Fast forward to new genes in mammalian reproduction

All species are dependant on reproduction to survive and evolve. Although there are a wide variety of reproductive methods, many basic principles of sexual reproduction are relatively conserved. The overall processes are well-defined, but the underlying molecular basis of each step, from the formation of germ cells and haploid gametes to the fertilization process, is less understood. Gamete production in animals is characterized by three major processes: (1) establishment of the germ lineage in the context of sexually differentiated gonads; (2) meiosis, which involves the haploidization of the number of chromosomes; and (3) development of the haploid gametes into cells capable of fertilization, thus restoring the diploid number of chromosomes in offspring.

Mammalian reproduction is dynamically regulated by the pituitary gonadotropins, luteinizing hormone (LH) and follicle-stimulating hormone (FSH). They are synthesized in the pituitary gland following stimulation by the gonadotropin-releasing hormone (GnRH) and act by stimulating steroid production and gametogenesis in both males and females. FSH acts on the Sertoli cells of the testes and is involved in spermatogonial maturation (Haywood et al. 2003; Meachem et al. 2005) and LH triggers the production of testosterone in the Leydig cells. Testosterone is essential for spermatid elongation and development of secondary sexual characteristics (O'Donnell et al. 1994; O'Donnell et al. 1996). In females, FSH is recognized as being essential for ovarian follicle growth and maturation. LH is involved in the production of androgens in the theca cells of the ovary that ultimately are transformed into oestrogen in the granulosa cells (Fortune & Armstrong, 1977; Fortune & Armstrong, 1978). Oestrogen functions in the regulation of gonadotropin release and is responsible for promoting events in folliculogenesis (Herbison, 1998).

The mammalian germ cell lineage is established early during development. Primordial germ cells (PGCs) are detectable at E7.5 in mice (Ginsburg et al. 1990). At E8.5 they migrate towards the genital ridge and reach their target by E10.5–11. Gonadal sex is determined between E10 and 12.5, triggered by expression of the Sry gene (Koopman et al. 1991). The presence of the Sry gene on the Y chromosome directs development into testes, while the absence of the Sry gene leads to formation of ovaries. Depending on whether the germline stem cells reside in gonads destined to become ovaries or testes, they undergo further maturation and their fate diverges.

In male embryos, the PGCs intermingle with Sertoli cells and the seminiferous cords are formed. Within the seminiferous cords, the PGCs differentiate into gonocytes which remain quiescent until after birth. Spermatogenesis is initiated soon after birth by the division and differentiation of the gonocytes into spermatogonia. The spermatogonia continue further mitotic division and differentiate into spermatocytes, which in turn undergo meiotic divisions and become round spermatids. Further development of the spermatids into elongated spermatids and eventually spermatozoa takes place before they leave the seminiferous epithelium, and enter the epididymis where they acquire motility.

In the developing female ovary, the PGCs eventually differentiate into oogonia. The oogonia continue mitotic divisions until they enter meiosis and become oocytes. The oocytes arrest in the diplotene stage of the first meiotic prophase around birth, and remain arrested until a luteinizing hormone surge induces ovulation of a subset of oocytes. Following completion of the first meiotic division and ovulation, the cells progresses to metaphase II at which point they once again arrest. The cells are only released from this second arrest following fertilization. In contrast to males where the spermatogonial stem cells continuously renew the germ line to produce unlimited numbers of sperm, the proliferation of oogonia occurs only prenatally. Hence the female reproductive lifespan is limited by the number of oocytes in the initial pool, and the rate at which they deplete, either by ovulation or atresia.

The molecular steps in meiosis are essentially the same for males and females. Meiosis consists of two divisions designated meiosis I and meiosis II. In prophase of meiosis I, the key molecular events are pairing of homologous chromosomes, initiation of recombination via the formation of genetically induced double strand breaks (DSBs), and recombinational repair of the breaks by either gene conversion (non-crossover) or reciprocal recombination between non-sister chromatids (crossing over), the latter of which results in chiasmata that physically hold together homologous chromosome bivalents until the first metaphase. In the first meiotic division, homologous chromosomes separate into separate cells (first polar body and oocyte in females; secondary spermatocytes in males). The final step in production of haploid cells is the separation of sister-chromatids in the second meiotic division. Thanks to extensive genetic and biochemical characterization of meiosis in fungi and other model organisms, many genes and their biochemical functions have been delineated. Furthermore, the evolutionary conservation of the meiotic process has enabled the use of reverse genetic approaches in mice (gene knockouts) to study and compare the function of orthologous genes (e.g. SPO11, DMC1) (Habu et al. 1996; Matsuda et al. 1996; Pittman et al. 1998; Yoshida et al. 1998; Keeney et al. 1999; Romanienko & Camerini-Otero, 1999). However, mammals clearly have many genes required for meiosis that are not present in simpler model organisms, and vice versa. Importantly, due to the complex milieu in which gametogenesis occurs in mammals compared to fungi or other simpler organisms, there are unique signalling pathways and biological networks existing in mammals that will only be elucidated with the aid of comparative genetics.

The search for infertility genes

For many years, genetic approaches have been successful in identifying a variety of human disease genes. The process is simplified when a particular disease has a distinct phenotype, and mutant alleles are present at a significant frequency in a human population(s) such that multiple families segregating the disease can be pooled for genetic linkage analysis. Because infertility has myriad causes, and its very nature works against the persistence of mutant alleles, progress in characterizing genes underlying human infertility has been minimal. An exception is the case of male infertility linked to Y chromosome deletions.

Because of these difficulties in studying human reproductive genetics, most of our knowledge comes from studies in the lab mouse. For the same reasons as humans, spontaneous mutations have not contributed significantly to the identification of mouse infertility genes. Rather, ES cell knockout technology has generated over 200 mouse models known for infertility, but it is expected that the number of genes involved in reproduction is much higher (Jorgez et al. 2005). Some have put the estimates as high as > 2300 genes that are male germ-cell specific (Schultz et al. 2003). Many of the mouse models were created serendipitously; that is, the targeted gene was initially chosen with the expectation that phenotypes other than infertility would arise, but the resulting animals were sterile in addition to, or exclusive of, other phenotypes. It is also clear that many genes that are required for viability of the animal also play roles in germ cell development. Therefore, null alleles generated by gene targeting fail to elucidate the germ cell function of such genes. These experiences underscore our limited knowledge of genes required for gametogenesis, and the need for unbiased functional searches for infertility genes and a collection of viable alleles in mouse models.

A forward genetic mutagenesis screen is a classic and powerful strategy to address this problem. This approach has been successfully used in other organisms and for the study of other phenotypes. In mice, forward genetic screens are most commonly and successfully conducted by N-ethyl-N-nitrosourea (ENU) treatment of the germ cells in male mice. Effective and reproducible protocols for ENU mutagenesis, as well as chemical mutagenesis of embryonic stem cells (ES) with ethylmethane sulphonate (EMS) for similar purposes, have been published and are readily available (Justice et al. 2000; Munroe et al. 2000). Both compounds are potent alkylating agents that cause random single base pair substitutions throughout the genome. The frequency of the induced mutations can be controlled by carefully regulating the concentration of the mutagens and the length of exposure. This lab, as well as the Reprogenomics program at the Jackson Laboratory (http://reprogenomics.jax.org, Bar Harbor, ME, USA), initiated large scale forward mutagenesis screens around the year 2000. Mutations were induced in whole live mice and ES cells using ENU and EMS, respectively. These experiments have been described in detail previously (Munroe et al. 2000; Ward et al. 2003; Lessard et al. 2004). This review describes the effort that was undertaken to discover new genes/alleles involved in reproductive physiology as well as a description of some of the genes that already have been identified and characterized.

Mutants with a wide variety of phenotypes

The mutants and corresponding phenotypes from the ENU mutagenesis screen are listed on http://reprogenomics.jax.org. Additionally, Ward et al. (2003) have published a summary of the mutants identified by EMS mutagenesis of ES cells. A striking observation is that the number of mutations affecting only males is much greater than the ones affecting only females. In the Reprogenomics programme, 26 of 34 mapped mutations affect males only, and only 2 of the 21 mutants display a female-specific phenotype (repro22 and repro5). Possible explanations for the dramatic difference include: (1) spermatogenesis is simply much more complex than oogenesis, and requires many more genes; (2) oogenesis has more genetic functional redundancy, and/or is more tolerant of certain (potentially deleterious) defects; (3) the ‘target’ size (number of genes and physical extent of DNA) is larger for genes involved in spermatogenesis than for those involved in oogenesis (although there is a similar bias in knockout mutants); and (4) genes required for oogenesis tend to be required also for other developmental events, so mutations resulting in lethality or other serious defects ‘cull’ them as a gametogenesis mutant.

Some of the mutants have been mapped to relatively narrow regions and positionally cloned. The mutations that have been mapped are randomly distributed across the genome, although none have yet to be mapped to the sex chromosomes, which is a drawback of ENU mutagenesis. This occurs because pedigrees are initiated with male founders that are sired by crosses of ENU-treated males to wild-type females. The founder males, then, do not contain a mutagenized X chromosome, and would be sterile if they carried a spermatogenesis mutation on the Y. In addition to the mutants described below, there is an ongoing effort to positionally clone and characterize some of the other mutants. The phenotypes observed can broadly be categorized in three classes: premeiotic, meiotic and postmeiotic. Here, we will review selected mutations from each of these stages (Fig. 1). These genes provide interesting examples of how forward genetics provides clues into reproductive genetics that might not have been revealed by gene targeting.

The germline stem cell mutant gcd2 (germ cell depletion 2)

The gcd2 mutation was initially identified as a germ cell depletion mutant with a phenotype similar to the FancC and Pog alleles, but with incomplete penetrance (Ward et al. 2003). It is the only mutation from a small scale EMS mutagenesis screen that affects premeiotic germ cell development. In order to time the disappearance of germ cells, embryos at various time points were stained with germ cell nuclear antigen antibody (anti-GCNA) (Reinholdt et al. 2006). Staining of gcd/gcd mutants was observed at time points E9.5, 10.5 and 11.5 but not on embryos at E15.5. The E11.5 time point corresponds to the time when the majority of the germ cells have migrated to the urogenital ridge. These observations support a model where the gcd2 mutation affects the PGCs after they migrate to the embryonic gonad as opposed to the migration itself, and it is rather unique in this respect.

The mutation initially was mapped to a 37 Mb region on chromosome 2. Reinholdt et al. (2006) further mapped the mutation to a region less than 1 Mb that contained five genes, Bdnf, Lin7c, Gpr48, Pou5f1 and Gus and one novel RIKEN gene. Based on prior knowledge, none of these genes were thought to be involved in any aspect of primordial germ cell development and none were subsequently found to be mutated in gcd2-carrying mice. Since no mutations were detected, it is possible that the mutation might be in a regulatory sequence controlling a gene outside of the candidate region. Alternatively, the mutation might be in transcribed but not translated DNA, such as microRNA. It is increasingly being recognized that microRNAs play important roles in a wide array of biological processes and the involvement of microRNAs in spermatogenesis has already been reported (Yu et al. 2005; Costa et al. 2006). Although sequences encoding microRNAs only make up a fraction of the total genome, they are still targets due to the random nature of the chemical mutagenesis screens. Indeed, the first evidence for the existence of functional microRNAs came from positional cloning of lin-4 in C. elegans, a germline defective mutation (Horvitz & Sulston, 1980; Lee et al. 1993). Either way, the gcd2 is a good example of both the strengths and weaknesses of ENU mutagenesis. On one hand, it is clear that a novel and very important element or gene has been mutated, which would not have been selected for targeted mutagenesis because of the lack of any annotation. One the other hand, single nucleotide mutations, especially those not in annotated genes, are difficult to identify. However, increasingly high throughput, less expensive DNA sequencing technologies will continue to mitigate this problem.

The gcd2 mutation may eventually prove to be an example of a point made earlier, that null mutations may obscure the roles of many genes in germ cell development, due to earlier lethal effects on embryonic development. Reinholdt et al. (2006) observed sub-Mendelian ratios of gcd2/gcd2 progeny in their crosses, and suspected early embryonic lethality of a percentage of homozygotes. It remains to be seen whether gcd2 is a null or hypomorphic allele, but if the latter is true, then a null mutation would not uncover the germ cell phenotype.

A final general point to make, which applies to both targeted and chemically induced mutations, is that some mutant phenotypes can be modified by genetic background. The penetrance of gcd2 is much higher in the CAST/EiJ background (90%) than in C57BL/6J (37%). Hence, the gcd2 mutant could be exploited to identify and characterize other, potentially novel genes in the germ cell development pathway.

The meiotic mutation Mei1

The Mei1 allele was the first mutation recovered from screens of mice derived from EMS-mutagenized ES cells to be positionally cloned and published (Libby et al. 2002; Libby et al. 2003). Mei1 homozygotes of both sexes are sterile due to arrest at the transition between the zygotene and pachytene stages of meiosis in spermatocytes. Meiotic chromosomes from mutants were devoid of RAD51 (required for recombinational repair of DSBs), and exhibited very low levels of γH2AX phosphorylation, which normally occurs at the leptotene stage of meiosis in response to DSBs (Mahadevaiah et al. 2001). In every species examined so far, the formation of DSBs is catalysed by the SPO11 protein in a type II topoisomerase-like fashion. This phenotype is similar to the one observed in Spo11^−/− mice (Romanienko & Camerini-Otero, 2000). These observations left two possibilities: (1) that MEI1 was required for responding to DSB formation; or (2) that MEI1 is needed for DSB formation. Two experiments suggested the latter. Treatment with cisplatin, a powerful alkylating mutagen that induces inter- and intrastrand crosslinks, restored RAD51 loading in mutant spermatocytes. Secondly, Reinholdt & Schimenti (2005) exploited the fact that arrest occurs later in Mei1 mutant oocytes, a pattern typical of mutations that affect chromosome synapsis but not DSB repair defects. By creating double mutants of MEI1 and the meiosis specific recombinational repair protein DMC1, which arrest oogenesis prior to entry into pachytene (earlier than the Mei1 mutant), the authors observed that the phenotype was identical to that of Mei1^−/− mutants, thus concluding that MEI1 acts upstream of DMC1, that is, at the stage of DSB formation. The current data strongly suggests that MEI1 is involved, directly or indirectly, with the formation of genetically programmed DSBs.

In yeast, several additional genes (RAD50, MRE11, XRS2, MER2, MEI4, REC102, REC104, REC114, SKI8, MER1 and MRE2) have been identified and are known to be necessary for initiation and processing of DSBs. The exact mechanism of many of these genes is still elusive but some of them are directly involved in the formation of DSBs by possessing nuclease activity, while others are structural or are involved in signalling and regulation of other proteins in the DSB initiation complex (Malone et al. 1991). The current knowledge on similar proteins in mammals is bleaker, but it is expected that it will be at least as complex as in yeast. Despite the critical importance of DSB formation – initiation of recombination – only three of these 12 yeast genes have orthologues in mammals: SPO11, RAD50 and MRE11. At this point, we do not know whether orthologues do indeed exist in mammals but have diverged beyond the point of recognition, if meiotic DSBs are induced in a very different way, or if convergent evolution of the yeast protein functions has occurred. MEI1, which is only the fourth gene known to be required for DSB formation in mammals, does not share homology with any of the yeast proteins, or to any proteins outside of vertebrates. Because of this gene's anonymity with respect to putative function or any other known protein structure domains, it would not likely have been subjected to functional analysis other than by its identification via phenotype-driven mutagenesis. Since its amino acid sequence is unique, MEI1 may be a functional, but not evolutionary homologue of one of the yeast proteins mentioned above. Alternatively, it might be involved in a biological process, such as the signalling between germ and somatic cells, that reflects the increased complexity of meiosis in vertebrates relative to yeast. Indeed, recent studies of MEI1 temporal and spatial localization in testis are hinting of such a role (B. Furnes and J. Schimenti, unpublished observations). Thus, it most likely represents a new class of proteins with an as yet unknown function or mechanism of action in vertebrates.

A recent study has implicated mutations in MEI1 in human infertility (Sato et al. 2006). The complete coding region of MEI1 was sequenced in a sample of 27 men of different ethnicities (European-American, Israeli and Japanese) diagnosed with azoospermia due to complete meiotic arrest. A total of four novel SNPs were identified, two of which were associated with azoospermia in European-Americans but not in individuals of Israeli background. Although based on a relatively modest number of individuals, these results suggest that some of the SNPs and corresponding haplotypes might be associated with azoospermia in certain ethnicities. It remains to be seen whether these SNPs are associated with reduced fertility levels in women. To our knowledge, this is the first example of a gene being first identified in a forward mutagenesis screen in mice that might affect human fertility.

The meiotic mutation Mei8, an allele of the meiotic cohesin Rec8

The Mei8 mutant is an example of ENU mutagenesis producing a mutation in a gene that has a known function in other species. It was produced in a novel two generation screen for infertility mutations produced by EMS-induced mutations of ES cells (Munroe et al. 2004). Homozygotes of both sexes are sterile. Male germ cells undergo meiotic prophase I arrest at late zygotene/early pachytene (Bannister et al. 2004). After mapping the mutation down to a relatively large 4.4 Mb region on chromosome 14, the meiotic cohesin gene Rec8, known to disrupt meiosis in yeast, stood out as an obvious candidate and was sequenced. This revealed a C → T nonsense mutation 1/4th into the protein coding region (Bannister et al. 2004). While this positional cloning effort was much simpler than that for Mei1, a novel gene that required genetic mapping to a < 200kb interval in order to identify Rec8 as Mei8 was not as exciting in that Rec8 was a known entity from work in other organisms (see below). Nevertheless, Mei8 was the first mutant allele of Rec8 produced in mice, and thus this was a productive exercise. Given that the number of targeted mutations in known genes is ever increasing, so is the chance that a new ENU-induced mutation will reside in one of these, likely to the dismay of the investigator who has devoted time and effort to mapping an ENU allele. For example, one of the meiotic mutations uncovered by the Reprogenomics group turned out to be an allele of another meiotic cohesin, SMC1b (Handel et al. 2006).

Interestingly, a Rec8 targeted allele was subsequently produced by another group, resulting in a mouse exhibiting identical meiotic phenotypes (Xu et al. 2005). However, that allele caused non-meiotic defects as well, which we suspect is due to effects of the insertion of the targeting cassette on nearby genes, a known problem with the selectable markers in targeting vectors that are under the control of strong promoters (Olson et al. 1996). Thus, the ENU allele appears to have produced a ‘cleaner’ null mutation.

REC8 was initially established as being necessary for meiosis specific cohesion in yeast (Lin et al. 1992; Molnar et al. 1995; Watanabe & Nurse, 1999). Mutations caused reduced meiotic recombination in centromeric regions. Later work established the yeast Rec8 protein as important in maintaining sister chromatid cohesion and orienting the kinetochores so as to direct sister chromatids to the same pole. In mammals, REC8 is also an integral part of the cohesion complex, necessary for the establishment of sister chromatid cohesion, formation of the synaptonemal complex and recombination. In addition to REC8, at least four to five other proteins are though to make up the cohesion complex: SMC1A, SMC3, RAD21 and either SA1 or SA2.

The postmeiotic mutations swm2, repro2 and repro3

Whereas induced mutations in meiotic genes have a substantial likelihood of having orthologues in lower model organisms, genes involved in mammalian-specific aspects of gamete development are less likely to have conserved orthologues with analogous function. Three spermiogenesis mutants discussed below illustrate this point.

The three mutants, swm2, repro2 and repro3, were initially characterized by being infertile but with seemingly normal morphology of the reproductive organs (Lessard et al. 2006). They all display low sperm numbers and abnormal sperm morphologies. The spermatogonia, spermatocytes and round spermatids are normal but defects begin to appear in later stage spermatids (stage 8–9). The swm2 mutant sperm exhibit a short and discontinuous midpiece. Upon closer examination, the spermatids do not have a correctly aligned mitochondrial sheath, but instead seemed to be organized in cytoplasmic droplets or clustered around the axoneme. The repro2 mutant sperm have a short tail and a dysplasia of the fibrous sheath. Finally, the repro3 mutant shows a range of variations in both tail and headshape and absence of a mitochondrial sheath.

In order to determine the molecular basis for these defects, genetic mapping has been performed. Swm2 has been localized to a 2.4 Mb region on chromosome 7 containing 51 genes. repro2 and repro3 were mapped to chromosome 5 (3.55 Mb, 88 genes) and chromosome 10 (2.75 Mb, 16 genes), respectively. Within the candidate regions, no known genes involved in sperm motility and morphology exist. This is encouraging in that these mutations will be in ‘novel’ genes and thus yield new insights into spermiogenesis. However, this complicates gene identification.

Importantly, the phenotypes of these mutations resemble that of sperm from humans with oligoastheno-teratozoospermia. These individuals generally have sperm with low motility and abnormal shape or size, low sperm count or a complete absence of sperm. Idiopathic oligoasthenoteratozoospermia is the most common cause of reduced fertility with an estimated 30% of infertility explained by oligoasthenoteratozoospermia (Cavallini, 2006). Although environmental factors are believed to be an important factor in defective spermiogenesis, genetic factors have also been identified (Juneja & van Deursen, 2005). Based on the fact that they share the same phenotypes as observed in humans, the swm2, repro2 and repro3 mutants may be good models for oligoasthenoteratozoospermia of familial origin.

Concluding remarks

The molecular basis of gametogenesis and fertility is inherently difficult to study due to the complexity and transient nature of the processes involved. In order to increase the number of mouse infertility mutants and subsequently facilitate the study of reproduction pathways we, our colleagues and others have carried out large forward genetic screens. The power of the forward mutagenesis screen lies in the fact that the chemically induced mutations are randomly distributed in the genome and no a priori knowledge about the genes involved is necessary. This allows for the detection and investigation of genes involved in a wide range of reproductive processes, ranging from germ cell development to fertilization. We have currently identified novel genes with no known function, known genes not previously thought to be involved in reproduction, as well as genes that already have an established role. We note that mutations that cause subfertility, or which have incomplete penetrance, may be missed in the screens. This is because the ‘phenotype’ that is initially screened is failure to produce progeny. Thus, any animals that may be homozygous for such a mutation will not be pursued if it yields even a single pup during the high-throughput phenotyping screens.

An important goal of this project has been to identify genes and to establish models to gain an increased understanding of human health and reproduction. It is estimated that approximately 13–18% of human couples are experiencing fertility problems and this seems to be a growing phenomenon. There is evidence that this is due to genetic as well as lifestyle and environmental factors. By increasing the number of genes known to affect mouse fertility, this better empowers human association studies designed to find genetic loci linked to infertility. SNPs implicated by such association studies may still need to be validated in functional studies, possibly by engineering mice with analogous point mutations.

In summary, forward mutagenesis screens for fertility mutants is a feasible and highly efficient approach to study novel aspects of reproduction as well as expand on the knowledge of current molecular pathways. Overall, our efforts have yielded over 30 mutations acting at various stages of gametogenesis, particularly in spermatogenesis (summarized in Fig. 1). Other groups have also identified infertility mutations in forward genetic screens (Clark et al. 2004; Kennedy et al. 2005). As these mutations are positionally cloned, the focus will shift towards functional studies of molecular pathways and mechanisms. The identification and characterization of novel proteins, such as MEI1, and reproductive mechanisms may then provide attractive targets for future drug development, especially contraceptive drugs.