Identification of spatial and temporal cues that regulate postembryonic expression of axon maintenance factors in the C. elegans ventral nerve cord

Postmitotic neurons display specific gene expression profiles that define the differentiated state of the neuron. These gene expression profiles generally show two distinct kind of temporal dynamics. Some genes are transiently expressed until shortly after the birth of a neuron and include transcription factors, axon pathfinding and target recognition cues, which serve to wire a neuron into specific circuits (e.g.Wadsworth et al., 1996;Sharma et al., 1998;Sarafi-Reinach et al., 2001). Other genes are constitutively expressed throughout the life of a postmitotic neuron and define several functional properties of the neuron such as its neurotransmitter phenotype (e.g. Eastman et al., 1999). Transient peaks of gene expression after the birth of a neuron are usually coupled with the exit from the cell cycle, while sustained expression of genes during the life of a neuron is often controlled by transcriptional autoregulatory feedback loops(Edlund and Jessell, 1999). By contrast, little is known about temporally defined, stereotyped changes in gene expression patterns within postmitotic neurons that are induced long after neuron generation and differentiation. Although neurons are able to respond transcriptionally to signaling inputs (e.g. in the course of learning and memory) (Bailey et al.,1996), those changes in gene expression patterns usually do not occur at pre-programmed times in the life of a neuron.

Instances in which neuronal phenotypes are modified in a developmentally programmed, stereotyped temporal manner that occurs significantly past the axon outgrowth and target recognition stage include sexual differentiation in the mammalian brain at defined postnatal stages(MacLusky and Naftolin, 1981),remodeling of the insect CNS during metamorphosis(Levine et al., 1995) and synaptic rewiring of postmitotic motor neurons in the ventral nerve cord (VNC)of C. elegans (White et al.,1978). In all three circumstances, temporally defined alterations in gene expression are presumably mediated through hormones and their effector transcription factors to eventually cause a structural remodeling of individual groups of neurons (Levine et al., 1995; MacLusky and Naftolin, 1981; Zhou and Walthall, 1998).

We describe our analysis of another neural gene expression program that is not only under tight spatial, but also under stereotyped temporal control. Unlike the cases described above, these dynamics in gene expression do not serve to induce remodeling of the neuron but provide the neuron with a novel and specific capacity at a defined time in its life. The case studied is the PVT interneuron located in the VNC of C. elegans(Fig. 1). Using cell ablation techniques, PVT was recently found to have two temporally separable roles in axonal patterning in the VNC. In mid-embryonic stages, PVT is required for axon attraction into the VNC (Ren et al.,1999). Surprisingly, at the postembryonic larval L1 stage, PVT was found to have an additional role in non-autonomously maintaining the intact axon scaffolding in the VNC, such that loss of PVT causes axon `flip-overs'across the ventral midline (Fig. 1) (Aurelio et al.,2002).

These two temporally separated functions of PVT correlate with the execution of a temporally controlled, biphasic gene expression program. In mid-embryogenesis, during the development of the VNC, PVT transiently expresses the UNC-6/Netrin cue to attract axons into the VNC(Wadsworth et al., 1996). Several hours later, in the larval L1 stage, PVT initiates the expression of several members of the zig gene family(Fig. 1)(Aurelio et al., 2002).zig genes encode small transmembrane (zig-1) or secreted(zig-2 to zig-8) proteins composed exclusively of two immunoglobulin domains. Six zig genes are expressed in restricted domains of the nervous system, while the two remaining family members are expressed outside the nervous system. The expression of all six neuronally expressed zig genes overlaps in the PVT interneuron. The onset of expression of several of the zig genes in the PVT neuron in the L1 stage precisely correlates with the requirement for PVT in maintaining axonal organization of the VNC (Aurelio et al.,2002). Consistent with the hypothesis that zig genes mediate the maintenance function of PVT, we found that a deletion in thezig-4 locus affects the maintenance of axonal position of a subset of those neurons that are affected by PVT ablation(Aurelio et al., 2002),suggesting that different zig family members are required to maintain the positioning of distinct subsets of VNC axons.

What are the molecular mechanisms that underlie the biphasic gene expression program in PVT and, specifically, what are the factors that confer the precise spatial and temporal aspects of zig gene expression? We used postembryonically expressed zig gene reporter gene constructs as tools to investigate these mechanisms. We define three factors that are required for the execution of the temporal and spatial aspect of ziggene expression and show that their absence not only disrupts ziggene expression but also leads to defects in the maintenance of axon positioning in the VNC. Furthermore, we find that spatiotemporal misexpression of one of the zig genes, zig-4, causes developmental defects in the VNC, suggesting that the normal spatiotemporal control of ziggene expression serves to dedicate them to a role in axon maintenance and to prevent them from interfering with axon outgrowth.

• OH110, lim-6(nr2073)X (Hobert et al., 1999b);

• OH170, ceh-14(ch3)X (derived from TB522(Cassata et al., 2000) through outcrossing of dpy-20);

• OH158, zig-4(gk34)X (Aurelio et al., 2002);

• OH636, lim-6(nr2073) ceh-14(ch3)X;

• DR441, lin-14(n179ts)X;

• MT1848, lin-14(n360ts)I;

• MT1388, lin-14(n355n679ts)I;

• MT1524, lin-28(n719)X (Ambros and Horvitz, 1984);

• MT2257, lin-42(n1089)X (Z. Liu, PhD thesis, Massachusetts Institute of Technology, 1990);

• NG83, fax-1(gm83)X (Much et al., 2000); and

• CB270, unc-42(e270)V (Baran et al., 1999).

PVT cell fate markers used in this study are syIs7: Is[gpa-2::lacZ;dpy-20(+)] (Zwaal et al.,1997), otIs90: Is[pin-2::gfp; rol-6(d)](Hobert et al., 1999a),otIs85: Is[srq-1::gfp; rol-6(d)] and otIs39:Is[unc47Δ::gfp; lin-15(+)]. The srq-1 reporter is a translational fusion constructed from a 4.4 kb genomic fragment, encompassing 2.2 kb of sequence upstream of the predicted ATG and 2.2 kb of exonic and intronic sequences of the predicted serpentine receptor gene F59B2.13 gene. The construct was generated using a PCR fusion protocol(Hobert, 2002). A corresponding F59B2.13::lacZ construct had been constructed and analyzed by Ian Hope in the course of a genome-wide expression pattern analysis (Lynch et al., 1995). The otIs39 integrant derives from juEx60(Eastman et al., 1999) and contains a promoter fragment from the unc-47 gene. The pan-neuronal marker evIs111 (integrated F25B3.3::gfp); the PVQ markeroyIs14 (integrated sra-6::gfp); and the zig::gfpreporter gene constructs otIs25 (zig-1::gfp), otIs7(zig-2::gfp), otIs14 (zig-3::gfp), otIs20(zig-4::gfp), otIs11 (zig-5::gfp) andotIs96 (zig-8::gfp) have been described previously(Altun-Gultekin et al., 2001;Aurelio et al., 2002). Thelim-6r::gfp expression construct is a full length translational fusion of the lim-6 locus to gfp and contains all the regulatory regions required for rescuing lim-6 mutant phenotypes(Hobert et al., 1999b);expression in PVT was scored in an extrachromosomal line, otEx406. The ceh-14::gfp construct was engineered using a PCR fusion protocol(Hobert, 2002) and encompasses 3.75 kb of promoter and parts of the first exons. A similar promoter fragment has previously been described to faithfully reflect the expression of the endogenous gene (Cassata et al.,2000). Expression of this construct was monitored in the chromosomally integrated line otIs126.

Reporter gene expression levels and neuroanatomical defects were assayed— with the exception of the extrachromosomal lim-6r::gfp array— using chromosomally integrated gfp or lacZexpression constructs (see above). The integrated arrays were crossed into the respective mutant backgrounds, thus allowing a direct comparison between otherwise isogenic backgrounds. If not indicated otherwise, reporter gene expression was scored in well-fed animals. zig-4::gfp expression was also scored in starved animals (see text). For these starvation experiments,eggs were prepared from otIs20 (zig-4::gfp) animals using a standard bleaching protocol, plated onto non-seeded plates and scored forzig-4 expression in PVT several days later.

To score neuroanatomical defects at the early L1 stage (when PVT has yet no maintenance role), eggs were prepared by bleaching and freshly hatched L1s scored 2 hours after the preparation of the eggs. With regard to neuroanatomical defects, we focused on the description of defects in the VNC. We found that in various mutant backgrounds (e.g. lim-6 ceh-14)neuroanatomical defects can also be observed in other axon fascicles.

The nucleus of PVT was identified based on its characteristic position and morphology by Nomarski optics. Ablation was performed using a LSI VSL-337 laser as previously described (Aurelio et al., 2002; Bargmann and Avery,1995).

A genomic fragment of the zig-4 locus ranging from its start to stop codon and including all introns was amplified with restriction sites on either end of the amplicon. The amplicon was cloned behind a 1.3 kb promoter fragment from the upstream regulatory region of the flp-1 gene(Nelson et al., 1998) and a 2.4 kb fragment from the muscle-specific myo-3 promoter(Okkema et al., 1993). When fused to gfp, the 1.3 kb flp-1 promoter (-1358 to -9 relative to ATG) is exclusively active in the AVKL and AVKR neurons. All expression constructs were injected into oyIs14 animals at 50 ng/μl using ceh-22::gfp(Okkema and Fire, 1994) at 50 ng/μl as an injection marker.

LIM homeobox genes regulate various aspects of postmitotic neural differentiation (reviewed by Bach,2000; Hobert and Westphal,2000; Jessell,2000; Lee and Pfaff,2001) and have been proposed to act through combinatorial codes of overlapping expression (`LIM-code')(Sharma et al., 1998;Thor et al., 1999;Tsuchida et al., 1994). A similar LIM code may exist in the tail of an L1 stage C. eleganslarva (Fig. 1). Specifically,two of the seven C. elegans LIM homeobox genes, theLmx-class gene lim-6(Hobert et al., 1999b) and theLhx3-class gene ceh-14(Cassata et al., 2000) are co-expressed in PVT (Fig. 1,Fig. 2A), yet their function in PVT development has not been addressed.

Six out of the eight zig gene family members are expressed in PVT and, in addition, in a few other cell types(Aurelio et al., 2002). We tested whether these six genes are under control of the lim-6 andceh-14 LIM homeobox genes by crossing previously described,chromosomally integrated gfp reporter genes that monitor the PVT expression of the six genes (Aurelio et al., 2002) into lim-6(nr2073) andceh-14(ch3)-null mutant animals, respectively. Neither LIM homeobox gene mutation had a significant effect on initiation of zig gene expression by itself. However, in lim-6(nr2073) ceh-14(ch3) double null mutants, the expression of all postembryonically expressed ziggenes was significantly affected in PVT, if not completely abolished(Fig. 2B,Fig. 3).

Each one of the PVT-expressed zig genes is also expressed in a restricted subset of other cell types(Aurelio et al., 2002), yetlim-6 ceh-14 double null mutants only show an effect on ziggene expression in PVT (Fig. 2Band data not shown). As shown in Fig. 3, the impact of lim-6 ceh-14 is strongest on thosezig genes that are expressed strictly postembryonically (zig-2,zig-3, zig-4, zig-8), although it is weaker, but still significant, on those zig genes that already show some expression in embryos(zig-5::gfp, completely penetrant embryonic expression in wild type;zig-1::gfp, incompletely penetrant embryonic expression in wild type)(Aurelio et al., 2002).

Given the significant loss of zig gene expression in PVT inlim-6 ceh-14 double mutants, we next examined whether the generation and overall cell fate of PVT is grossly affected in lim-6 ceh-14mutants. To this end, we monitored the expression of several PVT cell fate markers, namely the GTPase gpa-2(Zwaal et al., 1997), the LIM-only gene pin-2 (Hobert et al., 1999a), the serpentine receptor srq-1 (see Materials and Methods) and the unc-47 GABA transporter gene(Eastman et al., 1999). All of these markers are expressed normally in lim-6 ceh-14 double mutants(Fig. 4). In addition, as assessed with the unc-47Δ::gfp marker, PVT axon morphology appears unaffected (data not shown). Last, the embryonic,unc-6/Netrin-mediated role of PVT in attraction of axons from tail ganglia into the VNC (Ren et al.,1999) is unaffected in lim-6 ceh-14 double mutants, as visualization of several of these axons in lim-6 ceh-14 mutants reveals no defect (data not shown). Taken together, these data suggest that in the absence of the two LIM homeobox genes, lim-6 and ceh-14,PVT adopts its correct cell fate, i.e. displays all known features of its terminal differentiation program, but is incapable of initiating expression ofzig genes in the L1 stage.

Although both lim-6 and ceh-14 are jointly required for initiation of zig gene expression, they are not jointly sufficient to induce zig gene expression for the following two reasons. First, bothlim-6 and ceh-14 are already expressed embryonically(Fig. 2A) and hence significantly precede the onset of zig gene expression; second,pan-neuronal misexpression of lim-6 under control of theunc-119 promoter yields co-expression of lim-6 andceh-14 in several neurons (ceh-14 is expressed in a total of 18 neurons) (Cassata et al.,2000), yet does not cause ectopic zig-4 expression in any of these cells (data not shown).

We also tested whether other transcription factors previously shown to be expressed in PVT, namely the nuclear hormone receptor fax-1(Much et al., 2000) and the paired-type homeobox gene unc-42(Baran et al., 1999), affectzig gene expression. By crossing various zig::gfp reporter gene integrants with fax-1 and unc-42 single mutants and with lim-6; unc-42 double mutants, we found this not to be the case(data not shown).

A loss-of-function mutation in zig-4 or postembryonic laser removal of PVT, the cellular source of zig-4, causes a disorganization of the left and right axonal tracts in the VNC, characterized by a `flip-over' of embryonically established axonal tracts into the opposite fascicle (Fig. 1)(Aurelio et al., 2002). The disruption of zig gene expression in PVT in lim-6 ceh-14double mutant animals would thus be expected to cause similar disruptions of VNC architecture. To investigate this point, we crossed lim-6 ceh-14double null mutants with transgenic animals in which either the whole VNC or individual neurons within the VNC are labeled with gfp. Using a pan-neuronal marker, we indeed observed left/right axon flip-overs from one fascicle into the opposite fascicle in the VNC of lim-6 ceh-14 double mutant animals (Fig. 5,Table 1).

The VNC axon flip-over defect in lim-6 ceh-14 animals manifests itself only at stages past the late L1 stage, while freshly hatched animals show no mutant phenotype (Table 1). Hence, lim-6 ceh-14 mutants show axonal maintenance defects in the VNC that resemble those in PVT ablated and zig-4mutant animals not only in overall appearance but also in their temporal profile.

In summary, VNC axon flip-over defects are present in the lim-6 ceh-14 double mutant, but not in the single mutants; the expression oflim-6 and ceh-14 exclusively overlaps in PVT; PVT loses the postembryonic zig gene expression in lim-6 ceh-14 double mutants; and the VNC defects show a zig-4(—) and PVT(—)-like characteristic postembryonic profile. These points strongly suggest that lim-6 and ceh-14 act in PVT to affect VNC maintenance by regulating zig gene expression.

Examination of individual axons in the VNC in lim-6 ceh-14 double mutant animals lead to a surprising observation. Although the axons of the PVQL/R neurons flip into the opposite site of the VNC in zig-4 mutant animals, no such defect above background level can be observed in lim-6 ceh-14 double mutant animals, in which zig-4 expression is downregulated (Fig. 5B,Fig. 6). This observation could be explained by zig-4 expression levels and hence zig-4activity not being as significantly affected in lim-6 ceh-14 double mutants as in the zig-4 null mutant. Alternatively, lim-6and/or ceh-14 gene activity could be required for the PVQL/R flip-over event to occur in the absence of zig-4. To address this issue, we conducted a genetic epistasis test in which we analyzed PVQL/R neuroanatomy in lim-6(nr2073) ceh-14(ch3) zig-4(gk34) triple null mutant animals. We found that the zig-4 mutant phenotype is suppressed in these triple mutant animals(Fig. 6).

We next asked whether lim-6 ceh-14 act in PVT to suppress thezig-4 mutant phenotype. It could, for example, be envisioned that expression of a factor X is repressed by the LIM-6 and/or CEH-14 proteins in PVT and that its derepression in lim-6 ceh-14 mutants prevents an axon flip-over caused by the absence of zig-4. Alternatively,lim-6 and/or ceh-14 could act outside of PVT to suppress thezig-4 mutant phenotype. To distinguish between these possibilities we laser ablated PVT in lim-6 ceh-14 double mutant animals. While PVT ablation in wild-type animals causes PVQL/R axon flip-overs, the same ablation in lim-6 ceh-14 animals causes no PVQL/R flip-overs above background levels (Fig. 6). Hence,lim-6 and ceh-14 must act outside of PVT to suppress the PVT(—) or zig-4(—)-induced axon flip-over.

Although the above mentioned experiments were all made with a lim-6 ceh-14 double null mutant, it could be envisioned that one of the two LIM homeobox genes alone is sufficient to suppress the zig-4 induced axonal flip-over of PVQL/R. In contrast to lim-6, which in the L1 stage is not expressed elsewhere in the VNC than in PVT, ceh-14 is a good candidate as it is normally expressed in PVQL/R(Fig. 1)(Cassata et al., 2000). It was thus conceivable that ceh-14 is required for PVQL/R to flip into the opposite cord, e.g. through regulating the expression of homophilic adhesion molecules, which we have previously implicated in playing a critical role in the axon flip-over process (Aurelio et al.,2002). We indeed found that in ceh-14 zig-4 double null mutant animals, but not in lim-6 zig-4 double null mutant animals,the zig-4-mediated PVQL/R axon flip-over phenotype is suppressed(Fig. 6).

Taken together, our data suggests that lim-6 and ceh-14probably act in PVT to affect zig gene expression and thatceh-14 has an additional role outside of PVT, possibly within PVQL/R,to enable axon flip-over to occur in the absence of a functional axon maintenance mechanism.

Previous studies focused on postembryonic aspects of lim-6 andceh-14 expression (Cassata et al., 2000; Hobert et al.,1999b). We examined the temporal dynamic of lim-6 andceh-14 reporter gene expression in PVT in more detail, and found both genes to be expressed already in mid-embryonic stages(Fig. 2A) and maintained throughout the life of the animal (data not shown). Considering the postembryonic expression of most zig genes (zig-1, zig-2, zig-3,zig-4, zig-8) in PVT, the embryonic expression of lim-6 andceh-14 indicates that these genes are insufficient to provide the temporal trigger for onset of zig gene expression at the L1 stage. Although we can not exclude the possibility that the lim-6 andceh-14 reporter gene constructs do not accurately reflect endogenous gene expression, the reporter gene results provided us with sufficient motivation to search for other, temporal parameters possibly involved in timing zig gene expression.

An important temporal trigger for initiation and progression of larval development is the feeding state of the animal. In the absence of an external food source, animals arrest after hatching at the L1 stage. Hence, we considered the possibility that a food-dependent signal, possibly a hormonal signal, is involved in determining the timing of zig gene expression. By cultivating transgenic zig-4::gfp animals (otIs20) on plates that contain no food, we found that 20/20 L1 animals showzig-4::gfp in PVT under starvation conditions (see Materials and Methods). This results suggests that zig-4::gfp expression is not triggered through an extrinsic, food-dependent signal, but rather depends on an intrinsic timer.

In C. elegans and possibly other animals, timing of developmental events depends on well defined heterochronic genes(Ambros, 2000;Rougvie, 2001;Slack and Ruvkun, 1997). Heterochronic genes act at defined larval stages and their aberrant activity(either through loss or ectopic expression) causes either precocious or delayed execution of developmental events. The lin-14 gene codes for a ubiquitously expressed nuclear protein, which is so far the only heterochronic factor known to be required within the L1 stage to ensure the correct execution of L1-specific blast cell divisions(Ambros and Horvitz, 1984;Ambros and Horvitz, 1987;Ruvkun and Giusto, 1989) and neuronal rewiring events (Hallam and Jin,1998). To investigate whether the role of this heterochronic gene also extends to the timing of a postmitotic gene expression program, we crossed a representative zig gene reporter, chromosomally integratedzig-4::gfp, with three different loss-of-function alleles oflin-14. We focused on the zig-4 gene as it is the onlyzig gene for which a mutant phenotype is available so far and which would therefore allow us to compare its mutant phenotype with potentiallin-14 mutant phenotypes (see below). We found that the L1-specific onset of zig-4::gfp expression is abolished in all lin-14loss-of-function alleles tested (Fig. 7A, Table 2). Notably, onset of expression is not merely precociously executed or later initiated in the L2 stage, but absent throughout all stages. lin-14does not affect the generation or overall fate of PVT as the cell fate markersunc-47Δ::gfp and pin-2::gfp show normal expression in PVT in lin-14 mutant animals; moreover, the axon anatomy of PVT is unaffected in lin-14 mutants (Fig. 7B).

A lin-14 gain-of-function allele that causes LIN-14 protein to be present at all larval stages (Ambros and Horvitz, 1984; Arasu et al.,1991) has no effect on zig-4 expression(Table 1). A heterochronic gene that is required for L2 specific developmental events, lin-28(Ambros and Horvitz, 1984),also has no effect on zig-4 expression(Table 2). The periodhomolog lin-42, which affects timing of L4 stages (Z. Liu, PhD thesis, Massachusetts Institute of Technology, 1990), but whose expression has been shown to peak in each individual larval stage(Jeon et al., 1999) leaveszig-4 expression unaffected as well(Table 2). Taken together,while zig-4 is expressed throughout all larval stages well into adulthood, it is specifically the passing through the L1 stage but not any other larval stage that is required for initiation of zig gene expression.

Expression of zig-4::gfp is not affected by lin-14 in cells other than PVT. Consistent with this observation, none of the otherzig-4::gfp expressing cells (ASK, ASI, BAG, M2) show stage-specific regulation of zig-4 expression in wild-type animals.

Loss of zig-4 expression in lin-14 mutant animals would be expected to cause axon maintenance defects in the VNC of lin-14mutants. We analyzed VNC architecture by crossing three different loss-of-function alleles of lin-14 with transgenic animals that express a gfp reporter either in all VNC neurons or in a selected subset. As expected from the loss of zig-4 gene expression, we found that lin-14 animals display axon flipovers in the VNC(Fig. 8;Table 3). The defects can be observed with a pan-neuronal gfp marker as well as with the PVQL/R specific marker and are thus very similar to defects observed inzig-4 mutant animals. Heterochronic genes that affect later stages of larval development do not affect VNC architecture (data not shown).

The VNC defects in lin-14 mutant animals display temporally dynamic axonal defects. Freshly hatched lin-14 animals show no defects above background level, while animals at later larval and adult stages display significant defects (Table 3). These temporal dynamics precisely reflect the temporal dynamics observed in zig-4 mutants animals and are also consistent with the observation that PVT is specifically required in the L1 stage to affect VNC architecture (Aurelio et al.,2002).

lin-14 is ubiquitously expressed in neuronal and non-neuronal cells (Ruvkun and Giusto,1989). Its expression is initiated in embryogenesis, peaks right after hatching, then rapidly fades and is absent by the L2 stage(Ruvkun and Giusto, 1989). To affect VNC architecture, lin-14 could conceivably act in neurons or,alternatively, in the underlying hypodermis. The later possibility seems unlikely, because a null mutation in the lin-28 gene, which diminishes LIN-14 protein levels in the hypodermis, but not in neurons(Arasu et al., 1991), has no effect on VNC architecture (Table 2). As we have so far been unable to express lin-14 under control of a PVT-specific promoter, we sought to further narrow the focus of action of lin-14 action by conducting a laser ablation experiment. We reasoned that if lin-14 acts in a PVT-dependent manner, laser ablation of PVT in a lin-14 mutant background would not enhance thelin-14 mutant phenotype. By contrast, if lin-14 were to act independently of PVT in some other cell, laser ablation of PVT would enhance the lin-14 mutant phenotype. Laser ablation of PVT causes a 36.4%penetrant PVQL/R phenotype (Fig. 6). We found that 38.1% of lin-14(n179) animals in which PVT was ablated showed defects in PVQL/R axon positioning(Table 3). This number is virtually the same as the 40.8% defects observed in unoperatedlin-14(n179) animals (Table 3). Although this result does not conclusively prove thatlin-14 acts within PVT, it provides strong suggestive evidence thatlin-14 activity is mediated through PVT to affect VNC structure. Together with the effect of lin-14 on zig-4 expression, we conclude that the lin-14 defects are probably due to absentzig gene expression in PVT.

We have defined three factors, lim-6, ceh-14 and lin-14that are required for the correct spatiotemporal expression of zig-4. We investigated the consequences of overriding the spatiotemporal control ofzig-4 mediated through these factors by ectopically expressingzig-4 under the control of heterologous promoters at earlier time points and at different locations in the VNC. To this end, we made use of the promoter fragments from the flp-1 and myo-3 genes(Nelson et al., 1998;Okkema et al., 1993). Theflp-1 promoter fragment is exclusively active in the AVKL and AVKR neurons (see Materials and Methods), which send axons along both sides of the VNC, while the myo-3 promoter is active in body wall muscle cells which abut the left and right VNC and send muscle arms into the VNC. As assessed through analysis of the expression of corresponding gfpfusion constructs, both promoters are already embryonically active (data not shown). We find that adult animals that misexpress zig-4 under control of either of the two promoters show mispositioned PVQL/R axons in the VNC (Fig. 9). The observed defects could be caused through two distinct mechanisms. In one scenario,precise levels and/or a defined localization of the endogenous ZIG-4 protein is required in the L1 stage to ensure axon maintenance and ectopic ZIG-4 expression obscures this finely tuned distribution. In an alternative scenario, ectopically expressed ZIG-4 may already act in embryonic stages to affect PVQL/R axon outgrowth. In an attempt to distinguish between these two possibilities, we assessed PVQL/R axon anatomy right after hatching, that is,before PVT and zig-4 are required to maintain axon anatomy in the VNC. We find that animals that ectopically express zig-4 already display defects at this early stage (Fig. 9). We thus conclude that ectopically expressed zig-4affects development of the VNC, probably during the axon outgrowth stages.

Several of the transgenic lines revealed another intriguing aspect ofzig-4 function. A subset of the transgenic lines (flp-1promoter: line #2 and #3, myo-3 promoter: line #2) show no developmental defects, yet showed maintenance defects in the adult(Fig. 9). As levels of expression from independent extrachromosomal arrays can be highly variable, we consider it possible that these lines do not express enough zig-4 to cause embryonic defects, but enough zig-4 to interfere with the maintenance role of endogenous zig-4 at the L1 stage. In other words,ectopic zig-4 has two separable effects: it can interfere with axon outgrowth but it can also disrupt the normal function of zig-4 at postembryonic stages to affect VNC axon positioning. Taken together, the precise spatiotemporal control of zig-4 gene expression is necessary to prevent it from inappropriately interfering with VNC patterning at distinct stages.

The PVT neuron displays unusual gene expression profiles. Shortly after its birth, the expression of several genes that define the differentiated state of PVT is initiated. One of these genes, unc-6/Netrin is only transiently expressed (Wadsworth et al.,1996), while other genes, such as gpa-2, pin-2 andsrq-1 are expressed throughout the life of the animals(Fig. 1). Yet another set of genes, namely several zig genes, is expressed in a strictly postembryonic manner. The functional relevance of these genes has been demonstrated through laser ablation and mutational analysis(Aurelio et al., 2002) (T. B. and O. H., unpublished). We extend the previous finding of the requirement forzig genes in VNC patterning, by showing that ectopic expression ofzig-4 causes VNC patterning defects, thus underscoring the necessity of precise temporal and spatial control of zig gene expression. We have described separate sets of factors that define the spatial and temporal expression domains of zig genes in the PVT neuron.

The lim-6 and ceh-14 LIM homeobox genes define the spatial domain of zig gene expression. The well defined cooperative action of several LIM homeobox genes in cell specification in the vertebrate spinal cord (Sharma et al.,1998; Tsuchida et al.,1994; Jessell,2000; Lee and Pfaff,2001) and in Drosophila(Thor et al., 1999),collectively referred to as a LIM-code, and the biochemical evidence for hetero- and homodimerization of individual LIM homeodomain proteins, including the vertebrate LIM-6 ortholog LMX-1 and the vertebrate CEH-14 ortholog LHX3(Jurata et al., 1998;Thaler et al., 2002), suggest a model in which LIM-6 and CEH-14 bind as heterodimers either directly tozig gene promoters or regulate the expression of intermediary factors that directly control zig gene expression. In the absence of either LIM homeodomain protein alone, the other LIM homeodomain protein may still be capable of activating its target gene(s) as a homodimer, thus explaining why each single mutant has little observable effect on zig gene expression.

Although lim-6 and ceh-14 are both jointly required forzig gene expression, they are not sufficient to induce ziggene expression, as inferred from the observation that the embryonic expression of both genes is insufficient to yield embryonic zig gene expression and from the observation that supplying lim-6 in otherceh-14-expressing cells does not cause ectopic zig gene expression. The effects of lim-6 and ceh-14 are thus strictly dependent on the cellular context and strongly suggests the existence either of other activators required in PVT to induce zig gene expression or of repressors that prevent lim-6 and ceh-14from activating zig genes in embryonic stages and/or other cell types. The strict cellular context-dependence of target gene regulation bylim-6 and ceh-14 is reminiscent of several other C. elegans LIM homeobox genes, such as mec-3(Duggan et al., 1998) orttx-3 (Altun-Gultekin et al.,2001). Each of these genes is required for expression of all subtype characteristics of specific neuron classes, but incapable of inducing these characteristics in many other cell types when ubiquitously expressed. A similar context-dependent activity of the ceh-14 orthologLhx3 has also recently been described in vertebrates(Thaler et al., 2002).

Although the expression of all neuronally expressed zig genes overlaps uniquely in PVT, individual zig genes are expressed in cells other than PVT. In those cells, zig gene expression does not appear to be temporally regulated in a similar manner as it is in PVT. Moreover,there is hardly any overlap of expression of zig genes,lim-6 and ceh-14 in cells other than PVT. This is again consistent with the previously discussed case of the ttx-3 homeobox gene. All genes that are under control of ttx-3 in the AIY interneuron class, are also expressed in cells other than AIY and are under control of different transcription factors in those other cell types(Altun-Gultekin et al., 2001). This point again illustrates that zig gene regulation by LIM homeobox genes in PVT falls into the general paradigm of context dependent regulation of gene expression.

An interesting feature of lim-6 and ceh-14 activity is the specificity of their impact on mostly those genes that show a temporally regulated profile of expression: lim-6 and ceh-14 affect all postembryonically expressed zig genes, but exert a lesser effect onzig genes that show a certain component of embryonic expression and no effect on any other of the embryonically expressed PVT cell fate markers available. In classical developmental terms, lim-6 andceh-14 thus appear to define the competence of the PVT cell to respond to a temporal cue.

One co-factor that either directly or indirectly contributes tozig gene expression is the heterochronic factor LIN-14. This protein defines the temporal domain of zig gene expression. Recent experiments show that LIN-14 acts as a DNA-binding transcription factor (V. Ambros, personal communication), which prompts the speculation that LIN-14 directly binds to zig gene promoters, possibly in conjunction with LIM homeodomain proteins. Initially, C. elegans heterochronic genes were defined through their effect on developmental stage-specific cell division events (Ambros and Horvitz,1984). The first evidence that they may also act in postmitotic processes in the nervous system was provided by the observation that loss oflin-14 leads to a precocious re-wiring of D-type motor neurons, an event that is normally only observed in late L1-stage animals(Hallam and Jin, 1998). Our observation of a function of lin-14 in temporally controlled induction of zig gene expression is qualitatively different from the D-type motoneuron case. D-type motoneuron rewiring is executed inlin-14 mutants, though at an inappropriate time; hence, LIN-14 acts as a repressor of the rewiring process. By contrast, zig gene expression in lin-14 mutants is not merely observed at an inappropriate time, but largely absent; hence, LIN-14 acts as an activator in PVT. In addition, while the molecular targets of LIN-14 that cause the D-type rewiring defect are as yet elusive, the VNC axonal patterning that we observe in lin-14 mutants is consistent with the notion that the ziggenes are (direct or indirect) effectors of lin-14 in axonal patterning.

If LIN-14 is the temporal trigger for activation for zig gene expression, what is the temporal mechanism that restricts LIN-14 activity to the L1 stage? This question is particularly relevant as LIN-14 antibody staining can already be observed during late embryogenesis(Ruvkun and Giusto, 1989). Two models could be envisioned: LIN-14 protein levels may have to accumulate to a critical threshold level that does not occur until the L1 stage. Observable embryonic expression may still be at subthreshold levels. Alternatively,restriction of lin-14 activity to the L1 stage may be dictated through the presence of a co-factor that is temporally regulated, such as a nuclear hormone receptor; the secretion of its ligand may be coupled to hatching of the embryo. The feeding state of the animal cannot be a determinant as we found zig-4 expression to be normal in starved animals.

We have previously suggested that zig genes may be required specifically during the L1 stage as a stabilizing factor to ensure that already established axonal tracts can cope with mechanical stress and changes in the local molecular environment occurring in the VNC in the L1 stage(Aurelio et al., 2002). But why is it that zig genes are under control of an L1-specific timer and not just simply expressed throughout embryonic, larval and adult stages, like other PVT cell fate markers? We show that inappropriate expression of thezig maintenance factors during embryonic development of the VNC interferes with axonal patterning possibly through interfering with the axonal outgrowth machinery. For example, during embryonic axon outgrowth, the SAX-3/Robo protein is required to prevent PVQL/R axons from crossing the midline inappropriately (Aurelio et al.,2002; Zallen et al.,1998). It is conceivable that precocious expression of the Ig-domain containing ZIG-4 protein may interfere with the activity of SAX-3,an Ig domain-containing transmembrane protein, thus causing the midline crossover defects that we observe. The tight postembryonic temporal control ofzig gene expression in wild-type animals thus may serve to preventzig-4 from inappropriately acting to affect VNC development.

The observation that loss of a developmental timer, i.e. the LIN-14 protein, causes severe disruption of axonal organization in the VNC, presents a striking example for the importance of precise temporal orchestration of gene expression events in the nervous system. Does the concept of temporal control of neural gene expression programs by heterochronic genes, as described by Hallam and Jin for motoneuron rewiring and as described here in this paper for zig genes, apply to species other than C. elegans? The recent identification of temporally regulated microRNAs in vertebrates (Banerjee and Slack,2002; Pasquinelli et al.,2000), some of which are orthologous to C. elegansheterochronic microRNAs (Lagos-Quintana et al., 2002; Pasquinelli et al.,2000) and the sequence similarity between heterochronic genes and circadian clock genes (Jeon et al.,1999) (F. Slack, personal communication), suggests not just a conservation of general concepts of heterochronic regulation of gene expression but also a conservation on the mechanistic levels.

We thank members of the worm community for providing published mutant and reporter strains; the NIH sponsored CGC for providing strains; H. Bülow for the kal-1 construct and help with scoring animals; R. Townley for construction and initial analysis of lim-6 and zig::gfpstrains; Z. Altun for construction of the flp-1::gfp reporter; Q. Chen for expert technical assistance; I. Hope for providingF59B2.13::lacZ; members of the Hobert and Greenwald laboratories for discussion and support; V. Ambros and F. Slack for communicating unpublished results; and P. Sengupta, V. Ambros, F. Slack, I. Greenwald and members of the Hobert laboratory for comments on the manuscript. We gratefully acknowledge funding by the March of Dimes and Whitehall Foundations. O. A. was sponsored by a NIH postdoctoral fellowship and T. B. by a fellowship from the Boehringer Ingelheim fonds. O. H. is a Searle, Klingenstein Rita Allen, Hirschl and Sloan fellow.