Investigating C. elegans development through mosaic analysis

C. elegans derives its power for elucidating animal development from its suitability for mutant analysis(Brenner, 1974). A gene reveals itself as important for development when its loss of function causes a developmental phenotype. Genetic mosaics allow one to ask what the phenotypic consequences are when some cells in an animal carry a wild-type gene and the other cells are homozygous mutant. By analyzing a collection of genetic mosaics, one can ask which cell or cells in an animal must carry the wild-type gene to produce a wild-type phenotype. The responsible cell or cells are referred to as the anatomical focus of the gene's action or function with respect to the phenotype under study.

As the anatomical focus of a gene's function may well be in only a subset of cells in which the gene is expressed, knowing its complete expression pattern does not necessarily tell one where the gene normally functions with respect to a particular phenotypic effect. Furthermore, the anatomical focus of a gene's action may or may not be in cells that exhibit a mutant phenotype. A gene is said to act cell autonomously when the phenotype of a given cell is affected by the gene's mutation and when that phenotype in mosaic animals depends solely on whether or not the cell has the wild-type gene and is unaffected by the genotypes of other cells. A gene is said to act cell nonautonomously when a cell carrying the wild-type gene in a mosaic animal exhibits a mutant phenotype or when a homozygous mutant cell exhibits a wild-type phenotype. Cell non-autonomy implicates cell-cell interactions, and mosaic analysis can be used to identify the responsible interacting cells. Some genes are used repeatedly during development, and the loss of an early essential role of such a gene in a mutant can lead to developmental arrest that precludes analyzing the gene's role in later developmental events. This problem can be overcome through mosaic analysis, as mosaic animals may be able to complete the early stages of development to reveal a gene's later role.

In nearly all mosaic analyses that have been carried out in C. elegans, mosaic animals have been generated by the spontaneous mitotic loss of an extrachromosomal genetic element that carries the wild-type allele of a gene in an otherwise homozygous mutant background. When the extrachromosomal element – and the wild-type gene carried by it –is present in all cells, the worm exhibits a completely wild-type phenotype. But when, as occurs at low frequency, the extrachromosomal element fails to be transmitted to one of the daughters of a cell division, all the descendants of that cell, a clone, will be homozygous mutant. Cells within a mosaic animal that lack the wild-type gene can be independently identified as mutant if the extrachromosomal element also carries a marker gene whose absence from a cell affects the cell's appearance (i. e. acts cell autonomously). Because the C. elegans cell lineage is invariant(Sulston et al., 1983), a cell autonomous marker can allow one to determine precisely where in a worm's lineage the extrachromosomal element was lost, which helps to verify, by reference to the known lineage, which cells in the mosaic animal are homozygous mutant and which are not. The frequency of loss of an extrachromosomal element per cell division is approximately the same throughout development (Hedgecock and Herman, 1995; Yochem et al.,1998). However, extrachromosomal elements are occasionally lost at two or more consecutive cell divisions to give a pattern of mosaicism that is somewhat more complicated than that corresponding to a single clone of mutant cells (Hedgecock and Herman,1995; Yochem et al.,1998). Possible misinterpretations caused by this effect can be avoided by scoring for the presence or absence of the extrachromosomal element in more than a few cells in the lineage of interest.

Two kinds of extrachromosomal element have been used for mosaic analysis:free chromosome fragments (Herman,1984) and extrachromosomal arrays(Lackner et al., 1994; Miller et al., 1996). C. elegans chromosomes do not have localized centromeres(Albertson and Thomson, 1982)– they are said to be holocentric – which means that a suitably large fragment of any part of a chromosome, referred to as a free duplication,can retain some centromeric function, and can behave as a fairly stable mini-chromosome and be maintained in genetic stocks.

Extrachromosomal arrays, which are routinely generated to demonstrate transformation rescue (or complementation) of a mutant phenotype(Mello and Fire, 1995), are now often used in preference to free duplications for mosaic analysis. When a mixture of DNA that contains a wild-type gene and a gene that encodes a cell autonomous marker is microinjected into the syncytial germline of an adult hermaphrodite, a mini-chromosome composed of many copies of the DNAs tends to form spontaneously (Stinchcomb et al.,1985) (Fig. 1). One can then select animals that carry an array that shows good expression of both the gene to be analyzed, as judged by the rescue of the mutant phenotype, and the marker gene. Arrays are generally present in one copy per cell. Mitotic loss leads to mosaic animals in which a clone of cells lacks the array and therefore lacks both the marker gene and the wild-type gene under study(Fig. 2). Mitotic losses of free duplications or extrachromosomal arrays frequently involve non-disjunction, in which one daughter cell receives no duplication and the other daughter receives two copies(Hedgecock and Herman, 1995; Yochem et al., 1998). Some arrays are lost at a frequency that may be inconveniently high for mosaic analysis, say >1/50 per cell division, which means every animal will contain many independent mutant clones. Preliminary work in selecting a suitable array is therefore a good idea. A crucial assumption when using extrachromosomal arrays for mosaic analysis is that the expression of the wild-type gene on the array mimics its normal pattern of expression. [For more information on this and other issues that are important in C. elegansmosaic analysis, see also previous reviews(Herman, 1995; Yochem et al., 2000).]

To track the mitotic loss of an extrachromosomal element, one wants a cell autonomous marker that: (1) can be readily scored in virtually all cells– preferably in nuclei – of living animals, even when losses occur very late in the lineage; and (2) has no other phenotypic effect that could confound mosaic analysis. One excellent marker is a wild-type ncl-1gene in a mutant ncl-1 background; nearly all homozygous mutant ncl-1 cells exhibit enlarged nucleoli(Fig. 3) with no other apparent effects (Hedgecock and Herman,1995; Frank and Roth,1998). Both free duplications and extrachromosomal arrays carrying ncl-1(+) have been used for mosaic analysis. Alternatively, the gene that encodes GFP (green fluorescent protein) when expressed under the control of a strong promoter, as in sur-5::gfp, can be used to track the inheritance of extrachromosomal arrays in living animals(Fig. 4)(Yochem et al., 1998).

Additional genetic markers that can be scored in mutant animals at low magnification are sometimes used for rapidly identifying potential mosaics,which are then scored for their cellular phenotypes at high magnification. For example, consider zygotes of genotype unc-36(–); Ex100[unc-36(+)sur-5::gfp]. This nomenclature means that the zygotes are homozygous mutant for unc-36 and harbor an extrachromosomal array, Ex100, that carries both unc-36(+) and the marker sur-5::gfp. A low magnification stereomicroscope equipped with a lamp for exciting green fluorescence can be used to screen rapidly for uncoordinated animals that are at least partly green, and these animals can then be scored at single-cell resolution using a compound microscope at high magnification. The fully uncoordinated (Unc-36) phenotype in mosaics is caused by array loss at AB or ABp in the C. elegans cell lineage(Fig. 5) because the anatomical focus of unc-36 action is among the descendants of both daughters of ABp: ABpl and Abpr. Array loss by either ABpl or ABpr leads to a weaker uncoordinated phenotype (Kenyon,1986). Fig. 6illustrates how the Unc-36 phenotype can be used to pick out rapidly classes of animals that are mosaic for essential genes; in these examples, the extrachromosomal array carries the wild-type allele of an essential gene under study in addition to unc-36(+) and sur-5::gfp.

If one wants to delimit the focus of action of a gene that affects the whole animal when mutated, such as by causing uncoordinated movement,lengthened lifespan or inviability, then as many different types of mosaics should be collected as possible to see whether one can correlate the mutant phenotype with the absence of gene expression in a particular cell or set of cells. A good start in such cases is to look first for worms that contain large mutant clones, which have been generated by loss of the extrachromosomal element very early in the embryonic lineage. Loss at P₁(Fig. 7), for example, would give rise to an animal, which we denote as a P₁(–) mosaic, in which 94 of its 95 adult body wall muscle cells would lack the wild-type gene,whereas almost all of its neurons would have it. Thus, P₁(–)mosaics and their complement, AB(–) mosaics, can quickly distinguish whether a gene acts primarily in the development of muscles or neurons.

Certain mosaics can also exhibit novel phenotypes, which may be informative. For example, if the focus of a gene's action is broadly distributed among several cells, then mosaics in which only a subset of the cells are mutant might exhibit a mildly mutant phenotype, as in the unc-36 example noted above. Mosaics of a recessive lethal gene might be viable but show novel cellular abnormalities. Homozygous germline clones in an array-bearing hermaphrodite parent may give rise to a more severe homozygous mutant phenotype than that produced by a heterozygous parent germline, thus revealing a maternal effect. Conversely, germline mosaics can be used to indicate that a gene is fully zygotic in its action(Chen et al., 1994).

When a loss-of-function mutation causes a specific cellular abnormality,one can ask whether or not the gene behaves cell autonomously(Fig. 2). In this case, one wants to identify mosaics in which the cell of interest is genotypically different from its neighbors and other potentially interacting cells. If the gene of interest behaves cell non-autonomously, a careful mosaic analysis should be able to identify the interacting cells responsible for the effect.

Two potential problems or complications associated with the analysis of mosaic worms should be noted. The first is called perdurance, which refers to the persistence of a gene product in a cell that lacks the gene. Thus, a wild-type product synthesized in an ancestral cell prior to gene loss might persist and be transmitted to descendant cells even though they did not inherit the gene. Because the effect of perdurance would be to weaken the expected mutant phenotype, mosaics that exhibit a fully expressed mutant phenotype seem to be free of perdurance. Perdurance is also unlikely to be a problem when the loss of a duplication or an array occurs early in the embryonic cell lineage or when the gene product is involved in terminal cellular differentiation and is synthesized only late in development. Very late losses affecting either ncl-1 or sur-5::gfp give clear mosaic phenotypes, indicating that there is very little perdurance of the wild-type gene product for either of these marker genes.

The second complication concerns limitations that occur as a consequence of the nature of the worm cell lineage. This is generally not a problem when investigating a cellular phenotype because mosaics can usually be identified in which the cell under investigation is genotypically different from cells that are candidates for interaction. However, the worm cell lineage can be limiting when a gene's focus of action is diffuse and distributed among cells of disparate lineage. The fully mutant phenotype may then only be apparent in mosaic animals when the duplication or array is lost by a progenitor of all or nearly all of the responsible cells, in which case it may be difficult to pinpoint the responsible cell types. A particularly difficult tissue to investigate in mosaic worms is the hypodermis, called hyp7, which forms the skin for the main body of the animal. Hyp7 is a single syncytial cell that is formed by the fusion of many mononucleate cells that descend from both AB and P₁, the daughters of the very first embryonic cleavage(Sulston et al., 1983);therefore, no mosaic animal can contain a single, completely mutant hyp7 clone(Fig. 7). Mosaic analyses have,nonetheless, been used to implicate hyp7 as the focus of action of several genes, as we illustrate below.

More than 70 genes affecting worm development have been studied in genetic mosaics. Rather than attempt a comprehensive review of those genes here, we will discuss a few examples to illustrate how mosaic analysis has been used to elucidate certain mechanisms of development.

The foci of action of several genes implicated in cell-to-cell signaling during development have been determined by mosaic analysis. Two classic examples are glp-1 and lin-12, which encode members of the Notch family. The proteins encoded by these genes contain copies of a motif that resembles epidermal growth factor (EGF). Because EGF is an extracellular factor that affects other cells, and because it is cleaved from a membrane-bound precursor, it was natural to consider LIN-12 (the protein encoded by lin-12) and GLP-1 as possible sources of extracellular signal. However, mosaic analysis indicated that both glp-1 and lin-12 act cell autonomously in the cells whose fates require their function (Austin and Kimble,1987; Seydoux and Greenwald,1989), as would be expected if the genes encode receptors and not precursors for secreted signals.

By contrast, lin-44 encodes a Wnt signal that affects the polarity of a cell in the worm's tail, called T, and that was shown by mosaic analysis to act cell non-autonomously – not in T but most likely in tail hypodermal cells (Herman et al.,1995), the cells believed to secrete the Wnt signal. Another gene that has been shown by mosaic analysis to act cell nonautonomously is her-1 (Hunter and Wood,1992), a gene that is essential for determining male sexual fate(Hodgkin 1980). This result was consistent with the later molecular characterization of her-1(Perry et al., 1993), which suggested that HER-1 protein is secreted and acts as a signaling molecule in the worm sex determination pathway.

Loss-of-function mutations in unc-5 lead to defects in the dorsal-ward guidance of pioneering axons and migrating cells during worm development. Mosaic analysis has shown that unc-5 acts cell autonomously (Leung-Hagesteijn et al.,1992); for example, a migrating cell was found to be defective in its dorsalward migration if, and only if, it lacked unc-5(+). This result, together with the molecular characterization of unc-5(Leung-Hagesteijn et al.,1992), led to the proposal that UNC-5 is a transmembrane receptor that promotes the movement of migrating axons and cells away from a high ventral concentration of the extracellular matrix protein UNC-6(Wadsworth et al., 1996). UNC-6 is a member of the netrin family of proteins(Ishii et al., 1992), which have been shown to affect axon guidance in vertebrate embryos(Serafini et al., 1994). The vab-8 gene has also been shown by mosaic analysis to affect axon guidance cell autonomously; in this case, Wolf et al. were able to conclude that vab-8 must be expressed in certain neurons despite the fact that immunofluorescent staining was only able to detect VAB-8 expression in body muscle (Wolf et al.,1998).

The example of let-23, which encodes an EGF tyrosine kinase receptor (Aroian et al., 1990),is also instructive. Activation of LET-23 by an EGF-like ligand occurs repeatedly during C. elegans development to trigger diverse developmental events. One of these events is the induction of vulval development, in which three vulval precursor cells generate 22 cellular descendants that form the vulva. In the absence of an EGF signal from the gonadal anchor cell, none of the vulval precursor cells contributes to vulval development, and no vulva is formed. Because vulval development is not essential to a worm's survival, it can be analyzed in let-23 mosaics,even though let-23 is an essential gene. Such an analysis has shown that normal vulval development can occur when only one vulval precursor cell contains let-23(+) (Simske and Kim, 1995; Koga and Ohshima,1995). According to the picture that emerges from these studies,the vulval precursor cell closest to the signaling anchor cell is activated by the reception of the EGF signal and then induces its neighboring vulval precursor cells on each side (through LIN-12-mediated signaling) to embark on vulval development.

Downstream of LET-23 in the vulval signal-transduction pathway is the small G protein RAS, which is encoded by let-60(Beitel et al., 1990; Han and Sternberg, 1990). The lethality of a let-60 loss-of-function mutation has been traced by mosaic analysis to a single cell, the excretory duct cell(Yochem et al., 1997). An unexpected mosaic phenotype was encountered in this study: when descendants of ABpl (Fig. 7), which normally includes the excretory duct cell, were mutant for let-60, a let-60(+) cell assumed the duct cell fate.

Another example in which mosaicism has given an unexpected phenotype involves a natural ambiguity in the cell lineage between two cell fates. Either one of two cells defined by the lineage randomly assumes the anchor cell (AC) fate, with the other cell becoming VU(Kimble, 1981). The VU fate is specified by an AC-to-VU signal and requires lin-12 function: in a lin-12 loss-of-function mutant, both cells become AC, and in a lin-12 gain-of-function mutant, both become VU(Greenwald et al., 1983; Seydoux and Greenwald, 1989). Mosaic analysis has shown, as expected, that lin-12 function is required in VU and not in AC (Seydoux and Greenwald, 1989). But the collection of mosaics in which one of the two AC/VU cells was homozygous for a lin-12 loss-of-function mutation and the other was lin-12(+) gave an unexpected result: in every animal, the mutant cell became AC and the wild-type cell became VU. Thus, loss of lin-12 in one cell had a non-autonomous effect that forced the other cell to become VU. This result is explained by a model in which the two AC/VU cells signal to each other until the greater activation of LIN-12 receptor in one (VU) leads to the loss of LIN-12 activity in the other(AC) (Seydoux and Greenwald,1989).

Mosaic analysis also revealed an unexpected anatomical focus for the action of unc-52, which encodes the worm homolog of mammalian perlecan(Rogalski et al., 1993). Null unc-52 mutants are unable to assemble a myofilament lattice and undergo embryonic arrest typical of that caused by defective body muscle(Williams and Waterston,1994). UNC-52 localizes between muscle and hypodermis, and it had been concluded that the protein is synthesized by muscle. But mosaic analysis showed that unc-52 does not function in muscle, because P₁(–) mosaics (Fig. 7) were viable and fertile(Spike et al., 2002). It was therefore concluded that UNC-52 is synthesized and secreted by hypodermis.

The anatomical foci of other whole-animal, developmental phenotypes have been deduced from mosaic analysis. For example, analysis of mosaics of sma-3, which encodes a component of a TGFβ signaling pathway that regulates body size, lent weight to the conclusion that sma-3acts in the hypodermis to control body size(Wang et al., 2002). The daf-2 gene, which encodes a homolog of insulin-like growth factor I receptors (Kimura et al.,1997), affects two whole-animal phenotypes: lifespan and the decision of young larvae to enter a state of diapause (the dauer stage) rather than progressing to adulthood. A partial loss-of-function mutation in daf-2 leads to extended lifespan(Kenyon et al., 1993), whereas a stronger daf-2 mutation leads to the inappropriate formation of dauer larvae (Riddle et al.,1981). The characterization of many daf-2 mosaics with different mutant clones has led to the conclusion that the wild-type DAF-2 receptor acts diffusely in multiple cell lineages to regulate the production or activity of a secondary signal, which then affects lifespan and dauer formation by affecting the tissues of the whole animal(Apfeld and Kenyon, 1998).

Mosaic analysis in C. elegans requires a knowledge of worm anatomy and development, both to pinpoint where in the cell lineage a genetically-marked extrachromosomal element has been lost and to analyze the cellular phenotypes of mosaic animals. This can seem formidable at first, but knowledge of this organism's anatomy can be one of the joys of working with C. elegans. (To begin acquiring such knowledge, we recommend you go to http://www.wormatlas.org/index.htm)One prospect for the future is additional computer-assisted tutorials on worm anatomy. Another future development that would assist the field would be the creation of improved markers to allow for the even more rapid examination of thousands of worms at low magnification for specific mosaics.

Another method for determining the anatomical focus of a gene's action is to introduce transgenic copies of the wild-type gene under the control of cell- or tissue-specific promoters into otherwise homozygous mutant animals. The rescue of a mutant defect by a particular transgene indicates that the gene's expression in the indicated cell type can provide the required function. However, one cannot be certain that such promoters are absolutely specific in their effects, and one must make a new construct when testing each promoter with a gene of interest. In addition, rescue by expression from the transgene in this case (generally overexpression because of the multiple gene copies) does not prove that the wild-type gene normally acts in the same cell type or tissue. But this approach, especially in combination with mosaic analysis, can help to build a strong case (e.g. Zhen and Jin, 1999; Inoue and Thomas, 2000; Zhen et al., 2000; Wang et al., 2002). The inclusion of mosaic analysis provides greater confidence that one has identified the cell or tissue that requires the activity of a gene, and we expect future mosaic analyses to yield more unexpected insights into the mechanisms of C. elegans development.

We thank Leslie Bell for comments on the manuscript. Our work is supported by a grant from the NIH.