (green). An upward arrow (↑) indicates an increase or enhancement in the phenotype (green). Downward arrow (↓) indicates a reduced phenotype (gold). The blank spaces indicate that the traits were not tested in the listed papers. MWM = Morris water maze. PPI = pre-pulse inhibition. KO = knockout. WT = wild-type.

	Physical		Behavioural			Neurological		Learning and Memory
Mouse Strain	Macro- Orchidism	Dendritic Spine Abnormalities	Hyperactivity/ Anxiety	Seizure Susceptibility	Startle Response	PPI	MWM	Passive Avoidance Test	Fear Conditioning	Reference
C57BL/6	✓		↑				=	=		[12]
C57BL/6	✓	✓		↑				=		[15]
C57BL/6		✓	↑		↓	↑				[18]
C57BL/6	✓						=		=	[19]
FVB	✓						↓		=
C57BL/6			↑	↑	=	↑		↓	↓	[20]
C57BL/6			↑		↓	↑				[21]
FVB			=		↓	↑
FVB	✓		↑					↓		[22]
C57BL/6/FVB cross	✓			↑	=	=	=			[23]
C57BL/6		✓				↓				[24]

One limitation of mouse models for a neuronal developmental disorder is that mice mature rapidly, with a fully developed adult brain at roughly 3 months of age, and the brain volume largely stabilises in both mice and rats by three weeks of age [32,33]. In the first month, mice mature an estimated ~150 times faster than humans over the same period [34]. The postnatal brain development phase in FXS patients is potentially a pivotal time for treatment, especially with reagents designed to modulate neuronal differentiation, growth, and synapse formation. Therefore, a model with a more extended postnatal brain development phase could have better utility for testing treatments intended to normalise brain development.

In 2021, a marmoset model of FXS was made by knocking out the endogenous FMR1 [35]. The editing of exon 3 resulted in six marmosets carrying small deletions in the FMR1 coding sequence, ranging from a single base pair to 21 bp. Five animals were confirmed not to produce FMRP in fibroblasts or the brain. The sixth animal was determined to be a mosaic and still produced detectable FMRP in fibroblasts. Interestingly, all five complete FMR1 knockouts died within eight days of birth. No other KO animal model of FXS has reported neonatal death. The authors attributed the deaths to a phenotype specific to the marmosets [35]. In theory, a non-human primate model would be ideal for FXS, as it would better model the complexity of the human brain than a rodent model. However, working with primates involves significant ethical and cost barriers. Reproduction in non-human primates is also challenging, requiring large numbers of breeding animals, especially for X-linked disorders, to generate a testing population.

1.2. Justification for an Ovine Model

Here, we present the initial steps towards the development of an FXS ovine model, describing two founder animals. We also recognise that a large mammalian model will not replace the relatively high-throughput screening of therapeutics in mice. An ovine model provides a good compromise between rodent and primate models, as they are a large animal comparable in weight to humans, with a complex brain and a lifespan of approximately 12 years or more. Sheep are easy to care for, are accustomed to being handled, require little specialised care, and are relatively cheap to maintain in a typical farming environment. In comparison to mice, sheep have an extended brain growth and developmental phase (7–10 months), which provides a significantly longer therapeutic window for testing treatments targeting brain development. The ovine brain weighs approximately 120 g in an adult and is morphologically similar to the human brain. Accelerated reproductive methods have been developed and refined for sheep, facilitating the rapid production of experimental populations. For example, female lambs can be super-ovulated at six to eight weeks of age, in a process called juvenile in vitro embryo production and embryo transfer (JIVET), delivering ~20 oocytes that can then be used for in vitro fertilisation and implantation. Preimplantation screening is available, as is artificial insemination with cryopreserved sperm, allowing for the purpose-designed generation of research cohorts that minimise the use of animals. Lastly, a practical advantage in modelling disease in larger animals is the simple access to relatively large quantities of blood and cerebrospinal fluid (CSF) from live animals, as well as large amounts of tissue at postmortem.

Herein, we describe our first steps towards generating an ovine model of Fragile X Syndrome. We report the production of two FMR1 KO founder animals, their molecular characterisation, and subsequent breeding, and discuss the next step for further population expansion and behavioural phenotyping.

2. Materials and Methods

2.1. CRISPR-Cas9 Guide RNA Design

RNA guides were designed using CRISPOR (version 4.97) against the Ovis aries Domestic Sheep (Ensembl 76 (Oar_v3.1)). CRISPR guide RNA Exon1_A (5′ CAGGGCGGACGAGAAGATGG 3′) and Exon1_B (5′ CTCCAATGGCGCTTTCTACA 3′) were designed to target exon one of the endogenous ovine FMR1 gene. Guide RNAs were synthesised by Integrated DNA Technologies (IDT, Singapore). Embryos were injected with RNP complex to produce animals as described in [36]. One departure from the described method was the use of Cas9 as a protein, rather than introducing an mRNA. The Cas9 protein was included at a final concentration of 50 ng/µL, coupled with a final concentration of each crRNA (RNA guide complexed with tracr RNA (IDT) at 1:1 ratio) at 15 ng/µL in the final microinjection mix.

2.2. Animal Work

Animals were created and housed at the South Australian Research and Development Institute (SARDI), South Australia. All animal work was conducted under #14/20 PIRSA Animal Ethics Committee. Where applicable, the study aligned with the ARRIVE 2.0 use of animals in research guidelines. Tail tips were taken from animals shortly after birth to be used for nucleic acid and protein lysate extraction.

2.3. FMRP Western Blot Analysis

Using the TissueLyser II (Qiagen Hilden, Germany,), ~30 mg from fresh frozen tail tissues was homogenised, and 30 μg of whole protein lysate was loaded in a 4–20% ExpressPlus™ PAGE Gels (GenScript, M42012, Singapore) for separation via electrophoresis. Following electrophoresis, samples were transferred onto an Amersham™ Protran™ 0.2 μM HC membrane (Amersham, Buckinghamshire, UK). Membranes were probed with antibodies, including Anti-FMRP (ab259335) at 1:2500, Actin (A000702) at 1:2500, Secondary mouse HRP-conjugated (A4461) at 1:5000, and Secondary rabbit HRP conjugated (A6154) at 1:5000, in 5% BSA in TBS-T. The membrane was incubated in ECL™ Prime Western-blotting detection reagent (Amersham, Buckinghamshire, UK) before visualisation on the Amersham Imager 600 (GE Healthcare, Buckinghamshire, UK).

2.4. Whole-Genome Sequencing (WGS) and RNASeq

DNA was extracted from lamb tail tissue using the boiling lysis method. A small sliver (<1 mm²) was cut from a clean surface of tissue and transferred to a tube with 50 µL of a solution of 34 mM DTT and 0.2 mg/mL Proteinase K (Novex, Waltham, MA, USA). Using cut-off 20 µL pipette tip, 5 µL of a water washed suspension of Chelex-100 beads (BioRad, Hercules, CA, USA) was added to each sample and the tubes were places in a Thermocycler at 56 °C for one hour, then heated to 99 °C for 10 min and cooled to 4 °C. After spinning briefly, 1 µL of 1:10 dilution of the DNA was used as the template for PCR.

The extracted tail DNA was sent to Macrogen, South Korea, for genome sequencing. Sequencing was performed to an average depth of 30x across the genome. Reads were aligned to the Oar_4.0 reference genome using BWA (Burrows Wheeler Aligner) (V0.7.17) [37] on New Zealand eScience Infrastructure (NeSI) and manually examined in Integrative Genome Viewer (IGV) (version 2.12.2).

Extracted RNA was sequenced at Macrogen. Data returned was mapped to Oar_rambouillet_v1.0 using Spliced Transcripts Alignment to a Reference (STAR) on NESI and visualised using the IGV.

For cDNA analysis, RNA extracted from tail tissue was reverse-transcribed using SuperScript™ III Reverse Transcriptase (Thermo Fisher, Waltham, MA, USA) according to the manufacturer’s instructions, then amplified with specific primers described in Section 2.7.

2.5. MiSeq Sequencing

PCR amplicons were generated using primers with Illumina-compatible MiSeq adaptors to incorporate the adapters into the amplicon. Amplicons were sequenced at Auckland Genomics (University of Auckland) on an Illumina MiSeq Genome Sequencer. Outputs were returned as fastq files. Reads were mapped to exon one of the OAR_4.0 reference genome using BWA within NeSI. Aligned reads were visualised in IGV.

2.6. CRISPR-Cas9 Off-Target Cleavage Analysis

Potential off-target regions predicted by the CRISPOR software (Version 4.97) were extracted from WGS and manually examined to determine if editing had occurred. Potential off-target sites with two or three mismatches and off-target sites with four mismatches that sit within exons were examined. Fifty-eight off-target sites fit these criteria (Supplementary Table S1).

2.7. PCR Amplification over the Target Region

PCR amplification of genome target edit regions was performed using the KAPA2G Robust Hotstart PCR kit (Sigma Aldrich, St. Louis, MO, USA), according to the manufacturer’s instructions, with 30 ng of gDNA template in a 25 μL reaction. Primer sequences targeting exon 1 of the FMR1 genomic DNA were gFWD (5′ CTCATTCCGACAGACGCTGG 3′) and gREV (5′ CCTCACCGGAAGTGAAACCG 3′) to produce a 472 bp product from wild-type (WT) DNA. For cDNA amplification of FMR1 exon 1, primers were cFWD (5′ GGTTTCACTTCCGGTGAGGG 3′) and cREV primers (5′ GGCAGTTGGTGCCTTTTCTG 3′) to produce a 662 bp product in WT. The 3305 bp product was amplified using the primers FX351_FWD (5′ ACCACATTCTCACAGTGGTCA 3′) and FX351_REV (5′ TCGGTTTCACTTCCGGTGAG 3′). The AMEL gene was amplified using AMEL_FWD (5′ CTCCATGACTCCAACCCAAC 3′) and AMEL_REV (5′ ACTTCTTCCCGCTTGGTCTT 3′).

3. Results

3.1. Generation of an FMR1 Knockout Ovine Model

In the ovine genome, FMR1 is located on the X chromosome. The homology between the human and ovine FMR1 coding sequences is high, with a 91.7% similarity. One difference between the human and ovine gene is the mutation site 5′ CGG repeat length. Sheep have four repeats, which appear stable across multiple ovine species (Poll Dorset, Rambouillet, Texel, and SA merino) (Supplementary Figure S1) compared to the polymorphic human repeat expansion.

CRISPR-Cas9 guide RNAs were designed to target exon one of the endogenous FMR1 gene. In theory, the targeting approach used dual guides to induce two cuts, thereby increasing the likelihood of knocking out the gene. The predicted cut sites of these two RNA guides are located 45 bp apart (Figure 1A). If both guides induce a double-strand break (DSB), the 45 bp encompassed by the cut sites will be lost, resulting in the loss of the ATG start codon and 88% of exon one (Figure 1A). The guides were introduced as part of the RNP complex (recombinant Cas9 protein and guide RNA) into SA merino embryos via microinjection into the pronucleus or cytoplasm. Embryos that developed to the blastocyst stage were vitrified, generally as pairs, for later implantation into surrogate ewes.

Figure 1.

Confirmation of FMR1 editing in founder sheep. (A) Schematic of the dual guide RNA placement on the ovine FMR1 to knock out the gene. Exon one is shown in green. CRISPR guides, Exon1_A and Exon1_B (blue), recruit Cas9 to induce double-stranded breaks ~45 bp apart in exon one. (B) Screenshot from Intergrative Genomics Viewer (IGV) showing WGS confirmed two edited alleles in the female founder in exon 1 of FMR1. In the IGV screenshot from top to bottom, the blue bar is the gene track, the sequence, the coverage histogram, and then the reads. One allele shows a 45 bp deletion, resulting in a 50% reduction in the coverage histogram. The other alleles carry two small indels at the cut site (shown in red) of the CRISPR guide RNAs. (C) Screenshot from IGV showing WGS from the male founder revealed a 2706 bp deletion with cut sites in (D,E) with the ATG highlighted in green. WGS = whole-genome sequencing.

Between two rounds of editing, 31 injected embryos were implanted into 17 ewes (14 as pairs and three as singletons). These implantations resulted in nine pregnant ewes carrying 14 lambs (four sets of twins and six singletons) at day 60 post-implantation scanning. At birth, a small tail tip sample was taken from the 13 live-born lambs (four ewes and nine rams). Along with the DNA for the initial genotyping, RNA and a protein lysate were obtained from the tail tissue of the founder animals and age-matched controls. The FMR1 exon one was PCR amplified using primers flanking the editing target region, and the amplicons were Sanger-sequenced. Editing within the FMR1 gene was confirmed in the ewe lamb, FX301. Sanger sequencing confirmed all other animals as wild-type at the FMR1-targeted locus. Surprisingly, the same PCR assay did not generate a DNA amplicon from FX351, a ram lamb. Further testing showed this animal had a larger deletion that removed the primer binding sites. Both FXS founder animals, FX301 and FX351, had normal birth weights and appeared to be healthy.

3.2. DNA Sequence Analysis of the FMR1 Target Site in Founder Animals

Both founders were whole-genome-sequenced (to 30×-coverage Illumina paired-end reads), and an analysis confirmed the Sanger sequencing results showing successful editing at the FMR1 exon one locus and, importantly, no indication of mosaicism. The female founder (FX301) carries two edited alleles, a 45 bp deletion (X:87,739,091–135) (Assembly ARS-UI_Ramb_v3.0) on one allele and two small indels (a one bp deletion (X:87,739,137) and a two bp deletion (X:87,739,091–092) at the two CRISPR-guide-RNA-mediated cut sites) on the other (Figure 1B). The one- and two-base deletions were consistently observed on the same Illumina reads, indicating they had occurred on the same allele.

The genome sequence revealed that the male founder (FX351) had a considerably larger FMR1 deletion of 2706 bp (X:87,739,137–87,736,432), removing all of FMR1 exon one and roughly half of the first intron (2654 bp) (Figure 1C).

3.3. FMR1 mRNA Detected in Female Founder but Not in Male Founder

In the female founder, the RNASeq of the tail-derived RNA showed that, despite the edits, both alleles are still transcribed to some extent in full-length transcripts (Figure 2A,B). Deep MiSeq sequencing of cDNA showed an approximate 40:60 ratio in favour of the 45 bp deleted allele. This indicates a slight bias in transcription or cDNA production towards the allele carrying the 45 bp deletion, in contrast to the approximate 50:50 allele ratio observed for the genomic DNA. It is difficult to determine whether there is an underlying biological cause or whether the preferential amplification of one allele occurred during PCR. There appeared to be no reduction in the FMR1 transcript levels in FX301. The RNASeq from FX301 showed ~74.6 transcripts per million (TPM) mapped to FMR1, compared with the age- and sex-matched control animal, which showed 41.67 TPM.

Figure 2.

FMR1 mRNA production in the female KO founder. (A) Screenshot of IGV alignment of RNASeq from tail-derived RNA of a control animal and the female founder; coverage over all exons (grey peaks) shows the presence of a full length transcript in both the founder and wild-type control. (B) Closer examination of FMR1 exon one (with ATG highlighted in green) reads confirms that, despite full-length transcripts, both edits present in the genome are detected in the RNA, showing both alleles are transcribed. (C) cDNA PCR amplification of FMR1 exon one and five, and AMEL gene in FX351 and control animals (343 and 347) and the PCR negative control. (i) shows amplification of the cDNA from FX351 and controls over the exon of FMR1. No detectable band was observed in FX351. (ii) shows amplification over cDNA exon five of FMR1, with no band detectable in FX351. Together, the absence of amplification indicates that FX351 is not producing FMR1 mRNA. The cDNA amplification over the AMEL gene shows successful cDNA conversion of FX351 and controls (iii). WT = wild-type.

In the male founder, the targeted PCR of exon one and exon five of FMR1 cDNA showed no amplification (Figure 2C), indicating no FMR1 mRNA production in the ram FX351 tail tissue. A lack of FMR1 mRNA was confirmed, and no transcripts were observed in the RNASeq (Supplementary Figure S2). Given that the exon one splice site was removed by the editing in FX351 but not in FX301, it was hypothesised that the lack of FMR1 mRNA was due to the loss of the splice site, resulting in the incomplete or irregular splicing of FMR1 mRNA, which would most likely be degraded in the nucleus.

3.4. No FMRP Detected in Founder Animals

A Western blot analysis of protein lysate extracted from the tail of both founder animals revealed, as expected based on the edits, no evidence of FMRP (Figure 3).

Figure 3.

Confirmation of the successful knockout of FMRP in FX301 (A) and FX351 (B) compared to age-matched controls using Western blotting. FMRP was detected in tail lysate (30 µg) with Abcam anti-FMRP (ab259335) and secondary (A6154) antibodies using ECL to visualise on Amersham. Nonspecific banding at 90 kDa and actin (42 kDa) (primary: A000702, secondary: A4461) act as internal controls. FMRP = Fragile X Messenger Ribonucleoprotein.

3.5. Genome Off-Target Analysis in Founder Animals

Potential off-target cut sites were identified during the guide design process using the CRISPOR software package. Off-target sites with two or three mismatches and off-target sites with four mismatches that sit within exons were examined. Fifty-eight off-target sites fit these criteria for the FMR1-targeting guides used (Supplementary Table S1). These sites were examined in the WGS of both founders to detect unintended or off-target editing. Both founders had edits in an intron of the MAP2K5 gene. CRISPOR identified the off-target site in MAP2K5 as having three mismatches with guide Exon_1_A, two at the 5′ end and one at position 11 (from the PAM end) (Supplementary Table S1). In FX301, the edits observed were two single-bp indels on separate alleles, one a deletion (7:14,588,675) and the other an insertion (7:14,588,673–4), one bp apart at the predicted cut site (Supplementary Figure S3B). FX351 WGS showed a heterozygous 15 bp deletion at the site (7:14,588,668–14,588,682) (Supplementary Figure S3A). The variants are 12 kb from the nearest exon, within a 30 kb intron. An examination of the species homology shows that the similarity between sheep and human MAP2K5 is low. In the last 50 kb of the gene where off-target editing occurs, the similarity is 56.9%. The relatively low level of similarity may indicate that this region is not highly conserved and, hence, unlikely to be critical to the gene’s function.

Surprisingly, an unintended MAP2K5 edit was also detected in FX295, one of six FXS control animals also assessed via PCR and Sanger sequencing. FX295 was previously confirmed as wild-type at the FMR1 locus.

A second off-target edit was found only in the ram FX351 in the PCMTD1 gene. This was observed as a heterozygous 13 bp deletion (9:36,746,536–3,674,549) (Supplementary Figure S4). The deletion occurred in the sixth and final exon of PCMTD1. CRISPOR identified this off-target site as having four mismatches between guide Exon_1_A and the genomic sequence, three at the 5′ end of the guide and one at five bp from the 5′ end (Supplementary Table S1). The off-target edit’s predicted effect is a frameshift, resulting in a premature stop. Assuming the alternate codons are used, the resultant protein would then lose 38 amino acids at the C terminus. There are no known conditions associated with PCMTD1 that are documented on OMIM or pathogenic variants linked to a condition in ClinVar [38,39]. It will be important to track the transmission of these unintended editing consequences to select for a line of animals that does not carry these variants.

3.6. Preliminary Animal Phenotyping

Both founders and controls from the same cohorts were assessed by a veterinarian, blinded to the animals’ genotypes, using standard physical examinations, blood counts, and biochemistry tests when the founders were 16 months (ewe) and 11 months (ram) old. The physical assessment included the condition of the heart, lungs, teeth, eyes, and feet (scored on a scale of one to five), and the live weight at the time of assessment. The full blood panel tests the animals’ general health and organ function, including Alkaline Phosphatase (ALP), Gamma-Glutamyl Transferase (GGT), Glutamate Dehydrogenase (GLDH), creatinine, Calcium, and Phosphate. Overall, the blood biochemistry results were normal. Any slight deviation from expected values was deemed most likely due to population variation or from an alternative cause, such as mild dehydration, rather than indicating a clinical problem. No differences between the FXS founders and their cohort mates was observed, and the animals were deemed healthy. In addition to standard health checks, all animals were assessed for joint flexibility (on a scale of one to five, where five is normal). The female founder, FX301, had noticeable joint looseness and received the lowest score (3/5) in her cohort and in any other animals tested at the same time (35 in total) for joint flexibility.

Natural mating between the male founder, FX351, and eight wild-type ewes resulted in the birth of nine lambs, demonstrating that he is fertile, with all four female offspring obligate carriers of his FMR1 KO allele. The successful and stable transmission of the edited KO allele was confirmed via PCR amplification (Figure 4). The female offspring were then super-ovulated at 6 weeks of age to obtain oocytes which have been artificially fertilised and vitrified. In total, 55 embryos are ready for implantation to produce a new population of animals carrying the FMR1 KO allele.

Figure 4.

Pedigree of the offspring of FX351 and WT females (circles) and the successful transmission of the FMR1 edit into the female offspring (blue). Electrophoresis gel shows the PCR amplification of DNA extracted from tail tissue over the edited exon one region of FMR1 confirmed female offspring to be heterozygous carriers of the deletion, producing two bands, the WT 3305 bp and a 599 bp band due to the deletion. Male offspring show a single WT band at 3305 bp inherited from their WT mothers. WT = wild-type.

4. Discussion

We describe the first large animal model of FXS in the development of two founder sheep, a ram and a ewe, carrying loss-of-function mutations in the FMR1 gene. A CRISPR-Cas9 approach was undertaken using a dual guide to disrupt the translation start (MET) codon and delete 45 bp from exon one. The founder ewe carried the expected deletion on one allele and one- and two-base-pair deletions at the guide-mediated cut sites on the other allele. This, in effect, inactivated both alleles, resulting in no FMRP evident in the tail tissue. The ram carried a single deletion of 2706 bp, cleaving one bp upstream of the Met codon and a deletion incorporating all of exon one and 2653 bp into intron one. The large deletion resulted in no FMR1 mRNA and, consequently, no FMRP as measured in the tail tissue.

Interestingly, the female founder (FX301) produced mutant FMR1 mRNA despite lacking detectable FMRP. A similar phenomenon was reported in the original FMR1 knockout mouse model, in which mutant FMR1 mRNA was present despite the absence of FMRP, leading to the development of the KO2 model [12,13]. In contrast to the KO mouse model, there did not appear to be any reduction in the levels of the mutant mRNA in the ovine model. Both alleles were shown to be transcribed, but no mutated or truncated protein was detected from the tail tissue in FX301. The mRNA may be undergoing nonsense-mediated decay (NMD); however, given that the mutant mRNA levels are near wild-type levels, this is unlikely. If NMD is occurring, it may be incomplete, allowing the detection of the mutant mRNA in RNASeq. Alternatively, the mRNA may be translated, and the resulting peptides may be degraded due to misfolding or non-functionality. The predicted effect based on the sequence for the two indel alleles is a frameshift resulting in a premature stop codon, then the resulting 24 amino acid peptide most likely be degraded. In the 45 bp deletion allele, two of the three nucleotides comprising the translational start codon are deleted. As a result, any protein produced from this allele would be translated out of frame and would not yield functional FMRP.

Both founders were whole-genome-sequenced (30× coverage) to evaluate the site of the FMR1 edit and examine the predicted potential off-target sites based on the sequence similarity to the CRISPR RNA guides. This analysis identified an unexpected off-target edit in both founder animals within the MAP2K5 and PCMTD1 genes. The identification of unintended genome edits, including in a wild-type animal at the FMR1 locus, highlights the methodological considerations relevant to genome editing in animal models. These observations indicate that, even with careful guide RNA design and off-target prediction, unanticipated editing events may occur. Such events could complicate the genotype–phenotype interpretation, particularly when the functional consequences of the variants are unknown or when the condition being modelled is especially susceptible to genetic modifiers. Accordingly, these findings suggest the importance of the comprehensive genomic characterisation of new models generated using gene-editing technologies, which are often under- or unreported, alongside standard phenotypic and behavioural characterisation. In the ovine FXS model, the off-target variants identified in founder animals were heterozygous and not located on the X chromosome (chromosomes 7 and 9), allowing their exclusion through selective breeding, thereby removing potential confounding phenotypic effects.

The two founder animals underwent a blinded veterinary examination and standardised blood analysis for animal health. FX301 was identified as having the loosest joints in a cohort of 35 animals of various ages and genotypes. In human FXS cases, double-jointed and hyperflexible joints are observed in just over half (approximately 57%) of FXS individuals [40]. This may be a physical manifestation of FXS in the ovine model. No behavioural or health concerns were observed in the founders, a considerable advantage from an animal husbandry perspective. Further work is required to determine whether the ovine model exhibits behavioural, cognitive, or neuroanatomical phenotypes of FXS.

The male founder carrying the large 2706 bp deletion removing exon 1 was successfully used to generate nine offspring with four obligate female carriers. The female offspring from these matings underwent JIVET to extract oocytes at six weeks of age. This technology enables the production of more animals within 6 months of the founders’ birth, a significant advantage in a large mammalian model. These oocytes were fertilised in vitro, resulting in 55 embryos that have been vitrified ready for implantation. From the 55 embryos, it is expected that 50% will be male and half of those will carry the KO allele. Based on the previous successful implantation rate to live birth ratio, we expect 6–7 FMR1 KO male offspring.

The model presented here establishes foundational structural and molecular features but does not yet demonstrate behavioural phenotypes or therapeutic responsiveness. The next step is to expand the population for phenotypic and behavioural characterisation, allowing for a comparison between the current models and the human condition. Once fully established and characterised, the ovine FXS model offers several translational advantages across key therapeutic domains. The model could improve dose scaling and the pharmacokinetic assessment, which are often challenging to scale directly from rodents to humans. Because sheep can be studied longitudinally from early life onward, they provide a practical platform for testing the developmental timing of interventions, which is difficult to examine in human cohorts and where rodents’ neurodevelopmental windows are too short. Finally, the compatibility of sheep with clinically used EEG and neuroimaging modalities supports biomarker development, which aligns directly with human diagnostic and outcome measures.

Male offspring (F2 generation) can be used for further characterisation and eventual treatment trials. It is known that passive behavioural phenotypes in sheep are often lost when kept in a mixed-genotype flock [41]. Once an FXS-only flock is established, they will be subjected to behavioural testing. Although not a standard model like rodents, existing protocols exist for testing cognitive ability, learning, and memory in ovine models through maze trials [42]. GPS tracking has been used on sheep and could be implemented to test for hyperactivity, circadian rhythm, sleep abnormalities, and altered social interaction [41], all of which mirror FXS phenotypes observed in humans and the mouse model. Furthermore, ovine brain activity has previously been monitored via EEG [43]. While not all of these measures may directly apply to FXS, they show the versatility and utility of ovine neurological models. In addition, a large animal imaging (positron emission tomography (PET), computerised tomography (CT), and Magnetic Resonance Imaging (MRI)) and a good laboratory-practice-approved facility to facilitate preclinical safety and efficacy studies is located near the farm. These resources will enable future therapeutic studies, such as intracranial AAV gene therapy, to incorporate electrophysiological and neuroimaging endpoints, which are essential for validating the model’s relevance to human FXS.

5. Conclusions

In summary, the development of an ovine model for FXS presents a valuable opportunity to bridge the translational gap between rodent models and effective human treatments. The similar body size of sheep and humans, and the comparatively larger ovine brain, position sheep favourably for the preclinical testing of gene-therapy-based FMR1 gene replacement. The current study focuses on genetic engineering, genomic characterisation, and early physical assessment, forming the groundwork for subsequent investigations into behavioural, electrophysiological, and imaging phenotypes. Ongoing and planned studies will address these critical dimensions of FXS pathology, ultimately determining the model’s suitability for translational research and therapeutic testing.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/genes17020152/s1, Figure S1: The FMR1 5’ CGG repeat across multiple breeds of Ovis aries; Figure S2: RNASeq from FX351 showing no FMR1 mRNA transcripts; Figure S3: Off-Target guide mediated editing site analysis in MAP2K5 in FX301 and FX351; Figure S4: Off-target analysis of PCMTD1 in FX301 and FX351; Table S1: Potential off-target sites for CRISPR-RNA guides targeting ovine FMR1 exon one.

Author Contributions

V.H.: writing—original draft, conceptualisation, investigation, and formal analysis. J.C.J.: writing—review and editing, and supervision. R.R.H. and K.H.: supervision. S.R.R.: investigation, project administration, and writing—review and editing. C.J.M.: investigation, and writing—review and editing. J.M.K.: investigation, and methodology. K.L.: data curation. P.J.V.: conceptualisation. R.G.S.: conceptualisation, supervision, writing—review and editing, funding acquisition, and project administration. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

All animal work was conducted under approval #14/20 (approved on 1 July 2020) by the Department for Primary Industries and Regions (PIRSA) Animal Ethics Committee.

Informed Consent Statement

Not applicable.

Data Availability Statement

The original contributions presented in this study are included in the article/Supplementary Materials. Further inquiries or data requests can be directed to the corresponding authors.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research was funded by CureKids (3914). V.H. was funded by the Health Research Council of New Zealand (HRC) (20/259).