Genetic Analysis of the Caenorhabditis elegans pax-6 Locus

Abstract

PAX-6 proteins are involved in eye and brain development in many animals. In the nematode Caenorhabditis elegans the pax-6 locus encodes multiple PAX-6 isoforms both with and without a paired domain. Mutations in the C. elegans pax-6 locus can be grouped into three classes. Mutations that affect paired domain-containing isoforms cause defects in epidermal morphogenesis, epidermal cell fates, and gonad cell migration and define the class I (vab-3) complementation group. The class II mutation mab-18(bx23) affects nonpaired domain-containing isoforms and transforms the fate of a sensory organ in the male tail. Class III mutations affect both paired domain and nonpaired domain isoforms; the most severe class III mutations are candidate null mutations in pax-6. Class III mutant phenotypes do not resemble a simple sum of class I and class II phenotypes. A comparison of class I and class III phenotypes indicates that PAX-6 isoforms can interact additively, synergistically, or antagonistically, depending on the cellular context.

PROTEINS of the Pax6 family play critical roles in the development of many animals (Callaerts et al. 1997; Gehring and Ikeo 1999; Simpson and Price 2002). Pax6 proteins regulate transcription and contain two DNA-binding domains, an N-terminal paired domain (PD) and a C-terminal homeodomain (HD). Mutations in the human PAX6 gene cause defects in eye development, including aniridia (lack of iris), Peters anomaly, and autosomal dominant keratitis (van Heyningen and Williamson 2002). In mice loss of one copy of the murine Pax6 gene results in eye defects; complete loss of Pax6 function blocks eye development at an early stage (Hogan et al. 1986; Hill et al. 1991). In Drosophila, partial loss of function in the Pax6 homolog eyeless (ey) blocks eye development (Quiring et al. 1994); ectopic expression of eyeless is sufficient in some contexts to induce formation of ectopic eyes (Halder et al. 1995). Pax6 family members or related genes have been found to be expressed in eyes in many animal phyla, including mollusks (Tomarev et al. 1997), nemerteans (Loosli et al. 1996), annelids (Arendt et al. 2002), and cnidarians (Kozmik et al. 2003). The high sequence conservation in the Pax6 gene family and the conserved expression of Pax6 expression in eyes or photoreceptor cells suggests that Pax6 genes play a critical and evolutionarily conserved role in eye development (Gehring and Ikeo 1999) and, furthermore, that animal eyes have evolved from a single ancestral eye type (Arendt and Wittbrodt 2001).

Despite the potential conservation of Pax6 as a regulator of eye development, the diversity of eye morphology and physiology among animals raises the question of whether the cellular functions of Pax6 have been conserved (Harris 1997; Pineda et al. 2002). Indeed, analysis of Pax6 in a variety of organisms suggests that although Pax6 function is often closely associated with eye development, Pax6 by itself is neither necessary nor sufficient for eye development in all cases. In some species, such as sea urchins, Pax6 is not expressed in the visual organs (Czerny and Busslinger 1995). In planarian worms, Pax6 orthologs are expressed in developing eyes, but do not appear to be required for eye development (Pineda et al. 2002). Indeed, in most organisms Pax6 proteins are also expressed in a variety of tissues in addition to eyes. In vertebrates, Pax6 is widely expressed in the central nervous system (Walther and Gruss 1991; Puschel et al. 1992), where it regulates neuronal fate specification (Mastick et al. 1997; Takahashi and Osumi 2002; Talamillo et al. 2003). Outside the nervous system, Pax6 is expressed in the vertebrate pancreas, where it regulates pancreatic hormone expression (Sander et al. 1997). Pax6 proteins therefore specify very different cell fates in different cellular contexts.

We are studying Pax6 in the nematode Caenorhabditis elegans, a species that lacks eyes and has a vestigial light sense of unknown cellular origin (Burr 1985). The C. elegans genome encodes a single Pax6 ortholog. We previously showed that vab-3 mutations, which affect the paired domain of C. elegans PAX-6, cause homeotic transformations in the head region epidermis, suggesting that pax-6 specifies anterior epidermal fates (Chisholm and Horvitz 1995). vab-3 mutations have pleiotropic effects on development. vab-3 mutants display abnormal cell migration in the somatic gonad (Nishiwaki 1999), transformations of blast cell fates in the rectal epidermis (Chamberlin and Sternberg 1995), defects in head epidermal morphogenesis [the variable abnormal (Vab) phenotype] that result in incompletely penetrant lethality (this article), and aberrant axon guidance in the anterior nervous system (Zallen et al. 1999).

Previously reported vab-3 alleles affect exons 1–7 of the pax-6 locus, which encode the PD and part of the linker between the PD and HD. The pax-6 locus also encodes at least two isoforms that lack the PD (Zhang and Emmons 1995; Zhang et al. 1998). These non-PD-containing isoforms are encoded by transcripts arising from internal promoters and contain alternative 5′ exons (exons 8 and 9) that splice to a common set of 3′ exons (exons 10–14) that encode the HD and C terminus. The mutation bx23 deletes exon 8 and causes homeotic cell fate transformations in the peripheral nervous system of the male tail, defining the mab-18 complementation group. bx23 mutants have normal levels of PD-containing PAX-6 isoforms but reduced levels of non-PD-containing isoforms (Zhang et al. 1998). Thus, bx23 reduces, but may not abolish, the function of non-PD-containing isoforms. vab-3 mutations, affecting the PD-encoding isoform PAX-6A, complement mab-18(bx23), suggesting that the PD-containing and non-PD-containing isoforms of PAX-6 have nonoverlapping functions in development.

In an effort to more fully define the functions of pax-6 in C. elegans we here report the genetic and molecular analysis of a large set of new mutations in the pax-6 locus. Our genetic and molecular analysis confirms that strong vab-3 alleles likely eliminate function of PD-containing PAX-6 isoforms. We describe a set of mutations that affect both PD-containing and non-PD-containing isoforms, the strongest of which are candidate pax-6 null mutants. We find that these mutations cause phenotypes that do not resemble a simple sum of the separate phenotypes of vab-3 and mab-18 mutants. Comparison of the putative pax-6 null mutant phenotype to the phenotypes of isoform-specific mutants suggests that C. elegans PAX-6 isoforms can interact synergistically or antagonistically, depending on the cellular context.

MATERIALS AND METHODS

Genetic analysis:

Worms were maintained as described in Brenner (1974). Animals were raised at 20° unless stated otherwise. The following mutations were used: LGII—tra-2(q276), tra-2(q122dm), and rrf-3(pk1426); LG IV—him-8(e1489; LGX—lon-2(e678), unc-18(e81), dpy-6(e14), unc-27(e155), unc-115(e2225), dpy-7(e88), and unc-9(e111). Rearrangements used were szT1 (I; X). Transgenes used were jcIs1[AJM-1::GFP] (Koppen et al. 2001). Mutations are described in Hodgkin (1997). pax-6 mutations are listed in Table 1.

TABLE 1.

Origins ofpax-6 mutations

Allele	Class	Mutant screen	Mutagen	Reference
bc52	III	pkd-2::GFP expression	EMS	P. Grote and B. Conradt   (personal communication)
bx23	II	Male tail	EMS	Zhang and Emmons (1995)
cr2	III	?	EMS	P. Kuwabara   (personal communication)
e41, e648, e1022	I	Morphology	EMS	S. Brenner; Hodgkin (1983)
e1062	I	Male mating	EMS	Lewis and Hodgkin (1977)
e1178	I	Morphology	ICR-191	D. Riddle   (personal communication)
e1796	III	Cell lineage	EMS	Hedgecock (1985)
e2429	I	(Found in CB1452)	?	This article
ju5	I	Morphology	TMP/UV	This article
ju441	I	Morphology (clonal)	EMS	This article
ju468	III	e1796 noncomplementation	EMS	This article
ju478	III	Integration of array	TMP/UV	W.-M. Woo   (personal communication)
k109, k121, k143	I	Distal tip cell migration	EMS	Nishiwaki (1999)
ky112	III	Axon guidance	EMS	Zallen et al. (1999)
ky664	I	Axon guidance	EMS	J. Kennerdell and C. I. Bargmann   (personal communication)
n2550	I	Defecation	EMS	E. Jorgensen   (personal communication)
n3721, n3723	I	pkd-2::GFP expression	EMS	H. T. Schwartz and H. R. Horvitz   (personal communication)
sa503	I	Male tail cell lineage	EMS	H. Chamberlin   (personal communication)
sy66, sy281	I	Male mating	EMS	Chamberlin and Sternberg (1995)
we4	III	vab-7 enhancement	EMS	J. Ahringer   (personal communication)
zd91	I	Axon guidance	EMS	S. G. Clark   (personal communication)

Alleles are listed alphabetically. The mutant phenotype by which the mutation was isolated, if known, is listed under mutant screen.

Mutations in the pax-6 locus were isolated in a variety of genetic screens (Table 1). The vab-3 complementation group was defined by the reference allele e648. New alleles were assigned by mapping and complementation tests; map data are available from Wormbase. Mutations were initially assigned to the pax-6 locus by failure to complement the weak class III allele e1796 for the gonad migration (Mig) phenotype in hermaphrodites.

Most genetic screens that recovered pax-6 alleles were biased toward viable mutations. We used two approaches to recover pax-6 mutations from screens that could potentially recover lethal alleles of pax-6. From genome-wide semiclonal screens for morphogenetic mutants we recovered the strong class I allele ju441. We also took advantage of the viability of e1796/nDf19 heterozygotes to design a noncomplementation screen. Wild-type males were mutagenized with EMS and crossed with dpy-6 e1796 unc-9 hermaphrodites. F₁ cross-progeny hermaphrodites were screened for gonadal migration defects. From 20,000 chromosomes screened we isolated one class III mutation, ju468.

The pax-6 locus is X linked. To analyze male phenotypes of pax-6 mutants we generated pax-6/O hemizygous males either by crossing parental hermaphrodites with wild-type XO males or by use of the him-8(e1489) nondisjunction mutation. To examine male phenotypes in pax-6 heteroallelic combinations we generated XX males using the transformer mutation tra-2(q276) (Hodgkin 2002). To quantify penetrance of lethal phenotypes in heterozygous animals we used the dominant feminizing mutation tra-2(q122) to eliminate self-progeny (Barstead and Waterston 1991).

Phenotypic analysis of pax-6 mutants:

Penetrance of morphogenetic and lethal phenotypes (Table 3) was determined as described by George et al. (1998). At least three complete broods were scored for each genotype. For most mutations the penetrance was determined at 15°, 20°, and 25°; none of the alleles tested showed significant temperature sensitivity (data not shown). We analyzed the cell lineages of mutant animals as described previously (Sulston and Horvitz 1977). Gonadal migrations were scored using Nomarski optics in young adults or using the lag-2::GFP transgene qIs19 to visualize distal tip cells (Henderson et al. 1994). Head epidermal cell morphology was analyzed using immunofluorescence labeling with monoclonal antibody MH27, which recognizes the apical junction protein AJM-1, or with the AJM-1::GFP transgene jcIs1. AJM-1::GFP images in Figure 3 were taken using a Zeiss Pascal confocal microscope, on transgenic animals fixed in 2% paraformaldehyde for 20 min. To quantify B daughter nucleus size in early L2 males we used NIH Image.

TABLE 3.

Penetrance ofpax-6 morphogenetic and lethal phenotypes

Genotype	% embryonic lethal (n)	% larval lethal	% adult Vab	% adult non-Vab
Wild type	0 (1258)	0	0	0
Class I   e41	4.7 (545)	49	43.6	1.2
e648	3.2 (1258)	24.5	65.4	6.9
e1062	2.6 (947)	53.0	39.7	3.0
ju5	4.4 (630)	39.4	56.0	0
n2550	4.1 (551)	39.2	52.4	4.1
cr2	4.3 (509)	30.7	48.6	0
zd91	3.8 (557)	7.6	88.6	0
sa503	2.7 (331)	26.6	53.8	16.9
sy66	1.3 (760)	10.9	47.3	38.5
e2429	1.1 (555)	2.5	7.6	88.1
sy281	<0.1 (1340)	<0.1	1.2	98.3
n3723	1.6 (1210)	0	4.3	94.1
n3721	1.2 (1309)	0	0.6	98.2
k109	1.6 (1467)	1.6	0	98.4
k121	0 (1399)	0	0	100
k143	3.8 (1667)	0	0	96.2
Class II   bx23	0.1 (922)	0.1	0	99.8
Class III   ju478	2.7 (621)	87.2	9.9	0
ju468	6.0 (447)	50.3	43.6	0
we4	7.7 (875)	55.9	36.5	0
e1022	8.2 (672)	78.2	12.9	0.6
e1178	3.1 (1123)	63.5	33.4	0
ky112	5.0 (332)	3.7	38.9	52.4
bc52	2.3 (984)	0	1.3	95.9
e1796	0.4 (890)	0.1	0	99.5

Alleles are listed in order of phenotypic strength within each class. The penetrance of lethality and head epidermal morphogenetic defects (Vab) for each mutation was determined as described in materials and methods. The penetrance of larval lethality caused by strong class III alleles is significantly higher than that caused by the strongest class I alleles (e.g., ju468 vs. e1062, P < 0.01 by t-test). The penetrance of embryonic lethality is not statistically different in comparisons of strong class I and strong class III mutants.

Figure 3.—

Epidermal cell boundaries are altered in class I and class III pax-6 mutants. Epidermal cell attachments in pax-6 mutants is visualized using AJM-1::GFP. (A and B) Wild-type head epidermal cell boundaries. The patterns are dynamic and are shown in early L1 (A) and late L3 stage animals (B). In wild-type L1 animals the anterior end of H0 is attached to the hyp5 cell; H0 appears elongated and thin by AJM-1 localization. The hyp6/7 boundary connects to the H1 seam cell in early L1 animals, to H0 in later stages, and disappears during the L3 stage (Yochem et al. 1998). (C–J) AJM-1::GFP (jcIs1) expression in pax-6 mutants. All images are confocal micrographs of fixed transgenic animals. In class I (e648, E and F) and strong class III (ju468, I and J) mutants, H0 was frequently not connected to an epidermal ring in the early L1 stage (arrowhead, I); the H0 cell appeared rounder by AJM-1::GFP localization (arrowhead, J). It was unclear in these animals if hyp5 was present as a distinct cell or whether it had fused with hyp6 and hyp7. The hyp6/7 junction (arrows, C and G) was frequently absent in class I and class III mutant L1s (E and I). In 5/11 class I (e648) and 11/13 class III (ju468) mutants the junctions between anterior epidermal rings hyp4, hyp5, and hyp6 were absent (I and J), suggesting that these cells may have collectively fused with hyp7. In class II (bx23) mutants H cell morphology and junctions appeared normal (G and H). Bar, 10 μm (C, E, G, and I) and 20 μm (D, F, H, and J).

Molecular analysis of pax-6 mutations and gene products:

We determined the sequence of mutant genomic DNAs as described previously (Chisholm and Horvitz 1995).

Western blot of PAX-6 proteins:

Lysates of C. elegans were prepared as described by Woo et al. (2004) and analyzed by SDS-PAGE. Western blotting followed standard procedures (Harlow and Lane 1999). Blots were incubated with affinity-purified anti-PAX-6-C-terminal antibody (Zhang et al. 1998) at 1:500 dilution. Staining was visualized with the SuperSignal West Pico kit (Pierce, Rockford, IL).

RESULTS

Molecular lesions of class I and class III pax-6 mutations—structure of the C. elegans pax-6 locus and its isoforms:

The genomic structure of the pax-6 locus (gene model F14F3.1) is shown in Figure 1. The pax-6 locus encodes two major sets of isoforms (Zhang et al. 1998). The paired domain-containing isoform, PAX-6A (51 kD), is expressed throughout development, being more abundant in embryos (not shown). At least two nonpaired domain-containing isoforms (PAX-6B and PAX-6C) are expressed at roughly equivalent levels at all stages, with the exception of the dauer stage, where they are more abundant.

Figure 1.—

Structure of the C. elegans pax-6 gene and products. (A) Genomic structure of the pax-6 locus (F14F3.1) showing physical locations of mutations. The ATG of the PAX-6A transcript is in exon 2. The paired domain is encoded by exons 3–6 and the homeodomain by exons 10–12. The mutation bx23 deletes exon 8, affecting the F14F3.1b transcript. A third transcript (F14F3.1c) is supported by EST data in Wormbase. The probable extent of the ju478 rearrangement is shown on the basis of Southern blotting and PCR experiments. (B) Structure of PAX-6 isoforms and their expression in pax-6 mutant classes. The gel shows a Western blot of C. elegans lysates probed with antibodies specific to the C termini of PAX-6 isoforms. The top band corresponds to PAX-6A (PD-containing, 455-amino-acid residues, 51.1 kD); the bottom bands correspond to the PAX-6B (269 aa, 30.3 kD) and PAX-6C isoforms (296 aa, 33.2 kD). As shown previously, class I (vab-3) mutants such as e648 lack expression of PAX-6A isoforms; the class II mutation mab-18(bx23) reduces but does not eliminate expression of smaller isoforms. The strong class III mutant ju468 lacks detectable expression of PAX-6 isoforms. (Bottom) A control immunoblot using the monoclonal antibody MH35, which recognizes α-actinin ATN-1.

Class I mutations affect paired domain-containing isoforms:

vab-3 mutations were first isolated on the basis of their defects in head morphogenesis (Lewis and Hodgkin 1977). Subsequently, additional alleles were isolated by virtue of their effects on male mating (Chamberlin and Sternberg 1995), gonadal distal tip cell migration (Nishiwaki 1999), axon guidance (Zallen et al. 1999), and expression of pkd-2-GFP transgenes (H. T. Schwartz and H. R. Horvitz, personal communication; P. Grote and B. Conradt, personal communication; Table 1). We initially assigned these mutations to the vab-3 complementation group by mapping and complementation testing (see materials and methods). Our analysis shows that vab-3 mutations affect a subset of the functions of the C. elegans pax-6 locus; we have therefore classified these alleles as class I alleles of the pax-6 locus. For reasons described below we classify the mab-18(bx23) mutation as a class II pax-6 allele and a further group of mutations as class III alleles.

The molecular lesions of four class I alleles were previously described (Chisholm and Horvitz 1995). We analyzed the molecular lesions of 14 new class I alleles (Table 2). The new class I alleles result from 10 distinct lesions as 3 lesions have been found more than once. In total, eight mutations cause equivalent phenotypes and likely eliminate function of the PD-containing PAX-6A isoform; we refer to these as strong class I alleles. Three mutations (e1062, n2550, and zd91) are identical transitions in exon 6 resulting in an opal stop codon 17 residues C-terminal to the paired domain (R149opal). The R149opal alleles cause phenotypes similar to that of the reference class I allele e648 (W101opal, a change also found in ky664). e648 animals lack expression of the PAX-6A paired domain-containing isoform (Figure 1B). Three additional class I alleles, the rearrangement ju5 and the splice site mutations e41 and cr2, have strong class I phenotypes. We conclude that the strong class I mutations likely abolish function of PD-encoding isoforms of PAX-6.

TABLE 2.

Sequences of mutations in theC. elegans pax-6 locus

Allele		Wild-type sequence	Mutant sequence	Effect
Class I (vab-3)   cr2	strong	ttctag|GGT	ttctaa|GGT	Exon 5 acceptor
e41	strong	AGC|gtgagt	AGC|atgagt	Exon 5 donor
e648, ky664	strong	TGG	TGA	W101opal
e1062, n2550, zd91	strong	CGA	TGA	R149opal
ju5	strong	—	Rearrangement	Exons 3–6 deleted
e2429	intermediate	AAG|gtctgt	AAG|atctgt	Exon 4 donor
sa503	intermediate	ttgtag|ATG	ttgtaa|ATG	Exon 3 acceptor
sy66	intermediate	ATC	AAC	I103N
ju441	intermediate	CAG|gtacgt	CAG|atacgt	Exon 2 donor
sy281	weak	GGG	AGG	G19R
k143	weak	GGA	GAA	G13Q
k109	weak	TCA	TTA	S75L
n3723	weak	CGG|gtaagt	CGG|ataagt	Exon 3 donor
k121, n3721	weak	AGC|gtgagt	AGC|gtgaat	Exon 5 donor
Class II (mab-18)   bx23	—	556-bp deletion	Exon 8 deleted
Class III   e1022	strong	tttcag|AAT	tttcaa|AAT	Exon 11 acceptor
e1178	strong	CCC	CCCC	Frameshift at codon 323
ju468	strong	CAG	TAG	Q262amber
ju478	strong	—	Rearrangement	Exons 13 and 14 deleted
we4	strong	CAG	TAG	Q228amber
ky112	weak	tttcag|ATT	tttcaa|ATT	Exon 13 acceptor
bc52	weak	tttcag|ATG	tttcaa|ATG	Exon 10 acceptor
e1796	weak	ATT, CTT	ACT, CAT	I229T, L232H

Sequences of mutant DNAs were determined as described in materials and methods; mutations are underlined. Data for e41, e648, sy66, and sy281 are taken from Chisholm and Horvitz (1995); data for bx23 and e1796 are from Zhang and Emmons (1995). The k121 lesion was identified by K. Nishiwaki and confirmed by us; the bc52 lesion was identified by P. Grote and B. Conradt and confirmed by us. The newly reported EMS alleles are all GC-to-AT transitions. The e1178 frameshift is typical of the effects of the half-mustard ICR-191, in which additional G's or C's are inserted into runs of G's or C's (Skopek and Hutchinson 1984).

Ten additional mutations cause phenotypes similar to those of strong class I alleles but with lower penetrance; we have classified these as intermediate or weak class I alleles on the basis of the penetrance of morphogenetic defects (Table 3). These class I mutations result in missense alterations in the PD or affect splice sites of PD-encoding exons (Table 2). Another weak class I mutation, k143, results in a missense alteration (G13Q) at an invariant residue in the loop between the first two β-sheets of the PD N-terminal subdomain; this loop appears to stabilize the helix-turn-helix of the PD (Xu et al. 1995, 1999). The weak class I mutation k109 results in the missense alteration S75L, affecting a conserved residue in the linker between the N-terminal and C-terminal subdomains of the Paired domain; this residue contacts a DNA base and may play a role in binding site recognition. Six other hypomorphic class I mutations affect splice donor or acceptor sites of exons 2–5 and presumably allow reduced levels of normally spliced transcripts to be made. The weakest class I allele is an alteration in the +5 position of the splice donor of exon 5, found in the independent mutations k121 and n3721. These alleles cause highly penetrant distal tip cell migration (Mig) phenotypes but almost no morphogenetic or lethal phenotypes (Table 3).

Class III mutations affect exons common to paired domain and nonpaired domain-containing isoforms:

The pax-6 HD allele e1796, isolated by virtue of its cell fate transformations in the head epidermis (Hedgecock 1985), causes both a Mab-18-like ray 6 transformation phenotype and Vab-3-like distal tip cell migration defects (Zhang and Emmons 1995), yet does not affect head epidermal morphogenesis (Tables 3 and 5). We here describe seven additional mutations that, like e1796, cause both Mab-18-like ray 6 transformations and Vab-3-like distal tip cell Mig phenotypes. On the basis of their genetic and molecular characteristics we have named these class III pax-6 alleles. The class III mutations e1022, e1178, ju468, and we4 cause similarly penetrant defects, whereas ky112 is weaker. The bc52 mutation causes distal tip cell migration defects but does not affect male tail development or head morphology; we classify it as class III on the basis of its molecular lesion. The rearrangement ju478 causes phenotypes stronger than those of other class III alleles and likely also affects an adjacent gene (see below). Class III alleles failed to complement class I alleles for the morphogenetic and Mig phenotypes. Class III mutant males display ray 6 to ray 4 transformations similar to those of mab-18(bx23) males, and class III mutations failed to complement bx23 for this defect (see below). Thus, class III mutations affect both class I (vab-3) and class II (mab-18) functions. However, as we show below, class III mutant phenotypes are not a simple sum of the class I and class II phenotypes.

We determined the molecular lesions of the new class III alleles. Two strong class III mutations, ju468 and we4, result in premature stop codons in the PAX-6 homeodomain (Table 2). The class III mutation e1022 alters the splice acceptor of exon 11, which encodes part of the HD. e1178 causes a frameshift at codon 323, in the proline/serine/threonine (PST)-rich domain, and is predicted to result in a truncated and non-PST-rich C-terminal domain. Since in e1178 mutants the PD and HD are intact, we tested if the e1178 loss-of-function phenotype was due to nonsense-mediated decay of the mutant message (Mango 2001). We found that the nonsense-mediated decay mutant smg-1(r861) did not suppress e1178 phenotypes (data not shown), suggesting that the C-terminal PST-rich domain is essential for PAX-6 function in C. elegans.

The class III allele ju478 causes significantly higher larval lethality than do the four other strong class III alleles. The higher larval lethality of ju478 may be the result of disruption to genes 3′ to pax-6, as transgenes containing wild-type copies of pax-6 did not fully rescue the ju478 larval lethal phenotypes (not shown). From PCR and Southern blotting experiments (not shown) ju478 appears to be a rearrangement that results in the deletion of exons 13 and 14, which encode the PST-rich domain and the 3′-untranslated region. The weak class III mutations bc52 and ky112 affect the splice acceptors of exons 10 and 13, respectively (Table 2; P. Grote and B. Conradt, personal communication). Thus, all known class III alleles affect exons or intron-exon junctions common to PD-containing and non-PD-containing isoforms of PAX-6.

Several lines of evidence suggest that the strong class III mutations eliminate pax-6 function. First, four independent mutations (e1022, e1178, ju468, and we4) cause similar phenotypes and result from lesions predicted to truncate PAX-6 isoforms in the HD or in the PST-rich domain. The ju468 phenotype was not enhanced by pax-6 RNA interference experiments (see materials and methods). Finally, PAX-6 isoforms were undetectable in Western blot analysis of lysates from ju468 mutants (Figure 1B).

Class I and class III pax-6 mutations cause abnormal epidermal morphogenesis:

Class I and class III mutations cause defects in head epidermal morphogenesis leading to the “Notched head” phenotype (Figure 2). Morphogenetic defects in the epidermis also lead to incompletely penetrant embryonic and larval lethality (Table 3). The strong class I mutations e648, e1062, and n2550 cause similar morphogenetic defects and cause ∼50% of each self-progeny brood to arrest during embryonic or larval development. Thus, anterior epidermal morphogenesis requires PD-containing isoforms of PAX-6. The class II mutation mab-18(bx23) does not cause head morphogenesis defects and complements class I alleles for such defects, suggesting that non-PD-containing isoforms are not required for epidermal morphogenesis. However, the morphogenetic defects of strong class III mutant strains are often qualitatively distinct from those of class I alleles (Figure 2, compare C and E to G and I). In addition, strong class III mutations caused higher levels of embryonic and larval lethality than did the strongest class I alleles (Table 3). The differences between class I and class III phenotypes suggest that PD and non-PD isoforms may play partly redundant roles in epidermal morphogenesis.

Figure 2.—

Abnormal anterior epidermal morphology in pax-6 mutants. Anterior is to the left and dorsal is up. (A and B) Head morphology of wild-type adult hermaphrodite, visualized by Nomarski DIC microscopy, medial section (A) and lateral view, showing the normal anterior end of lateral alae (arrowhead, B). (C–J) Head morphogenesis in representative pax-6 mutants. The epidermal morphology of typical strong class I (C and D, e648; E and F, cr2) and strong class III (G and H, ju468; I and J, e1022) adults is shown. Both class I and class III mutants display a range of morphogenetic defects, from slight displacement of the mouth (C) to pronounced deformations of the head epidermis (E). The spectrum of morphological defects appears different in class I vs. class III animals, in that class III mutants more frequently displayed narrowed or pinched epidermis (G and I) and apparent degeneration of head musculature on both ventral and dorsal sides. Lateral epidermal alae usually extend to the tip of the head in class I mutants (arrowhead, D) or pass through the ventral notch to meet up with contralateral alae (arrowhead, F). In class III mutants, lateral alae usually terminate at the level of the pharyngeal isthmus or anterior bulb (arrowheads, H and J), reflecting the detachment of H0 from the hyp5 cell. Bar, 20 μm.

The epidermal morphogenesis defects in pax-6 mutants might reflect aberrant specification of head epidermal cell fates, as PAX-6 isoforms are expressed in anterior epidermal cells (Chisholm and Horvitz 1995). We examined the cellular connections of the anterior epidermis in pax-6 mutants (Figure 3). In wild-type animals the most anterior seam cell H0 forms junctions with the hyp5 ring syncytium at the tip of the head (Figure 3, A–D). The more posterior epidermal ring hyp6 forms a distinct syncytium in L1s (boundary indicated by arrow, Figure 3C) and fuses with the main epidermal syncytium hyp7 at about the L3-L4 molt (Yochem et al. 1998). In general, class I and class III mutants showed similar defects in epidermal cell junctions. In both class I and class III mutants the hyp6/7 junction was often not visible during larval stages (Figure 3, E and F), suggesting that hyp6 prematurely fused with hyp7. In contrast, class II mutants displayed normal anterior epidermal cell junctions (Figure 3, G and H). In both class I and class III mutants more anterior epidermal cell junctions (hyp5/6 and hyp4/5) were frequently not visible, suggesting that these anterior cells may also have fused with hyp7; in such animals H0 was abnormal in morphology (Figure 3, I and J, arrowheads). In animals where anterior epidermal rings were present, H0 often did not form a junction with them, possibly reflecting a transformation in H0 fate (see below). The penetrance of anterior epidermal junction defects was higher in class III mutants than in strong class I mutants (Figure 3), consistent with non-PD-containing isoforms playing redundant roles in anterior epidermal cell fates.

Class I and class III mutants display different cell fate transformations in the anterior epidermis:

Class I (vab-3) mutations cause transformations in cell fates in anterior lateral epidermal blast cells, also known as seam cells. In the wild type, the three anterior seam cells present in the L1, known as H0, H1, and H2, develop differently (Figure 4A). In class I (vab-3) mutants the anterior-most seam cell H0 takes on the fate of H1 (Figure 4B; Chisholm and Horvitz 1995). Lateral alae are cuticular ridges normally formed by the seam cell descendants of H1, H2, V1-6, and T, but not by H0. The transformation of H0 fate in class I mutants leads to ectopic alae in the most anterior part of the head. We quantified the lineage transformation in class I mutants by analysis of the extent of anterior alae in adults (Table 4). All class I mutations caused transformation of H0 to H1, although to different extents, correlating with their molecular lesions. The strong class I alleles e648, e1062, and n2550 caused fully penetrant H0-to-H1 transformations. In contrast, the class II mutation mab-18(bx23) does not affect head seam cell fates (Figure 4E). Thus, PD-containing isoforms of PAX-6 are required for the normal H0 fate.

Figure 4.—

Cell fate transformations of head lateral epidermis in pax-6 mutants. (A) Lineages of head lateral epidermal blast cells in the wild-type L1 stage. H0 does not divide in the wild type and becomes a lateral seam cell (open circle) that does not make alae. H1 divides once in the L1; the anterior daughter H1.a gives rise to two alae-making seam cells and two hyp7 nuclei; H1.p fuses with the hyp7 syncytium (solid circle). H2.a divides in the late L1 stage to give rise to the deirid socket cell (so) and a hyp7 nucleus; H2.p gives rise to a seam cell and three hyp7 nuclei. (B) The deirid socket cuticular substructure is visible in dauer larvae (arrowhead). (C) In pax-6 class I (vab-3) animals H0 is transformed to H1 (Chisholm and Horvitz 1995); class I mutants make a single deirid socket cuticular substructure (D, arrowhead). (E) In pax-6 class II (mab-18) animals, head epidermal fates are normal; class II mutants make a single deirid socket cuticular substructure (F, arrowhead). (G) In strong class III pax-6 mutants both H0 and H1 take on H2-like fates. L1 cell lineages in ju468 (n = 3) and e1178 (n = 2) showed the same range of defects. In 2/5 animals H0 divided once, and its posterior daughter divided once to yield a compact nucleus (H0.pa) that we infer to be an ectopic anterior deirid socket cell; divisions in the H0 lineage were sometimes delayed up to 3 hr relative to H1 and H2. We note that the polarity of the initial asymmetric division of H0 is reversed, such that the posterior daughter divides to give rise to the ectopic deirid socket; the reason for this polarity reversal is not clear, as transformed H0 cells show normal asymmetry in class I mutants. In 3/5 animals H0 divided once in the L1 stage. (H) Approximately 50% of strong class III mutants generate two ectopic substructures (arrows) anterior to the normal deirid substructure (arrowhead). (I) In the weak class III mutants (e1796) H1 is fully transformed to H2, whereas H0 takes on an abnormal fate. Head epidermal lineages were followed completely in 4 animals (8 sides). H1 executed an H2-like lineage in 8/8 sides. H0 and its descendants divided in the L1, L2, and L3 stages and variably in the L4 stage, on 8/8 sides; the descendants of H0 appeared to form a syncytium. (J) e1796 dauer larvae make one ectopic cuticular substructure (arrow) in the region of the H1 descendants, anterior to the normal deirid substructure (arrowhead).

TABLE 4.

H lineage transformations inpax-6 mutants

Genotype	% anterior alaea (n)	Ectopic deirid socketsb (n)
Wild type	0	0
Class I   e41	100 (64)	0 (45)
e648	94 (56)	0 (33)
e2429	83 (52)	0 (24)
n2550	100 (22)	0 (32)
sa503	100 (26)	0 (29)
sy66	100 (32)	0 (37)
sy281	9 (21)	0 (31)
Class II   bx23	0 (24)	0 (37)
Class III   e1022	57 (52)	1.65 (range 1–2) (85)
e1178	65 (23)	1.59 (range 1–3) (73)
e1796	ND	1.0 (24)
ju468	ND	1.54 (range 0–2) (83)
ju478	ND	1.83 (range 0–3) (67)
ky112	ND	1.3 (range 0–2) (40)

^aHead alae were scored in young adult hermaphrodites using Nomarski DIC optics. Patterns classified as abnormal included: (1) anterior alae extending to tip of nose (e.g., Figure 2D), (2) anterior alae passed laterally through the “notch” of a notched head animal (e.g., Figure 2F), or (3) alae in the H cell region that are branched, broken, or looped.

^bEctopic deirid socket cuticular substructures were scored in dauer larvae using Nomarski DIC; only animals in which the normal deirid socket was visible were scored. n, number of sides examined; ND, not determined.

Class III pax-6 mutations cause transformations in epidermal cell fates that are qualitatively distinct from those seen in class I mutants. In strong class III mutants both H0 and H1 divide in H2-like patterns (Figure 4G), although the transformation of H0 is not always complete. In animals carrying the weak class III allele e1796, H1 is transformed to H2 and H0 divides unlike any other epidermal cell lineage in the wild type (Figure 4I). The H0 descendants in e1796 mutants typically do not generate alae.

We quantified the transformation of H0 and H1 to H2 in class III mutants by scoring the presence of deirid socket cells, which are normally generated by H2. The H2.aa cell is born in the early L2 stage and becomes the anterior deirid socket cell (Sulston and Horvitz 1977). In dauer larvae the deirid socket cell forms a structure known as the cuticular substructure (Albert and Riddle 1983). We quantified the H cell transformation to H2-like fates by scoring the number of ectopic cuticular substructures in mutant dauer larvae. Consistent with cell lineage analysis, strong class III mutant dauer larvae had up to two ectopic cuticular substructures (Table 4) in the regions of the H0 and H1 descendants (Figure 4H); rare animals had three ectopic substructures, perhaps due to extra divisions in the H0 or H1 lineages. e1796 dauer larvae had on average one ectopic cuticular substructure (Figure 4J). Class I and class II mutant dauer larvae did not have ectopic cuticular substructures (Table 4; Figure 4, D and F).

In summary, class III mutations cause more extensive transformations of anterior to posterior fates than class I mutations. In strong class I mutants H0 is transformed to H1, whereas in strong class III mutants both H0 and H1 are transformed to the H2 fate. The differences between the class I and class III phenotypes suggest that non-PD-containing PAX-6 isoforms promote the H1 fate. However, H1 appears normal in class II mutants. PD-containing and non-PD-containing isoforms might act redundantly in H1 fate specification; alternatively, the bx23 mutation may not affect non-PD isoforms that promote H1 fates.

Class I and class III mutations cause aberrant migration of the distal tip cell:

pax-6 mutant hermaphrodites display highly penetrant defects in migration of the gonadal distal tip cell (dtc; Nishiwaki 1999; Figure 5; Table 5). The most common defect is that the distal tip cells fail to cease migration at their normal stopping points above the center of the gonad and instead migrate along additional reflexed paths (the type I migration defect, as classified by Nishiwaki). We have classified these aberrant migrations as type A (one extra turn; Figure 5B) or type B (more than one extra turn; Figure 5C). Some pax-6 mutants also show premature dorsalward migration of the dtc (classified as PT in Table 5; Figure 5D). In addition to these defects in dtc migration, pax-6 mutations cause other less penetrant defects in gonadal development and oogenesis, possibly secondary to the aberrant dtc migration. The self-progeny brood sizes of pax-6 mutants are not significantly different from the wild type (data not shown).

Figure 5.—

Abnormal gonadal cell migration in pax-6 mutants. (A) Position of distal tip cells in wild-type young adult hermaphrodite; ventral view. The inferred migration paths of the distal tip cells are indicated by dotted lines; arrowheads indicate distal tip cell positions. (B) Ventral view of class I mutant (e648). Both distal tip cells fail to stop at the normal position, make a 180° turn, and continue their migration (Mig phenotypic class A). (C) Lateral view of class III mutant (ju468). After making a turn similar to that shown in B, the distal tip cell migrates posteriorly and makes a second turn (Mig phenotypic class B). (D) Lateral view of ju468, L4 stage. The distal tip cells undergo a premature dorsalward migration from the ventral side. Bar, 10 μm.

TABLE 5.

Gonadal distal tip cell migration defects inpax-6 mutants

	% no extra turnsa		% Mig class A, one extra turn		% Mig class B, more than one extra turn
	NPTb	PT	NPT	PT	NPT	PT
Wild type (50)	100	0	0	0	0	0
Class I (n)
e648 (16)	6.2	0	18.8	6.2	62.5	6.2
ju5 (15)	6.7	0	33.3	0	46.7	13.3
ju441 (14)	14.3	0	50	7.1	28.6	0
ky664 (16)	0	0	12.5	0	87.5	0
n3721 (40)	40	2.5	32.5	7.5	10	7.5
n3723 (28)	17.9	0	64.3	7.1	3.6	7.1
Class II
bx23 (53)	100	0	0	0	0	0
Class III
ju468 (26)	0	0	23.1	11.5	50	15.4
ky112 (20)	0	0	65	0	35	0
e1796 (37)	67.6	27.0	2.7	2.7	0	0
bc52 (29)	72.4	0	10.3	13.8	3.4	0
RNAic
PD (84)	15.3	0	45.2	0	20.2	19.3
3′-UTR (40)	0	0	5	0	72.5	22.5

The penetrance of distal tip cell (dtc) migration (Mig) defects was determined as described in materials and methods.

^aMig class A, the distal tip cell migrated normally to its stopping point and then turned 180° and migrated anteriorly or posteriorly along the dorsal midline (see Figure 5B). Mig class B, same as class A except that the dtc underwent more than one extra turn along the dorsal midline (e.g., Figure 5C).

^bIn some pax-6 mutants the distal tip cell undergoes a premature dorsalward turn in the L3 stage (PT, see Figure 5D); animals with no premature turn are denoted NPT. Thus, the first column of the table corresponds to wild-type distal tip cell migration (no premature dorsal turn, no extra turns).

^cDtc migrations were scored in the F₁ progeny of rrf-3 hermaphrodites injected with dsRNAs corresponding to the pax-6 paired domain (PD) or the 3′-UTR (see supplementary materials at http://www.genetics.org/supplemental/).

This dtc Mig phenotype was previously described for the weak class I alleles k121 and k143 and the strong class I allele e648 (Nishiwaki 1999). We found that all strong class I alleles cause similar distal tip cell Mig phenotypes (Table 5). In contrast, the class II mutation mab-18(bx23) does not cause a Mig phenotype and complements the Mig phenotype of class I alleles, suggesting that non-PD-containing isoforms do not play a role in dtc migration. Consistent with this interpretation, we found that the Mig defects in class III mutants were indistinguishable from those of strong class I mutants, in both penetrance and type of defect (Table 5; Figure 5). Class III mutants failed to complement class I mutants for the Mig phenotypes. Because missense mutations in the PD (e.g., k143) and in the HD (e1796) both cause Mig phenotypes, normal dtc migration requires function of both the PD and the HD of PAX-6 PD-containing isoforms.

Analysis of the penetrance of the Mig phenotype in pax-6 mutants suggests that dtc migration is very sensitive to reduction in pax-6 function. For example, animals mutant for the very weak class I allele k121 display highly penetrant dtc Mig defects and essentially normal epidermal morphogenesis and male tail development. Additionally, RNA interference (Table 5; see supplementary methods at http://www.genetics.org/supplemental/) was most effective at phenocopying the pax-6 Mig phenotype. Thus, the differential effect of weak pax-6 mutations on dtc migration may reflect different thresholds for PAX-6 function rather than qualitatively different functions of PAX-6 PD isoforms in different tissues.

Class I and class III pax-6 mutations cause defects in asymmetric cell divisions in the male tail:

Males mutant for strong class I or class III pax-6 mutations cannot mate because the male-specific spicules fail to develop (Figure 6; Table 6). The lack of male spicules is due to defects in the asymmetric division of the male-specific B ectoblast (Chamberlin and Sternberg 1995). In wild-type males the B cell divides unequally along the antero-posterior axis, with the larger daughter cell B.a dividing transversely to generate two similar daughters, B.al and B.ar. B.al and B.ar normally give rise to most of the spicules; B.p generates two neurons and four ectodermal cells of the proctodeum. In class I (vab-3) mutant males B.al and B.ar execute B.p-like cell lineages, leading to the absence of spicules (Chamberlin and Sternberg 1995). We confirmed these observations by direct lineage analysis of the class I mutant e2429 (data not shown). We quantified B lineage transformations in other alleles by scoring spicule morphology in adults (Table 6).

Figure 6.—

Ray and spicule development in pax-6 mutant males. (A) Wild-type adult male tail ray morphology, ventral view. Rays 4 and 6 are numbered; ray 6 is thicker than other rays and extends to the edge of the fan. (B) Wild-type spicule morphology (arrowheads), lateral view. (C) Ray 6 morphology and position are normal in the class I mutant e648. (D) Lack of spicules in an e648 male. (E) mab-18(bx23) males display almost fully penetrant transformation of ray 6 to ray 4 and fusion of the two rays in the normal position of ray 4 (4+6). (F) Spicule morphology is normal in bx23 males. (G) In strong class III (ju468) males ray 6 is thinner and frequently mispositioned, closer to but not fused with the normal ray 4. (H) Defective spicule formation of ju468 males. (I) Ray 6 transformation in the weak class III mutant e1796, showing mispositioned ray 6 that has not fused with ray 4. (J) Normal spicule morphology of e1796 males. Bar, 10 μm.

TABLE 6.

Penetrance of male tail spicule defects and ray transformations inpax-6 mutants

Genotype	% normal spiculesa (n)	% ray 6 fused to ray 4b	% ray 6 thin, adjacent to ray 4	% ray 6 thin, normal position	% ray 6 normal (n)
Wild type	100 (20)	0	0	0	100 (40)
Class I
Strong class I	0	0	0	0	100
ju441	43.2 (51)	0	0	0	100 (51)
k143	48 (50)	0	0	0	100 (50)
n3723	80 (40)	0	0	0	100 (40)
k121	100 (41)	0	0	0	100 (41)
n3721	100 (40)	0	0	0	100 (40)
Class II
bx23	100 (67)	96.4	0.8	2.6	0 (113)
Class III
e1022	0 (81)	49.3	47.1	0	3.6 (140)
e1178	0 (55)	43.3	55.8	0.8	0 (120)
e1796	51.8 (110)	75.2	21.2	3.5	0 (113)
ju468	0 (39)	46.4	53.6	0	0 (28)
ju478	0 (20)	68.8	31.2	0	0 (16)
ky112	0 (42)	90.5	9.5	0	0 (42)
bc52	97.5 (42)	0	0	0	100 (42)
we4	0 (42)	50.0	50.0	0	0 (42)
Class I/II
e648/bx23	84 (50)	0	0	1	100 (96)
Class I/III
e648/e1796	16 (24)	4	0	0	96 (24)
Class II/III
bx23/e1178	100 (29)	80.3	19.6	0	0 (51)
bx23/e1796	100 (32)	85.7	14.3	0	0 (63)

^aThe percentage of males with well-formed spicules was estimated by scoring spicule morphology in L4 or young adult males. Both spicules were scored in each animal and classified as normal or abnormal (short, crumpled, rudimentary, or completely absent). n, number of spicules scored. “Strong class I” includes the mutations e41, e648, e1062, e2429, ju5, k109, n2550, sa503, sy66, and sy281, all of which result in fully penetrant spicule defects and normal ray 6 morphology.

^bThe penetrance of ray morphology defects was scored in adult males, expressed as percentage of sides. Only sides that were fully in view were scored. Ray 6 was classified as (1) normal (i.e., fat, tapering, and posterior to ray 5), (2) thin and in normal ray 6 position, (3) thin and adjacent to ray 4, or (4) fused with ray 4. n, number of sides.

We also observed cell fate transformations in the Y lineage in pax-6 males. In pax-6(e2429) males the cells Y.p(l/r) failed to divide and instead took on a compact nuclear morphology. By analogy with the loss of asymmetry in the B lineage, these defects suggest a transformation of cell fate of Y.p(l/r) to that of Y.a, which differentiates into the neuron PDA. Thus PAX-6 PD-containing isoforms function to promote asymmetry in both the B and the Y cell lineages. We also observed polarity reversals in the F and U lineages in pax-6 class I males (data not shown); these may be secondary to the defects in B and Y as such reversals have been seen in other situations where B and Y are absent or abnormal (Chisholm and Hodgkin 1989).

mab-18(bx23) mutant males have normal spicules and mate efficiently; moreover, mab-18(bx23) complements class I and class III alleles for spicule defects (Table 6), suggesting that non-PD-containing isoforms do not function in B and Y lineage asymmetry. However, class III mutant males appear to display slightly more complete B.a-to-B.p transformations than class I mutant males. In the strong class III mutant e1178 B.a usually divides like B.p, rather than dividing transversely to yield two B.p-like cells (Chamberlin and Sternberg 1995). In 20% of e1178 males (but not in e648 males) B.a and B.p were equal in size, showing that PAX-6 is required for the unequal division of B (Chamberlin and Sternberg 1995). To ask if class III mutants displayed more severe B.a transformations than class I mutants we quantified the relative sizes of the B daughters in early L2 males. In wild-type males B.a is on average 1.3 times larger in diameter than B.p. In both class I mutants (e648) and class III mutants, B.a was on average 1.1–1.2 times larger than B.p. Although class III mutants occasionally showed more dramatic defects in B daughter size than class I mutants, these defects were not statistically significant. We conclude that the function of PAX-6 in asymmetric cell division in the B and Y lineages is largely provided by PD-containing isoforms. Because missense mutations in the PD and in the HD both result in spicule defects, both the PD and the HD may be required for this function. Furthermore, because class III mutants display a more complete loss of B division asymmetry than do class I mutants, non-PD-containing isoforms may act synergistically with PD-containing isoforms in this process. As in the specification of H cell fates, the lack of spicule defects in mab-18(bx23) mutants suggests that either non-PD isoforms are partly redundant with PD-containing isoforms or the bx23 mutation does not eliminate the non-PD-containing isoforms that function in B and Y cell fate specification.

Class II and class III alleles cause transformation of the fate of ray 6:

The class II mutation mab-18(bx23) causes a fully penetrant transformation of the fate of ray 6 in the male tail to that of ray 4 (Zhang and Emmons 1995). Ray 6 has a unique tapering morphology and can also be distinguished from other rays by its position and its ultrastructural characteristics (Sulston et al. 1980; Figure 6A). In mab-18(bx23) mutant males ray 6 is almost always fused with the normal ray 4, suggesting that ray 6 is transformed to a ray 4 fate (Figure 6G). The weak class III allele e1796 causes a similar transformation in ray 6 morphology (Figure 6I) and fails to complement bx23 for this phenotype (Zhang and Emmons 1995). In most (75%) e1796 males ray 6 fuses with ray 4; in the remaining animals ray 6 is morphologically transformed (i.e., thinner) and is situated adjacent to ray 4 but does not fuse with it (Table 6).

All strong class III alleles cause transformations in ray 6 morphology and fail to complement bx23 for this phenotype (Table 6). In contrast, class I mutant males display normal ray 6 morphogenesis; class I mutations complement class II and class III mutations for this phenotype (Table 6; Zhang and Emmons 1995). Thus, ray 6 specification does not require PD-containing isoforms of PAX-6.

Paradoxically, class III mutations classified above as strong on the basis of penetrance of morphological defects and H cell transformations (e1022, e1178, ju468, ju478, and we4) result in apparently less complete ray 6 transformations than do the weak class III alleles e1796 and ky112 or the class II allele bx23. In males bearing such strong class III mutations, ray 6-4 fusions are observed in 40–70% of sides, compared to 75% in e1796 and 90% in ky112 (Table 5). In the remaining class III males, ray 6 is morphologically transformed and adjacent to ray 4 but is not fused. The penetrance of the 6-4 fusion phenotype is increased in bx23/e1796 heterozygous males (i.e., class II/weak class III) relative to bx23/e1178 males (i.e., class II/strong class III; Table 6). Thus, the strength of the ray 6 transformation defect in class III mutants does not correlate with the strength of the defects in epidermal morphology, H cell specification, or B cell specification.

Why should a more complete loss of gene function result in a less complete transformation of ray 6 fate? We assume that fusion of ray 6 to ray 4 reflects a complete fate transformation and that morphological transformation without fusion reflects an incomplete transformation; thus bx23 mutants display the most complete transformation of ray 6 fate and e1796 males display slightly less penetrant transformations. One explanation is therefore that the strong class III alleles are not null mutations as they retain some function in ray 6 specification. However, this fails to explain why weak class III alleles (e1796, ky112) should cause more penetrant ray 6 fusion phenotypes. If the strong class III alleles are nulls, the less penetrant fusion of rays 6 and 4 in these mutants might reflect antagonistic functions of PD-containing isoforms and non-PD-containing isoforms in ray development. We therefore tested whether reducing the function of PD-containing pax-6 isoforms by RNAi could suppress the ray 6-to-4 fusion defects of mab-18(bx23) mutants. We found that RNAi of PD-containing isoforms (paired domain template) or of all isoforms (3′-UTR template) suppressed the ray 6-4 fusion phenotypes of mab-18(bx23) mutants (Table 7). Thus, reduction of function of PD-containing isoforms reduces the penetrance of ray 6-to-4 fusion in mutants with reduced function of non-PD-containing isoforms, consistent with our genetic analysis of class III pax-6 alleles.

TABLE 7.

Suppression ofbx23 ray 6 transformations bypax-6 RNA interference

RNAi treatmenta	% spicules normal	% ray 6 fused to ray 4	% ray 6 thin, adjacent to ray 4	% ray 6 WT (n)
None (HT115)	100	89	11	0 (46)
oig-1	100	92	8	0 (53)
pax-6 exons 4–6	72.7	79.5	20.5	0 (44)
pax-6 exon 9	97.2	86.1	13.9	0 (36)
pax-6 exon 8	94.4	91.6	8.4	0 (36)
pax-6 3′-UTR	90.0	78.0	22.0	0 (50)

^aRNAi experiments were performed by feeding, using the strain rrf-3; him-8; mab-18(bx23).

We envisage two possible interpretations of these data. In one model, PD-containing PAX-6 isoforms directly antagonize non-PD-containing isoforms in specification of ray 6 identity, such that reduction of function in all isoforms leads to a less penetrant fusion phenotype. Alternatively, reduction in function of all PAX-6 isoforms might trigger the activation of a parallel antagonistic pathway, resulting in only partial loss of ray 6 identity.

DISCUSSION

As part of an effort to define the functions of pax-6 in C. elegans we have analyzed an extensive series of mutations in the C. elegans pax-6 locus. The variety of mutant hunts in which pax-6 mutations have been recovered reflects both the pleiotropic function of the pax-6 gene and the apparent sensitivity of some developmental pathways, such as distal tip cell migration, to partial reduction in pax-6 function. Another reason for the frequent recovery of pax-6 alleles is the viability of strong loss-of-function mutations as homozygous strains due to incompletely penetrant lethality.

One goal of this work was to determine the null phenotype for the C. elegans pax-6 locus. The arguments that we have eliminated function of either specific isoforms or the entire locus are based largely on the allelic series of pax-6 mutations. Strong class I alleles likely eliminate function of the PAX-6A isoform, as these mutations cause indistinguishable phenotypes and result from molecular lesions predicted to truncate the PD. Class I mutations fully complement the class II allele mab-18(bx23), indicating that class I alleles do not affect the function of non-PD-containing isoforms. Because the single class II mutation does not affect all non-PD-containing isoforms, remaining non-PD-containing isoforms may have functions that we have not yet uncovered. RNA interference experiments targeting known non-PD-containing isoforms failed to reveal additional phenotypes (see supplementary data at http://www.genetics.org/supplemental/); however, these experiments are not conclusive because RNAi only weakly phenocopies the mab-18(bx23) phenotype. Nevertheless, our analysis suggests that strong class III alleles eliminate pax-6 function. A caveat to this conclusion is that class III mutants could express truncated PAX-6 proteins that lack the homeodomain, which have been shown to be functional in Drosophila (Punzo et al. 2001). However, attempts to enhance the phenotypes of class III mutants by RNAi were unsuccessful, using RNAi treatments that consistently reduce pax-6 function in control experiments. We conclude that strong class III alleles likely represent the amorphic condition of the C. elegans pax-6 locus. Knowledge of the pax-6 null phenotype will be important for future analysis of the functions of pax-6 regulators and targets in C. elegans.

The C. elegans pax-6 locus expresses multiple transcripts that encode both typical PAX6-like proteins and smaller isoforms lacking the paired domain. Both previous work and this analysis of the pax-6 allelic series are consistent with these products having both independent and redundant functions in development. PD-containing isoforms have unique functions in gonadal distal tip cell migration, head epidermal morphogenesis, specification of the H0 ectodermal cell fate, and specification of B.a and Y.p ectodermal blast cell fates. Nonpaired domain-containing isoforms have unique functions in specification of the ray 6 fate. Both PD and non-PD isoforms may function redundantly to specify the H1 cell fate. Our comparison of the phenotypic strength of class I and class III mutants suggests that non-PD-containing isoforms may synergize with PD-containing isoforms in head epidermal morphogenesis. Non-PD-containing isoforms may also weakly synergize with PD-containing isoforms in B division asymmetry. Finally, PD-containing isoforms appear to antagonize the role of non-PD-containing isoforms in ray 6 fate specification. pax-6 has also been shown to function in cell fate specification in nondividing head epidermal cells; both PD-containing and non-PD-containing isoforms appear to contribute to regulation of transcription of the sine oculis class homeobox gene ceh-32 (Dozier et al. 2001). An important avenue for future work will be to define additional targets of pax-6 regulation in the many tissues affected in pax-6 mutants.

The existence of multiple pax-6 isoforms that share a common C terminus explains an apparent paradox of the C. elegans pax-6 allelic series, in which C-terminal mutations can cause stronger phenotypes than can N-terminal mutations. In this context it is interesting to compare the C. elegans allelic series with those known for other pax-6 family members. The most extensive series of pax-6 mutations available is that for the human PAX6 locus (Prosser and van Heyningen 1998), listed in an online database (http://pax6.hgu.mrc.ac.uk/; Brown et al. 1998) of >300 allelic variants. Multiple mutations are also known for the mouse Pax6 locus (Favor et al. 2001) and Drosophila eyeless (Kronhamn et al. 2002).

Our knowledge of an extensive set of mutations in human PAX6 can be attributed to the dominant effects of PAX6 loss of function. Most human PAX6 mutations have been found in aniridia patients; a minority have been found in patients with other eye disorders. Mutations associated with strong aniridia phenotypes likely cause complete loss of PAX6 function due to premature stop codons generated by base substitution, frameshift, or splice site mutations; transcripts bearing premature stop codons may be subject to nonsense-mediated decay and thus may not express truncated proteins. These data support the model that the dominant inheritance of aniridia reflects PAX6 haplo-insufficiency (van Heyningen and Williamson 2002; Vincent et al. 2003). Mice and rats also display haplo-insufficiency for Pax6, as small-eye phenotypes are seen in animals heterozygous for Pax6 deletions (Glaser et al. 1994). Deletion of both copies of Pax6 in mice leads to embryonic lethality and defects in the development of the central nervous system (Hogan et al. 1986). Mutation of both copies of human PAX6 is likely to result in lethality (Glaser et al. 1994), and heterozygotes also display brain malformations (Sisodiya et al. 2001). Thus, Pax6 proteins play a dosage-sensitive role in eye development in vertebrates, but are also essential for the development of other brain regions and tissues.

C. elegans pax-6 function appears haplosufficient for the processes described in this article, even in cells that are extremely sensitive to reduction in pax-6 function, such as the distal tip cell. pax-6 gene dosage may in general be less critical for C. elegans development. In Drosophila, eyeless is haplosufficient (Kronhamn et al. 2002), although the interpretation of complete loss of Pax6 function in Drosophila is complicated by the presence of other Pax6 family members such as twin of eyeless (Czerny et al. 1999), which has partial redundancy with eyeless (Kronhamn et al. 2002). The Drosophila genome also contains two genes that encode Pax-6-related proteins with truncated Paired domains, eyegone (Jang et al. 2003) and twin of eyegone (Aldaz et al. 2003); eyegone functions in eye development independently of eyeless (Dominguez et al. 2004). The C. elegans genome encodes a single eyegone-like gene (Hobert and Ruvkun 1999), of unknown function. Thus, the apparent lack of sensitivity of Drosophila or C. elegans to pax-6 gene dosage effects might reflect compensation by other pax-6 family members in these species. Our observations do not exclude the possibility that C. elegans pax-6 might be haplo-insufficient in contexts not examined here.

Some C. elegans pax-6 mutations cause identical or very similar changes to those of human PAX6 mutations, underscoring the evolutionary conservation of Pax6 structure and function. A change equivalent to the strong class I allele e41 has been found in several independent sporadic aniridia patients and appears to be a hotspot for PAX6 mutation (Prosser and van Heyningen 1998). The strong class III HD mutation ju468 (Q262amber) is equivalent to that found in aniridia patient 3.1 (Q255amber; Sisodiya et al. 2001); the strong class III allele we4 is equivalent to a Q221amber mutation found in another aniridia patient (Zumkeller et al. 2003). A missense change equivalent to the weak class I allele sy281 (G19R) has been found in a human patient with Peters anomaly (G18R; cited by van Heyningen and Williamson 2002). Missense mutations in human PAX6 tend to cause either milder forms of aniridia or other eye disorders, such as displaced pupils or Peters anomaly (Azuma et al. 1996, 2003; Hanson et al. 1999; Hanson 2003). In humans it is not certain whether these missense mutations cause partial loss of PAX6 function or whether they result in gain of function; we find that sy281 and the other missense alleles behave as a partial loss of function.

Non-PD-containing PAX-6 isoforms have been reported in several species. Non-PD-containing isoforms of PAX-6 were first reported in quail (Carriere et al. 1993). In mice, non-PD-containing isoforms arise by alternative splicing (Mishra et al. 2002) or by transcription from an internal promoter, as in C. elegans (Kleinjan et al. 2004). Non-PD-containing isoforms promote transcriptional activation by full-length PAX-6 proteins in vitro (Mikkola et al. 2001). However, the in vivo roles of non-PD-containing PAX-6 isoforms have in general not been characterized. If non-PD-containing PAX6 isoforms have roles in the development of other species, one might expect to see, as in C. elegans, that the effects of N-terminal (PD) truncation mutations are distinguishable from those of C-terminal truncation mutations. At present non-PD-containing isoforms have not been detected in humans, and nonsense mutations in the human PAX6 HD have been reported to have the same effects as nonsense mutations in the PD (van Heyningen and Williamson 2002; Vincent et al. 2003). However, interpretation of the human allelic series is complicated by potential variability of the genetic background and potential variability in the clinical diagnosis. Similarly, murine Pax6 truncation mutations result in similar phenotypes both as homozygotes and as heterozygotes (Favor et al. 2001), suggesting that non-PD-containing isoform function may not contribute to the Pax6 phenotype. As in C. elegans, the function of non-PD-containing isoforms may be revealed only by rare isoform-specific mutations or by a detailed comparison of N-terminal and C-terminal mutant phenotypes.