Abstract
The heterochronic gene lin-14 controls the temporal sequence of developmental events in the C. elegans postembryonic cell lineage. It encodes a nuclear protein that is normally present in most somatic cells of late embryos and LI larvae but not in later larval stages or adults. Two lin-14 gain-of-function mutations cause an inappropriately high level of the lin-14 nuclear protein late in development. These mutations delete 3′ untranslated sequences from the lin-14 mRNAs and identify a negative regulatory element that controls the formation of the lin-14 protein temporal gradient. The 21 kb lin-14 gene contains 13 exons that are differentially spliced to generate two lin-14 protein products with variable N-terminal regions and a constant C-terminal region. No protein sequence similarity to any proteins in various databases was found.
The temporal and cellular expression patterns of lin-14 protein accumulation is altered by mutations in the heterochronic genes lin-4 and lin-28. The lin-4 gene is required to down-regulate lin-14 protein levels during the mid-Ll stage. The lin-4 gene product could be the trans-acting factor that binds to the negative regulatory element in the lin-14 3′ untranslated region. In contrast, the lin-28 gene activity positively regulates lin-14 protein levels during early LI. Thus, these genes act antagonistically to regulate the lin-14 temporal switch.
The normal down-regulation of lin-14 within 10 h of hatching is not determined by the passage of time per se, but rather is triggered when feeding induces post-embryonic development. Loss of lin-28 gene activity causes precocious down-regulation of lin-14 protein levels before feeding, whereas loss of lin-4 gene activity does not affect the level of lin-14 protein before feeding. These data suggest that to trigger the lin-14 temporal switch, the lin-4 gene is up-regulated after feeding which in turn down-regulates lin-14 via its 3’ untranslated region.
We speculate on the evolutionary implications of dominant mutations in pattern-formation genes.
Pattern-formation genes
During the ontogeny of multicellular animals, diverse cell types are generated from the zygote in an intricate series of cell divisions and differentiations. Over the last 20 years, genes that control this process have been discovered and isolated from Drosophila and Caenor-habditis elegans (Sternberg and Horvitz, 1984; Ingham, 1988; Scott et al. 1989). Molecular analyses of these pattern-formation and cell-specification genes have shown that in many cases these genes encode transcription or RNA processing factors (Costa et al. 1988; Finney et al. 1988; Scott et al. 1989; Bell et al. 1988) or genes involved in extracellular or cell-cell signal transduction pathways (Greenwald, 1985; Hafen et al. 1987). Genes homologous to these invertebrate patternformation genes have been isolated from mammals, and are also thought to control development (Graham et al. 1989; Balling et al. 1988).
In many cases, these pattern-formation genes are expressed or become active in spatial domains that presage the generation of differentiated cells and structures (Akam, 1987). Mutations in these genes that disrupt their normally asymmetric pattern of expression or activation have been shown to lead to homeotic changes (Ingham, 1988; Schneuwly et al. 1987; Costa et al. 1988; Ruiz i Altaba and Melton, 1989).
Just as pattern-formation genes define and interpret spatial information during development, in C. elegans the temporal sequence of developmental events has been shown to be explicitly controlled by heterochronic genes (Ambros and Horvitz, 1984, 1987; Ruvkun and Giusto, 1989). Heterochronic mutations cause many blast cells to undergo patterns of cell division and differentiation normally observed at distinct developmental times, suggesting that these mutations perturb the definition or interpretation of developmental time (Ambros and Horvitz, 1984). Thus, both the time and space dimensions during development are defined by pattern-formation genes.
These discoveries about the molecular nature of pattern-formation genes allow us to ask very direct questions about the molecular mechanisms operating during development. We must explain how these genes or their products come to be activated or expressed non-uniformly, and we must explain how these gene activities that affect groups of cells interact to specify particular cell fates.
Here we present our analysis of how developmental time is specified by the heterochronic gene lin-14 to form a temporal developmental switch, and how other heterochronic genes regulate the lin-14 temporal switch.
Genetic and molecular studies of heterochronic genes in C. elegans
After 12 h of embryogenesis, the 550-cell LI stage larvae of C. elegans hatch. About 80 of these cells will divide again over the next larval stages, to generate cell lineages that are in most cases distinct from those generated by blast cells at different times and locations. Many of these blast cells are shown in Fig. 1. Heterochronic mutations cause most of these blast cells (for example, the intestinal cells (E lineage), the mesoblast (M lineage) and the hypodermal cell lineage (HO, Hl, H2, VI to V6, and T lineages)) to adopt fates normally associated with cells at earlier or later stages of development.
The lin-14 gene
The lin-14 heterochronic gene plays a central role in this temporal regulation of the cell lineage. Loss-of-function lin-14 alleles cause the precocious execution of cell lineages normally observed in descendent cells one or two larval stages later. Gain-of-function lin-14 alleles affect the same cell lineages but cause the opposite transformations in cell fate: early cell lineages are normal, but later cells reiterate the early cell lineages, normally associated with their ancestor cells. For example, in the development of the lateral hypodermis, in lin-14 loss-of-function mutants, the blast cell T skips its characteristic larval stage 1 (LI) sublineage and instead expresses a sublineage normally associated with its granddaughter blast cell T.ap (Fig. 2). In lin-14 gain-of-function mutants, the blast cell T expresses its normal Ll-specific sublineage, but its granddaughter T.ap reiterates this Ll-specific sublineage normally associated with blast cell T (Fig. 2) (Ambros and Horvitz, 1984). The lin-14 gain-of-function mutations have been shown by genetic criteria to cause excess lin-14 gene activity, while the lin-14 loss-of-function mutations are due to loss of lin-14 gene activity (Ambros and Horvitz, 1987). These data suggest that during normal development, a relatively high lin-14 gene activity during early larval stages, for example in cell T, is reduced later in development, for example in cell T.ap, to form a temporal developmental switch. In an analogous way the lin-14 gene coordinately controls the postembryonic fates of cells in a number of lineages: the E, M HO, Hl, VI to V6, and T lineages.
A regulatory hierarchy of heterochronic genes
Other genes have been shown to act in the same pathway as lin-14 to control the temporal fates of these same cells or a subset of them, and their epistatic interactions have been determined (Ambros and Horvitz, 1984; Ambros, 1989). A recessive loss-of-function mutation in the gene lin-4 causes the same retarded phenotype as lin-14 gain-of-function mutations (Fig. 2). In addition, the lin-4 phenotype depends on a functional lin-14 gene: lin-14 loss-of-function mutations are epistatic to lin-4 (Ambros, 1989). These data show that the lin-4 gene negatively regulates the lin-14 gene, though they do not suggest whether this is direct or at what level the interaction occurs.
lin-28 mutations are not as pleiotropic as lin-14 or lin-4 mutations: these mutations cause precocious expression of later fates in P, HO, Hl, VI to V6, and T hypodermal lineages but not in the M lineages or E lineages (muscle or intestine) (Fig. 2). Mutations in lin-28 are epistatic to lin-4 or lin-14 gain-of-function mutations in these hypodermal lineages (Ambros, 1989). Thus the lin-28 gene is necessary for the lin-14 gene activity in the hypodermal lineages.
Loss-of-function mutations in the lin-29 gene only affect larval stage 4 cell fates in the hypodermal lineages: these cells continue to express L4 fates during adult stages in lin-29 mutants, lin-29 mutants cause this phenotype in double mutant combinations with any of the other heterochronic genes, suggesting that it is farthest downstream in the pathway and directly regulates an L4/adult switch.
The lin-14 gene encodes a nuclear protein that forms a temporal molecular gradient
We cloned the lin-14 gene by genetically mapping in parallel many linked DNA polymorphisms (RFLPs) associated with the transposon Tel to find the two closest to the gene (Ruvkun et al. 1989). We isolated lin-14 cDNA clones and fused these to the E. coli lacZ gene to produce ant\-lin-14 antibodies so that the temporal, spatial, cellular and subcellular distribution of lin-14 protein could be followed. These antl-lin-14 antibodies detect specific somatic nuclei in wild-type preparations but not in strains bearing lin-14 null alleles, showing that the antibody is specific for lin-14 protein (Ruvkun and Giusto, 1989).
During wild-type development lin-14 protein was first observed in embryos about half-way through embryogenesis and staining was most intense in nuclei of late embryos just before hatching, and in newly hatched LI animals, lin-14 protein was present in the nuclei of all postembryonic blast cells affected by lin-14 mutations: the hypodermal blast cells Hl, H2, VI to V6, and T, the intestinal (E) cells, both neuroblasts QR and QL, the mesoblast M cell, and in the P cells. No staining was observed in the somatic or germ line gonadal blast cell nuclei Z1 to Z4, which are not affected by lin-14 mutations.
The blast cells containing lin-14 protein execute Ll-specific lineages and their daughter cell nuclei initially contain lin-14 protein but the level falls before the next cell division (Fig. 2). Late in the LI stage, the lin-14 protein staining in all nuclei rapidly falls. By L2 and in subsequent larval stages, only barely detectable lin-14 protein staining remains in some neuronal nuclei (Arasu et al. 1991). Western blotting has shown that the level of the lin-14 protein decreases by a factor of more than 25 from LI to L2.
Gain-of-function lin-14 mutations dramatically alter the temporal regulation of lln-14 protein levels
Gain-of-function lin-14 mutations cause reiterations of early larval cell lineages at late larval and adult stages in a number of cell lineages and tissues (Ambros and Horvitz, 1984). Embryonic and LI stage staining with the anti-lin-14 antibodies in these mutants was equivalent to wild type in amounts, nuclear localization and cellular distribution of the protein. However, unlike wild type, these mutants showed lin-14 protein at high levels in many nuclei during larval stages 2, 3, 4 and in adults. All of the postembryonic hypodermal blast, intestinal blast, neuroblast and mesoblast cells known to be affected by gain-of-function lin-14 mutations inappropriately accumulate the lin-14 protein at these late stages. For example, in the case of the T lineage, during normal development, the lin-I4 nuclear protein is present at high levels in the T cell but is not observable in cell T.ap (Fig. 2). In lin-14 gain-of-function mutants, this temporal gradient is disrupted: the lin-14 protein is now observed at all stages of development (Ruvkun and Giusto, 1989), for example in both cells T and T.ap (as well as T.apap, etc.), and these cells reiterate Ll-specific cell lineages (Fig. 2).
Thus the normally sharp decrease in lin-14 protein levels during the LI stage causes cells to switch from Ll-specific cell lineages to L2-specific cell lineages, and in lin-14 gain-of-function mutants, the inappropriate presence of the lin-14 nuclear protein late in development prevents this temporal switch in cell fate.
The lin-14 gain-of-function mutations are located in the 3′ untranslated region of the lin-14 mRNA
The n355 mutation is an insertion or inversion of at least 10kb of unknown DNA sequences 256 bases 3’ to the termination codon of the lin-14 protein coding region common to both lin-14 transcripts (Fig. 3). The other lin-14 gain-of-function mutation, n536, is a 607bp deletion, 300 bases downstream from the 3’ end of the lin-14 open-reading frame (Fig. 3) and overlaps the region which is rearranged in the n355 mutation.
The location of both lin-14 gain-of-function mutations in the 3′ untranslated region suggests that they do not affect the lin-14 protein. Immunoblot analysis of the lin-14 proteins from wild-type and both the gain-of-function mutants confirmed this prediction; no difference in the size of the lin-14 proteins was observed in these mutants (Wightman et al. 1991).
Computer analysis of the lin-14 mRNA sequence identified a stem-loop structure with 21 out of 22 matching pairs in the main stem and side stems of 9 and 10 base pairs located about 60 bases downstream from the end of the lin-14 protein coding region and about 80 bases upstream of the n355 breakpoint. This structure would be expected to have a free energy of −38.3 kcal mole−1, about 3× that of a randomized sequence of the same base composition. The location of this structure suggests that it would not be directly affected by the lin-14 gain-of-function mutations.
Thus the 3′ region of the lin-14 transcripts contains an element that encodes the down-regulation of lin-14 protein levels after larval stage 1. Posttranscriptional regulation of translation or transcript stability has been demonstrated in the 3′ UTR of other eukaryotic genes (Casey et al. 1988) and proteins have been identified that bind to those elements (Leibold and Munro, 1988). In HIV, the rev transactivator protein has been shown to interact with an RNA structure in the 3′ UTR of the viral transcript (Zapp and Green, 1989) and to mediate export of the mRNA from the nucleus (Emerman et al. 1988). Thus, sequences in the 3′ UTR can regulate export of a transcript from the nucleus, the half-life of the transcript, or translation of that transcript.
Temporal down-regulation of lin-14 expression is initiated by a developmental cue
When newly hatched C. elegans LI animals are starved, they do not begin postembryonic development, but instead can suspend development for up to 5 days, when upon feeding, they will reinitiate postembryonic development. Because the level of the lin-14 protein normally falls within 12 h of hatching, we investigated whether the passage of time in these suspended LI s causes the lin-14 protein levels to fall. We found that both the lin-14 protein levels and the lin-14 transcript levels are maintained at the high level of normal early LI animals in these starved Lis (Arasu et al. 1991). Thus the lin-14 down regulation that is necessary for normal developmental timing does not respond to clock time, but instead must first respond to developmental cues.
The high level of lin-14 protein expression in these starved Lis could reflect a very stable lin-14 protein or mRNA that is only destablized upon feeding, perhaps via induction of a protease or nuclease. We addressed this by inhibiting translation with cycloheximide and found that the level of the lin-14 protein quickly fell to zero, even though myosin levels remained constant. The level of the lin-14 mRNA also remains at normal levels in starved Lis and cycloheximide-treated starved Lis. These data suggest that lin-14 translation continues in these starved L1s.
How other heterochronic genes interact with lin-14 to generate the temporal gradient or interpret it
Other heterochronic genes could control developmental timing of the C. elegans cell lineage by participating in the generation or reception of the lin-14 temporal gradient. The lin-4 gene is necessary for the down regulation of lin-14 protein levels: inappropriate lin-14 protein staining at late stages is observed in this mutant (Fig. 2) (Arasu et al. 1991). Thus, directly or indirectly, lin-4 negatively regulates lin-14. It is possible that the lin-4 protein directly interacts with the lin-14 3’ regulatory sequences.
Mutations in lin-28 lead to precocious expression of L3-specific hypodermal cell lineages during the L2 stage, and so are similar to lin-14 mutations in these hypodermal cell lineages only (Ambros and Horvitz, 1984). The level of Un-14 protein is decreased in lin-28 mutants initially in the hypodermal cell lineages but in all cell lineages of starved LI animals, suggesting that the gene acts upstream of lin-14 (Arasu et al. 1991). We have not yet determined at what level the lin-28/lin-14 gene interaction is taking place, lin-28 could activate lin-14 transcription or translation, or stabilize the lin-14 protein by, for example, a heterodimer interaction.
The lin-14 gene encodes two protein products
The lin-14 DNA sequence revealed 13 exons in the lin-14 gene, with introns ranging in size from about 12 kb to 44 bp. Differential splicing of exons 1,2, and 3 (the Un-143 transcript) or exon 4 (the lin-14A transcript) to the common exons 5 to 13 was observed in lin-14 cDNAs (Fig. 3). RNAase protection experiments using probes to exons 1,2, and 3 and exons 4 and 5 confirmed this differential splicing and showed that the two transcripts are expressed in wild type at similar levels and show the same temporal regulation of transcript levels.
The longest open reading frame in the Un-143 transcript yields a protein of 539 amino acids. The lin-14A. transcript encodes a 537 amino acid lin-14 A protein. This protein has a 63 amino acid N-terminal domain that is distinct from the 65 amino acid N-terminal domain of the Un-143 protein. No similarity between the two amino-termini of the lin-14 proteins could be detected.
Databank searches using the amino acid sequences of both lin-14 proteins revealed no significant sequence similarity to any proteins. A 20 amino acid region (positions 417 to 436), common to both lin-14 proteins, could potentially form an amphipathic a-helix with two basic regions separated by an acidic region on the hydrophilic face (Fig. 4) (Wightman et al. 1991). This motif could encode a DNA- or RNA-binding domain.
The lin-14 proteins also contain a high proportion of prolines, serines, threonines and glutamates, or PEST sequences, as has been observed in various unstable proteins (Rogers et al. 1986; Nash et al. 1988). Given that the protein appears to be quickly degraded late in larval stage 1 to control an L1/L2 switch in cell fates (Ruvkun and Giusto, 1989), these sequences could mediate that instability.
Discussion
Generation of the lin-14 temporal gradient
The DNA sequence of the lin-14 gene from wild-type and three lin-14 mutants, and its expression pattern in wild type and various mutants, has revealed aspects of how the lin-14 protein gradient is generated (Fig. 5). First, during normal development the down regulation of lin-14 protein levels begins only after a feeding signal initiates postembryonic development. Both the lin-14 and the lin-28 gene activities are necessary to maintain the level of lin-14 protein high before this food signal. After initiation of postembryonic development, down-regulation of lin-14 is triggered. The lin-4 gene activity is necessary for this down-regulation, and may in fact be up-regulated by feeding or postembryonic development. This leads to a 25-fold decrease in lin-14 protein level from LI to L2 that causes cells to switch to L2-specific fates, lin-14 gain-of-function mutations abrogate the negative regulation of lin-14 protein levels. This failure to reduce markedly the lin-14 protein levels prevents or delays the normal LI to L2 switch in cell fates in these mutants.
The stability of the lin-14 protein is also relevant to the formation of the lin-14 temporal gradient. The half-life of the previously synthesized lin-14 protein must be less than 1 h to account for this observed rate of disappearance. The lin-14 protein levels decrease at the same rate in both dividing cells and non-dividing cells. This suggests that, at least in the non-dividing cells, breakdown of the nuclear membrane is not necessary for the degradation of the lin-14 protein and that nuclearly localized proteases must control this process (Ruvkun and Giusto, 1989). The presence of PEST sequences in the lin-14 protein supports the notion that the rate of lin-14 protein degradation is relevant to the formation of the lin-14 protein gradient.
While the DNA sequence of the lin-14 gene did not reveal any homology that would suggest its molecular mechanism, the nuclear localization of the lin-14 proteins suggests that they may regulate the pattern of gene expression of the cells that accumulate them. The observation that the lin-14 protein is normally present only in embryos and larval stage 1 animals suggests that it either activates early genes or represses late genes so that the disappearance of lin-14 after larval stage 1 causes a transition from the expression of early cel) lineage genes to late cell lineage genes.
The particular early or late cell fate specified by the level of lin-14 gene activity is distinct for many of the postembryonic cell lineages affected by lin-14 mutations, although the fates inappropriately executed in these mutants are always fates normally executed by a closely related descendent or ancestor cell (Ambros and Horvitz, 1984). The lin-14 gene product may function to convey general temporal information to these cell lineages. The specific response made by each cell must be caused by unique properties of that cell, defined by other developmental control genes, that either modify the lin-14 signal or cause cells to interpret it differently. It is possible that this difference between cells is reflected in the distinct spectra of previously specified or partitioned nuclear factors which interact with the lin-14 nuclear protein to cause a distinct response to this temporal signal in nearly every cell. The best current model for this type of signal integration in cell fate specification is from yeast: the three yeast cell types are specified in a combinatoric manner by the MATal and MATn-2 nuclear proteins (Goutte and Johnson, 1986).
Pattern-formation genes and evolution
The spatial and temporal asymmetries in the patterns of developmental control gene activity during ontogeny has been shown to cause cells or groups of cells to become different from each other (Ingham, 1988). Mutations that change the expression or activation pattern of patternformation genes in genetically studied animals such as Drosophila and C. elegans lead to major changes in the morphology of the organisms, in many cases deleting structures and/or duplicating groups of cells and structures (Sternberg and Horvitz, 1986; Akam, 1987). For example, the Drosophila Antennapedia mutant adds a nearly complete leg where an antenna is normally located (Dennell, 1973), and particular bithorax mutants can add another pair of wings to a thorax segment that normally has no wing (Lewis, 1982). The lin-14 gain-of-function heterochronic mutant duplicates entire sets of muscle, hypoderm, endoderm and neuronal cells, and is missing other such cells.
These are the same sorts of homeotic and heterochronic variations that have been observed in phylogeny and point to mutations in pattern-formation genes as a major cause of the variation necessary for evolutionary change (Raff and Kauffman, 1983). For example, the C. elegans heterochronic mutations are analogous to the heterochronic variation between species noted in phylogenetic studies (Gould, 1977). This heterochronic variation in evolution could be due to mutation in one or a few heterochronic genes like lin-14, rather than many mutations that incidentally change developmental timing. More generally, mutations that change the spatial, temporal, or cellular asymmetries in patternformation gene activities may be the underlying cause of the many forms of metazoans and may be a significant force in evolutionary change.
While the most frequent class of mutations are those that lead to a decrease in gene activity, many mutations that cause an increase or inappropriate activity of pattern-formation genes have also been isolated. In the case of the lin-14 heterochronic mutations, deletion of the lin-14 3’ untranslated regulatory region causes inappropriate expression of the lin-14 protein at late developmental stages, leading to reiteration of early cell lineages. The Drosophila Antennapedia mutation has been shown to arise by a chromosomal inversion that causes more promiscuous expression of the gene due to fusion of the protein-coding regions to a novel regulatory region (Schneuwly et al. 1987). This causes the Antp protein to be uniformly expressed rather than expressed only in the thorax region.
The misregulated expression of pattern-formation genes does not completely disrupt all cell identity specifications during development. Rather, only a limited number of structures or cells are affected. For example, uniform expression of the Antp protein transforms only the antenna to a leg, but does not affect other segments (Schneuwly et al. 1987). Similarly, uniform expression of one of the bithorax complex proteins causes segmental transformations of only particular cells within particular segments (Mann and Hogness, 1990). Uniform expression of the int-1 oncogene in Xenopus causes production of two organizers, leading to a duplication of the embryonic axis (McMahon and Moon, 1989). Thus only some cells are competent to respond to the presence of these pattern-formation gene products so that uniform expression causes a nonuniform response. This is most likely due to the combinatorial nature of pattern-formation genes: a spectrum of proteins is necessary to specify particular cell fates so that only cells already containing the interacting proteins will be responsive to the inappropriate presence of one such gene product. These data suggest that discrete developmental variation can result from mutations that lead to uniform expression of a particular pattern-formation gene.
Gain-of-function mutations that lead to misregulated gene expression or activity can also arise in any gene or gene product that contains a negative regulatory domain. For example, gain-of-function mutations in the C. elegans sex determination genes tra-2 and fem-3 that transform hermaphrodites into pseudomales or females, are located in a negative regulatory region of the 3′ end of the transcript, downstream of the protein coding regions (Rosenquist and Kimble, 1988; Okkema and Kimble, personal communication). Similarly, 3’ mutations in the vertebrate oncogene c-fos lead to activation of this protooncogene, possibly by stabilizing the mRNA (Curran et al. 1985). A gain-of-function mutation in the Drosophila sex determination gene Sxl has been shown to disrupt a region that negatively regulates one of the spliced mRNAs, leading to expression of the normally XX(female)-specific Sxl protein in XO males (Bell et al. 1988). Gain-of-function mutations in the C. elegans lin-12 (Greenwald and Seydoux, 1990) and Drosophila Notch (Kelley et al. 1987) genes that control particular cell-cell signaling during development have been shown to be point mutations in protein coding regions, thus identifying a domain which negatively regulates the activity of these proteins. Negative regulation is a common form of feedback control, and those regions of genes, mRNAs and proteins that mediate down regulation would be the targets for such mutations. Dominant mutations have also been shown to result in mutant gene activities that interfere with the wild-type gene activity (Muller, 1932; Park and Horvitz, 1986), or that result in entirely new gene activities (Muller, 1932), or that reduce or eliminate gene activity (Muller, 1932).
Such gain-of-function mutations in pattern-formation genes lead to genetically dominant morphological transformations. The genetically dominant nature of such pattern-formation gene mutants has enormous evolutionary implications. Unlike recessive mutations, dominant mutations need not be homozygous to affect the phenotype. The dominant mutation will be segregated to and cause a phenotype in 1/2 of the progeny of a heterozygous carrier, independent of effective breeding populations. In this way, evolutionary change can arise and a selectively advantageous allele can become more frequent in a large non-isolated population.
The evolutionary implications of gain-of-function mutations seem to have been overlooked by population biologists and evolutionary theorists. While subtle recessive mutations are sufficient to explain the observed gradual variation between races and related species, sudden morphological changes have been noted in phylogeny and have been a major problem for evolutionary theories that suppose gradual change via recessive mutations. One camp of paleontologists resolves this conflict by stressing that the sudden changes are artefacts of the incomplete geological record and that evolutionary change indeed proceeds by incremental steps (Grant, 1985). On the other hand, advocates of punctuated evolution have stressed that bursts of major structural changes in the paleontological record are not due to gaps in the record but in fact accurately present phylogenetic history (Eldredge and Gould, 1972). For these rapid evolutionary changes, they argued that small isolated populations would be able to generate morphological variants due to a relatively high frequency of production of homozygous recessive mutants, and that the theory of allopatric speciation could explain the sudden appearance of such a new form in the paleontological record: the variant would arise and define a new species in a geographically isolated region and then suddenly (in geological time) appear in larger areas.
In contrast, we propose that because genetically dominant mutations in pattern-formation genes have been shown to be relatively common and to cause significant morphological changes, sufficient variation for sudden evolutionary change could take place within large breeding populations. Thus, due to mutations in patternformation genes, even a large outbred population would produce a constellation of morphological variants, most of which are not viable. Rarely, perhaps due to recombination of two such mutations or due to a change in environment, viable and fertile mutants would be selected and either displace or speciate from the original form. Thus there is a molecular genetic mechanism to generate the much maligned ‘hopeful monsters’ of Goldschmidt (Raff and Kaufmann, 1983). These major variants could be the vanguard of major evolutionary events, like the flowering of metazoans during the Cambrian Period or the mammalian radiation.