We show here that the lin-11 LIM homeobox gene is expressed in nine classes of head, ventral cord, and tail neurons and functions at a late step in the development of a subset of these neurons. In a lin-11 null mutant, alllin-11-expressing neurons are generated. Several of these neurons, however, exhibit neuroanatomical as well as functional defects. In the lateral head ganglion, lin-11 functions in a neural network that regulates thermosensory behavior. It is expressed in the AIZ interneuron that processes high temperature input and is required for the function of AIZ in the thermoregulatory neural network. Another LIM homeobox gene, ttx-3, functions in the antagonistic thermoregulatory interneuron AIY (Hobert et al., 1997). Thus, distinct LIM genes specify the functions of functionally related antagonistic interneurons within a neural network dedicated for thermoregulatory processes. Both ttx-3 andlin-11 expression are maintained throughout adulthood, suggesting that these LIM homeobox genes play a role in the functional maintenance of this neural circuit. We propose that particular LIM homeobox genes specify the distinct features of functionally related neurons that generate patterned behaviors.
Genetic analysis of animals showing behavioral or neuroanatomical defects has identified pathways that control the generation and functional specification of neurons (Desai et al., 1988; Ruvkun, 1997). Such studies have revealed sequential steps to neurogenesis that are mediated by conserved regulatory factors. Homeobox genes represent a class of regulatory factors that control various stages of neurogenesis in Caenorhabditis elegans and other organisms (Chalfie, 1993; Manak and Scott, 1994). For example, the unc-86 POU homeobox gene acts early in neural development to couple cell lineage to neuroblast identity (Finney and Ruvkun, 1990). After determination of neuroblast fate inC. elegans, certain neuroblast migrations are regulated by a homeobox gene (Salser and Kenyon, 1992). The determination of the pattern of presynaptic input or neurotransmitter identity is also under control of homeobox genes (Miller et al., 1992; White et al., 1992; Jin et al., 1994).
Dedicated neural networks that subserve specific behaviors have been identified in C. elegans. The cellular components of the neuronal circuitry evoking the touch and tap responses and the neuronal circuitry that regulates the response to temperature have been characterized in detail, revealing the existence of dedicated sensory neurons, interneurons, and motor neurons (Chalfie et al., 1985; Mori and Ohshima, 1995; Wicks and Rankin, 1995). A genetic analysis of these behaviors has identified several regulatory factors that are required to specify functionally these neural circuits. For example, themec-3 LIM homeobox gene is required for the specification of touch neurons within the touch circuit (Way and Chalfie, 1988), whereas the ttx-3 LIM homeobox gene acts to control the function of an interneuron in the thermoregulatory circuit (Hobert et al., 1997).
LIM homeobox genes represent a particular subclass of homeobox genes that encode proteins characterized by the presence of a DNA-binding homeodomain and two LIM domains, each of which bears two Zn-finger-like motifs (for review, see Dawid et al., 1995). LIM homeobox genes have been identified across phylogeny and can be grouped into defined subclasses (Fig. 1). The analysis of LIM homeobox genes in C. elegans and Drosophila has suggested a critical role for this class of genes in neurogenesis. TheDrosophila apterous and islet genes control axon pathfinding of particular neurons in the CNS (Lundgren et al., 1995;Thor and Thomas, 1997). The C. elegans LIM homeobox genesmec-3 and ttx-3 are each necessary for the differentiation of particular neuronal cell types; the continuous expression of these genes throughout adulthood suggests that they are also required for neural maintenance (Way and Chalfie, 1991; Hobert et al., 1997). Similarly, two other C. elegans LIM homeobox genes identified from the genome sequence display a neural restricted expression pattern that is maintained throughout adulthood (O. Hobert and G. Ruvkun, unpublished observations).
The lin-11 LIM homeobox gene was first identified by the defects in ectodermal cell lineage asymmetry caused bylin-11 loss-of-function mutations (Ferguson and Horvitz, 1985; Freyd et al., 1990). Given the involvement of other LIM homeobox genes in neurogenesis, we sought to determine whether lin-11also plays a role in neural development or neural function. We find that lin-11 is expressed in a specific set of postmitotic neurons. lin-11 is not required for the generation of these neurons but has a role in their functional specification. One of these neurons has a function that is related to the function of a neuron specified by the LIM homeobox gene ttx-3. The interneurons AIZ and AIY act antagonistically in the neural network that subserves thermoregulatory behavior, and the LIM genes lin-11 andttx-3, respectively, are necessary for events late in the development of each neuron. These findings indicate that LIM homeobox gene paralogs may serve homologous regulatory functions in the development of related neurons.
MATERIALS AND METHODS
Strains and genetic procedures. The strains used in this study are wild-type C. elegans Bristol strain (N2), CB1467 [him-5(e1467)], MT633 [lin-11(n389); him-5(e1467)], GR1052 [unc-86(n846)], and CB1372 [daf-7(e1372)]. lin-11(n389) represents a null mutation in which the entire coding region of lin-11 is deleted (Freyd et al., 1990). Double mutant strains were constructed as follows: the daf-7(e1372); lin-11(n389) double mutant strain was constructed by mating lin-11(n389); him-5(e1467) males with daf-7(e1372) hermaphrodites. Nondauer lin-11, daf-7 heterozygous cross-progeny was identified by raising animals at 25°C. F2 animals homozygous for daf-7 were identified at 25°C and recovered at 15°C, and homozygous lin-11adults were scored for the lin-11-associated egg-laying defect.
Expression constructs and generation of transgenic animals.The plin-11-ABCDE-GFP reporter gene was constructed by subcloning a genomic 10 kb fragment into the pPD95.75 vector, creating a translational fusion of a BamHI site within the fifth exon of lin-11 just upstream of the homeobox with the green fluorescent protein (GFP) coding region. Theplin-11-CDE-GFP, plin-11-DE-GFP, andplin-11-E-GFP constructs were injected as purified restriction fragments obtained from the originalplin11-ABCDE-GFP construct using the ApaLI site at nucleotide position −212 of the ATG start codon (forplin-11-CDE-GFP), the BstEII site at position −1466 (for plin-11-DE-GFP), or the BglII site at position −2564 (for plin-11-E-GFP) plus the EagI site downstream of the GFP 3′-untranslated region. The plasmid or restriction fragment DNA was injected at 50 ng/μl into N2 wild type using pRF-4 (harboring a dominant rol-6 mutation) as an injection marker (100 ng/μl). Multiple independent transgenic lines were examined for expression. The plin-11-ABCDE-GFP andplin-11-DE-GFP extrachromosomal arrays were integrated using a Stratalinker 1800 UV light source at 300 J/m2 and were back-crossed several times. Neuroanatomical defects (see Fig. 4) were identified by crossing the integrated plin-11-DE-GFPreporter array from wild-type into lin-11(n389) null mutant animals. Because this allowed the comparison of reporter strains that only differed in lin-11 gene activity and otherwise were completely isogenic, artifacts induced by the reporter array could be excluded.
Cell identifications and scoring of neuroanatomical defects.Identification of neurons that express lin-11-GFP was done by monitoring cell body position and axon morphology in larval stage L1 and adult animals using light microscopy. Additionally, staining with the fluorescent dye DiI, which fills defined subsets of head and tail neurons, was undertaken (as described by Hedgecock et al., 1985) on transgenic GFP reporter gene-expressing animals to compare green GFP fluorescence with red dye fluorescence using different microscope filter sets. The amphid neuron ADL and the phasmid neuron PHA showed overlapping GFP and dye staining; the relative positions of adjacent GFP reporter gene-expressing neurons as well as Nomarski optics were used to identify neighboring cells in the head and tail ganglia. The wild-type position of cells and the axon morphology of neurons have been described by Sulston et al. (1983) and White et al. (1986). Neuroanatomical defects (see Fig. 4) were identified by crossing the integrated plin-11-DE-GFP reporter array from wild-type intolin-11(n389) null mutant animals. VC neuroanatomical defects (see Fig. 6) were identified with a plin-11-B-GFP; rol-6extrachromosomal array; four independent transgenic lines were examined and showed similar defects. Neuroanatomical defects were quantified by picking a random amount of transgenic, reporter gene-bearing animals (numbers given in the text) and were scored according to the categories described in the text.
Behavioral assays. Single worm thermotaxis assays were performed as described previously (Mori and Ohshima, 1995). Briefly, single animals were placed on 8 cm agar plates, which were then inverted, and a vial of frozen acetic acid (2.7 cm in diameter) was placed in the center on top of the agar plate to create a radial temperature gradient from 16 to 25°C. Animals were allowed to migrate in this radial temperature gradient for 1.5 hr, and their position and tracks on the agar plate were scored. For dauer formation assays, the synchronized brood of the respective genotypes was raised at different temperatures and scored for characteristic dauer features (Riddle and Albert, 1997). Because of the variability of dauer formation caused by subtle changes in the environment, comparisons of different genotypes were only made with assays that were done in parallel. Each assay was performed in triplicate.
Expression pattern of a lin-11-GFP reporter gene
The neural specific expression of many LIM homeobox genes inC. elegans (Freyd, 1991; Way and Chalfie, 1991; Hobert et al., 1997) (Hobert and Ruvkun, unpublished observations) prompted us to investigate the temporal and spatial control of expression of alin-11-GFP reporter gene construct in the nervous system ofC. elegans. A genomic DNA fragment bearing 6.5 kb of 5′-upstream sequences as well as the first five exons and four introns of lin-11 was fused to GFP just upstream of the homeobox, transformed into C. elegans, and integrated into chromosomes to yield stable reporter gene expression. lin-11-GFPexpression can be observed from late embryonic stages throughout larval and adult stages. Because of their characteristic position (Sulston et al., 1983), the lin-11-GFP-expressing cells were identified in early larval stages. At the L1 stage, lin-11-GFPexpression is exclusively confined to neurons in the head ganglia and the lumbar ganglion in the tail (Fig. 2). The neurons that express lin-11-GFP in the head ganglion are the sensory neurons ADF and ADL and the interneurons AIZ and RIC (Fig.3). The expression of lin-11in head sensory neurons correlates with the expression of the vertebrate lin-11 homolog Lhx1 in head sensory structures (Barnes et al., 1994). Weak lin-11-GFP expression can be observed in the interneuron AVG that sends a process along the ventral cord. A role for lin-11 in AVG is suggested by the observation of Baumeister et al. (1996) that in lin-11mutant animals unc-86 expression is aberrantly activated in AVG, suggesting that lin-11 normally repressesunc-86 expression in AVG. The identity of another head neuron pair that exhibits lin-11-GFP expression could not be unambiguously determined, but because of its characteristic axonal morphology in the ventral cord (White et al., 1986), we tentatively assigned this pair of neurons as either the AVH or AVJ interneuron. The expression pattern of lin-11-GFP in the head ganglia is summarized in Figure 3.
Motor neurons within the vertebrate spinal cord express a specific set of LIM homeobox genes (Tsuchida et al., 1994). We find thatlin-11-GFP is also expressed in motor neurons of theC. elegans ventral cord (Fig. 2). lin-11-GFP is activated in the six VC ventral cord motor neurons after they are generated during the L1 stage. Although their pattern of connectivity suggests a function in egg laying (White et al., 1986; Li and Chalfie, 1990), it has not been reported what kind of defects are caused by microsurgical removal of the VC ventral cord motor neurons. Adultlin-11 null mutant animals display uncoordinated backward locomotion (Freyd, 1991). It is unclear, however, whether this defect can be related to a function of lin-11 in the VC motor neurons.
The lin-11-GFP-expressing tail neurons were identified as the PHA phasmid sensory neurons and the PVQ ventral cord interneurons, one of which (PVQL) extends a process on the left side of the ventral cord (Fig. 2). PVQ is a pioneer neuron of the ventral cord (Durbin, 1987) and required for correct axon fasciculation of follower neurons such as the HSN motor neuron (Garriga et al., 1993). HSN fasciculation also requires lin-11 gene activity (Garriga et al., 1993). The expression of lin-11 in PVQ but not HSN suggests that PVQ requires lin-11 to provide fasciculation cues for HSN. It is interesting to note that all the neurons that expresslin-11 in the ventral cord are directly connected to otherlin-11-expressing neurons; PHA forms synapses with AVG, PVQ, and AVH, and the VC neurons synapse onto PVQ (White et al., 1986). The functional significance of this observation is unclear, however (see Discussion).
Outside the nervous system, lin-11-GFP is expressed in cells of the developing vulva (Fig. 2), which correlates with the function oflin-11 in these cells (Freyd et al., 1990).lin-11 expression in the vulval precursor cells is dynamic and disappears once the vulva is formed. lin-11-GFPexpression is also observed in the spermatheca (data not shown). Alin-11-LacZ-reporter gene fusion reveals a similar expression pattern (Freyd, 1991).
Temporal control of lin-11 expression
The vertebrate lin-11 homolog Lhx1 has been shown to regulate embryonic inductions in pregastrulation stages (Taira et al., 1994; Shawlot and Behringer, 1995). In this regard, it is remarkable that no embryonic defects can be observed inlin-11 mutants (Ferguson and Horvitz, 1985; Freyd et al., 1990). We corroborated the apparent absence of a role forlin-11 in early embryogenesis by determining at which embryonic stage lin-11-GFP expression is activated. We found no detectable expression of lin-11-GFP in embryonic blastula and gastrula stages (Fig. 2; data not shown). lin-11-GFPexpression is first observed in embryonic stages (∼400 min) long after the beginning of gastrulation (at 100 min) (Sulston et al., 1983). The observed ∼30 min delay between the onset of GFP protein expression and GFP fluorescence observed in C. elegans (G. Seydoux, personal communication) does not seem to be sufficient to account for the delay. To exclude any GFP maturation artifacts of our GFP reporter construct, we performed anti-GFP antibody staining onlin-11-GFP-expressing embryos and confirmed the postgastrulation embryonic expression of the lin-11-GFPreporter (data not shown). Moreover, a lin-11-lacZ fusion, the enzymatic activity of which is not subject to maturation, shows a similar onset of expression (Freyd, 1991). Thus, lin-11expression is activated significantly later in development than is that of its vertebrate homologs. The lack of early embryonic defects in alin-11 null mutant (Ferguson and Horvitz, 1985; Freyd et al., 1990) is consistent with this observation.
lin-11 expression is activated either significantly after the birth of the respective neuron (ADF, ADL, AIY, AVG, AVH/AVJ, or VC) or at approximately the time the respective neuron is born (RIC). In the case of the VC motor neurons, this temporal control is particularly obvious; the VC motor neurons are formed in the L1 larval stage (Sulston et al., 1983) and extend their axonal projections along the ventral cord in L3 larval stages (Li and Chalfie, 1990).lin-11-GFP expression can first be observed in late larval stages once the VC motor neurons have sent out their axonal projections (Fig. 2). These observations demonstrate that activation oflin-11 is a postmitotic event.
The continuous expression of a transcription factor gene throughout the life of an animal suggests a role in maintenance of cellular function. In adult C. elegans, we find that lin-11expression is maintained at similar levels in all classes of head neurons throughout adulthood (Fig. 2). In contrast, expression in the tail neurons fades at late larval stages and cannot be observed in adult animals. These observations suggest a role for lin-11in the maintenance of neural function of head neurons. The perdurability of lin-11-GFP is not attributable to an artifactual stability of the reporter gene construct for several reasons. First, lin-11-GFP expression fades in tail neurons in postlarval stages, whereas it perdures in head neurons. Second,lin-11-GFP expression is rapidly turned off whenunc-86, an upstream transcriptional regulator oflin-11, is conditionally turned off (J. Sze and G. Ruvkun, unpublished observations). Third, a ttx-3-GFP fusion gene is similarly expressed throughout the life of the animal, and in this case the expression is dependent on ttx-3(+) activity, showing that the maintenance of GFP expression is an active process in this neuron (Hobert et al., 1997).
Defined cis-regulatory elements are responsible forlin-11 gene expression
Deletion mutants of the original lin-11-GFP reporter gene define the control elements for lin-11 expression (Fig.4). A 3.5 kb region upstream of the putative ATG start codon (B element) mediates expression oflin-11 in the developing vulva, the spermatheca, and the ventral cord motor neurons (Fig. 4). The cis-regulatory elements that mediate lin-11 expression in neurons of the head and tail ganglion (DE elements) are localized to a region downstream to the putative translational start site (Fig. 4). Transgenes bearing an even shorter cis-regulatory element (E element) express only weakly in head neurons (Fig. 4). Interestingly, although the DE elements mediate initiation of lin-11expression in head neurons, they do not maintain lin-11expression throughout adulthood in all neurons (data not shown), suggesting that initiation and maintenance regulatory elements are separable.
It is remarkable that the regulatory elements responsible for drivinglin-11 expression in different classes of head neurons that are not closely related by lineage (Sulston et al., 1983) are confined to a small genomic 1.1 kb region downstream of the putativelin-11 translational start site. The location of transcriptional regulatory elements within introns has been observed in various neuronally expressed genes, such as nestin in vertebrates (Zimmerman et al., 1994), eyeless in Drosophila(Quiring et al., 1994), or unc-86 (Baumeister et al., 1996) and ttx-3 (Hobert and Ruvkun, unpublished observations) in C. elegans. This observation underscores the compactness of the C. elegans genome, both in terms of gene structure and the structure of gene regulatory elements.
The paired-homeobox gene Pax-6 is involved in neuronal patterning (Stoykova et al., 1996) and influences the combinatorial code of LIM homeobox genes that is thought to determine neuronal identity in the vertebrate brain and spinal cord (Tsuchida et al., 1994; Ericson et al., 1997; Osumi et al., 1997). Because the C. elegans Pax-6 ortholog vab-3 is also involved in head patterning (Chisholm and Horvitz, 1995), we determined whether expression of the Lhx1/5 ortholog lin-11 in either the head ganglion or in the ventral cord is under control ofvab-3. However, the neuronally expressedplin-11-DE-GFP construct shows normal expression in animals with a loss-of-function vab-3(e648) mutation (data not shown).
LIN-11 does not autoregulate its expression
The only two other LIM homeobox genes in C. elegans for which loss-of-function phenotypes have been described are mec-3 (Way and Chalfie, 1988) and ttx-3(Hobert et al., 1997). In both cases, the respective LIM homeobox gene is expressed throughout adulthood and is required for the maintenance of its own expression (Way and Chalfie, 1991; Hobert et al., 1997). We similarly observe continuous lin-11-GFP expression throughout adulthood, but we find that lin-11 gene activity is not required for maintenance of its expression; no changes in the intensity of lin-11-GFP reporter gene expression were observed when comparing wild-type with lin-11 null mutant animals (compare Figs. 2 and 5). This comparison was performed with an integrated transgenic array that was crossed from wild-type to mutant animals, excluding artifacts attributable to different transgenes. Factors other thanlin-11 must be required for maintenance of lin-11expression. In fact, J. Sze and G. Ruvkun have recently shown that continuous unc-86 gene activity is required for maintenance of lin-11 expression in the AIZ interneuron, the only overlap between the unc-86 and lin-11 expression patterns (Sze and Ruvkun, unpublished observations).
LIN-11-expressing neurons are generated but defective inlin-11 null mutants
The lin-11 gene was identified based on defects in vulval cell lineages (Ferguson and Horvitz, 1985). The neuronal expression pattern of lin-11, which we have described above, led us to investigate further what role lin-11 might play in neurogenesis or neural function. We first wanted to determine whetherlin-11 gene activity is required for the generation of thelin-11-expressing neurons or whether lin-11affects neuronal development after the initial generation of neurons, as suggested by the temporal regulation of lin-11expression. We visualized the lin-11-expressing neurons inlin-11(n389) null mutant animals using alin-11-GFP reporter gene construct. lin-11(n389)represents a null mutation in which the entire coding region oflin-11 is deleted (Freyd et al., 1990). Because theplin-11-ABCDE-GFP reporter gene construct (Fig. 4) exhibits partial rescuing activity (data not shown), we used theplin-11-DE-GFP intronic promoter construct that has nolin-11 gene activity and that entirely recapitulates the head neuronal expression pattern of lin-11. We crossed the chromosomally integrated plin-11-DE-GFP reporter gene array into lin-11 null mutant animals. As judged by cell number and characteristic axon morphologies, we observed that alllin-11-expressing neurons are present in lin-11null mutant animals (Fig. 5), arguing that lin-11 plays no major role in the generation of the lin-11-expressing neurons.
We do, however, observe subtle but readily detectable neuroanatomical defects in >60% of lin-11 mutant animals. Representative examples are shown in Figure 5. Because the main axonal trajectories and the approximate cellular position are not severely disturbed, these types of defects are indicative of neural specification defects rather than lineage defects. The most obvious defects are (1) the presence of additional posteriorly directed processes that do not follow any specific path in the ventral or in the dorsal nerve cord and that terminate at random positions and (2) posteriorly displaced cell positions of some neurons, the axonal projections of which take unusual paths (Fig. 5). The identities of the posteriorly displaced neuronal cell bodies are most likely AIZ and RIC. It is possible that the posterior displacement of AIZ and RIC represents a migration defect of these neurons. The axons of the posteriorly displaced AIZ and RIC neurons often follow an aberrant path into the ventral cord (Fig.5 I,V, arrow a) that is distinct from the amphid commissure along which AIZ and RIC normally send their axonal projections into the ventral cord (see amphid commissure in Fig. 3). Examples of posteriorly directed processes are shown in Figure 5(III–V, arrow p). Their origins are often hard to trace, but we observed cases in which they originate either from the posteriorly displaced neurons or, alternatively, from neurons in the more anterior head ganglion. These abnormalities are not observed in wild-type animals expressing the same, integratedplin-11-DE-GFP reporter gene construct. Occasionally, one of the processes in the ventral cord (AVG or AVH/AVJ) terminates prematurely in the lin-11 mutant (Fig.5 II). Aberrant branching of posteriorly directed processes can also be observed (Fig. 5 III,IV). Neuroanatomical defects of the ADF and ADL sensory neurons could not be observed; however, the axonal processes of ADF and ADL are difficult to follow because they fasciculate and run along with most otherlin-11-expressing neurons in the amphid commissure and the nerve ring (Fig. 3).
Because lin-11 is also expressed in some neurons of the ventral cord, we examined their fate in lin-11 null mutant animals. The plin-11-B-GFP reporter gene construct (Fig. 4) was used to monitor VC motor neuron anatomy in lin-11 mutant animals. This reporter gene construct allows visualization of the VC motor neuron axonal projections without the confounding axonal processes that emanate from the lin-11-expressing head or tail neurons. We found that lin-11 gene activity is not required for the VC motor neurons to be generated or to extend their processes along the ventral cord (Fig.6). However, in >50% of the animals examined, the VC motor neurons displayed a defasciculated phenotype (Fig. 6). Normally, the six VC motor axons run parallel to one another and form multiple connections to each other as well as to other neurons in the ventral cord (White et al., 1986). Note that the head neurons that send processes along the ventral cord also show fasciculation defects (Fig. 5). The fasciculation defects we observe in the VC ventral cord motor neurons might be caused by a cell-autonomous requirement for lin-11 in the VCs. Alternatively, these defects could be caused by defects in the lin-11-expressing AVG motor neuron, which is a pioneer neuron in the ventral cord (Durbin, 1987) and the laser ablation of which affects the bundling of neurons in the ventral cord, causing aberrant transfer and crossing of bundles (Durbin, 1987). In particular, several V- and D-type motor neurons display axonal defects if AVG is laser-ablated (Durbin, 1987). We addressed whether lin-11 might have a role in AVG in guiding follower neurons by visualizing in lin-11 null mutant animals several GABAergic V- and D-motor neurons in the ventral cord with an anti-GABA antibody (kindly provided by H. R. Horvitz;McIntire et al., 1993). No obvious axonal defects could be detected (data not shown). The expression of lin-11 in the PHA phasmid neurons also prompted us to examine their neuroanatomy inlin-11 mutant animals. Labeling the PHA sensory neuron with the fluorescent dye DiI, which fills amphid and phasmid sensory neurons (Hedgecock et al., 1985), we observed no obvious neuroanatomical defects of this pair of neurons in lin-11 null mutant animals (data not shown).
lin-11 mutants display thermotactic defects
Some of the head neurons that expresslin-11 have been assigned functions by laser ablation and behavioral assays (Fig. 3). This type of analysis has revealed that the head interneuron AIZ plays a pivotal role in thermotaxis (Mori and Ohshima, 1995). Thermotaxis represents a learning paradigm in which the animal displays the capacity to sense and memorize its cultivation temperature so that it will migrate to its cultivation temperature when placed on a thermal gradient (Hedgecock and Russell, 1975). The thermotactic response relies on the output of two balanced and opposing temperature-processing pathways, one governed by the AIY and the other by the AIZ interneuron (Fig.7 A). Laser ablation of the AIY interneuron causes a cryophilic phenotype because of the unregulated activity of the other unaffected pathway, whereas laser ablation of the AIZ interneuron causes a thermophilic phenotype (Mori and Ohshima, 1995). Because lin-11 is expressed in AIZ and becauselin-11 mutations cause AIZ neuroanatomical defects, we tested whether the absence of lin-11 gene activity affects thermotaxis. Thermotaxis assays on lin-11 null mutant animals revealed a thermophilic phenotype (Fig. 7 B), which is similar to laser ablation of AIZ. The assignment of thermotactic function with AIZ is underscored by the similar thermophilic phenotype of unc-86 mutant animals (Fig. 7 B) (Mori and Ohshima, 1995), which do not generate the AIZ interneuron because of a cell lineage defect (Finney and Ruvkun, 1990). Because AIZ is formed but apparently not functional in lin-11 mutant animals (Fig.5), we conclude that lin-11 is required for the functional specification of AIZ.
Another temperature-controlled behavior is the execution of the dauer developmental program (Golden and Riddle, 1984). Specific sensory inputs that are indicative of harsh environmental conditions, such as low amounts of food, crowding, and high temperature, are processed by the C. elegans nervous system to cause arrest of C. elegans larvae at the dauer stage (Riddle and Albert, 1997). Considering the thermotactic defect of lin-11 (Fig.7 B), we asked whether lin-11 loss of function also affects temperature modulation of dauer arrest. We observed a small but significant suppression of daf-7-induced dauer formation at 25°C (for daf-7(e1372); lin-11(n389), 15 ± 4% suppression; n = 928; fordaf-7(e1372), 0% suppression; n > 100). However, because a variety of unrelated genes also display weak suppression of dauer formation, we cannot conclude thatlin-11 affects dauer formation specifically through the thermosensory neural circuit.
Specification of a neural network by two LIM homeobox genes
C. elegans thermotactic behavior represents one of the few animal behaviors the neuronal components of which have been well defined, revealing the existence of a specific neural regulatory network that subserves this behavior (Mori and Ohshima, 1995). The organization of this network into two parallel, warm- and cold-processing thermoregulatory pathways is remarkably similar to that of thermocontrol in vertebrates (Boulant and Dean, 1986). The molecular basis for the generation and function of thisC. elegans pathway may therefore be related to the mechanism by which vertebrate thermoregulation is achieved.
We have shown previously that the AIY interneuron that is necessary for thermophilic responses requires a LIM homeobox gene, ttx-3, for its functional specification (Hobert et al., 1997). In the case ofttx-3, it is expressed only in AIY, and the lack ofttx-3 gene activity affects AIY structure and the thermoregulatory function assigned to AIY by laser ablation. Our assignment of lin-11 function in thermoregulation to the AIZ interneuron involves a similar argument. (1) A lin-11-GFPfusion gene is expressed in AIZ, (2) the neuroanatomy of AIZ is affected in a lin-11 null mutant, and (3) the thermoregulatory function assigned to AIZ by laser ablation is affected similarly in a lin-11 null mutant. Although we cannot exclude the possibility that the thermotactic defect oflin-11 mutants is because of defects in otherlin-11-expressing neurons, we consider this unlikely for the following reasons. First, none of the otherlin-11-expressing neurons has been implicated in thermotaxis (Mori and Ohshima, 1995); second, unc-86 causes similar thermotactic defects (Fig. 7 B), and the only overlap betweenunc-86 and lin-11 expression in postmitotic head neurons is in AIZ. Thus we suggest that the lin-11 LIM homeobox gene acts in the AIZ interneuron that is functionally antagonistic to the AIY ttx-3-expressing interneuron (Fig.7 A).
The entry of C. elegans into the developmentally arrested dauer stage is also under thermosensory control (Golden and Riddle, 1984; Hobert et al., 1997). Our finding that lin-11 affects dauer formation could be explained with a defect in thermoregulation of dauer formation. This hypothesis is consistent with the expression oflin-11-GFP in the thermoregulatory AIZ interneuron. However, the effect of lin-11 on dauer formation is considerably weaker than the effect of ttx-3 on dauer formation (Hobert et al., 1997) and could be an effect of lin-11 unrelated to thermocontrol of dauer formation [e.g., lin-11 could affect sensory function of the ADF neuron, which controls the dauer program (Bargmann and Horvitz, 1991a)]. Thus, although ttx-3 andlin-11 are clearly both required for thermotactic behavior, it remains an open issue as to how much the neural and genetic components that couple thermosensory information to a motor output are the same as the neural components that couple to dauer development. Perhaps the major output to neuroendocrine control is via AIY and not AIZ.
The major thermoregulatory organ of vertebrates, the hypothalamus, contains distinguishable warm- and cold-sensing temperature-processing units (Boulant and Dean, 1986) that may be homologous to the antagonistic high and low temperature-sensing pathways of the C. elegans thermotactic response pathway (Fig. 7 A).lin-11 and ttx-3 in C. elegans, and their homologs in mammals, may thus mediate the development of two components of this phyletically conserved thermal-processing network. In support of this hypothesis, the vertebrate ttx-3 homologLhx2 and the lin-11 homologs Lhx1 andLhx5 are expressed in the diencephalon, which gives rise to the thermoregulatory hypothalamus (Fuji et al., 1994; Porter et al., 1997; Sheng et al., 1997).
The AIY and AIZ neurons are formed in animals bearing null mutations inttx-3 and lin-11, respectively, but these neurons show subtle anatomical defects suggesting that they fail to execute late events in neurogenesis. For example, TTX-3 and LIN-11 may regulate genes that are directly involved in axonal pathfinding, so that the patterns of neuroanatomical defects observed in both ttx-3and lin-11 mutant animals can be explained on the basis of direct disturbances in axonal pathfinding. Alternatively, these genes may regulate the expression of genes that mediate synaptic connectivity or synaptic activity, and these neural defects may be secondary consequences of the absence of synaptic signaling. Synaptic connectivity-dependent sprouting of additional processes has been observed in other systems (e.g., Henderson et al., 1983).
The expression and function of two genes from the same gene family in two parallel neural pathways provokes the question as to whetherttx-3 and lin-11 are differentially instructive to determine specific AIY and AIZ features, respectively. Alternatively, they could act permissively to regulate the same set of downstream target genes in order to determine features that are common to AIY and AIZ [e.g., their connection to the thermotactic RIA interneuron, which is a postsynaptic target of both AIY and AIZ and required for thermotactic behavior (Mori and Ohshima, 1995)]. InDrosophila, the paired-class homeobox genes prd,gsb, and gsbn are differentially expressed but can functionally substitute for one another (Li and Noll, 1994). Their divergent roles may have simply evolved by changes in deployment. The compactness and modularity of the cis-regulatory elements for lin-11 (Fig. 4) and ttx-3 (Hobert and Ruvkun, unpublished observations) would facilitate the evolution of distinct developmental function by acquisition of distinct regulatory elements. In summary, our data suggest that distinct LIM homeobox genes may determine the features of functionally related neurons of neural circuits that generate specific behaviors.
LIN-11 function in the ventral cord
We find that lin-11-expressing motor neurons in the ventral cord are formed in lin-11 null mutant animals but display neuroanatomical defects. There are six VC motor neurons in the ventral cord, the axonal projections of which fasciculate and run in parallel in the ventral cord to form chemical and electrical synapses with each other and with other neurons and target muscles. We find thatlin-11 gene activity is not required for the generation or the axonal outgrowth of these neurons but required for them to fasciculate correctly. It is possible that LIN-11 regulates downstream target genes that are required for homophilic interaction of the VC motor neurons; alternatively, these downstream target genes might be required for correct fasciculation with other neurons in the ventral cord. It is also possible that the VC motor neuron fasciculation defects are caused by absence of lin-11 function in the AVG neuron, which is a pioneer in the ventral cord (Durbin, 1987); however, because we do not observe defects in other ventral cord motor neurons that require AVG to fasciculate, such as the V- and D-type motor neurons (Durbin, 1987), we consider this possibility less likely.
The HSN motor neuron requires the PVQ ventral cord pioneer to fasciculate correctly (Garriga et al., 1993). Moreover, Garriga et al. (1993) have shown that lin-11 gene activity is required for proper HSN fasciculation. Our identification of PVQ, but not HSN, as one of the lin-11-expressing neurons is consistent with a hypothesis in which lin-11 acts in PVQ to affect HSN morphology. The function of lin-11 in the VC motor neurons and PVQ neurons might thus be similar in that LIN-11 could to be required to provide cues for selective fasciculation of neurons in the ventral cord. For example, LIN-11 could regulate the expression of adhesive cell surface receptors that mediate fasciculation between VC neurons and between PVQ pioneer and HSN follower axons. It is also intriguing to note that all the lin-11-expressing neurons in the ventral cord have at least one synaptic partner that also expresseslin-11. VC synapses with PVQ and HSN; PHA synapses with AVG, AVH, and PVQ; and PVQ synapses with VC and HSN. Thus, another hypothesis posits that lin-11 is required to determine synaptic connectivity and that the fasciculation defects of the VC and HSN neurons are a secondary consequence of missing synaptic inputs. Although we do not observe fasciculation defects of thelin-11-expressing head neurons in lin-11 mutants, the process outgrowth and migration defects of the head neurons could also be explained by misregulation of cell surface molecules required for correct axonal guidance or synaptic activity.
The vertebrate lin-11 ortholog Lhx1 is also expressed in motor neurons of the spinal cord. Lhx1 null mutant mice die because of early embryonic defects (Shawlot and Behringer, 1995); thus the importance of Lhx1 in motor neurons could not be addressed to date. Regarding the function oflin-11, it is conceivable that Lhx1 might act in a similar manner in determining axonal morphology of spinal cord motor neurons.
A conserved neural function for LIM homeobox genes
Loss-of-function studies in C. elegans andDrosophila have shown that LIM homeobox genes display a postmitotic function late in neural development. C. elegans mec-3 acts in touch cell receptor specification (Way and Chalfie, 1988, 1991); ttx-3 (Hobert et al., 1997) andlin-11 as shown here act in thermotactic interneuron specification. All three genes continue to be expressed at high levels throughout adulthood. Two recently identified C. elegans LIM homeobox genes are also expressed in a very restricted subset of postmitotic neurons, and their expression is also maintained throughout adulthood (Hobert and Ruvkun, unpublished observations). Axonal pathfinding defects have been described in Drosophilamutants of the LIM homeobox genes apterous andislet (Lundgren et al., 1995; Thor and Thomas, 1997);Drosophila islet also controls neurotransmitter identity (Thor and Thomas, 1997). Also, vertebrate LIM homeobox genes from all LIM homeodomain subclasses (Fig. 1) have been shown to be expressed continuously throughout adulthood, mostly in the brain (Thor et al., 1991; Taira et al., 1992; Xu et al., 1993; Barnes et al., 1994; Fuji et al., 1994; Gong et al., 1995; Sheng et al., 1997). Taken together, these observations suggest a common theme for the function of LIM homeobox genes late in neural differentiation, potentially in the establishment of features required for synaptic signaling as well as for the maintenance of the differentiated state of neurons. Additional and relatively specialized functions of LIM homeobox genes, such aslin-11 function in vulva formation (Ferguson and Horvitz, 1985; Freyd et al., 1990), apterous function in wing development (Cohen et al., 1992), Lhx3 function in pituitary development (Sheng et al., 1996), or Lim1 function in organizer activity (Taira et al., 1994; Shawlot and Behringer, 1995), might have been co-opted by specific phyla at different stages of evolution.
This work is supported in part by Hoechst AG to G.R. O.H. is supported by a postdoctoral fellowship from the Human Frontiers Science Program. We thank Cori Bargmann, Gwen Acton, members of the Ruvkun laboratory, and anonymous reviewers for helpful suggestions on this manuscript, G. Seydoux for comments and communicating unpublished observations, the Caenorhabditis Genetics Center (funded by the National Institutes of Health Center for Research Resources) for supplying strains, A. Fire for the GFP vector pPD95.75, Bob Horvitz for anti-GABA-antibody, and A. Coulson and the C. eleganssequencing centers for their updates on the status of the genome sequencing project. We are particularly grateful to Cori Bargmann for the initial identification of lin-11-lacZ-expressing neurons and to Hitoshi Sawa and Bob Horvitz for providing the plin-11-B-GFP construct shown in Figure 4.
Correspondence should be addressed to Dr. Gary Ruvkun, Massachusetts General Hospital, Department of Molecular Biology, Wellman 8, Boston, MA 02114.