The C. elegans tailless/Tlx homolog nhr-67 regulates a stage-specific program of linker cell migration in male gonadogenesis

The importance of spatial cues in the development of multicellular organisms is well established, and much is known about the signaling pathways that provide spatial information within developing tissues (Towers and Tickle, 2009). Less understood, but probably equally important, are temporal cues, which organize cells into organs and coordinate the development of each organ with the developmental program of the whole organism. At least some organs use transcriptional regulation as a mechanism to respond to timing cues. For example, hormones, such as ecdysone in insects or those produced by the pituitary gland in mammals, serve as global regulators of timing by activating organ-specific transcription factors that respond to these cues (Randall et al., 1997). In another example, neuroblast differentiation in mouse and Drosophila depends on a series of transcription factors that are expressed in a defined order within a cell lineage (Pearson and Doe, 2004). The importance of proper timing is underscored by the fact that the same transcription factors required during development can lead to disease when expressed at an inappropriate stage (Robson et al., 2006).

Cell migration, which is integral to many types of organogenesis, clearly depends on spatial cues (Cardoso and Lu, 2006; Ghysen and Dambly-Chaudiere, 2007; Montell, 2003; Sauka-Spengler and Bronner-Fraser, 2006). Because cell migrations occur at specific times during organogenesis and often involve stage-specific changes in migratory behavior, they are likely to also be temporally regulated; however, there are few systems tractable for studying temporal regulation in migration. C. elegans is particularly conducive for studying developmental timing because it develops with both invariant timing of cell divisions and invariant cell positions, making aberrations easy to identify (Kimble and Hirsh, 1979; Sulston et al., 1980; Sulston and Horvitz, 1977). This feature has allowed the identification of heterochronic mutants, in which the developmental timing of some tissues is altered relative to the rest of the organism (Ambros and Horvitz, 1984). Some of these genes have since been shown to control vertebrate development, but because the effect is more subtle in vertebrates, the studies in C. elegans have been pivotal in identifying the role of heterochronic genes (Moss, 2007). The precise timing of C. elegans development has also been used to identify the logic of transcriptional regulation of temporal information during the development of organs such as the pharynx (Gaudet and Mango, 2002; Gaudet et al., 2004).

We are studying the stage-specific regulation of migration in the linker cell (LC), an individual leader cell whose migration defines the shape of the developing male gonad and ultimately connects the gonad to the cloaca, enabling sperm release (Kimble and Hirsh, 1979; Klass et al., 1976). The LC clearly requires regulation at different stages as it navigates a complex trajectory during three larval stages (L2-L4) of development. The LC migration route consists of several linear segments and two turns (see Fig. 1). The first turn occurs in L2 larvae, when the LC travels from the ventral to the dorsal bodywall, while changing direction from anterior to posterior. The second turn occurs in mid-L3 larvae, when the LC travels from the dorsal bodywall back down to the ventral bodywall as it migrates posteriorly. The LC finishes migrating in mid-L4 larvae and is engulfed by the U.lp/U.rp cell near the cloaca, undergoing cell death. Some data suggest that timing cues, rather than physical landmarks along the migration route, induce specific LC behaviors at different points of migration. For example, the second turn executed by the LC is regulated by daf-12, a nuclear hormone receptor that regulates developmental timing and dauer diapause in response to a hormonal ligand, dafachronic acid (Antebi et al., 1998; Motola et al., 2006; Su et al., 2000). In addition, LC death occurs in the L4 molt, regardless of engulfment by the U.lp/U.rp cell or its position in the worm (Abraham et al., 2007).

Our characterization of the migrating LC has revealed that it changes gene expression and cell shape, in addition to the position along the migratory route. We show that changes in LC gene expression and cell shape in L4 larvae are primarily controlled temporally. We have identified the tailless/Tlx homolog nhr-67 (Gissendanner et al., 2004), a nuclear hormone receptor, as a stage-specific regulator of LC migration during the L3 and L4 stages, including the negative regulation of the unc-5 netrin receptor. In particular, nhr-67 is required for executing LC developmental changes at their proper time during the L3 and L4 stages. We propose that nhr-67 is a stage-specific regulator that modifies a basal LC migratory program to execute L3 and L4 stage changes with normal timing.

C. elegans strains were cultured at 20°C using standard protocols (Brenner, 1974) unless indicated otherwise. All strains used carry the him-5(e1490) mutation, except nuIs9 [unc-5::GFP], which carries the him-8(e1489) mutation. The alleles and transgenes used in this study are: zmp-1(cg115)III; unc-119(ed4)III (Maduro and Pilgrim, 1995); him-8(e1489)IV (Hodgkin et al., 1979); him-5(e1490)V (Hodgkin et al., 1979); him-4(e1267)X (Hodgkin et al., 1979); syIs128 [lag-2::YFP]II; syIs49 [zmp-1::GFP]IV (Wang and Sternberg, 2000); nuIs9 [unc-5b::GFP] (Su et al., 2000); muIs27 [mig-2::GFP] (Zipkin et al., 1997); qEx454 [gon-1::GFP] (Blelloch and Kimble, 1999); syEx925 [nhr-67::GFP] (Fernandes and Sternberg, 2007); him-4::YFP; nhr-67::unc-5; rde-1(ne215); znex338[lag-2::rde-1, lag-2::mRFP] (Lucanic and Cheng, 2008).

An RNAi screen of 508 known and putative transcription factors was conducted on him-5 animals. Adult males were scored for incomplete gonad migration under the dissecting microscope. A complete list of transcription factors tested can be found in Fernandes and Sternberg (Fernandes and Sternberg, 2007). A previously described RNAi protocol was used (Kamath et al., 2001) with a few modifications. Eggs were harvested from gravid adults by bleaching and incubated on plates containing RNAi bacteria at 22°C. The RNAi bacteria were obtained from the Ahringer Library (Geneservice).

nhr-67 expression levels were modulated during gonad migration by first growing animals on nhr-67 RNAi bacterial plates from egg to the late L1/early L2 stage over ~41 hours. The animals were then washed with M9 solution and placed on plates with OP50 bacteria and scored at the late L4 stage. This experiment was repeated by switching plates in the mid-L1 and mid-L2 stages.

The early L3 stage was identified under Nomarski optics by the ‘10-cell stage’ configuration of the B lineage (Chamberlin and Sternberg, 1993) and by the yet undivided P10.p and P11.p cells. Late L3 stage animals were identified by there being at least ten P10.p and P11.p progeny cells. The L4 stage was divided into four time categories based on hook and tail retractions. A small hook retraction just anterior to the cloaca was defined as the early L4 stage, and a large retraction characterized by jagged, receding edges, was defined as the early/mid-L4 stage. The beginning of the tail retraction was defined as the mid/late L4 stage, and a large tail retraction to the base of the tail taper was defined as late L4 stage.

lag-2::YFP plasmid was generated by PCR amplification of a 3.3 kb sequence of the lag-2 5′ region using primers GAAACTGCAGTGCCACTCATTATTTTGGACG and GAAGGATCCCTAGCAAAGCTCAAGGTCGAC, and then cloning into the PstI and BamHI sites of vector pPD136.64 (a gift from Andy Fire), modified from Siegfried and Kimble (Siegfried and Kimble, 2002). him-4::YFP was generated by a similar strategy using primers TTTAACTGCAGCTGTATGTCCGCTAGGGTCGATT and TTCCGGATCCGATATTGAAGGATTGATGTGCTG to amplify 4 kb of 5′ region, and restriction sites PstI and BamHI. nhr-67::unc-5 plasmid was generated by first PCR amplifying a 5.3 kb sequence of 5′ nhr-67 using primers CCCAAGCTTCAATAACTTCCGTTTTGCCAGATC and CCCAAGCTTTCTTGGCGCCAGATTC ATTAATTC, and the UNC-5a coding region from cDNA using primers CGGGGTACCATGGACGAAATCACAATCACAACA and CGGGGTACCTTATGGGGACACAATTTGTGGAAA; then, the two fragments were sequentially inserted into plasmid pPD49.26 (a gift from Andy Fire) using HindIII and KpnI restriction sites, respectively.

The plasmids were injected into unc-119(ed4); him-5(e1490) hermaphrodites at 60 ng/μl for lag-2::YFP and him-4::YFP and at 15 ng/μl for nhr-67::unc-5, along with plasmids unc-119(+) (Maduro and Pilgrim, 1995) and pBSK+ (Stratagene).

Fluorescence images of unc-5::GFP animals were captured on a Hamamatsu ORCA ER camera and the mean pixel intensity of the LC was calculated using Openlab software (Improvision). DIC and fluorescence images were overlaid using Adobe Photoshop 7.0.

The known nhr-67 binding site, AAGTCA, was searched for in the regulatory regions of unc-5 and zmp-1. In the 4.6 kb promoter region of unc-5 used for the reporter strain (Su et al., 2000), four potential binding sites (underlined) were found: AAGCACGGATAAAGTCACTTTTTTCTGTGATTA; AAAAAATTTTAAAGTCATTTTTGAGCTCTGTAG; TTGAAAAGAAAAAGTCACAACTAATTGTAAGTTAT; and ACAAATTGAAAAAGTCAAAATACTTTTCAAATTTAT. In the 2.8 kb promoter region of zmp-1, four potential binding sites were also found: AAATCAGATTTTTAAAAAAGTCAATTTTGTAATTACTCATGT; ACGTAATAATTTTTGGAAAGTCATGTTTTGACGTGTTTCAAA; TTAAAAATGCTGTAAAAAGTCATTATGCATAGAATATGAA; and TTCTTCAATTGTTTTAAAAGTCATTACTCATCTTCTTTTTAT.

Because a particular hexamer can occur randomly in the sequence at a high rate (1 in 4000 bases), the non-gap alignment program MUSSA (Kuntz et al., 2008) was used to computationally determine whether these hexamers are in conserved regulatory regions. The C. elegans unc-5 promoter was aligned with sequences from C. briggsae and C. brenneri (5′ region plus introns). A similar analysis was performed of the 5′ region of zmp-1 in the three species. We used 15/15 and 18/20 conserved bases as thresholds for the MUSSA alignment. In addition, we searched by eye for any conserved sequences adjacent to each of the hexamers in the three species.

To find genes that stage-specifically affect LC migration, we performed an RNAi screen of 508 known and putative C. elegans transcription factors. Because gonad shape is a trace of the LC migratory path, we searched for genes, the reduction-of-function of which caused abnormal adult male gonad morphology. We identified a number of transcription factor genes, including lin-29, nhr-25 and egl-5, already known to have roles in gonadogenesis (Abraham et al., 2007; Asahina et al., 2000; Chisholm, 1991; Euling et al., 1999; Ferreira et al., 1999), as well as others with yet unknown gonadal functions, including the tailless homolog nhr-67. tailless has been shown to be required for patterning the anterior and posterior poles of the Drosophila embryo (Pignoni et al., 1990), for development of the brain and eye in Drosophila and vertebrates (Hollemann et al., 1998; Kitambi and Hauptmann, 2007; Monaghan et al., 1995; Roy et al., 2004; Younossi-Hartenstein et al., 1997; Yu et al., 1994), and for proper vulval morphogenesis and neuronal development in C. elegans (Fernandes and Sternberg, 2007; Sarin et al., 2009). We were interested in nhr-67 because the male gonad shape in nhr-67(RNAi) animals suggested a defect specific to the later stages of gonadal development. We thus investigated whether nhr-67 is a regulator of stage-specific LC development.

We compared the gonadal morphology of wild-type and nhr-67-deficient animals at different stages (Fig. 2A-D). Because nhr-67 deletion mutants have an embryonic lethal phenotype, we used RNAi feeding on newly hatched worms to study the function of nhr-67. nhr-67(RNAi) males have a fully penetrant gonadal defect, with none of the LCs completing the migration by the L4 molt (n=87). At the early L3 stage, no gonadal defects in nhr-67(RNAi) males were observed: 85% of LCs in both nhr-67(RNAi) (n=31) and wild-type (n=22) males executed the turn from the ventral to the dorsal side and migrated posteriorly approximately one-third the length of the ventral gonadal arm (Fig. 2Aa,Ab,D). By the L3-to-L4 molt, however, several gonadal defects had become apparent in nhr-67(RNAi) males. First, the LC did not migrate as far in nhr-67(RNAi) males as in the wild type, with only 17% (n=23) of LCs in nhr-67(RNAi) males reaching the P7.p hypodermal cell, compared with 93% (n=14) in wild-type males (Fig. 2Ba,Bb,D). Second, by the L3-to-L4 molt, 100% (n=14) of LCs in wild-type males turned back down from the dorsal to the ventral bodywall, whereas only 4% (n=24) of LCs in nhr-67(RNAi) males turned ventrally. Most LCs of nhr-67(RNAi) males remained on the dorsal bodywall despite having migrated past the position at which they normally turn. Later during the L4 stage, however, most LCs in nhr-67(RNAi) males turned ventrally (95%, n=87; Fig. 2Ca,Cb,D). Although LCs in nhr-67(RNAi) males continued to migrate during the L4 stage, they were even farther behind their normal position than at the L3-to-L4 molt. As a result, by the late L4 stage, when 100% (n=28) of the LCs in wild-type males had finished migrating and been engulfed by the U.lp/U.rp cell, none (n=87) of the LCs in nhr-67(RNAi) males had completed their migration. LCs in nhr-67(RNAi) males might therefore have a defect in executing the correct developmental programs at the appropriate stage.

We found that nhr-67 functions cell-autonomously in the LC to regulate gonadal migration. We used rde-1 mutants, which are refractory to RNAi effects, rescued for rde-1 function cell-specifically in the LC by a lag-2::rde-1 transgene (Lucanic and Cheng, 2008). When these animals were treated with nhr-67(RNAi), 18 of 21 L4 stage males had the same gonadal defects as males receiving systemic nhr-67(RNAi). Most rde-1 mutant males without the rescuing construct did not have gonad migration defects (29/34 were as wild type).

When we examined the expression of a transcriptional nhr-67::GFP reporter in the male gonad, we found that it was expressed only in the LC. Furthermore, the duration of nhr-67::GFP expression in the LC, from the early L3 stage until LC death (Fig. 2E,F), is consistent with the timing of the gonadal defects observed during the L3 and L4 stages. nhr-67 is also expressed in the intestine, Pn.p cells, and in a few head and tail cells.

We also used nhr-67::GFP-expressing animals to test the efficacy of our nhr-67(RNAi) feeding method. Feeding GFP(RNAi) to nhr-67::GFP-expressing animals abolished GFP expression in all tissues except for a few refractory head neurons (n=43/44 no LC GFP).

Most genes utilized for proper gonad migration by the male LC are also used by the hermaphrodite for gonad migration (Lehmann, 2001). The best-studied hermaphrodite gonad has two leader cells: the distal tip cells (DTCs) that start migrating from the midbody in opposite directions to create a two-armed gonad. Although nhr-67 is important for male gonad migration, hermaphrodite DTC migration is not regulated by nhr-67. Twenty-nine of thirty hermaphrodites treated with nhr-67(RNAi) had normal DTC migrations, compared with 36/37 wild-type hermaphrodites. Consistently, we did not observe the expression of nhr-67::GFP in the hermaphrodite DTCs. nhr-67 is thus a rare instance of a gene used male-specifically in gonadal leader cell migration.

We examined whether the failure of the LC in nhr-67(RNAi) animals to execute the second turn during the L3 stage involves the regulation of netrin receptors by nhr-67. Netrin receptors are known to be necessary for specific C. elegans gonadal turns (Su et al., 2000). Two netrin receptors, UNC-5 and UNC-40, are required by the LC for the first turn in the L2 stage, and for a similar, dorsal turn by the hermaphrodite DTCs at the L3 stage. In both migratory cell types, the UNC-40 netrin receptor alone is expressed by the gonadal leader cells in the early stages of migration, causing them to be attracted to netrin molecules on the ventral bodywall and to migrate along it. At the time of their ventral-to-dorsal turn, the leader cells begin to coexpress the UNC-5 netrin receptor, which then changes their response to netrin from attraction to repulsion and guides their migration away from the ventral bodywall.

We examined unc-5::GFP expression in wild-type and nhr-67(RNAi) males at the L3 and L4 stages. Although UNC-5 expression is visible in the LC by antibody staining at the start of the ventral-to-dorsal turn (Su et al., 2000), the unc-5::GFP reporter only becomes visible during the turn in the L2-to-L3 molt. Levels of LC UNC-5 expression remained constant during the L3 and L4 stages (Fig. 3A,B). By contrast, in nhr-67(RNAi) animals, LC expression of unc-5::GFP became progressively brighter during the L3 and L4 stages, indicating elevated levels of unc-5 expression (Fig. 3A,B). NHR-67 is therefore required to negatively regulate unc-5 expression during the L3 and L4 stages.

We further investigated whether the failure of the LC to turn ventrally at the mid-L3 stage in nhr-67(RNAi) animals was due to overexpression of UNC-5 past the mid-L3 stage. We ectopically expressed unc-5 in the LC during both the L3 and L4 stages in wild-type animals to see whether this prevented the LC from turning ventrally in the mid-L3 stage. We used an nhr-67::UNC-5 construct because the nhr-67 promoter drives gene expression in the LC during the L3 and L4 stages. Fifty-seven percent (n=30) of nhr-67::unc-5-expressing animals showed continued dorsal LC migration throughout the L3 stage, and often into the L4 stage (Fig. 3C). In these animals, unlike in nhr-67(RNAi) animals, the LC reached the posterior region of the body before the L4-to-adult molt, indicating that the defects that we observed were due to unc-5 overexpression and not to an nhr-67 deficiency resulting from the sequestering of factors that bind the nhr-67 promoter. Ectopic expression of unc-5 is therefore sufficient to prevent the LC from turning ventrally in the mid-L3 stage. We conclude that the negative regulation of unc-5 by NHR-67 is required in the LC for it to execute the mid-L3 stage turn from the dorsal to ventral side.

We then investigated the effect of nhr-67 on the LC in L4 larvae. Because little is known about LC development during its migration, we began by identifying changes in LC gene expression, shape and migration during the L4 stage. One change we found is that zmp-1, which encodes a zinc matrix metalloprotease, is expressed by the LC from the early L4 stage until LC death in the L4-to-adult molt (n=40/41; Fig. 3D). When we examined zmp-1::GFP expression in nhr-67(RNAi) animals, we found that it is never expressed in the LC (n=0/35; Fig. 3E). zmp-1 mutants, however, had a wild-type gonad migration (22/22 animals with complete LC migration by mid/late L4).

We could not find any evidence of nhr-67 directly regulating either unc-5 or zmp-1 based on sequence conservation. Although the promoter regions of unc-5 and zmp-1 used in the reporter strains contain the hexameric nhr-67 binding sequence (DeMeo et al., 2008), the sequences adjacent to the hexamers are not conserved in C. briggsae and C. brenneri (see Materials and methods).

nhr-67 must regulate other genes besides unc-5 and zmp-1 because animals that are both mutant for zmp-1 and overexpress unc-5 resemble the unc-5-overexpression phenotype, rather than the nhr-67 phenotype. We examined 85 animals with both zmp-1 and unc-5 perturbations at various times during the L4 stage and found that most LCs reached the posterior region of the animal by the mid/late L4 stage (n=60/61) and became polarized (n=31/36).

Having found that NHR-67 negatively regulates unc-5 and positively regulates zmp-1, we examined whether NHR-67 specifically regulates genes that become expressed during the L3 and L4 stages, or whether it regulates gene expression in the LC more broadly. We tested the effect of nhr-67(RNAi) on transcriptional and translational fluorescent reporters of a selection of genes that are expressed throughout LC migration (Blelloch and Kimble, 1999; Henderson et al., 1994; Lundquist et al., 2001; Vogel and Hedgecock, 2001; Zipkin et al., 1997). None of the genes that we tested, which were lag-2::YFP (n=30/30), gon-1::GFP (n=16/16), mig-2::GFP (n=29/29) and him-4::YFP (n=32/32), was affected by nhr-67(RNAi) (Fig. 3F). These observations suggest that nhr-67 is not generally required to maintain gene expression in the LC, but instead regulates expression specifically during the mid-L3 through L4 stages.

We also identified morphological changes in the LCs of L4 larvae. Using a marker for LC cytoplasm, lag-2::YFP, we observed that the LC changes from a round to a polarized shape during the L3 and L4 stages (Fig. 4Aa-Af). We quantified these changes (Fig. 4C) in the early and late L3 stage and at four time points during the 10 hours of the L4 stage, classifying LC shape by the degree of polarization as ‘round’, ‘pointed’, ‘elongated’ or ‘extended’. LCs with ‘completed’ migrations were typically round and engulfed. Wild-type L3 stage animals had predominantly round LCs that changed from being oblong in the vertical to horizontal direction (Fig. 4Aa,Ab). In L4 larvae, the rounded leading edges of the LCs became pointed, and LCs became more elongated (Fig. 4Ac-e). However, LCs rarely became more polarized than the elongated stage before they finished migrating (Fig. 4Af).

We examined whether nhr-67 regulates LC morphology in L3 and L4 larvae. In nhr-67(RNAi) animals, the LC maintained a round shape, typical of the early L3 stage, throughout all of the L3 stage and most of the L4 stage (Fig. 4B,C). Over 90% of LCs in the first half of the L4 stage still had a round shape. However, these LCs converted to an extended shape, typical of late L4 stage larvae, at the normal time. Thus, there might be other factors besides NHR-67 that regulate the late L4 stage LC program. Our results show that nhr-67 is required for LC shape polarization through the L3 and most of the L4 stages. nhr-67 induces LC shape change by regulating other genes besides unc-5 and zmp-1, as LCs in animals both overexpressing unc-5 and mutant for zmp-1 were able to undergo shape change (n=31/36).

We have demonstrated that LCs of L4 stage nhr-67(RNAi) larvae resemble those of normal L3 larvae in their morphology, gene expression and position along the migratory path. NHR-67 might be required by the LC to correctly interpret temporal cues and synchronize its developmental program with the rest of the organism. Alternatively, because the LC has not reached the posterior body in nhr-67(RNAi) males by the L4 stage, the LC might be failing to receive spatially restricted cues from its environment. In this latter case, the changes in morphology and gene expression at the L4 stage would not be directly regulated by nhr-67, but would be secondary effects of LC position. To distinguish between these possibilities, we used a migration-defective mutant, him-4 [which encodes the extracellular matrix (ECM) protein hemicentin], to test whether an incorrectly positioned L4 stage LC can still undergo changes in morphology and gene expression. In the absence of HIM-4, the LC meanders from its normal path; by the L4 stage, it is in either the head or mid-body region instead of in its normal posterior location (Vogel and Hedgecock, 2001). We found that LCs in him-4(e1267) mutants undergo the L4 stage-specific morphology change with correct timing, despite not having reached the posterior body: 100% (n=20) of the LCs in L3 larvae had a round shape, but by the mid-L4 stage 100% (n=14) of the LCs had a polarized shape (Fig. 5A-D). This is the same timing as observed for wild-type males, with 100% (n=35) having a round LC shape in the L3 stage and 100% (n=73) having a polarized shape in the mid-L4 stage. We defined a polarized LC shape to include all non-round shapes. We also found that the expression of ZMP-1 in L4 larvae is not regulated by positional cues. In him-4; zmp-1::GFP animals, zmp-1 was correctly expressed in the LC at the L4 stage, despite the anterior positioning of the LC [95% (n=38) for him-4 versus 98% (n=41) for wild type; Fig. 5E,F]. We conclude that nhr-67 primarily regulates the timing of both L3 and L4 stage events by responding to either a cell-intrinsic clock or a global timing signal, rather than to spatial cues that originate from a posterior source in the animal.

However, positional cues might also contribute to nhr-67(RNAi) LC migration defects. Because it has been shown that MIG-2 localization in motile cells depends on ECM components and is required for cell shape change (Ou and Vale, 2009; Ziel et al., 2009), we examined MIG-2::GFP localization in the LC in wild-type and nhr-67(RNAi) males (Fig. 5G-J). In the LCs of L3 stage wild-type males, MIG-2::GFP was evenly distributed throughout the membrane, but during the L4 stage MIG-2 became polarized to the adherent, ventral side. In L4 stage nhr-67(RNAi) males, however, only five of 26 LCs showed MIG-2::GFP membrane polarization to the adherent side. It might be that MIG-2 does not localize properly in nhr-67(RNAi) males because the LC is not positioned in L4 to access spatially restricted ECM components.

We investigated whether nhr-67 functions only at the start of the L3 stage to start a cascade of other genes that regulate later events in the L3 and L4 stages, or whether NHR-67 itself regulates events throughout these stages. We grew animals on nhr-67(RNAi) bacteria and then shifted them to non-RNAi-containing bacteria, such that they would be exposed to the effects of nhr-67(RNAi) during the L3, but not the L4, stage. If NHR-67 only acts at the early L3 stage, we would expect to see males with the full nhr-67(RNAi) gonadal defect, even if they had been removed from nhr-67(RNAi) after its initial effects. By contrast, if nhr-67 acts at different times throughout the L3 and L4 stages, we would expect to see males with a hybrid phenotype of early nhr-67(RNAi) defects but later wild-type characteristics. Indeed, we found hybrid phenotypes in the L3 and L4 stages when the animals were shifted from the nhr-67(RNAi) to non-RNAi-containing bacteria during the late L1/early L2 stage. In 11/44 males, the gonad failed to perform the turn from the dorsal to the ventral side during the mid-L3 stage, but properly expressed zmp-1::GFP by the late L4 stage (Fig. 6). These migrations were slower than in the wild type, but faster than in nhr-67(RNAi) animals. The hybrid phenotype further indicates that a shift in the levels of nhr-67 can occur within ~6 hours. Our results support a role for NHR-67 at different times throughout the L3 and L4 stages in the execution of the LC developmental program.

Here we have described how the C. elegans male LC relies on timing cues to execute the various stages of its migration. Closer examination of LC migration revealed that the LC displays many complex behaviors, including migration over different body surfaces, the execution of two turns, and changes in cell shape and migration speed. Moreover, these behaviors occur at specific times during the migration, suggesting that dynamic, stage-specific gene regulation is involved. Three genes that we found to exhibit stage-specific expression in the LC are the tailless homolog nhr-67, the netrin receptor unc-5 and the zinc metalloprotease zmp-1. nhr-67 is expressed by the LC during the L3 and L4 stages and is discussed below. unc-5 is expressed from the late L2 stage (Su et al., 2000) to the mid-L3 stage, and its downregulation is necessary for the LC to turn ventrally in the mid-L3 stage. unc-5 is likely to be just one of several ECM receptors dynamically regulated in the LC, as judged by the fact that other migratory cells, such as the hermaphrodite DTCs in C. elegans and the neural crest and primordial germ cells in vertebrates, express several different ECM receptors required to navigate their complex course (Henderson and Copp, 1997; Meighan and Schwarzbauer, 2008). zmp-1 is expressed only in the L4 stage. We found no defect in LC migration in zmp-1 deletion mutants, but this might be due to redundancy of matrix metalloproteases in the LC (Meighan et al., 2004). In addition to these three genes, we uncovered stage-specific cell behaviors, such as the L4 stage LC shape change, that are not mediated by UNC-5 or ZMP-1, suggesting that the LC expresses many other genes stage-specifically.

One way to modulate complex cell behaviors is through instructional cues that are spatially restricted, such that a cell modifies its behavior upon reaching specific sites along its migratory route. Instead, the LC appears to implement these behaviors as part of its developmental program based on a temporal cue. We show, for instance, that changes in cell shape and zmp-1 expression occur in L4 larvae even when the LC has not reached its normal position in the posterior body.

We have revealed a new function for nhr-67 in controlling a time-dependent subset of events during LC migration. nhr-67 acts cell-autonomously to regulate all the LC changes that we have identified within a specific time-window of the L3 and L4 stages, but none of the migratory changes either before or afterwards. This observation suggests that LC migration is assembled from temporal subprograms that might each be independently regulated by factors such as nhr-67.

In addition, nhr-67 controls the timing of events during the subprogram it regulates. For each of the developmental events in the LC during the L3 and L4 stages, nhr-67(RNAi) animals display a phenotype indicative of the continuation of the early L3 stage and the delay of succeeding stages. There is evidence that Tlx, the mammalian homolog of nhr-67, regulates timing in mouse neurogenesis (Roy et al., 2004). In nhr-67(RNAi) animals, UNC-5, which in wild-type animals is downregulated in the LC by the mid-L3 stage, continues to be highly expressed in the late L3 stage and during the L4 stage, whereas zmp-1, a gene normally expressed in the LC at L4 stage, is never expressed. These roles of NHR-67 are consistent with previous findings that tailless acts as both a positive and negative regulator of gene expression (Hoch et al., 1992; Margolis et al., 1995; Pankratz et al., 1992). Also, in nhr-67(RNAi) animals, the LC does not turn from the dorsal to the ventral bodywall during the mid-L3 stage as in wild-type worms. We show that this abnormal guidance is due, at least in part, to the continued expression of UNC-5 in nhr-67(RNAi) animals past the mid-L3 stage. Finally, in contrast to the increasingly polarized morphology of the LC in wild-type worms, the LC of nhr-67(RNAi) males remains in the round, early L3 stage shape throughout most of its migration. Although an nhr-67 null mutant might have a more severe phenotype than nhr-67(RNAi), the nhr-67(RNAi) phenotype is both penetrant and consistent in each animal and across trials.

An interesting question is how NHR-67 regulates events in the LC at diverse times during the L3 and L4 stages. One possibility is that NHR-67 confers specificity by binding to a heterodimeric partner or co-regulator (DeMeo et al., 2008; Nettles and Greene, 2005). Another possibility is that a temporal gradient is set by the level of expressed NHR-67, which would gradually accumulate over time. An example of this is seen with PHA-4, the master regulator of pharynx formation, which first activates genes whose regulatory regions have the highest affinity for PHA-4, and later activates genes whose regulatory regions have lower affinity (Gaudet and Mango, 2002).

We propose that LCs have both a basal migration program, which begins at LC specification and is used through the life of the LC, and at least three stage-specific programs that modify the basal program (Fig. 7). NHR-67 regulates one such program from the early L3 to mid-L4 stage. Since in nhr-67(RNAi) animals, LC migration before the early L3 stage and after the mid-L4 stage occurs with normal timing, there must be other transcriptional regulators acting in the LC. We hypothesize that each of these stage-specific regulators acts on a basal migration program consisting of genes that are expressed throughout LC migration. Some of the genes that we examined are part of the basal migration program, including lag-2, gon-1, him-4 and mig-2; the latter three are required for normal gonadal migration (Blelloch and Kimble, 1999; Henderson et al., 1994; Lundquist et al., 2001; Vogel and Hedgecock, 2001; Zipkin et al., 1997). As the expression of these genes was unaffected in nhr-67(RNAi) animals, we propose that the basal migration program continues even without the execution of stage-specific programs.

Cell migrations are often required during organogenesis and contribute to the shape and function of the final organ. In many of these cell migrations, spatially graded cues not only provide guidance but also regulate motility, cell morphology and gene expression. For example, in border cell migration in Drosophila, the migratory behavior is mediated through EGF and PDGF/VEGF receptors that bind signaling molecules from a spatially restricted origin (Duchek et al., 2001). Similarly, tracheal outgrowth in mouse or Drosophila requires FGF signaling from surrounding tissue, with the cells closest to the tissues being activated (Affolter and Caussinus, 2008; Cardoso and Lu, 2006). Temporal regulation of migration is less well characterized, perhaps because spatial cues have such a visible role in most cell migrations. Unlike other well-studied instances of cell migration, we have shown that the migration of the LC depends heavily on timing cues to execute different stages of its migration. We found that even when the LC is mispositioned in the body of the animal, and hence removed from its normal spatial environment, it is still capable of migrating and undergoing stage-specific programs. The delayed development of the LC in nhr-67(RNAi) animals also suggests that the LC uses NHR-67 to interpret timing information, the source of which might be a cell-intrinsic clock or a global timing cue. One hint that a global signal may be involved comes from the previous finding that the DAF-12 nuclear hormone receptor, which binds the global hormone dafachronic acid, is required by both the male LC and hermaphrodite DTCs for executing gonadal turns in L3 larvae and for arresting development in dauer larvae (Antebi et al., 1998; Motola et al., 2006).

Understanding temporal regulation during organogenesis should shed light on morphological differences between species, and between normal and diseased states. Since the gonads of different nematode species have different shapes (Chitwood and Chitwood, 1950), it is interesting to speculate that dramatic changes in gonadal shape could arise from subtle changes in temporal regulation of the LC. The existence of stage-specific regulators suggests the possibility that dramatic changes in organ shape can be achieved through the regulation of relatively few temporally regulated genes.

We thank the Andrew Fire laboratory for reagents; Gladys Medina and Barbara Perry for technical assistance; members of our laboratory for helpful discussions; and, in particular, Jolene Fernandes, Jennifer Green, Amir Sapir, Erich Schwarz, Adeline Seah, Jagan Srinivasan, Cheryl Van Buskirk and Allyson Whittaker for critically reading the manuscript; Elyse Blum for help with scoring LC death; and Sumeet Sarin for discussions on nhr-67 binding site sequence alignment. Some nematode strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources. Confocal images were acquired at the Caltech Biological Imaging Center. P.W.S. is an investigator with the Howard Hughes Medical Institute and M.K. was supported by an NIH postdoctoral fellowship. Deposited in PMC for release after 6 months.