An AAV-CRISPR/Cas9 strategy for gene editing across divergent rodent species: Targeting neural oxytocin receptors as a proof of concept

A major issue in neuroscience is the poor translatability of research results from preclinical studies in animals to clinical outcomes. Comparative neuroscience can overcome this barrier by studying multiple species to differentiate between species-specific and general mechanisms of neural circuit functioning. Targeted manipulation of neural circuits often depends on genetic dissection, and use of this technique has been restricted to only a few model species, limiting its application in comparative research. However, ongoing advances in genomics make genetic dissection attainable in a growing number of species. To demonstrate the potential of comparative gene editing approaches, we developed a viral-mediated CRISPR/Cas9 strategy that is predicted to target the oxytocin receptor (Oxtr) gene in >80 rodent species. This strategy specifically reduced OXTR levels in all evaluated species (n = 6) without causing gross neuronal toxicity. Thus, we show that CRISPR/Cas9-based tools can function in multiple species simultaneously. Thereby, we hope to encourage comparative gene editing and improve the translatability of neuroscientific research.


INTRODUCTION
Nature provides an abundant variety of organisms using neural systems to interact with their environment to ensure survival and maximize fitness in diverse ways (1). Comparative neuroscience takes advantage of this diversity among species to find general principles of neural circuit functioning as well as mechanisms giving rise to variation in neural function and behavior (2). One efficient method to investigate the functional role of neural circuits is genetic dissection, a family of molecular approaches that allows for the manipulation of targeted genes (e.g., gene knockdown or exogenous expression) in specific cell populations (3)(4)(5). Traditionally, genetic dissection has been tractable in only a handful of model species, most commonly in laboratory mice, and this has limited our ability to leverage this powerful comparative approach to help improve treatments for neurobehavioral psychiatric disorders (6). However, ongoing efforts to construct whole-genome assemblies of nontraditional model species (7), along with the recent development of efficient genome editing tools (8,9), which can be used in conjunction with versatile viral vectors (10), have made it increasingly feasible to use genetic dissection strategies in a wide range of species (11)(12)(13)(14)(15). We believe that the application of genetic dissection approaches across a range of species has great fundamental and translational value, as it will greatly enhance our ability to take advantage of nature's remarkable diversity to bolster our understanding of species-specific and general mechanisms of complex brainbehavior relationships. Here, we demonstrate the feasibility of designing viral-mediated CRISPR/Cas9 tools for targeted gene editing across a wide range of model organisms. While we focus on neural oxytocin receptors (OXTRs) in rodents, this strategy should be effective for targeting any gene in any tissue across model organisms.
As a proof of concept, we developed a strategy for disrupting OXTR signaling in a wide range of rodents, as diverse members of this order are used in OXT research, particularly in relation to social behavior (16)(17)(18)(19)(20)(21)(22). OXT is a highly conserved peptide that has been studied in many species (14,(23)(24)(25)(26)(27), and this research has highlighted its role in the regulation of numerous social behaviors, including parental care, social bonding, mate preference, social recognition, and social vigilance (28,29). The OXT system also has translational potential as a target for psychiatric disorders with disruptions in the social domain (30)(31)(32). A remarkable feature of the oxytocinergic system is that Oxt expression patterns are highly conserved across species, with Oxt being expressed primarily in paraventricular and supraoptic hypothalamic nuclei across vertebrates (33). This is in sharp contrast to its receptor (Oxtr), which shows strong intra-and interspecific variation even among closely related species (28,29,34,35). Variation in OXTR density in various brain regions affects the processing of social information, which is thought to contribute to diversity in social behaviors (20,28,36). To be able to explore OXTR function from a comparative perspective, tools are needed to enable the genetic dissection of OXTR signaling in multiple species. While systemic Oxtr knockout lines exist for a few species (13,14,23,37), a single, cross-species tool to robustly disrupt functional OXTR signaling using targeted gene disruption in a spatially and temporally controlled manner in multiple rodent species has not been available. The development of such a tool is valuable to the comparative neuroscience field as it negates the need for extensive validation in each individual species. It will facilitate the applicability of gene editing in species that are studied for their interesting behavior, but for which genetic tools are not readily available, and thereby diversify the investigation of behaviors that are known to be modulated by OXT, but are not easily studied in mice (e.g., biparental care, pair bonding, partner loss, and female aggression) (15,20,35,38).
Viral-mediated CRISPR/Cas9 genome editing has emerged as an efficient and versatile strategy to induce indels (insertions/deletions) in protein-coding sequences and perturb associated functional protein levels (11). We adopted an adeno-associated virus (AAV)-based strategy to deliver the CRISPR/Cas9 components to adult neural tissue (10) and reduce OXTR protein levels in vivo. The system consists of a Cas9 endonuclease from Streptococcus pyogenes (spCas9) that induces double-strand breaks in the DNA, and a guide RNA (gRNA), whose specificity determines the genomic location of Cas9-mediated mutagenesis. By targeting regions in the Oxtr coding sequence that are conserved across rodent species, we demonstrated the efficacy of this AAV-CRISPR/Cas9 vector at reducing OXTR density in six rodent species used in behavioral neuroendocrine research, and we predict that this approach will be effective in more than 80 rodent species. Furthermore, we demonstrated selectivity of the AAV-CRISPR/Cas9 approach by assessing its effect on the expression of a close match in the rodent genome, the arginine vasopressin receptor 1A (Avpr1a). Thereby, this tool will facilitate the comparative study of OXT signaling and advance our understanding of general principles of this ancient neuropeptide's function as well its role in modulating a diverse range of social behaviors in a species-specific manner. These results also provide a proof of principle for a strategy that can be widely applied to other genes in comparative neuroscience research.

Design of viral strategy to target Oxtr in multiple rodents
Our goal was to develop a viral vector-mediated tool to reduce OXTR density in brains across a wide range of rodent model species. We aimed to validate the efficacy of our tool using six model species that are used in OXT research: Acomys cahirinus (spiny mouse), Mesocricetus auratus (golden hamster), Microtus ochrogaster (prairie vole), Mus musculus (house mouse), Peromyscus californicus (California deer mouse), and Rattus norvegicus (Norway rat). We previously developed an AAV-CRISPR/Cas9 strategy using a gRNA targeting the prairie vole Oxtr to efficiently reduce functional OXTR in the brain (11,39); however, the gRNA target sequence was not conserved across rodent species. Here, we used a similar approach using gRNAs targeting Oxtr sequences that are conserved across rodent species. To identify conserved Oxtr coding sequences that were amenable for mutagenesis by AAV-CRISPR/Cas9, we used the ClustalW algorithm of the msa package in R/Bioconductor to align the Oxtr coding sequences of species with available RefSeq sequence data (golden hamster, house mouse, Norway rat, and prairie vole) (40). Two conserved regions were found to be of sufficient length for gRNA targeting (>19 nucleotides) and to contain a permissive protospacer adjacent motif (PAM) sequence (5′-NGG-3′) (Fig.  1A). The gRNA(ΔOXTR)s target sequences in or just before the second transmembrane domain.

Validation of the viral strategy in the prairie vole
We produced AAV-gRNA(ΔOXTR) vectors to target these two regions and first assessed their effectiveness and specificity in the prairie vole. AAV-gRNA(ΔOXTR) vectors were combined with AAV-Cas9 vectors and injected unilaterally in the nucleus accumbens (NAc), while the contralateral side received an AAV-CTRL mix (Fig. 1B). The gRNA(CTRL) targets a sequence of the bacterial LacZ gene, which is not expressed in vertebrates. The NAc expresses high levels of Oxtr in prairie voles and is thought to mediate monogamous pair bonding behaviors (41). Two weeks after surgery, AAV-CRISPR/Cas9-mediated genomic editing was validated through a T7 endonuclease assay (Fig. 1C). DNA was isolated from enhanced green fluorescent protein (eGFP)-positive tissue (Fig. 1D), and the AAV-CRISPR/Cas9-targeted Oxtr coding regions were polymerase chain reaction (PCR)-amplified. T7 endonuclease restriction was observed in PCR amplicons, which were of the expected length from AAV-ΔOXTR-infected tissue, but not in AAV-CTRL-infected tissue, indicating that specific mutagenesis had occurred in the Oxtr coding sequence. The digested fragments were of the expected size based on the position of the gRNA target sequence. To test whether these mutations translated into reduced OXTR protein levels, we performed I 125 -OVTA autoradiography on brain tissue ( Fig. 1, D and E). Both AAV-gRNA(ΔOXTR) vectors strongly reduced I 125 -OVTA binding in the NAc, indicating that the generated Oxtr indels disrupted functional OXTR production. Reduction in OXTR levels was not caused by neuronal cell death because AAV-ΔOXTR-injected and sham-injected hemispheres contained a similar number of cells that were positive for the neuronal marker NeuN. In addition, we performed RNAscope in situ hybridization and found faithful expression of spCas9 in AAV-injected hemispheres (N = 2) (Fig. 1G).

The viral strategy does not affect AVPR1A levels in the prairie vole
Then, we sought to assess potential off-target effects of the gRNA(ΔOXTR)s. The vasopressin receptor 1A (Avpr1a) and vasopressin receptor 1B (Avpr1b) genes are closely related to Oxtr and share high sequence homology (42). While both gRNA target sequences do not align to the prairie vole Avpr1b sequence [>10-base pair (bp) mismatches], the gRNA(ΔOXTR.2) target sequence has the highest sequence homology to Avpr1a (3-bp mismatches) of all genes, which makes it the most likely off-target gene in the prairie vole genome for this gRNA. The gRNA(ΔOXTR.1) target sequence differs from the prairie vole Avpr1a sequence by 8 bp. To functionally assess the specificity of the gRNAs, we used the injection strategy as described above and targeted a region in the prairie vole in which Avpr1a is expressed: the ventral pallidum. AVPR1A autoradiography revealed no disruption in AVPR1A levels in the targeted area, demonstrating the selectivity of our gRNA(ΔOXTR)s (Fig. 2).

Efficacy of the viral strategy in five more rodent species
Next, we tested the efficacy of AAV-gRNA(ΔOXTR.1) in five more rodent species (Fig. 3). As OXTR expression varies extensively among species, a range of brain areas was selected for injection based on the distribution of OXTR in each species (43)(44)(45). The NAc was targeted in spiny mice, the endopiriform cortex in golden hamsters, the ventromedial hypothalamus (VMH) in house mice, the lateral septum in California deer mice, and the central amygdala (CeA) in Norway rats. AAV-gRNA(ΔOXTR.1) significantly reduced OXTR levels across species and areas. In spiny mice, however, the efficacy was lower than in other species ( fig. S1). During the preparation of this manuscript, the genome assembly of spiny mice became available, and we found no PAM sequence next to the target sequence of gRNA(ΔOXTR.1), which could explain the decreased effectiveness of gRNA(ΔOXTR.1) in spiny mice. However, the gRNA(ΔOXTR.2) target sequence is located next to a PAM site, so we injected a second batch of animals with this vector and observed strongly reduced OXTR levels. In sum, for all species, we achieved a significant reduction in OXTR levels.

The viral strategy targets a wide range of rodent species
We subsequently used the blastn algorithm from the National Center for Biotechnology Information (NCBI) BLAST+ software suite to search available whole-genome data and identify species in which either of the two gRNA(ΔOXTR) target sequences aligns perfectly and is thus predicted to work (46). We first aligned all available rodent RefSeq Oxtr sequences (N = 30) and found gRNA(ΔOXTR.1) to be functional in 15 species and gRNA(ΔOXTR.2) to be functional in 23 species (Fig. 4). Next, we searched for perfect alignment of the gRNA target sequences in all publicly available rodent genomes (N = 231). In this search, we found gRNA(ΔOXTR.1) to perfectly align in 43 species, and gRNA(ΔOXTR.2) in 80 species. Together, these viral vectors are likely to function in at least 81 rodent species (Table 1). Of note is that most of the rodent genomes are not annotated (e.g., spiny mouse and California deer mouse), so a perfect alignment does not necessarily indicate that the Oxtr coding sequence is targeted, and could indicate genes that closely resemble Oxtr, such as the vasopressin receptor genes. However, if interested in using this tool in a species for which no annotated genome is available, one could determine the sequence homology of the regions surrounding the gRNA target sequence with other Oxtr sequences to ensure specific targeting of the Oxtr gene.

The viral strategy targets a wide range of nonrodent mammalian species
Last, we performed a BLAST search on all available RefSeq sequences in mammals and found gRNA target sequences in the Oxtr gene in many more mammalian species (Table 2). This search predicted that gRNA(ΔOXTR.1) specifically targets the Oxtr gene in five nonrodent mammalian species and that gRNA(ΔOXTR.2) is effective in 67 nonrodent mammals. This search further indicated that gRNA(ΔOXTR.2) targets vasopressin 1b (Avpr1b) in 127 nonrodent mammalian species (Table 3).

DISCUSSION
We here developed a specific, highly efficient tool to reduce functional OXTR densities in multiple rodent species used in sociobehavioral studies. We targeted the Oxtr coding sequence in five different brain regions across six rodent species and reduced OXTR levels in all cases. Injection of AAV-ΔOXTR did not result in gross neurotoxicity, as demonstrated by unaffected NeuN expression, nor did it reduce the protein density of the most probable off-target gene product, AVPR1A. Thereby, we validated this method for use in comparative OXT research.
Our tool represents a major advancement over other techniques that are used to genetically manipulate OXTR levels. It is more effective than short hairpin RNA (shRNA)-mediated knockdown (36) and more versatile and selective than systemic knockouts. The ability to reduce OXTR density in multiple species is of great value to OXT research, which has a strong comparative tradition. One of the guiding hypotheses that have emerged from comparative OXT research is that species-specific Oxtr expression patterns influence the regulation of social behaviors (29,33,47,48) and that OXTR in different circuits can exert different effects on behavior (49). For example, socially monogamous species of voles have higher densities of OXTR in the NAc than do nonmonogamous species, and manipulating OXTR density in that region affects pair bonding behaviors (20). OXTR signaling in the NAc is critical for pair bond formation and maintenance (38), by modulating the salience of social stimuli (50), while genetically mediated variation in OXTR in the NAc modulates pair bonding and the effects of early-life experience on pair bonding (34,35,51). In addition, OXTR levels tend to be high in brain areas that associate with the primary sensory modality of a species (52). In rodents, the olfactory system is densely populated with OXTR, while in primates the visual system has high levels of OXTR. Therefore, variation in OXTR distribution is thought to underlie differences in social strategies and contribute to diversity in social behaviors (53,54). In theory, future comparative genome editing strategies could facilitate testing of this hypothesis by allowing direct comparison of OXTR function across species.
A limitation of our strategy is that while it targets many species, it does not target all species. Part of this limitation stems from the evolvability of the target gene, as it will be easier to design comparative gene editing strategies to target conserved genes, rather than for genes of which many variants exist. For genes with abundant sequence diversity, only the use of multiple gRNAs can ensure the targeting of all gene variants. However, the relative ease with which gRNAs can be multiplexed in AAV vectors should make it possible to target a wide variety of species with one single strategy.

S C I E N C E A D VA N C E S | R E S E A R C H R E S O U R C E
It is also important to consider that while two-thirds of the mutated gene products are likely to be nonfunctional because of frameshift mutations, one-third of mutated gene products might retain some functionality (e.g., receptor dimerization) (55). While our tool induces a near-complete loss of a specific functionality of the OXT receptor (i.e., ligand binding), we cannot exclude the possibility that a small part of the mutated gene products retains some form of residual activity. However, the tool is easily adapted to target other protein domains for functional characterization. Perhaps the greatest challenge to the efficient applicability of comparative genome editing is the design of gRNAs that target the widest range of species with preservation of specificity and efficiency. To simplify the design of comparative gene editing, gRNA-specific computational tools will have to be developed.
This work demonstrates the feasibility of designing genetic tools that are functional in multiple species. Although our tool was designed to target rodent Oxtr coding sequences, it is predicted to target many nonrodent mammalian species as well. This shows that the design of genetic tools that work in multiple species is not limited to a single order but can be designed to target a much wider range of species. One of the main advantages of widely used neuroscience techniques, like chemo-and optogenetics, is that they can be used in a wide array of species (19,(56)(57)(58). This not only has given tremendous insight into the functioning of neural circuits but also has direct translational value, precisely because these techniques function in many species and thus allow the elucidation of general and species-specific principles of gene-brain-behavior relationships. Therefore, we believe that the comparative design of genetic tools will greatly enhance the translatability of future genetic techniques, both when used in research as well as in the clinic. In sum, we hope that this work will encourage the application of genetic dissection in comparative neuroscience and thereby advance our understanding of the general and species-specific principles of neural circuit functioning.

Mus spicilegus Yes Yes
Mus spretus Yes Yes

Chrysochloris asiatica No Yes
Cricetulus griseus Yes Yes

Globicephala melas No Yes
Grammomys surdaster Yes Yes

Mus musculus Yes Yes
Mus pahari Yes Yes

Mustela erminea No Yes
Myodes glareolus Yes Yes

Intracranial surgeries
For all species, anesthesia was induced by exposure to a 2 to 4% isoflurane/oxygen mix and maintained at 1 to 3%. Three daily doses of meloxicam or carprofen (2 to 5 mg/kg, depending on species) were administered after surgery. Using a stereotaxic apparatus, animals were unilaterally injected with a 1:1 mix of AAV9-RSV-Cas9 and AAV9-gRNA(ΔOXTR), while the contralateral side received a 1:1 mix of AAV9-RSV-Cas9 and AAV9-gRNA(CTRL). Stereotaxic coordinates and injected volumes are summarized in Table 4. . After 7 days, films were developed and imaged using an MCID core system (Interfocus Co., UK). Mean gray values of viral-targeted regions, corrected for background, were determined in ImageJ. I 125 -activity (disintegrations per minute) was calculated using an I 125 -standard and taken as a proxy for OXTR or AVPR1A density. Differences in OXTR or AVPR1A density were determined by comparing protein density levels in AAV-ΔOXTR-injected regions to protein density levels in contralateral AAV-CTRL-injected regions.

Immunohistochemistry
Animals were deeply anesthetized and transcardially perfused with PBS, followed by PBS supplemented with 4% paraformaldehyde. Brains were sectioned on a cryostat (Cryostar NX-70) at 40 μm and stored in cryoprotectant buffer at −20°C until use. Sections were thawed, washed in PBS, and blocked and permeabilized in PBS supplemented with 0.1% Tween (PBST) and 5% normal donkey serum for 1 hour at RT. Next, sections were incubated in PBST supplemented with 0.5% rabbit anti-NeuN (1:1000, AB104225, Abcam, UK) at 4°C overnight. After PBS washes, sections were incubated in anti-rabbit Alexa Fluor 568 antibodies (1:500, Molecular Probes, OR, USA) for 1 hour. Sections were mounted and coverslipped in Fluoromount-G containing 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific). Z-stacks with ×60 magnification of eGFP-injected regions and the corresponding sham-injected contralateral region were imaged using a Keyence microscope. Z-stacks were projected with maximal intensity, and NeuN-positive cells were manually counted in ImageJ by a blinded observer.

BLAST
We used the blastn algorithm in the NCBI BLAST+ software suite to search for perfect alignment of the target RNA sequences plus the permissive PAM sequence (5′-NGG-3′) to rodent (taxid 9989) and mammalian (taxid 40674) genomes to identify species in which our tool is predicted to be functional.

Supplementary Materials
This PDF file includes: