A framework for identifying fertility gene targets for mammalian pest control

Abstract

1. Fertility-targeted gene drives have been proposed as an ethical genetic approach for managing wild populations of vertebrate pests for public health and conservation benefit.

2. This manuscript introduces a framework to identify and evaluate target gene suitability based on biological gene function, gene expression, and results from mouse knockout models.

3. This framework identified 16 genes essential for male fertility and 12 genes important for female fertility that may be feasible targets for mammalian gene drives and other non-drive genetic pest control technology. Further, a comparative genomics analysis demonstrates the conservation of the identified genes across several globally significant invasive mammals.

4. In addition to providing important considerations for identifying candidate genes, our framework and the genes identified in this study may have utility in developing additional pest control tools such as wildlife contraceptives.

1. INTRODUCTION

Invasive vertebrates pose a major global threat to natural ecosystems (Norbury et al., 2013), economies (Latham et al., 2020), and human health (Tait et al., 2017). Pest control can either be achieved through lethal measures or birth control. Traditional physical and chemical pest control approaches typically rely on lethality, often in ways that raise welfare concerns (Littin et al., 2004; Kirk et al., 2020). Lethal control approaches also raise concerns about off-target impacts (Eason et al., 2002, 2011), and efficacy (e.g., resistance to control measures; Anderson et al., 2016^; Campbell et al., 2015). Thus, new pest control technology targeting the fertility of a population may gain a higher degree of social acceptance, particularly when combined with genetic engineering techniques (Kirk et al., 2020; Wilkinson & Fitzgerald 2006). This has motivated the development of novel molecular biocontrol approaches, such as meiotic drives that bias the inheritance of particular genes (Burt, 2003). There are several types of meiotic drives, including those that exist in nature (e.g. thaplotype sex-ratio distorter in mice; Hermann et al., 1999) and others that may be engineered such as homing (Kyrou et al., 2018) and toxin-antidote (Champer et al., 2020) gene drive systems.

Engineered homing gene drives harness guide RNA (gRNA) molecules to direct germline activated endonucleases (e.g., CRISPR-associated Cas9; Pfitzner et al., 2020) to sequences on the wildtype chromosome where they induce a double-strand break (DSB; McFarlane et al., 2018). For population eradication, a homing suppression gene drive can be designed to disrupt genes critical for survival or reproduction. With the drive allele serving as a repair template, homology-directed repair (HDR) allows the drive to be directly inserted into the target gene (Fig. 1A), resulting in both chromosomes and consequent gametes carrying an independent gene drive cassette. Alternatively, disruption through end-joining repair can introduce insertions and deletions at the DSB site to render the target gene non-functional (Fig. 1A), though such alleles often reduce the overall efficiency of the drive. If the transmission rate is sufficiently high, gene drives should spread to high frequency (Fig. 1B), resulting in eventual population collapse if the drive targets genes critical to fertility or fitness (Kyrou et al., 2018).

Figure 1.

(A) Two mechanisms that homing suppression gene drives use to disrupt genes critical for survival and reproduction: homology-directed repair (HDR; left) and disruption via end joining (EJ; right). In both systems, gRNAs initially target nucleases to a specific genomic locus to induce a double-strand break. If a repair template is provided, HDR leads to insertion of the entire drive construct, including the nuclease, gRNAs, and cargo gene (optional), into the targeted site on the homologous chromosome. Activation of end joining repair, an error prone process, will usually knock-out the gene. EJ repair may thus produce resistance alleles that either conserve or disrupt gene function. Resistance alleles that conserve gene function are problematic as they may prevent the success of future homing drive activity (due to mutations disrupting gRNA target site recognition). (B) Perfect gene drive transmission biases drive allele inheritance in offspring by 100%. Figure from Clark (2023) (CC BY 4.0).

The most important challenges of homing drive systems are the potential evolution of resistance and need for population containment. Resistance to the gene drive can evolve either through standing genetic variation (e.g. mutations in the gRNA target sequences), or drive-induced mutations (e.g. introduction of indels via end joining where HDR was required to transmit the gene drive construct to the homologous chromosome). Function-preserving resistance alleles are particularly concerning in the context of a suppression drive as such alleles can escape cleavage by the drive allele (Champer et al., 2018). Containment is another challenge for homing gene drives as even a small number of drive-carrying individuals are theoretically capable of spreading the drive to an entire population, even if there is a fitness cost to the drive carrier (Esvelt et al., 2014). This means that a drive would likely impact non-target populations, creating significant ecological, social-cultural and political challenges (Taitingfong et al., 2022).

Toxin-antidote systems are a promising method to localise drive spread and reduce the evolution of drive-induced resistance (Champer, Lee, et al., 2020; Champer, Kim, et al., 2020; Champer, Champer, et al., 2021). These drives work by cleaving a gene critical for survival or reproduction (representing the “toxin” portion of the drive), while providing a recoded but functionally equivalent version of the critical target gene (representing the “antidote”), which is linked to the drive construct. The toxin-antidote mechanism thus prevents the persistence of resistance alleles because only individuals that inherit the drive construct can successfully reproduce. Toxin-antidote drives additionally must reach a threshold frequency in the population before they are able to spread (e.g. 2-locus TA gene drives must reach ~20% frequency; Champer, Champer, et al., 2021). These high rates of initial release could act as a barrier to the drive escaping to non-target populations.

One challenge with toxin-antidote drives is that they can usually only be used for population suppression when used as a tethering system for a paired homing suppression drive (Dhole et al., 2019; Metzloff et al., 2022). One exception to this requires the use of haplolethal target genes, which poses substantial engineering difficulties (non-viability based on resistance allele formation and cell mosaicism caused by CRISPR cleavage during the transformation process (Champer, Yang, et al., 2020)). In either case, toxin-antidote drive versions with the most stringent confinement require relatively high rates of initial release (e.g. >61% initial introduction frequency required by a TADE underdominance suppression drive; Champer, Champer, et al., 2021). Eradication of ship rats from New Zealand using a drive system that requires even just a 5–10% initial inoculation rate would still require the breeding, engineering, and release of tens of thousands of drive carriers, which is an ecologically and economically unrealistic proposition, though it may be possible to release the drive in a limited area and allow it to spread over the landscape.

However, regardless of the type of suppression drive system, there is a research gap in the identification of biologically and ethically suitable candidate genes that could be used to regulate mammalian invasive species. Identifying suitable target fertility genes is also a critical step in investigating and understanding the complex molecular challenges that can arise during gene drive development (e.g., genomic resistance and off-target mutagenesis, and predicting successful in vivo application; Rode et al., 2019), as well as to inform parameterisation in predictive gene drive models.

Transgenic expression of sex-determining genes (e.g., Sry) has been discussed as a potential method to create a pool of infertile intersex phenotypes, reducing the proportion of fertile females in mammalian populations (Akiyama et al., 2002; Prowse et al., 2017). While promising in theory, experiments have proven challenging and feasibility concerns have been raised (Esvelt & Gemmell, 2017). A more practical approach is to identify and disrupt a functionally important gene region with a specific role in fertility. An ideal target would be a haplosufficient gene essential for fertility in one sex, such that the other sex can transmit the drive further. Targeting a haplosufficient gene essential for fertility in both sexes is not desirable, as such a drive is more likely to prematurely self-extinguish in a large population that is distributed across a heterogenous landscape, particularly if the species dispersal rate and population density is low (Champer, Kim, et al., 2021). Such a drive would instead need to have a reduced genetic load (e.g. targets a gene that only partially reduces fertility), which likely requires higher transmission efficiency to have sufficient power to eradicate the target population (e.g. higher dispersal propensity of drive carriers and high rates of perfect HDR (Fig. 1A)).

Here, we established rigorous criteria involving biological gene function, gene expression, and results from mouse gene knockout (KO) models to identify potential fertility target genes that could be biologically suitable (i.e., those with minimal known off-target effects) for rodent population management using gene drives. These criteria were subsequently used to identify 16 genes indispensable for male fertility and 12 for female fertility that are strong targets for a population suppression drive. We further demonstrated that several of these genes have sequence conservation, which may permit these genes to be investigated and targeted for the development of gene drive systems in other invasive mammalian species. The complexity of gene drive modelling and the expense of laboratory experimentation means that the use of this framework for rigorous prescreening of potential fertility targets will represent an efficiency multiplier for researchers seeking to develop gene drive technology for the management of invasive species.

2. MATERIALS AND METHODS

2.1. Gene identification and evaluation

2.1.1. Initial gene identification

We developed and applied a gene evaluation framework to systematically identify candidate fertility genes for mammalian gene drives (Appendix 1). Using the Online Mendelian Inheritance of Man (OMIM) database, we tested and refined keyword searches for genes relating to mammalian gametes or zygote-specific fertility. We used OMIM for initial gene identification because it is a comprehensive resource containing published literature on disease states and loss-of-function mutation causes of mammalian infertility (Hamosh et al., 2005). Izumo1, a gene previously identified as specifically essential for male fertility, was utilised as a positive control. The final optimised search yielded a list of 172 genes with the following keyword string:

(infertil* OR sterility OR sterile OR inviable OR inviability) AND (sperm OR ovum OR egg OR gamete OR zygote) AND (genetic OR gene OR allele) AND (human OR mammal OR mouse OR rat)

We also added a list of 19 genes identified as indispensable for fertility in mice by the Reproductive Genomics Program at the Jackson Laboratory for Mammalian Genetics (hereafter referred to as the Schimenti and Handel gene list; Schimenti and Handel 2018). Gene names, OMIM access number, phenotype, key reference and target sex were recorded. For uniformity, reported gene names and symbols are consistent with nomenclature approved by the HUGO Gene Nomenclature Committee for the gene ortholog in the respective species (Yates et al., 2017). Where there were doubts about the suitability of a gene (e.g., suggestive evidence of functional redundancy or conflicting phenotyping results), genes were excluded.

2.1.2. Inclusion and exclusion criteria

Following the initial gene identification, the gene list was further systematically checked against three criteria defined below in order to identify suitable gene targets for mammalian gene drives. Descriptions and key references for each gene that satisfied our inclusion criteria were further revised and updated in response to new literature.

1. Sex-specific fertility phenotype:

Candidate genes must affect gamete and pre-implantation phenotypes to limit instances of individuals carrying moribund embryos and ethical issues surrounding terminating embryonic development and the production of organisms with significant health defects. Evidence of a complete sex-biased fertility phenotype is required to permit the opposite sex to transmit the drive. Summaries of published literature provided by OMIM and additional peer-reviewed literature were reviewed to exclude genes leading to sub-fertility or minor fertility phenotypes in the non-targeted sex.

It is important to note that there are biological similarities between infertility and cancer pathologies, where components of both conditions result from failure to maintain control of somatic and germ cell proliferation (Nagirnaja et al., 2018; Tarín et al., 2015). In cancer, these associated genes are significantly overexpressed, so drives removing function and eliminating gene activity are unlikely to be associated with increased cancer susceptibility. Nevertheless, we excluded genes with evidence of tumour complications in knock-out mice, as this additional source of mortality could have implications on drive performance and ethical acceptability.

2. Tissue-specific expression pattern:

Gene expression (assessed via RNAseq data) was required to be restricted or highly overexpressed in specific reproductive tissue (testis or ovary), although minor expression elsewhere was permitted. This criterion reduces the potential for unanticipated phenotypic effects in off-target tissues, which could reduce the fitness of drive-propagating individuals and consequently limit the persistence and spread of the gene drive and may raise ethical concerns. Transcriptome data from the Mouse ENCODE Project via the National Centre for Biotechnology Information (NCBI) online database and Gene Expression Atlas from 23^rd September 2018 to 10th October 2018 was used to assess gene expression patterns in different adult tissues in relation to functional pathways, cell type, and developmental stage (Geer et al., 2010; Yue et al., 2014). To construct a quantitative threshold for expression outside the central reproductive tissue, we compared the gene expression results from Dmrt1 and Amh, two genes with strong effects on sexual phenotype (SI Appendix 2; Petryszak et al., 2016^; Soumillon et al., 2013). Based on this, genes with non-target tissue expression levels exceeding 5% of the expression levels in target reproductive tissue were excluded to reduce complications of pleiotropic genes.

3. Knock-out mouse model:

Raw data from in vivo mouse knock-out experiments for each gene of interest were compiled to validate gene expression profiles and support gene function predictions, particularly where it was difficult to isolate and quantify if gene activity in a specific tissue corresponded to an observable phenotype. In vitro data was also collected to support the phenotypic observations from in vivo experiments. To detect potential sub-fertility phenotypes and intermediate fecundity phenotypes segregating with heterozygotes for each gene, t-tests were employed to interrogate differences in litter sizes between genotypes (SI Appendix 3). We investigated this because drive-propagating individuals carrying a single null gene copy are required to be viable with no overt somatic developmental phenotypes to ensure subsequent drive inheritance. The histological impact of the gene knock-outs in specific reproductive tissue (e.g., reduced testis size) was noted. Still, only genes that failed to produce healthy knock-out (KO) offspring were excluded. The absence of potential somatic phenotypes, overt development, abnormal behaviour, and viability between genotypes was confirmed via the literature or through direct correspondence with the original studies’ authors.

2.1.3. Female fertility genes

A further 23 female candidate genes were identified by collaborators using the Mouse Genome Informatics (MGI) between August 2017 and April 2019. Various search terms specifically relating to “fertility defect” were used. Following initial identification, these genes were put through a series of filters, optimising for genes where homozygous mutant female mice displayed near or complete sterility, while heterozygous females and carrier males displayed mostly wild-type fertility. Other exclusion criteria included parturition defects (death of pups during childbirth) and abnormal somatic phenotypes resulting in a significantly reduced fitness of that genotype. When compared to our gene list, many of these female fertility genes had been excluded due to broad somatic expression profiles (i.e., failing our ‘tissue-specific expression pattern’ criteria). To examine the impact of ‘relaxing’ this specific criteria, we instead evaluated these 23 female fertility genes based on female-specific, non-syndromic sterility and the availability of a mouse KO model, ultimately resulting in a list of female target loci (hereafter referred to as the female candidate gene list).

2.2. Evolutionary conservation

Invasive vertebrates exist in many terrestrial ecosystems. However, the majority are non-model organisms for which a reference genome has only recently been assembled (stoat and brushtail possum; unpublished data, D. Bond) or are yet to be sequenced. Where reference sequences are available, conservation within coding regions in well-studied mammalian fertility genes may indicate functional conservation that can be further interrogated once population datasets are available (i.e. identification of variation related to species divergence, and investigation of sites that are near fixed in the population of interest, which may enable population-specific gRNA design for multiple species). Identifying conserved genes would also provide a starting point for empirical experiments, and subsequent tailoring of drive type to the requirements of a particular species. Targeting the same, well characterised gene in multiple species will also be useful to reduce overall resource expenditure for drive development across a number of vertebrate pests. As an island nation with multiple pest species, we use Aotearoa New Zealand (NZ) as an example system to explore the utility of our framework for potentially assisting with gene drive development in the future.

2.2.1. Sequence extraction and processing

Genome assemblies and corresponding gene feature files (GFF) of nine important invasive species in NZ: house mouse (Mus musculus), Norway rat (Rattus norvegicus), ship rat (Rattus rattus), rabbit (Oryctolagus cuniculus), western European hedgehog (Erinaceus europaeus), domestic cat (Felis catus), domestic ferret (Mustela putorius furo), stoat (Mustela erminea), and common brushtail possum (Trichosurus vulpecula), were retrieved from NCBI on 5th November 2020 (SI Appendix 4). Genomic coordinates for each Gnomon-predicted coding sequence (CDS) region within our identified candidate genes were extracted from the GFFs for each species (SI Appendix 5; Souvorov et al., 2010). The extracted coordinates were converted to zero-based BED formatted coordinates using BEDTools (v2.29.2; Quinlan 2014) and used to extract the corresponding nucleotide sequence. As it is optimal for gRNA multiplexing to have multiple adjacent target sites, CDS sequences less than 80 nt were omitted, to allow an optimal minimum of 3 gRNAs of 20 nt each spaced 10 nt apart (CRISPR HDR-based gene drives exhibit optimum drive conversion efficiency with 2–4 gRNAs, and there must be a sufficient number of gRNAs to avoid functional resistance alleles; Champer et al., 2018). Where genes were not present in their respective GFF, a manual search of NCBI for synonymous gene names was conducted to verify the absence of the gene ortholog from the respective target species or to determine the species-specific gene name. Subsequently, the gene feature coordinates for the Zpbp ortholog in the brushtail possum were manually added to our dataset.

2.2.2. Multiple sequence alignment

We used LASTZ (1.04.03; Harris 2007) to perform pairwise alignments of coding regions within our candidate genes of interest (Fig. 2; SI Appendix 6). Gap opening was allowed to capture insertions and deletions that may result from species divergence, and the default seed pattern was used. Nucleotide variation was limited to five mismatches per 20 nt during the gapped extension of seed hits to high-scoring pairs (HSP). Back-end filtering of the resulting pairwise alignment blocks set a minimum sequence identity of 75%, consistent with the tolerance of CRISPR-Cas9 gRNAs to no more than five nucleotide mismatches per 20 nt gRNA target sequence (Zheng et al., 2017).

Figure 2.

The sequence of steps used to generate and process pairwise alignments, which were subsequently processed into the multiple alignment format. Gene feature files were used to extract coding sequences from whole-genome assemblies. LASTZ (1.04.03) was then used to generate pairwise alignments with at least 75% identity and to limit mismatches to 5 per 20 nt. MULTIZ (v10.6) then filtered the resulting output to eliminate alignments less than 80 nt, and generated the final multiple species local alignments.

The subsequent alignments were then processed by MULTIZ (v10.6; sourced from https://github.com/multiz/multiz.git) to generate a multiple species local alignment of sequence blocks of more than 80 nt (SI Appendix 7). Each conserved sequence block was visualised using tidyverse (v1.3.0; SI Appendix 8; Wickham et al., 2019). Gap openings, including the ends of the alignments, were processed as mismatch penalties, due to the impact of a “missing sequence” on the efficacy of gRNAs designed for these sequence blocks.

3. RESULTS

3.1. Gene identification

3.1.1. Evaluation and characterisation

Our initial OMIM gene search generated a list of 172 results (SI Appendix 9), to which we added 19 genes from the Schimenti and Handel list to broaden the scope of our investigation (SI Appendix 10; Schimenti & Handel 2018). From the total of 191 results, eight were found to be duplicates (e.g., alternative allelic variants). These entries were merged with their respective genes, generating a list of 183 genes. Application of our filtering criteria identified that 103 genes demonstrated incomplete evidence of a sex-specific phenotype and were subsequently excluded. Sixteen of the remaining 80 genes are specifically critical for male fertility, contributing to several important prefertilisation processes, including gametogenesis, gamete recognition, and chromatin remodelling (Table 1; see SI Appendix 10 and 11 for excluded genes). Gene names are consistent with the Mus musculus orthologs reported by HGNC. However, variation in nomenclature usage across bodies of literature and databases is evident.

Table 1.

The 16 genes essential for male fertility identified via our framework. Gene name, OMIM access number, a brief functional description, chromosome number in relation to house mouse (Mus musculus; GRCm39: RefSeq assembly accession = GCF_000001635.27) and key functional reference are shown. References: 1. Geyer et al., 2009; 2. Ren & Xia, 2010; 3. Pierre et al., 2012; 4. Miki et al., 2004; 5. Inoue et al., 2005; 6. Zhang et al., 2009; 7. Tokuhiro et al., 2009; 8. Liu et al., 2020, Miyata et al., 2015; 9. Rode et al., 2012; 10. Fujihara et al., 2012; 11. Shao et al., 1999; 12. Zheng et al., 2007; 13. Shang et al., 2017; 14. Spiridonov et al., 2005; 15. Lin et al., 2007.

Gene	OMIM	Phenotype	Chr	Ref
Capza3	608722	Testis-specific actin-capping protein that controls actin polymerisation during spermiogenesis. Expressed in spermatids (Shima et al., 2004).	6	1
Catsper1	606389	Two of four primary subunits in a sperm-specific Ca2+ ion channel located in the flagellar, required for the hyperactivation of sperm. Expression is restricted to testes, specifically in developing spermatids, however, expression is not detected in mature sperm (Carlson et al., 2005).	19	2
Catsper2	607249	“   “   “   “   “   “	2	2
Dpy19l2	613893	Spermatid structure between acrosomal and nuclear membranes. Expression is specific to spermatids.	9	3
Gapdhs	610712	Enzyme localising to the principal piece of the flagellar and involved in energy supply for sperm propulsion. Expression restricted to spermatogenesis.	7	4
Izumo1	609278	Gamete recognition protein on the sperm cell surface that interacts with the oocyte Izumo1r receptor.	7	5
Meig1	614174	Involved in protein recruitment for sperm flagellar formation.	2	6
Oaz3	605138	Mammalian antizyme localising to developing spermatids that inhibits polyamine uptake and ornithine decarboxylase activity. Expressed in haploid germ cells. Deficiency results in a separated head and tail of sperm.	3	7
Ppp3r2	613821	Protein phosphatase regulatory subunit that is expressed throughout sperm development. Deficiency results in inflexible sperm midpiece, sperm motility and morphological defects.	4	8
Slc26a8	608480	Chloride/sulfate exchange in germ cells and required for sperm motility. Proteins localize in spermatocyte membranes and naive spermatids.	17	9
Spaca1	612739	Transmembrane protein localising in the equatorial segment of mature sperm. Expression associated with the inner and outer acrosomal membranes.	4	10
Spag4	603038	Works with KASH proteins to couple sperm nuclear structures to the cytoskeleton. Expressed in haploid male germ cells.	2	11
Spem1	615116	Plays a role in detaching cytoplasm from the spermatid nucleus and flagellum neck region required for straightening and stretching of the sperm head and neck. Expressed late in spermatid development.	11	12
Sun5	613942	Located in the inner membrane of the sperm nuclear envelope and is the most basal protein for flagellar anchoring. First expressed during spermiogenesis (haploid spermatids; Shang et al., 2017).	2	13
Tssk6	610712	Serine/threonine protein kinase that is required for post-meiotic chromatin remodelling. Expression in spermatids, not spermatocytes (Jha et al., 2017). Another member of this family, Tssk3, has been demonstrated to produce a more dramatic phenotype in KO mice (Nayyab et al., 2021).	8	14
Zpbp	608498	Secondary binding between acrosome-reacted sperm and the zona pellucida. Expressed in spermiogenesis (haploid stages of sperm development). Heterozygous males have an intermediate phenotype (increased levels of abnormal sperm, however, fecundity is comparable to WT sires).	11	15

3.1.2. Evaluation of the female fertility genes

The externally contributed list of 23 genes with female-specific phenotypes was evaluated using mouse knock out that demonstrated evidence of a sex-biased sterility phenotype – no tissue-specific expression criteria were applied (see Methods). Eight genes were found to demonstrate either a subfertile or potentially lethal phenotype and were subsequently removed from further analysis. Many of the remaining 12 genes (Table 2) have functional roles in maternal effect, gamete recognition, cell signalling, and gene regulatory processes at various stages before embryonic implantation. Two of these genes, Afp and Pgr, emerged as strong candidates for HDR-based gene drives, despite expression patterns outside the germline (Table 2).

Table 2.

The 12 genes essential for female fertility identified through our evaluation. Asterisks label two genes with somatic expression, which are therefore strong candidates for HDR-based gene drives controlled by a germline promoter. Gene name, OMIM accession number, a brief functional description, chromosome number with relation to house mouse (Mus musculus; GRCm39: RefSeq assembly accession = GCF_000001635.27) and key functional reference are shown. References: 1: Gabant et al., 2002; 2: Mulac-Jericevic et al., 2000; 3. Tsai et al., 2008; 4. Soyal et al., 2000; 5. Dong et al., 1996; 6. Bianchi et al., 2014; 7. Tong et al., 2000; 8. Rajkovic et al., 2004; 9. Yurttas et al., 2008; 10. Yu et al., 2014; 11. Liu et al., 2017.

Gene	OMIM	Phenotype	Chr	Ref
Afp*	104150	Not required for embryonic development, but KO female mice demonstrate anovulation due to the collapse of the hypothalamic/pituitary system (De Mees et al., 2006). Heterozygote intercrosses produce viable offspring with normal sex ratios at expected Mendelian ratios. Appears to produce defeminised maternal behaviour in Afp null mice (not concerning as female null mice are sterile; Bakker et al., 2006; Keller et al., 2010). It is also a growth promoter through suppressing apoptosis (biomarker for hepatocellular carcinoma; Chen et al., 2020), but the physiological/functional impact of this in KO mice is not well understood.	5	1
Pgr*	607311	Ovulation failure due to lack of response to progesterone. Defeminisation behaviour in KO females (isoform A specifically involved in female fertility). KO males have enhanced sexual behaviour, and pharmacological inhibition of Pgr produces results consistent with gene KO (Schneider et al., 2005). Suitable for drive spread but not appropriate for baiting as interference may increase male reproductive behaviour. Preliminary work shows that Pgr KO mice have deficits in spatial memory. Yet to be determined how significant this would be to the lifetime reproductive success of drive homozygous males (could potentially have a reduced tendency to disperse and defend home territory; Joshi et al., 2023).	9	2
Dlgap5	617859	Defective proliferation of endometrial stroma tissue ultimately leads to implantation failure (Tsai et al., 2008). Additional roles in oocyte maturation. Female KO has defective oocyte divisions due to microtubule bipolarity establishment and maintenance (Breuer et al., 2010).	14	3
Figla	608697	Female KOs lack primordial follicle development, and demonstrate a reduction in oocyte mass and ovaries at birth due to failed oocyte-development gene activation and suppression of sperm-development genes. Expression is detected through follicular development and in mature oocytes, ceasing in preimplantation embryos (Hu et al., 2010).	6	4
Gdf9	601918	Failure of follicle maturation and differentiation past the primary one-layer stage. Expressed in oocytes across follicle development (except primordial follicles).	11	5
Izumo1r	615737	Gamete binding receptor that pairs with the sperm cell-surface protein Izumo1. Failed fertilisation due to loss of gamete recognition.	9	6
Nlrp5	609658	Embryos fail to proceed beyond the 2-cell stage due to defects in the subcortical maternal complex (SCMC) - essential for zygotic development.	7	7
Nobox	610934	Atrophic ovaries, characterised by failed maturation of follicles and the reduction in ovarian mass. Also important for zygote genome activation and preimplantation embryo development (Madissoon et al., 2016).	6	8
Padi6	610363	Abnormal cytoplasmic lattice development and failed genome activation following fertilisation, consequent of failed post-translational modification during oocyte maturation (highly expressed in oocytes).	4	9
Ooep	611689	Asymmetrical zygotic cell division, ultimately terminating development at the 2-cell stage, resulting from the loss of structural stability of the SCMC (plays important roles in RNA metabolism and zygote genome activation).	9	10
Tle6	612399	“   “   “   “   “   “	10	10
Zp3	182889	Defective zona pellucida (ZP) development in oocytes, resulting in failure to trigger acrosomal reaction in sperm before fertilisation. Heterozygotes only have a ZP half as thick as ZP observed from WT oocytes. Expression appears to be from multiple structures, including oocyte (strong expression in nucleus during prophase; Gao et al., 2017). Needs further RNA seq work: may be useful for homing drives if expression ceases before recombination.	5	11

3.2. Evolutionary conservation

3.2.1. Overall patterns of sequence conservation

Our multiple species alignments suggest sequence conservation across the nine evaluated mammalian species, including a marsupial representative (brushtail possum; unpublished data, D. Bond). Sequences of more than 80 nt and at least 75% identity were identified across 10 of 16 evaluated male target genes (Catsper1, Catsper2, Gapdhs, Meig1, Oaz3, Slc26a8, Spaca1, Spag4, Tssk6 and Zpbp; SI Appendix 12), and 4 of 12 evaluated female candidates (Dlgap5, Pgr, Nobox, and Zp3; SI Appendix 13). However, not every species was necessarily represented in each of these conserved blocks (Appendices 14 and 15).

4. DISCUSSION

Our systematic gene assessment and subsequent evaluation identified 16 male fertility genes that are critical for mammalian fertility and are strong candidates for genetic population suppression approaches. Using a modified version of our initial evaluation approach (in which we allow minor expression beyond reproductive tissue), we additionally identified 12 female fertility genes. These genes are involved in various reproductive pathways, including structural and gamete recognition, which are processes preceding fertilisation, making these targets a more ethical choice than those with post-fertilisation functions. Many of these genes have previously been suggested as targets for non-hormonal human contraceptives (e.g., Ppp3r2; (Castaneda & Matzuk 2015)), and their significance to mammalian fertility has additionally been identified in reviews of the monogenic causes of mammalian fertility (Jiao et al., 2021; Oud et al., 2019). In the process of identifying these genes, we have established a systematic gene identification framework with broad utility, including identifying biological targets for developing species-specific toxins across various taxa, including insects, plants and mammals.

Male fertility bias and the shortfall of using expression data

During our systematic search and filtering, we observed a complete bias towards genes that are specifically indispensable for male fertility, which highlights that most female fertility factors appear to have phenotypic effects on males, with the reverse being less common due to the extensive and unique set of spermatogenic genes (Matzuk & Lamb 2002). This is consistent with biases presented by the Reproductive Genomics Program (finding a 15:2 ratio of male-specific to female-specific infertility phenotypes; Matzuk & Lamb 2008^, Schimenti & Handel 2018) and the Mouse Genome Informatics (finding a 59:21 ratio of male-specific to female-specific infertility phenotypes; Schimenti & Handel 2018). Through our systematic evaluation process, we discovered that most genes specifically indispensable for female fertility demonstrated gene activity in non-target tissue exceeding our tissue-specific expression threshold of 5% (activity of genes essential to female fertility is broad; Dean & Mank 2016). Such genes were subsequently omitted such as to prevent the selection of any genes that may have multiple biological roles and generate multiple pathological phenotypes in allelic-null individuals.

Excluding genes with high expression in somatic tissues also filtered out genes that are somatically expressed but are deposited in the germ cells where they play a significant role in gamete development, maturation and the zygotic activation processes (e.g., maternal and paternal effect genes). Future searches should consider refining the keyword string to probe for genes with isolated expression in granulosa and Sertoli cells as initial drive spread would likely benefit from heterozygotes of both sexes spreading the drive. Alternatively, expression of target fertility genes must precede drive initiated disruption (i.e., meiosis I; Weitzel et al., 2021) so that heterozygous drive carriers remain fertile. Fine-scale spatial characterisation of these expression windows during gamete development remains a major challenge and is an emerging field of new research as technological capabilities (e.g. single cell sequencing) advance.

The substantial male target bias identified through our systematic search is also problematic for suppression gene drive development and application because homing gene drives are most effective at inducing population suppression when targeting female fertility (Burt 2003; Deredec et al., 2008; Prowse et al., 2017). This is because female reproductive success is the major limiting factor in population growth and viability in most species (Galizi et al., 2016; Robertson et al., 2006; Schliekelman et al., 2005; Wedekind 2002). To alleviate the challenges of using gene expression profiles to characterise function for female fertility genes, we utilised fertility and viability data from established mouse knock-out models to verify that gene function predictions are germlinespecific. The production of healthy, viable knock-out mice affirms the absence of developmental consequences to the homozygous mutant mice, which implies the absence of any additional phenotypic effects if the gene is coupled to a gene drive. Validating the fitness impact of the candidate gene in both sexes is essential for predicting drive fitness to inform laboratory and computational experiments.

Functional characterisation

The variations in methods, sample sizes, and the way in which results are reported across the literature are challenging to manage when verifying knockout phenotypes. Where data were available, we found insignificant differences in the litter size between heterozygotes and wildtype individuals of the target sex, validating the absence of a significant intermediate fitness cost in heterozygotes arising from the disruption of one copy of the target gene. However, it should be noted that these results are limited by the experimental environment and genetic background that was tested. Additionally, both female fertility genes identified here as HDR-based gene drive candidates (Afp and Pgr) appear to play a role in sexual differentiation of the brain early in development and thus have the potential to produce behavioural differences. Further work is necessary in order to characterise how significant this is to the fitness of homozygous null males and confirm the absence of dosage effects in heterozygous gene drive females to understand behavioural limitations of targeting such genes with a drive. Although variations in genetic background and behavioural phenotypes may be relatively small, their effects may be amplified by the dynamics of a gene drive system in large heterogeneous wild populations. Future work should consider using individuals from target populations of interest in representative microcosm environments to control for local genetic and behavioural variation.

Evolutionary conservation

Several identified fertility genes demonstrate sequence conservation across multiple mammalian lineages, often spanning millions of years of evolutionary divergence. This suggests functional conservation, which may also limit the evolution of functional resistance alleles. These findings separate the genes into two distinct lists – those that are species or lineage-specific, and those that are more broadly conserved and thus may have general utility across a range of target species with species-specific gRNAs. Where resources are limited, targeting research into such general utility genes could substantially reduce the overall research and development expenditure across a cohort of invasive species. It will also ensure confidence that the conserved genic regions have a critical function and therefore are excellent targets.

The functional role of top male fertility genes is heavily biased towards sperm structural development and motility before fertilisation, except Tssk6, which regulates the expression of other spermatogenic genes through managing post-meiotic chromatin remodelling in maturing spermatids (Spiridonov et al., 2005). This is likely the result of stronger sexual selection and therefore faster evolutionary rates in sperm-associated genes, which is highlighted in our finding that evolutionarily conserved regulatory genes functioning in fertility are more abundant in the female candidate gene list. Targeting these key regulatory proteins that function in gametogenesis presents a method to create knockout redundancy in the gene drive system.

Additional considerations when screening potential candidate genes include testing gRNA cutting efficiency at such conserved target sites. Further, whole genome assessment is necessary to mitigate the risk of off-target mutagenesis where conserved sequences have been evolutionarily duplicated, giving rise to gene families (a set of genes with a homologous sequence origin but divergent molecular function). Variant assessments of target populations will also enable high-resolution characterisation of target sequences and further facilitate assessments of inter- and intra-species conservation, important for considering drive containment and the evolution of resistance to gene drives.

5. CONCLUSION

This stringent systematic analysis has identified a list of sex-specific fertility genes that have the potential as candidate genes for population suppression gene drives in multiple mammal species. The identified candidates have some level of evolutionary conservation and may also have utility for further applications, including developing human and wildlife contraceptives. Additionally, we establish a systematic gene identification framework, including key considerations for candidate gene selection, which may also have utility in other applications, including identifying biological targets for developing species-specific toxins. Further work, including tissue expression profiling, gene knockouts, and population analysis for each target species, will be essential to confirm functional specificity and conservation across mammalian reproductive systems.

DATA AVAILABILITY

Supplementary information and additional data available from the Figshare Digital Repository: https://doi.org/10.6084/m9.figshare.21905646.v1