“‘They asked him’—he paused and shivered—‘to tell the Truth, the Whole Truth and Nothing but the Truth…. Only, don't you see? They gave him too much of the truth drug…much too much too much…’ ‘…Oh, I can't remember any of it now… except most of the good bits were about frogs, I remember that.’ Suddenly he was hooting with laughter again and stamping his feet on the ground. ‘You would not believe some of the things about frogs,’ he gasped.”
-Douglas Adams, “Life, The Universe and Everything” (Adams,1982)
Xenopus oocytes and embryos have been essential for elucidating the molecular basis of vertebrate axis formation, embryonic inductions, signalling pathways in organogenesis, and the cell cycle (Amaya et al.,1991; Lamb et al.,1993; Philpott and Yew,2005; Whitman,1998; Wilson and Hemmati-Brivanlou,1995). Mass action biochemical assays, as well as sophisticated tissue manipulations such as grafts and explants, can be allied with gain-of-function overexpression via injection with synthetic mRNA. Until recently, the conspicuously empty drawer in the frog toolbox was genetics. In addition to their classic use for analyzing gene function, mutations are invaluable for diagnosing off-target effects in other loss-of-function approaches such as morpholino antisense oligonucleotide (MO) and siRNA knockdowns (Scacheri et al.,2004). Loss-of-function genetic backgrounds may also be combined with more conventional Xenopus approaches such as gain-of-function mRNA injection for enhancer and suppresser screens to dissect complex processes.
The last five years have seen dramatic advances in Xenopus genetics (Harland and Grainger,2011), including the first screens for induced mutations in amphibians (Goda et al.,2006), cloning of novel mutations (Abu-Daya et al.,2009,2010; Geach and Zimmerman,2010), generation of a meiotic map (Wells et al.,2010) and a genomic sequence assembly (Hellsten et al.,2010), reverse genetics (Goda et al.,2006; Young et al.,2011), and stock centers on either side of the Atlantic (see Pearl et al in this issue). Recent breakthroughs in high throughput sequencing, together with improved genomic resources, are revolutionizing both forward and reverse genetic strategies. Mutation identification, formerly a rate-limiting step relying on tedious positional mapping, can now be accomplished rapidly by exome sequencing (Fairfield et al.,2011) (see below).
Genetic screens in invertebrates, principally Drosophila and C. elegans, underpin contemporary understanding of animal development (Brenner,1974; Kemphues et al.,1988; Nusslein-Volhard and Wieschaus,1980). Large-scale developmental screens in vertebrates have focused primarily on zebrafish (Driever et al.,1996; Haffter et al.,1996; Mullins et al.,1994), with more limited screens in other teleosts (Loosli et al.,2000) and mouse (Garcia-Garcia et al.,2005; Wansleeben et al.,2011). While zebrafish offer many experimental advantages, the teleost lineage includes an ancient whole-genome duplication, with subsequent sub- and neo-functionalization diversifying roles of duplicated genes (Cresko et al.,2003; Force et al.,1999; Postlethwait et al.,2000; Ravi and Venkatesh,2008). Systematic comparison with a tetrapod bearing a canonically organized diploid genome is crucial for a broad understanding of vertebrate gene function and applications to biomedical research. Here, we discuss the features that make X. tropicalis ideal for combining genetic tools with Xenopus' broad range of functional assays, and survey recent progress.
Xenopus Genetics and Evolution
Although the X. laevis mutation anucleolate aided pioneering animal cloning studies a half-century ago (Elsdale et al.,1958; Gurdon et al.,1958) and albino animals are in wide use, genetic studies in X. laevis have been limited to small-scale gynogenetic screens for naturally occurring mutations (Krotoski et al.,1985) and comparative mapping of immune loci (Nonaka et al.,1997). Systematic genetic studies in Xenopus are hindered by the long generation time (1–2 years) and the peculiar speciation mechanism, where rare hybridization events can be accompanied by retention of both parental species' gene sets (allopolyploidy). Extant species form an allopolyploid series from the single known diploid X. tropicalis (haploid chromosome number N = 10) to tetraploids including X. laevis (N = 18) and X. epitropicalis (N = 20), octoploids (e.g. X. vestitus, X. wittei etc, N = 36) up to dodecaploids (X. ruwenzoriensis, X. longipes etc, N = 54) (Tinsley and Kobel,1996). In many cases, both parental species of higher ploidy hybrids survive and can be identified (Bewick et al.,2010). While the putative hybridization event giving rise to X. laevis is estimated at 40 MYA (Tinsley and Kobel,1996), divergence of its dual “alloallele” gene sets dates even earlier, from the last common ancestor of X. laevis' two diploid parental species, sometime after split from the X. tropicalis lineage 50–65 MYA (Evans et al.,2004). Functional redundancy between many X. laevis alloalleles is retained, impeding genetic analysis and necessitating targeting both genes in MO knockdowns, but other alloalleles show evidence of sub- and/or neo-functionalization (Chain and Evans,2006; Hellsten et al.,2007).
Conveniently, one species in the genus, Xenopus (Silurana) tropicalis, is a canonical diploid vertebrate. Its genome, at 1.5 × 109 bp in ten chromosomes, is among the smallest among tetrapods, and is highly syntenic to those of amniotes, including humans (Hellsten et al.,2010). The shorter generation time (4–6 months versus 1–2 years for X. laevis) also makes it more amenable to genetic analysis. In 1990, Marc Kirschner, then at UCSF, first brought X. tropicalis into the US for developmental biology research. Over the next decade several groups, particularly those of Rob Grainger at Virginia and Enrique Amaya at Cambridge, began piloting genetic approaches.
Since screening, phenotyping, and mapping can all be time consuming, it is important to balance inducing abundant mutations with obtaining interpretable single gene phenotypes and simplifying mutation identification. A variety of chemicals, radiation, transgene insertions, and most recently targeted gene editing can induce sequence lesions. The stage of germline targeted (mature sperm or mitotic spermatogonia) also has practical implications. As described below, the labour of positionally cloning mutations has been greatly reduced by improvements in the X. tropicalis genome assembly, gynogenetic shortcuts, and whole exome sequencing.
Since development of transgenesis in Xenopus (Kroll and Amaya,1996), insertional mutagenesis has been attractive for directly cloning disrupted loci. Indeed, this approach has already yielded a dramatic forelimbless phenotype (Abu-Daya et al.,2010) (see below). Remobilization of transposon-based transgenes (Yergeau et al.,2010) shows promise for efficiently generating new insertions, and is compatible with attractive gene-trap and enhancer-trap schemes (Bronchain et al.,1999; Yergeau and Mead,2007). However, recovery of phenotypes from dedicated insertional screens in X. tropicalis has yet to be reported; whether the rate of gene disruption approaches that of chemical mutagenesis remains a key question.
To date, published screens for induced mutations in X. tropicalis have used chemical mutagenesis, specifically in vitro alkylation of post-mitotic sperm using N-nitroso-N-ethylurea (ENU) (Goda et al.,2006). ENU creates point mutations forming both nulls (Abu-Daya et al.,2009) and milder hypomorphic alleles (Geach and Zimmerman,2010) more typical of human disease variants. Since dosage of the toxic and labile ENU is difficult to control, in vitro mutagenesis of sperm offers a practical opportunity to select a dose producing both high rates of dominant lethal abnormalities visible in the days after in vitro fertilization as well as viable F1 carriers to raise. However, this founder generation will be mosaic for induced mutations, which affects choice of screen schemes (see below). Since a given in vitro ENU adduct forms on only one strand of the mature sperm double helix, it is not “fixed” on the complementary strand until DNA replication and repair post-fertilization. 25% or less of the F1 germline carries any given mutation induced in vitro, and different germ cells in an F1 individual may carry different constellations of mutations. Alternatively, mutagenizing mitotic spermatogonia by injecting adult males with ENU produces non-mosaic F1 offspring, but animals need several months or longer to recover after treatment before breeding.
FORWARD GENETIC SCREENS: INBREEDING AND GYNOGENESIS
Gynogenesis, or generation of viable offspring without contribution from a paternal genome (Tompkins,1978), has been used extensively in zebrafish screens and mapping (Johnson et al.,1995,1996; Streisinger et al.,1986) and is an enormous advantage for Xenopus genetics. Simple and efficient, it reduces time and colony space to breed mutations or transgenes to homozygosity, and also provides a useful rapid mapping shortcut (Khokha et al.,2009) (see below). The first screens in both X. laevis and X. tropicalis used gynogenesis to uncover naturally occurring mutations in outbred animals (Krotoski et al.,1985; Noramly et al.,2005); several mutations were also uncovered via orthodox inbreeding (Grammer et al.,2005).
In a conventional forward screen scheme, phenotypes are uncovered in the third generation following mutagenesis (Fig. 1a) (Driever et al.,1996; Haffter et al.,1996). After in vivo mutagenesis of G0 males and two outcross generations, F2 siblings are randomly paired and their offspring inspected for recessive phenotypes in simple Mendelian ratios (25% mutant offspring from 1/4 of the F2 random sibling pairs). In addition to the colony space and time required to raise two generations to sexual maturity, this design is not practical with the mosaic F1 produced by in vitro mutagenesis: as <1/4 of F2 offspring of mosaic F1 are carriers of a given recessive allele, <1/16 random F2 sibling pairs will produce phenotypic clutches.
In contrast, gynogenesis allows recessive phenotypes to be uncovered in the offspring of heterozygous carrier females (Fig. 1b) (Goda et al.,2006; Khokha et al.,2009; Krotoski et al.,1985; Noramly et al.,2005). As with other amphibians and fish, Xenopus oocytes do not complete second meiosis by ejecting the second polar body until shortly after fertilization (Fig. 2). UV-inactivation of the sperm genome followed by in vitro fertilization yields haploid embryos, which typically develop for several days with posterior defects and truncations, but relatively complete anterior structures. Haploids can be efficiently rescued to viable diploids with a simple icewater cold shock to suppress polar body formation and retain both maternal sister chromatids (Goda et al.,2006; Khokha et al.,2009). Alternatively, haploids can be rescued by allowing the first round of zygotic DNA replication, then blocking cytokinesis with cold shock or pressure (Tompkins and Reinschmidt,1991), but with significantly lower efficiency. Androgenetic haploids, with no maternal genetic contribution, can also be made by UV-irradiation of eggs and fertilization with untreated sperm (Elsdale et al.,1960).
Gynogenesis minimizes colony space and generation time required for screens. With X. tropicalis' large number of eggs (up to 9,000 from one ovulation), screening and much preliminary mapping, or even cloning by direct sequencing, are possible using individual adult female carriers rather than families. Gynogenetically derived embryos (gynogenotes) also provide extremely useful rapid low-resolution map information (see below). In the first X. tropicalis gynogenetic screen for naturally occurring mutations, Grainger and coworkers isolated 10 heritable defects affecting diverse developmental processes from wild-caught females (presumably outbred with high allele diversity)(Noramly et al.,2005).
Unsurprisingly, chemical mutagenesis greatly increased the yield of phenotypes. A pilot gynogenetic screen of 192 mosaic F1 females derived from in vitro mutagenesis yielded 77 candidate post-neurulation phenotypes (Goda et al.,2006). More than 35 of these selected for further analysis have proven to be heritable, with only a handful not recovered despite mosaicism of the females used for gynogenesis. Nonspecific gastrulation defects were filtered out by focusing on post-neurulation organogenesis and differentiation phenotypes up to Nieuwkoop & Faber stage 45 (4 days post-fertilization). Phenotypes affecting a variety of processes including neural crest migration and differentiation, cardiogenesis and heart function, primitive haematopoiesis, eye and ear development, axis elongation, left-right asymmetry, and others were described. Nearly all similar phenotypes sorted into different complementation groups, with the exception of two independently isolated non-complementing alleles (mrs lotmh187 and mrs lotmh19). Interestingly, these map to an interval containing the giant protein titin, a large enough target to hit twice even in a small screen.
MAPPING FROG MUTATIONS
In addition to positionally cloning forward mutations, linkage analysis helps prove that loci known by specific sequence or transgene insertion, rather than secondary mutations, are responsible for specific phenotypes. Meiotic recombination events are the currency of genetic mapping, and relative to other vertebrate genetic models, X. tropicalis' abundant eggs translate to unparalleled wealth: a single mating can produce enough meioses to clone a mutation by conventional positional methods. In the past few years, addition of an X. tropicalis genome assembly (Hellsten et al.,2010) and meiotic map (Wells et al.,2010), accompanied by a series of breakthroughs in genomics technologies, have greatly simplified mutation identification.
Gynogenetic Centromere Linkage for Low-Resolution Linkage Assignment
Rather than scan the entire genome for linked polymorphisms, a simple shortcut has been developed to take advantage of the small number of chromosomes and ease of gynogenesis in X. tropicalis. In a small number of PCR reactions, this method can rapidly locate most X. tropicalis mutations to a genetic region (Khokha et al.,2002), which could then be inspected for candidate genes.
As seen in Figure 2, gynogenotes contain the sister chromatid products of meiosis I, held together during recombination with non-sister chromatids at their centromeres, meaning that nearby regions are always homozygous, while progressively more distal regions are more likely to be heterozygous due to intervening recombination. Likewise, phenotypically mutant embryos are homozygous at the mutant locus. Two useful pieces of information fall out of this observation. First, the phenotypic ratio in gynogenetically derived embryos reflects the gene-centromere distance (Johnson et al.,1995,1996; Reinschmidt et al.,1985; Streisinger et al.,1986; Thiebaud et al.,1984), with centromeric loci giving up to 50% mutant gynogenotes, decreasing to a plateau of ∼10% (∼40 cM) beyond which multiple intervening crossovers predominate (Khokha et al.,2009). Second, even distal mutations usually appear linked to one of the 10 X. tropicalis centromeres when analyzed in pools of gynogenetic embryos. With the exception of very telomeric loci, most homozygous mutant gynogenotes will have no crossovers between the mutant locus and their chromosome's centromere, and even when multiple intervening crossovers are present, half of these restore parental linkage. Linkage to a specific centromere is only disrupted when a locus is so distal that multiple intervening recombination events are present in the majority of mutant gynogenotes. In practice, assigning a mutation to a specific chromosome usually requires only interrogating two pools of mutant and wild type gynogenote DNA with polymorphic markers corresponding to the 10 X. tropicalis centromeres (Fig. 3e). The genetic locations of the 10 X. tropicalis chromosomes have been identified and confirmed cytogenetically (Khokha et al.,2009) and used to map mutations to specific chromosomes in a small number of PCR reactions (Geach and Zimmerman,2010).
Higher resolution mapping to specific genes can be accomplished using conventional positional cloning methods as well as candidate gene approaches (Abu-Daya et al.,2009; Geach and Zimmerman,2010). Efficiency of these strategies has greatly increased with improved genome contiguity and annotation and meiotic map marker density, and they remain useful for evaluating candidate genes. However, since these approaches are shared by other systems, and because they are being replaced by sequence-based direct identification/mapping, conventional high-resolution mapping will not be discussed further here.
Reverse Genetics and Rapid Mapping by High Throughput Sequencing
In contrast to conventional “forward” genetics (isolate a phenotype, then identify the responsible gene), reverse genetics refers to engineering or identifying a specific mutated sequence for subsequent functional analysis. While antisense morpholino oligonucleotides can effect knockdown of specific gene functions in early development, later embryonic events are less accessible, off-target effects can be problematic, and it can be challenging to obtain reproducible and complete depletion of gene products.
Next generation sequencing technologies have revolutionized mutation identification for reverse genetic strategies. In one strategy, Targeting Induced Local Lesions In Genomes (TILLING), animals carrying mutations in genes of interest are identified from mutagenesis-derived populations. TILLING was initially developed using non-sequence based methods (McCallum et al.,2000; Till et al.,2003), and adapted for X. tropicalis using capillary sequencing of specific amplicons (Goda et al.,2006). However, PCR amplification introduces allele bias and other errors, hindering high-throughput screens (Showell et al.,2011).
“Exome sequencing” has provided a critical breakthrough. Focus on the genome's protein-coding space is obtained using solution hybridization of sheared genomic DNA to a set of synthetic baits representing the entire exon set (“exome,” Fig. 3i) or a subset of specific genes of interest (Fairfield et al.,2011; Gnirke et al.,2009). Coding regions of individual mutagenesis-derived carrier animals can then be subjected to high throughput sequencing and bioinformatically inspected to identify the mutations they carry for either forward or reverse genetic applications. With collaborators, our groups have used this method to identify multiple chemically induced nonsense and essential splice mutations per F2 frog, and to confirm heritability of these lesions in the next generation. In addition to reverse genetic applications, mutations found by forward genetics can often be directly identified by exome sequencing. One variation involves simultaneous direct identification and mapping, by exome sequencing pools of phenotypically mutant embryos (Fig. 3h–k). Even if assignment of a responsible mutation is ambiguous, linkage information is obtained from the increasing SNP homozygosity in the mutation-containing interval.
Gene Editing In Vivo with Sequence-Specific Nucleases
Although TILLING by exome sequencing is now recovering large numbers of X. tropicalis mutations, there has also been a breakthrough in targeted mutation of specific genes (“genome editing”) using synthetic sequence-specific enzymes such as Zinc Finger Nucleases (ZFNs). The conventional ZFN design is a fusion of the non-specific nuclease domain of Fok1 and zinc-finger motifs engineered to bind and digest a specific genomic target (Miller et al.,2007; Urnov et al.,2010). The resulting double strand breaks are usually imperfectly repaired, with insertions and deletions causing frameshift mutations (Bibikova et al.,2002).
Young et al. (2011) used this method to heritably mutate the noggin locus as well as a GFP transgene in X. tropicalis. After careful design and precharacterization of ZFNs in yeast reporter assays, they injected ZFN mRNA into fertilized eggs, resulting in the visible mosaic deletion of the GFP reporter and recovery of mutated noggin alleles in the next generation. While building and validating these ZFN constructs is laborious, several variations in sequence-specific nuclease design are under active development (Christian et al.,2010; Kim et al.,2009; Maeder et al.,2009; Meng et al.,2008; Miller et al.,2011) and show promise for targeted mutagenesis in Xenopus.
Insights from Mapped Mutations
A number of groups have now mapped genes underlying mutations from X. tropicalis forward screens, and it is possible to make some general observations, illustrated with vignettes. First, forward screens efficiently recover disease models: >70% of mapped mutations (10/14) correspond to orthologs of human disease genes. Second, efficient gynogenetic screens recover mutations throughout the genome, including very telomeric regions (e.g. komimi, ∼40cM or 70 Mb from the centromere). Third, X. tropicalis screens are not redundant with zebrafish screens: fewer than half of the identified phenotypes have fish correlates (5/14).
One process where fish screens offer few insights is limb development. Limb formation in metamorphosing frogs is unique among vertebrate organogenesis processes, which otherwise occur during a brief coordinated period in early development. In contrast, X. tropicalis tadpoles grow for several weeks as feeding, swimming larvae with all other organ systems functioning before the first limb buds appear, eventually giving rise to appendicular skeletal structures homologous to those of mammals and other tetrapods. The xenopus de milo (xdm) mutation is caused by a random reporter transgene insertion disrupting the nephronectin (npnt) locus (Abu-Daya et al.,2010), resulting in a specific and complete abrogation of forelimb bud formation in homozygous tadpoles (Fig. 4a–d). This effect was upstream of the earliest known marker of the forelimb field, tbx5, but did not perturb hindlimb formation. In mouse, npnt is a secreted α8β1 integrin ligand required for development of the metanephros, a renal organ not found in frog, but mouse limbs were not affected by the knockout (Brandenberger et al.,2001; Linton et al.,2007), and xdm pronephros and mesonephros appear normal. This striking frog phenotype identifies a novel step upstream of the earliest previously-known gene function in limb subtype initiation.
Muscle and Cardiac Formation
Screens have identified a number of X. tropicalis mutations disrupting skeletal and/or cardiac muscle function both because paralysis phenotypes are easily scored, and because sarcomere assembly requires many unusually large proteins, providing a good genetic target. Xenopus embryos are particularly well suited to cardiogenesis studies, because in the absence of heart function they survive to swimming tadpole stages on diffused oxygen, unlike mammals where contractility defects cause early lethality. The three-chambered amphibian heart also more closely resembles that of mammals than the two-chambered fish heart.
The muzak (muz) mutation completely lacks heartbeat, and is caused by a truncation of the cardiac specific myosin heavy chain gene myh6 (Fig. 4e–h); skeletal muscle is unaffected (Abu-Daya et al.,2009). Muzak permits study of the roles of heartbeat, bloodflow, and myofibrils in cardiac morphogenesis. While cardiac myh mutants have been identified in zebrafish, teleost subfunctionalization has produced distinct atrial and ventricular myosin heavy chains, and awkward double mutants are required for analysis of a comparable phenotype. Despite loss of a major structural component and complete ablation of beating, early steps in cardiac development such as looping and chamber formation occur surprisingly normally, but later features such as valves and ventricular trabeculation were not detected. By reconstructing serial plastic sections of cardiac regions at different stages, it was possible to create 3-D projections of heart development in normal and non-contractile mutant hearts. In the absence of myh6, the mutant heart showed dilated cardiac chambers and malformed endocardium, which is compressed with drastically reduced lumen (Fig. 4g,h).
Likewise, Xenopus tadpoles excel in studies of skeletal muscle formation. As somites are added sequentially during development, multiple stages in sarcomerogenesis can be observed in a single embryo, conveniently imaged in the transparent tadpole tail. Paralysis phenotypes specifically affecting muscle structure are easily diagnosed by reduced birefringence under polarized light (Fig. 4i,j). The dicky ticker (dit) mutation, characterized by complete paralysis and lack of heartbeat (Goda et al.,2006), maps to the unc45b muscle specific myosin chaperone (Geach and Zimmerman,2010). At early stages of myofibrillogenesis in dit, myh protein can be detected in poorly ordered fibres, and a few thick filaments are present (Fig. 4k,l), but appearance of Z-discs and associated α-actinin staining was substantially delayed. This suggests modifications to current models of myofibrillogenesis, which assume that Z-discs form early and independent of sarcomeric myosin, are needed.
Inner Ear Development
At tailbud stages, the embryo's head becomes transparent as yolk is metabolized, facilitating observation of sensory structures such as the inner ear. Screens have identified many mutations affecting formation of the otoconia, calcitic crystals attached to inner ear stereocilia which provide mass for detection of balance and acceleration (Fig. 4m–p); otoconial phenotypes usually also result in behavioral abnormalities in swimming and orientation. Correct formation of otoconia reflects proper otic vesicle patterning as well as extensive intracellular trafficking to secrete proteins required for crystal formation. The seasick (ssk) mutation (Fig. 4n) is characterized by overgrowth of individual crystals, and is caused by a mutation in the vesicle transport adaptor protein subunit ap3δ1, part of the human disease complex Hermansky-Pudlack Syndrome; no zebrafish models for the HPS complex are currently known. Consistent with a vesicle transport disorder, ssk embryos also show punctate distribution of pigment-bearing melanosomes in melanocytes, although the stellate melanocyte morphology is not affected. In another mutant, komimi (kom), a single large otoconium forms over the two otic sensory epithelia (Fig. 4o). Komimi results from a splicing mutation in otoconin 90, previously thought not to be a major protein in frog otoconia (Poteet al., 1993), but here shown to be critical for proper formation. Another phenotype, bunny (bun, Fig. 4p), displays a smaller otic vesicle with a single sensory epithelium, probably reflecting an earlier defect in patterning the otic vesicle.
Cilia have emerged as a key cellular structure for receiving and integrating intercellular signals, in addition to their other medically important roles as motile structures lining the mammalian airway, kidney, and elsewhere. Ciliogenesis is an ideal genetic target in that it requires hundreds of genes, many of which have poorly characterized functions, and is uniquely accessible in the tadpole's external mucociliary epidermis (Hayes et al.,2007). Functional motile epidermal cilia generate an easily detected fluid flow, causing even anaesthetized embryos to “glide” across surfaces, or can be assayed using microscopic beads. The grinch mutation (Grammer et al.,2005) abrogates epidermal ciliary flow, and produces stubby cilia which are reduced in number in epidermal multiciliate cells (Fig. 4q–t).
While X. tropicalis genetics effectively recovers novel gene functions and disease models, it is clear that few Xenopus groups have the resources to initiate large-scale screens. However, NIH support for TILLING projects and stock centers for strain distribution means that the frog community will soon have access to a large number of identified null mutations as well as lines made by targeted gene editing with synthetic sequence-specific nucleases. Attractive focused applications of these loss-of-function genetic backgrounds include rescue with engineered BAC transgenes, for example to create conditional floxed alleles for analysis of specific later steps in development, tagged proteins for biochemical analysis, or gain-of-function suppresser/enhancer for interacting factors in complex processes (Bai et al.,2010). Stock center-based TILLING programs may also be able to give other genetic projects a one- or two-generation head start, by making excess mutagenized stock available for other types of screens. The simplicity of haploid genetics and gynogenesis, together with direct sequencing to identify mutations, means that even small-scale screens can recover useful phenotypes for studying vertebrate gene function.
For animated discussions and seminal contributions, authors would especially like to thank Richard Harland, Amy Sater, Enrique Amaya, Rob Grainger and Takuya Nakayama, and Derek Stemple and Samantha Carruthers, as well as Holly Ironfield for unpublished inner ear mutant data and other past and present members of our laboratories.