Cell interactions control the direction of outgrowth, branching and fasciculation of the HSN axons of Caenorhabditis elegans

A major goal in developmental neurobiology is to under-stand how neurites navigate to find their synaptic partners. What types of cells guide neurites to their targets? What is the molecular nature of the interactions between extending growth cones and the cells with which they interact?

The nematode Caenorhabditis elegans has been used in two different ways to study axonal guidance. One approach has been genetic. By analyzing the morphologies of indi-vidual axons in mutants with behavioral deficits, investiga-tors have identified genes that are required for axonal out-growth (Hedgecock et al., 1985, 1987, 1990; Desai et al., 1988; Siddiqui and Culotti, 1991; McIntire et al., 1992). An alternate approach has been to define the cell interactions required for axonal outgrowth by removing specific cells and asking if the growth of particular axons is affected. In this way, Durbin (1987) showed that the AVG and PVP cells act as pioneer neurons in the development of the ven-tral nerve cord, the major longitudinal nerve bundle of the animal. In two additional studies, the cellular requirements for axonal branching were investigated. First, the BDU axon was shown to guide or stabilize the branch of the AVM touch neuron (Walthall and Chalfie, 1988). Second, epithelial cells of the vulva were found to be required for the branching of the VC neurons (Li and Chalfie, 1990).

We have been studying axonal outgrowth of the C. elegans HSN neurons, a pair of serotonergic motor neurons that innervate the vulval muscles and stimulate egg laying by the hermaphrodite (Trent et al., 1983; White et al., 1986). We have identified ten genes required for the outgrowth of the HSN axons (Desai et al., 1988; McIntire et al., 1992). Some of these genes are needed for pioneering HSN out-growth along epithelial cells, while others are needed for HSN outgrowth along other axons within the ventral nerve cord. All ten of these genes are also required for the out-growth of the axons of other neurons. In principle, these genes could act either within the neurons altered in axonal outgrowth or within other cells that interact with these neu-rons. Genetic mosaic analysis suggested that one of these genes, unc-6, is required in epithelial cells for the outgrowth of motor neurons across these cells (cited by Hedgecock et al., 1990). Although genetic mosaic analysis should also be useful in defining the sites of action of many of the other genes required for axonal outgrowth, this type of study will be more difficult for genes required for growth of axons in nerve bundles, where the number of potential interacting cells is large. Knowledge of which cells are involved in HSN axonal guidance should help identify the cells in which these genes function.

In this paper, we identify cell interactions that are required for HSN axonal guidance. By analyzing HSN axonal morphology in animals missing specific cells, we find that both vulval cells and neurons that contribute axons to the ventral nerve cord are required for normal HSN axonal guidance. Vulval cells, which are specialized types of epithelial cells, are needed for specifying the initial direc-tion of HSN axonal outgrowth, for HSN axonal branching and for two distinct aspects of HSN axonal fasciculation. The axons of the PVP and PVQ neurons are also needed for normal fasciculation of the HSN axons. We propose that interactions between the HSN growth cones, the PVP and PVQ axons and vulval cells are required for normal pathfinding by the HSN axons.

Strains were grown at 20°C unless otherwise stated and were maintained as described by Brenner (1974). Most of the mutant strains used in this work were described by Greenwald et al. (1983), Trent et al. (1983), Ferguson and Horvitz (1985), Hedge-cock et al. (1985), Desai et al. (1988), Manser and Wood (1990) and Thomas et al. (1990). The weak lin-39 allele 709 was described by Fixsen et al. (1985), and the stronger n1490 and n1760 alleles were isolated and characterized by S. Clark (per-sonal communication), as was lin-2(n1602). The sem-2(n1343) allele was isolated and characterized by M. Stern. This paper uses standard C. elegans genetic nomenclature (Horvitz et al., 1979).

In addition to the standard wild-type strain (N2), strains with the following mutations were used in this work.

LGI: lin-10(n1390), lin-11(n382 and n389), mig-1(e1787 and n1354), sem-2(n1343).

LGII: clr-1(e1745ts), egl-27(n170), egl-43(n997), lin-7(e1413). LGIII: dig-1(n1321), lin-12(n137sd, n302sd and n302sd n865), lin-39(n709, n1490 and n1760), mig-10(ct41) and unc-32(e189). LGIV: egl-18(n162 and n474), egl-20(n585), lin-1(n1777), lin-3(e1417).

LGX: egl-15(n484), lin-2(n1602), lin-15(n765ts).

lin-39(n709) double mutants were constructed by crossing lin-39(n709)/+ heterozygous males with hermaphrodites from a strain containing the second mutation. Potential doubly heterozygous animals were picked; from those animals segregating progeny with ectopic cell deaths in the VNC (the VC neurons variably undergo programmed cell death in lin-39(n709) mutants, and this pheno-type can be scored using Nomarski optics; Fixsen et al., 1985), animals homozygous for lin-39(n709) were isolated. Animals homozygous for the second mutation were isolated from the prog-eny of these animals. Similarly, mig-1; lin-15 double mutants were constructed by crossing mig-1 homozygous males with lin-15 her-maphrodites. Non-Lin doubly heterozygous animals were picked from the F₁ progeny, and Lin F₂ animals were then isolated. From the F3 progeny, Mig animals were identified by their displaced HSNs using Nomarski optics.

Indirect immunofluorescence histochemistry was used to stain ani-mals for serotonin, essentially using the procedure described by McIntire et al. (1992). Adult hermaphrodites were fixed for 24 hours at 4°C in PBS (10.4 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 137 mM NaCl, 4.4 mM KCl, pH 7.2), 4% paraformaldehyde fixative. After being washed three to four times in PBS, the worms were rocked gently overnight at 37°C in a solution of 5% β-mercap-toethanol, 1% Triton X-100 in 0.125 M Tris-HCl, pH 6.9. The worms were washed three more times in PBS, and 20-30 μl of worms were shaken vigorously in 0.4 ml of 100 mM Tris-HCl pH 7.5, 1 mM CaCl₂ and 1,000-2,000 units/ml collagenase type IV (Sigma, St. Louis, MO). After 20-50% of the worms were frag-mented (30 minutes to 3 hours), the worms were washed with PBS three to four times and incubated overnight at room temper-ature in 20 μl of 1% rabbit anti-serotonin antiserum (provided by J. Steinbusch, Free University, Amsterdam, The Netherlands) in PBS, 1% Triton X-100, 1 mM sodium azide. The worms were then washed three times in PBS and incubated for 2 hours at 37°C in 20 μl of 5% FITC-conjugated goat anti-rabbit antiserum (Cappel Inc., West Chester, PA) in PBS, 1% Triton X-100, 1 mM sodium azide. The worms were then washed three times in PBS and incubated for 2 hours at 37°C in 20 μl of 5% FITC-conju-gated rabbit anti-goat antiserum in PBS, 1% Triton X-100, 1 mM sodium azide. The worms were then washed three times in PBS. 5 μl of stained worms were mixed with 5 μl of 1 mg/ml p-phenylenediamine in 10% PBS, pH 8.0, and 90% glycerol and viewed either by standard fluorescence microscopy (illuminated with a mercury vapor lamp, and viewed through an FITC filter; Zeiss 48705, excitation range 395-440 Å) or by confocal microscopy (Bio-Rad MRC-500 scanning confocal microscope; illuminated with an argon ion laser, excitation 488 Å).

Double staining animals with both anti-serotonin and anti-FMRFamide antisera to visualize the defasciculation of the HSN axons at the vulva was done using either of two procedures. One approach was to follow the protocol described above for anti-sero-tonin staining with the addition of a 5% rabbit anti-FMRFamide antiserum (provided by C. Li, Boston University, Boston, MA) with the anti-serotonin antiserum during the primary antisera incu-bation step. An alternative approach was to permeabilize the ani-mals using a Dounce homogenizer, as described by Li and Chal-fie (1990). Following the douncing, all antibody incubation steps were carried out as described above. Permeabilization by dounc-ing resulted in slightly increased staining but higher background levels. Double staining usually was preformed in double mutants that carried the mutation lin-39(n709). Since the only cell bodies in the VNC that obviously stain with the anti-FMRFamide anti-serum are those of the VC neurons, lin-39(n709) animals missing all six VC neurons were identified as those animals with no stained cell bodies in the VNC.

Procedures for using Nomarski optics microscopy to observe living animals were described by Sulston and Horvitz (1977).

Nuclei were destroyed as described by Avery and Horvitz (1987) in early L1 larvae shortly after hatching. The Pn.a or Pn.p cells were killed later during the L1 stage shortly after the P cells had divided. Verification of the kills were made the next day using Nomarski optics, except for the kills of the somatic gonad pre-cursor cells Z1 and Z4, which were confirmed by the resulting sterility. Once mature, laser-operated animals were placed in microtiter wells containing M9 buffer (Brenner, 1974) and chilled on ice for 20 minutes. The M9 was then replaced with ice cold 4% paraformaldehyde fixative in PBS and fixed overnight at 4°C. The worms were washed three times with PBS and incubated overnight at 37°C in 100 μl of a solution of 5% β-mercap-toethanol, 1% Triton X-100 in 0.125 M Tris-HCl, pH 6.9. The worms were washed three times in PBS and incubated for 1-3 hours in 100 μl of 100 mM Tris-HCl pH 7.5, 1 mM CaCl₂ and 1,000 units/ml collagenase type IV (Sigma, St. Louis, MO). The worms were then washed with PBS three to four times. Two pre-cautions were taken to reduce the background of nonspecific stain-ing that resulted when staining small numbers of worms. First, before incubating the worms in the primary antiserum, 20-30 μl of worm shards were added to the microtiter wells. Because of their small sizes, these pieces of worms are easily distinguished from intact laser-operated worms. The worm shards were prepared by exhaustively digesting worms that had been fixed and treated with β-mercaptoethanol with collagenase (see above) until only small fragments of worms remained. Second, the primary, sec-ondary and tertiary antisera used in these experiments were pre-viously used in the staining of bulk quantities of worms as described above. These antisera were pooled from staining exper-iments and centrifuged to remove residual worms. The super-natants were then stored at 4°C. The worms and worm shards were incubated overnight at room temperature in 10 μl of 1% rabbit anti-serotonin antiserum in PBS, 1% Triton X-100, 1 mM sodium azide. The worms were then washed three times in PBS and incubated for 2 hours at 37°C in 10 μl of 5% FITC-conju-gated goat anti-rabbit antiserum in PBS, 1% Triton X-100, 1 mM sodium azide. The worms were then washed three times in PBS and incubated for 2 hours at 37°C in 10 μl of 5% FITC-conju-gated rabbit anti-goat antiserum in PBS, 1% Triton X-100, 1 mM sodium azide. As the worm shards were lost in the PBS washes, new shards were added. The worms were then washed three times in PBS. 5 μl of stained worms were mixed with 5 μl of 1 mg/ml p-phenylenediamine in 10% PBS, pH 8.0, and 90% glycerol and viewed by fluorescence microscopy as above.

We found that growing animals at 25°C instead of 20°C helped increase the intensity of anti-serotonin staining. For the cell killing experiments in which antisera staining tended to be marginal, ani-mals were raised at 25°C, except in the case of lin-15(n765ts) ani-mals, which were grown at 20°C. lin-15(n765ts) animals are sick and stain poorly at 25°C.

We also attempted to determine HSN axonal morphology in animals missing their germline gonadal cells. These experiments were attempted either by killing, in newly hatched L1 stage ani-mals Z2 and Z3, the germline precursor cells (Kimble and Hirsh, 1979), or by killing, in the embryo P4, the precusor to Z2 and Z3 (Sulston et al., 1983). In these experiments, the HSNs either failed to stain or stained poorly with anti-serotonin antisera, making it impossible to reliably score HSN axonal morphology. We do not know the reason for this poor HSN staining.

To observe the ventral growth of the HSN axons, we used the temperature-sensitive mutant clr-1(e1745ts; Hedgecock et al., 1990). The boundaries of cells and axons can be visualized when clr-1(e1745ts) animals are grown at the nonpermissive tempera-ture of 25°C. To view HSN axonal morphology, we shifted clr-1(e1745ts) animals grown at 15°C to 25°C for 4 hours and then scored HSN axonal morphology in these animals using Nomarski optics.

The components of the C. elegans egg-laying system (White, 1988) are illustrated in Fig. 1. Eggs are stored in the uterus and are laid through the vulva upon contraction of the 16 uterine and vulval muscle cells. Eight uterine muscle cells (not shown) run circumferentially around the uterus and contract to squeeze eggs from the uterus. Eight vulval muscle cells (only four are shown) attach to the vulva and contract to open the vulva and allow eggs to pass. Gap junctions interconnect the vulval and uterine muscle cells and presumably cause all 16 muscle cells to contract syn-chronously.

Two classes of neurons synapse directly with the vulval muscles: the HSNs and the VCs (White et al., 1986). The cell bodies of the two HSN neurons are located on either lateral side of the animal just posterior to the vulva. (Fig. 1 shows only the left HSN cell body and its axon.) The HSN neurons are required for normal egg laying: her-maphrodites missing or with functionally impaired HSNs are egg-laying defective and bloat transiently with late-stage embryos (Sulston and Horvitz, 1981; Trent et al., 1983; Desai et al., 1988; Desai and Horvitz, 1989). The cell bodies of the six VC neurons are located along the ventral side of the hermaphrodite. (Fig. 1 shows the cell bodies of only the fourth and fifth VC neurons.) The VCs have yet to be assigned a role in egg laying, since hermaphrodites that are missing VCs appear to lay eggs normally (G.G., C. Trent and H.R.H., unpublished observations).

Fig. 2A illustrates four aspects of HSN axonal outgrowth which will be the focus of this study: (1) ventral growth of the HSN axon to the ventral nerve cord (VNC), (2) growth of the left and right HSN axons anteriorly along other axons in the left and right bundles of the VNC, respectively, (3) defasciculation of the HSN axons from other axons of the VNC at the vulva, and (4) branching of the HSN axon at the vulva. Below we describe in more detail these aspects of HSN axonal outgrowth and the techniques used to ana-lyze them.

As we described previously (Desai et al., 1988), the mature HSN can be visualized in the adult hermaphrodite by staining animals with antiserum to serotonin (Fig. 2B). A single axon emerges from each HSN cell body and extends to the ventral nerve cord, which consists of two parallel bundles of axons that extend the length of the animal. The axons from the left and right HSNs enter the left and right bundles of the VNC, respectively. Once in the VNC, each HSN axon turns anteriorly and skirts the vulva, at which point it forms varicosities and, usually, a small branch. The synapses with the egg-laying muscles are made at the varicosities and along the branch (White et al., 1986). Each HSN axon continues anteriorly on its respec-tive side of the VNC to the head, where it enters the nerve ring, a dense neuropil that runs circumferentially around the pharynx (not shown).

The HSN and the VC axons separate from the other processes of the VNC at the vulva and run in a distinct fas-cicle just dorsal to the main VNC bundle (White et al., 1986; J. White, personal communication; C.D., unpublished results). Originally observed in electron micrographs of serial sections, this defasciculation of the HSN axons from other axons in the VNC can also be seen and analyzed using the light microscope by staining the HSNs with an anti-serotonin antiserum and staining the VNC with an anti-FMRFamide antiserum (Fig. 2C). Anti-FMRFamide anti-serum recognizes the AVK axon on the left side of the VNC and several axons on the right side of the VNC (Li and Chalfie, 1990; Schinkmann and Li, 1992). However, anti-FMRFamide antiserum also recognizes the VC neurons, which defasciculate along with the HSNs. Thus, to analyze HSN defasciculation we had to develop a method in which the axons of the HSNs would not be confused with those of the VCs. We accomplished this objective by eliminating the VCs using the lin-39 mutation n709, which results in a variable loss of the VCs (Ellis, 1985; Fixsen et al., 1985; Li and Chalfie, 1990). Fig. 2C shows a lin-39(n709) animal that lacked all six VC neurons stained with both anti-sero-tonin and anti-FMRFamide antisera. Near the vulva, the left HSN axon can be seen to be separated from the left VNC bundle, which is labeled by the AVK axon. We find that it is where the HSN axons are separated from the main VNC fascicle that they form larger varicosities and branch. The HSN axonal morphology seen in lin-39(n709) animals is the same as the morphology of wild-type animals, deter-mined by electron microscopy.

Because only the adult HSN contains detectable serotonin (Desai et al., 1988), staining with anti-serotonin antiserum allows an analysis of HSN axonal morphology only after the axons have completed outgrowth. Based upon these anti-serotonin staining patterns, we have inferred the temporal sequences of events that lead to the final morphologies of the HSNs in wild-type and mutant animals (Desai et al., 1988; McIntire et al., 1992). In particular, we would not have observed dynamic aspects of axonal outgrowth, such as probing by and retraction of the HSN axons.

Using a method developed by Hedgecock et al. (1990), we now can observe HSN morphologies during the early stages of HSN axonal outgrowth. The boundaries of cells and axons can be visualized when animals carrying the tem-perature-sensitive mutation clr-1(e1745ts) are viewed using Nomarski optics after growth at the restrictive temperature (Hedgecock et al., 1990). Fig. 2D shows a photomicrograph of a late second larval (L2) stage clr-1 animal. The HSN can be seen to extend two neurites, which are typical of the HSNs in L2 or third larval (L3) stage animals (some late L2 stage animals do not appear to have yet extended HSN axons). One neurite extends dorsally, terminating at the excretory canal. This neurite presumably would later retract or rearrange, since such morphology is not seen in older animals. We do not know if this dorsal extension plays a role in HSN development. A second neurite extends ante-riorly and ventrally (see below). Unfortunately, once near the VNC, the HSN axon cannot be distinguished from the other axons in this bundle.

Thus, the HSN axons begin to grow during the L2 or L3 stage. These axons apparently do not terminate growth until after the late fourth larval (L4) stage, since they have not reached the nerve ring by this stage (White et al., 1986). The relatively long period of more than 20 hours of HSN outgrowth in the VNC indicates either that the HSN axons grow quite slowly (approximately 10-15 μm/hour) or that there is a period when the HSN axons slow or arrest their outgrowth.

To identify cell interactions involved in HSN axonal guid-ance, we focused on cell types that we considered likely to contact the extending HSN axons: cells of the egg-laying system and neurons that contribute axons to the VNC. These cells were removed either by mutation or by killing with a laser microbeam, and the mutant or operated adult animals were stained with anti-serotonin antisera to reveal HSN axonal morphology. The HSN axons initially extend along a hypodermal syncytium that constitutes the epithe-lial covering of the animal; the syncytial nature of this epithelium has prevented us from directly testing its role in HSN axonal outgrowth. We scored animals for HSN axonal outgrowth ventrally to the VNC, for the separate outgrowth of the two HSN axons on opposite sides of the VNC, for HSN axonal branching and for HSN axonal outgrowth ante-riorly along the VNC. Cells required to guide the HSN axons ventrally were also removed in a clr-1 mutant back-ground, and the early stages of HSN axonal outgrowth were examined directly in these animals. In addition, to score defasciculation of the HSN axons from the VNC at the vulva, we constructed double mutants carrying a mutation that removes a specific cell that we wished to test for a role in HSN defasciculation and the lin-39(n709) mutation, which removes the VC neurons; we stained these double mutants with anti-serotonin and anti-FMRFamide antisera (see above).

The vulval cells or their precursors play key roles in HSN axonal guidance (see below). To describe how we analyzed this function of the vulval cells, we will first briefly sum-marize aspects of what is known about how different types of vulval cells develop. A more detailed summary of vulval development has been recently published (Horvitz and Sternberg, 1991).

Six epithelial cells (P3.p-P8.p) located along the ventral side of the animal have the potential to generate cells of the vulva. The fates of these Pn.p cells can be distinguished by the numbers and cell types of the progeny they gener-ate. The fate of each of these cells depends on its proxim-ity to a signal emanating from the anchor cell of the somatic gonad, with the closest cell (P6.p) adopting the primary fate, the next two distal cells (P5.p and P7.p) adopting the secondary fate and the most distal cells (P3.p, P4.p and P8.p) adopting the tertiary fate. The vulva is formed by the 22 descendants of P5.p-P7.p, the cells that adopt the pri-mary and secondary fates. Cells that adopt the tertiary fate generate nonvulval progeny. In the absence of the signal from the somatic gonad, all six of these Pn.p cells adopt a tertiary fate. No vulval cells are formed, and the animal is said to be vulvaless. The Pn.p cells divide in the L3 stage, and vulval morphogenesis occurs during the L4 stage.

Vulvaless animals, generated either by mutation or by killing cells required for vulva formation with a laser microbeam, exhibit four types of HSN axonal defects: in the initial direction of HSN axonal outgrowth, in position-ing within the VNC, in defasciculation and in branching of the HSN axons at the vulva. Each of these defects is described below.

In vulvaless animals, the HSN axons sometimes fail to grow ventrally and instead extend anteriorly along a lateral path-way, eventually entering the nerve ring, their normal des-tination (for phenotype see Fig. 3A; Table 1). This failure to enter the VNC occurs much more frequently in vulva-less animals generated by eliminating the somatic gonad or the Pn.p cells by laser microsurgery (23-25%) than in vul-valess animals generated by mutations in the lin-2, lin-3, lin-7 and lin-10 genes (0-1%). We suspect this difference reflects the fact that some Pn.p cells retain partial vulval cell identity in these mutants (vulvaless animals can gen-erate some vulval cells but still not generate a functional vulva; Ferguson and Horvitz, 1985). Consistent with this interpretation, the HSN axons also fail to enter the VNC more frequently in lin(n300), lin-12(n302sd), lin-39(n1490) and lin-39(n1760) vulvaless mutants (8-22%); these mutants only infrequently generate vulval cells (Ferguson and Horvitz, 1985; S. Clark, personal communication). (In contrast to the weak lin-39 mutant n709, which is variably missing VC neurons and has a vulva, the stronger lin-39 mutants n1490 and n1760 lack all six VC neurons and are vulvaless; Li and Chalfie, 1990; S. Clark, personal com-munication.) This defect in HSN axonal morphology is a consequence of a defect in the initial direction of HSN axonal outgrowth, since we have directly observed the HSN axons extend aberrantly during the L3 stage in a clr-1; lin-39(n1760) vulvaless double mutant (20%, n=15). The abnormally extending HSN axons of clr-1; lin-39 animals grow along the canal-associated fascicle, which normally contains the processes of the ALA, BDU, CAN and PVD neurons and the excretory canal (White et al., 1986). In addition, in those vulvaless animals in which ventral out-growth occurs, the HSN axons often extend anteriorly or posteriorly a short distance before growing ventrally and entering the VNC (see Fig. 3C).

Vulval cells also appear to play a minor role in the direc-tion of HSN axonal outgrowth within the ventral nerve cord. In vulvaless animals, the HSN axons can enter the VNC normally and then grow posteriorly to the tail or grow a short distance posteriorly in the VNC before reversing direction and growing anteriorly (see left HSN axon in Fig. 3B). However, this defect occurs infrequently. For exam-ple, only 5% of the HSN axons that enter the VNC in lin-39(n1760) mutants grow to the tail.

In vulvaless mutants, the HSN axons usually fasciculate with one another in the VNC instead of growing separately on either side of the VNC (Fig. 3B, Table 1). In these ani-mals, the left HSN axon grows across the left nerve bundle of the VNC and enters the right nerve bundle of the VNC, where it runs with the right HSN axon to the nerve ring (see Fig. 3C). At the level of resolution of immunocyto-chemical staining, the two HSN axons appear to run as a single fascicle that separates only infrequently.

In vulvaless animals, the HSN axons fail to defasciculate from the VNC bundle at the vulva. This failure to defasci-culate was suggested by the relatively direct path the HSN axons take once they enter the VNC in vulvaless mutants (see Fig. 3B) and has been confirmed by double staining of the lin-39(n1490) vulvaless and VC-deficient mutant with both anti-FMRFamide and anti-serotonin antisera (Fig. 3C, Table 2).

In vulvaless animals, the HSN axons fail to branch (Table 1, Fig. 3B). Moreover, extra HSN branches are occasion-ally seen at the positions of ectopic vulval protrusions in the multivulva mutants lin-1 and lin-15 (Li and Chalfie, 1990; Fig. 4A and unpublished results). These results sug-gest that vulval cells determine the position of the HSN branch. Vulval cells are also required for branching of the axons of the VC neurons, the other class of neurons that innervate the egg-laying muscles (Li and Chalfie, 1990).

To determine which types of vulval epithelial cells are required for HSN axonal outgrowth, we analyzed HSN axonal morphology in mutants that produced only sec-ondary or only primary vulval cells. This analysis suggests that primary vulval cells play the major role in HSN axonal guidance. The dominant lin-12(n137sd) mutant, which gen-erates secondary but not primary vulval cells (Greenwald et al., 1983), was defective in ventral-directed growth of the HSN axons, in the positioning of the left HSN axon within the VNC, in defasciculation of the HSN axons from the VNC and in HSN branching (Fig. 3A, Tables 1, 2). By contrast, the recessive lin-12(n302 n865) mutant, which generates primary but not secondary vulval cells (Green-wald et al., 1983), had normal HSN axonal morphology (Fig. 4B, Table 2). These results strongly suggest that the primary vulval cells or their precursors control many aspects of HSN axonal guidance. An alternative interpreta-tion of these results is that the lin-12 gene functions in the HSNs and the HSN axonal defects of dominant lin-12 mutants is caused by intrinsic defects in the HSNs rather than by defects in vulval cells.

There might also be some influence of the secondary vulval cells on HSN axonal outgrowth. In lin-12(n137sd) mutants, the HSN axons entered the VNC and then grew posteriorly to the tail more frequently than in vulvaless mutants. Of those HSN axons that entered the VNC in lin-12(n137sd) animals, 21% extended to the tail (Fig. 2D), as opposed to only 5% in the lin-39(n1760) vulvaless mutant. This increased frequency of posterior-directed outgrowth in lin-12(n137sd) mutants might be a consequence of the absence of primary vulval cells combined with the presence of ectopic secondary cells posterior to where the HSN axons enter the VNC. If so, growing HSN axons within the VNC might be attracted toward secondary as well as primary vulval cells (see below).

Since the HSN axons begin to grow ventrally during the L2 and early L3 stages, which is before the vulval cells divide, it appears that vulval precursor cell P6.p is required for the ventral-directed outgrowth of the HSN axons. How-ever, it is not known when the HSN axons grow anteriorly within the VNC, defasciculate from the VNC or branch, so either the Pn.p cells or their progeny could influence these aspects of HSN axonal outgrowth.

The HSN neurons are generated in the tail of the embryo and then migrate anteriorly to their final positions near the middle of the body (Sulston et al., 1983). Mutations in sev-eral genes result in the HSNs either failing to migrate or migrating only partially, leading to posterior displacement of the HSN cell bodies (Desai et al., 1988; see Fig. 5). All of these mutants with posteriorly displaced HSNs display defects in HSN axonal morphology that are in many ways similar to those seen in vulvaless animals (Table 3). First, the axons of the displaced HSNs often extend anteriorly or posteriorly a short distance before entering the VNC. How-ever, these HSN axons never extend laterally all the way to the head as in vulvaless mutants: laterally extending axons of posteriorly displaced HSNs invariably enter the VNC at the vulva, presumably in response to P6.p or its descendants. As in the vulvaless mutants, the HSN axons can enter the VNC and grow posteriorly to the tail (e.g. 2% of the time in mig-1(n1354) mutants), which is again con-sistent with the vulval cells or their precursors playing a minor role in establishing the direction of HSN axonal extension in the VNC. Second, the HSN axons are posi-tioned abnormally in the migration mutants: the left HSN axon passes over the left nerve bundle of the VNC, enters the right bundle and runs alongside the right HSN axon. Finally, the HSN axons fail to branch in the migration mutants. Unlike vulvaless animals, the HSN axons defas-ciculate at the vulva in the HSN migration mutant mig-1(n1354) (Table 2).

The severity of the defects in HSN migration and in HSN axonal morphology are generally correlated. For example, mutant strains with relatively severe defects in HSN migra-tion tend to have higher frequencies of HSN axonal defects (Table 3). There is also a correlation between the severi-ties of the HSN migration and axonal defects within indi-viduals from a given mutant strain. For example, as shown in Fig. 5 for mig-1(n1354) animals, 77% of the HSN axons are normally positioned within the VNC and 80% of the HSN axons branch when the HSN cell bodies are found near or at their normal positions (located more than 70% of the distance from the anus to the vulva); by contrast, 13% of the HSN axons are normally positioned within the VNC and 28% of the HSN axons branch when the HSN cell bodies are displaced more posteriorly (located less than 70% of the distance from the anus to the vulva).

One interpretation of the existence of both migration and axon defects in these mutants is that the genes they define are separately involved in the two processes of HSN cell migration and HSN axonal outgrowth. Alternatively, these genes could be required primarily for HSN migration, with the HSN axonal defects occurring as a secondary conse-quence of the abnormal positions of the HSN cell bodies. To distinguish between these two possibilities, we con-structed a double mutant containing mutations in the mig-1 and lin-15 genes. In this double mutant, ectopic vulval cells are positioned near the displaced HSN cell bodies, allowing us to determine if the HSN axonal defects of the mig-1 mutant are suppressed by the nearby vulval cells. Indeed, both the fasciculation and branching defects were suppressed, with the HSN axon branching ectopically at the nearby vulval protrusion (Fig. 5). Thus, the HSN axonal outgrowth defects of the HSN-migration mutants are a sec-ondary consequence of the abnormal positions of the HSN cell bodies. In other words, it is not the correct position of the HSN cell body per se that is required for normal HSN axonal outgrowth, but rather it is the proximity of the HSN cell bodies to the vulval cells that is important.

Animals defective in the gene dig-1 have the same HSN axonal defects as the HSN-migration mutants (Table 3). In dig-1 animals, the gonad is anteriorly displaced so that the vulva develops in a more anterior position than it would normally; the HSN cell bodies are in their normal positions (Thomas et al., 1990). The existence of HSN axonal defects in dig-1 animals further supports the hypothesis that it is the relative positions of the HSN cell bodies and the vulva rather than the absolute positions of the HSNs that are important for normal HSN axonal guidance.

The defects in ventral growth and VNC positioning of the HSN-migration and dig-1 mutants seem likely to reflect the lack of vulval cells in the vicinity of the HSN axons as they begin to extend. However, the basis of the branching defect in these mutants is less obvious, since the HSN axons eventually grow past the vulval cells, and even appear to respond to cues for defasciculation at the vulva (Table 2). Perhaps the HSN axons are capable of branching only for a short distance or for a short period of time after their ini-tial outgrowth.

While analyzing the effects of abnormalities in vulval development on HSN axonal outgrowth, we discovered that lin-11 mutants are defective in HSN branching and HSN defasciculation at the vulva (Fig. 6, Tables 1, 2). The axons of the VC neurons also fail to branch and defasciculate at the vulva in lin-11 animals (C. Li, personal communica-tion; G.G., unpublished results). In lin-11 animals, the fates of the secondary vulval cells are altered, but the fates of the primary vulval cells are not obviously abnormal (Fer-guson et al., 1987). Since lin-12(n302 n865) mutants, which lack secondary vulval cells, appear to be normal in HSN branching and defasciculation (see above), the HSN abnor-malities in lin-11 animals were unexpected. Perhaps it is the presence of abnormal secondary vulval cells (as opposed to the absence of normal secondary cells) that is responsible for these HSN and VC outgrowth defects in lin-11 animals. Alternatively, lin-11 could affect another cell type involved in HSN and VC outgrowth, such as the pri-mary vulval cells or the HSNs and VCs themselves.

It is interesting that lin-11 is required for HSN branch-ing and defasciculation at the vulva but not for ventral- or anterior-directed outgrowth or for positioning of the left HSN within the left bundle within the VNC. That the lin-11 mutations affect only some aspects of HSN axonal out-growth suggests that lin-11 acts differently from the other genes required for vulval development and is consistent with the hypothesis that lin-11 functions in cells other than the primary vulval cells or in only a subset of these cells.

The egg-laying muscles, the VC neurons and cells of the somatic gonad do not appear to play a role in HSN axonal outgrowth. HSN axonal morphology appeared normal both in mutants with displaced (egl-15) or missing (sem-2) egg-laying muscles (Trent et al., 1983; Stern and Horvitz, 1991; M. Stern, personal communication), and in animals in which the precursor to the egg-laying muscles (M cell; Sul-ston and Horvitz, 1977) had been killed with a laser microbeam (Tables 2, 4). Similarly, the VC neurons, the other class of neurons that innervate the egg-laying mus-cles and branch at the vulva, also are not necessary for normal HSN axonal morphology. First, HSN axonal mor-phology is normal in lin-39(n709) animals missing all six VC neurons (Fig. 2C). Second, animals in which the neu-roblasts that give rise to the VCs were killed with a laser microbeam still formed apparently normal HSN axons (Table 4).

In addition, the somatic (non-germline) cells of the gonad do not appear to be required for normal HSN axonal mor-phology. We could not test the direct involvement of these cells simply by killing the somatic gonad precursor cells Z1 and Z4 (Kimble and Hirsh, 1979) in wild-type animals, since the somatic gonad signals the vulval precursor cells to develop as vulval cells and thus is indirectly needed for HSN axonal outgrowth (see above). To circumvent this problem, we killed the somatic gonad precursor cells in a lin-15 mutant (Table 4). Because the vulval cells develop independently of the gonadal signal in lin-15 mutants, lin-15 animals missing a somatic gonad still generate vulval cells (Ferguson et al., 1987). In these animals, the HSN axons extended to the VNC and branched in the absence of the somatic gonad. For technical reasons, we could not determine if the HSN axons fasciculated abnormally in the VNC (see Table 4 legend). We cannot absolutely exclude a role for the somatic gonad in HSN axonal outgrowth from this experiment, since it is conceivable that lin-15 causes gonad-independent HSN axonal outgrowth just as it causes gonad-independent vulval development.

Our methods would not have detected subtle changes in HSN morphology, such as defects in the extent or shape of the HSN branch or alterations in the positions of the HSN axons relative to other axons within the VNC. Thus, it remains possible that the egg-laying muscles, VC neurons or cells of the somatic gonad play a role in these more subtle aspects of HSN axonal guidance.

The C. elegans VNC is bilaterally asymmetric, with a large bundle of axons on the right side (50-60 axons) separated by a ridge of ventral hypodermal epithelium from a small bundle of axons on the left side (3-5 axons; Fig. 7A; White et al., 1976, 1986). In the region of the vulva in the left bundle of the VNC there are three axons besides that of the left HSN, those of the AVKR, PVPR and PVQL neurons. The cell bodies of the two AVK neurons, AVKL and AVKR, are located in the ventral ganglion of the head, and the cell bodies of the two PVP neurons, PVPL and PVPR, are located in the pre-anal ganglion of the tail. Both the AVK and PVP neurons send axons to the side of the VNC contralateral to their cell bodies. For example, the AVKR cell body on the right side of the head has an axon on the left side of the VNC. The cell bodies of the two PVQ neu-rons, PVQL and PVQR, are located in the lumbar ganglia of the tail and each sends an axon to the ipsilateral side of the VNC. To determine if axons within the VNC are required for extension of the left HSN axon along the left nerve bundle of the VNC, we killed AVKR, PVPR and PVQL, let the animals grow to become adults and stained them with an anti-serotonin antiserum. When all three of these neurons were killed, the left HSN axon crossed the hypodermal ridge and extended along the right bundle of the VNC with the right HSN axon, as in vulvaless animals (Fig. 7B, Table 5). Thus, the left HSN axon is not capable of pioneering the left side of the VNC. Electron micro-graphs of the VNC of animals in which the AVKR, PVPR and PVQL neurons were killed in newly hatched animals confirmed that the left side of the VNC was absent by the late L2 and early L3 stages, the time when the HSN axons begin to extend (E. Hartwieg and G.G., unpublished results).

To determine which of these three axons are required for growth of the left HSN axon on the left side of the VNC, we killed neurons individually and in pairwise combina-tions (Table 5). In experiments in which a single neuron was killed, only killing of the PVQL neuron had any obvi-ous effect on HSN axonal morphology: the left HSN axon extended along the right bundle of the VNC with the right HSN axon in two out of seven operated animals. However, in all animals missing both PVPR and PVQL, the left HSN axon extended along the right bundle of the VNC. These results suggest that the PVP and PVQ axons play partially redundant roles in guiding the outgrowth of the left HSN axon along the left VNC fascicle, perhaps by expressing similar molecules on their surface that are recognized by the HSN growth cone. These results also suggest that the left HSN growth cone does not interact with the left AVK axon.

When we killed the PVPL and PVQR neurons, which contribute axons to the right side of the VNC, we saw no obvious HSN axonal defects. In particular, the right HSN axon did not run in the left bundle of the VNC. Presum-ably, the right HSN axon can recognize one of the many axons on the right side of the VNC. We note that the appar-ently normal morphology of the right HSN axon in animals missing the PVPL and PVQR axons could be misleading. Our procedures would not have detected changes in the pre-cise positioning of the HSN axon within the VNC, and it is possible that the right HSN axon fasciculated in a dif-ferent region of the VNC in these animals. When both PVP neurons and both PVQ neurons were killed, the left HSN axon extended along the right bundle of the VNC and ran with the right HSN axon.

By analyzing HSN morphology in both mutants and laser-operated animals, we have discovered that many aspects of HSN axonal outgrowth are controlled by interactions with other cells. Epithelial cells of the vulva, in particular the primary vulval cells or precursors that generates these cells, play a major role. Our experiments, summarized in Table 6, lead to the models for HSN axonal outgrowth illustrated in Fig. 8. First, the vulval precursor cell P6.p attracts the growth cone of each HSN, helping its axon to extend ven-trally to the VNC (Fig. 8A). There the HSN growth cones interact with other axons, including those of PVP and PVQ, to determine which of the two fascicles each will join (Fig. 8B). The primary vulval cells or their precursors help direct the left HSN axon to the left VNC fascicle. Each HSN axon then elongates anteriorly for a short distance, after which interactions with the primary vulval cells or their precur-sors cause it to defasciculate from the VNC in the region of the developing vulva (Fig. 8C), where it will branch, form varicosities and make synapses with the vulval mus-cles. Developing vulval cells also induce or stabilize HSN axonal branching at the vulva (Fig. 8D). The HSN axons return to the VNC and fasciculate with the axons of the PVP and PVQ neurons, which guide them anteriorly toward the head. We discuss these steps in more detail below.

In wild-type animals, ventral guidance cues for the HSN growth cones appear to compete with another guidance system provided by laterally positioned processes: if the ventral guidance system is perturbed, the HSN axons extend along the lateral processes. The vulval precursor cell P6.p, which constitutes a part of the ventral epithelial suface of the animal (J. White, personal communication) and gives rise to the primary vulval cells (Sulston and Horvitz, 1977), provides a significant but minor component of this ventral guidance system. Since the HSN axons grow across lateral epithelial cells and are not in obvious contact with the vulval precursor cells (which are located along the ventral side of the animal), P6.p could be producing a diffusible signal that attracts the HSN growth cone (Fig. 8A). Alter-natively, HSN filopodia, not visible in the light microscope, might contact P6.p and draw the HSN axons toward the VNC. Both chemotropic guidance (Lumsden and Davies, 1983, 1986; Tessier-Lavigne et al., 1988; Heffner et al., 1990) and contact between filopodia and ‘guidepost cells’ (Bentley and Keshishian, 1982; Taghert et al., 1982; Bent-ley and Caudy, 1983) are known to direct neurite out-growth.

The major component of the ventral guidance system for HSN-axonal outgrowth is provided by the genes unc-6 and unc-40, two of three genes required for the circumferential outgrowth of many axons (Desai et al., 1988; Hedgecock et al., 1990; McIntire et al., 1992). Hedgecock et al. (1990) proposed that these genes encode products involved in the interaction between the axonal growth cones and the epithe-lium over which they extend. If so, the vulval cell P6.p and the lateral epithelium might act together to define the ven-tral direction of initial HSN outgrowth. Alternatively, unc-6 and unc-40 might act within P6.p to control the outgrowth of the HSN axons. However, defects in ventral-directed HSN outgrowth are much more frequent in unc-6 and unc-40 mutants (only 10% of the HSN axons extend ventrally in the unc-6(e7) mutant; Desai et al., 1988; McIntire et al., 1992) than in vulvaless animals (about 75% of the HSN axons extend ventrally in laser-generated vulvaless animals; Table 1). Thus, unc-6 and unc-40 are unlikely to act only within P6.p to control ventral-directed outgrowth of the HSN axon, although they still could act both within P6.p and other cells, including those of the lateral epithelium. Our attempts to resolve whether vulval cells and the unc-6 gene product act in the same process (e.g., P6.p secretes the unc-6 protein) or in separate processes for ventral-directed growth of the HSN axons by analyzing HSN mor-phology in unc-6 mutants that lack vulval cells resulted in an ambiguous finding: the ventral-growth defect was slightly enhanced in these animals relative to the unc-6 mutants, but still incomplete.

In animals that fail to generate the primary vulval cells, the HSN axons do not defasciculate from the VNC at the vulva. Because our light microscopic techniques do not permit visualization of the HSN axons at the time they are grow-ing along the VNC, we do not know when this defascicu-lation normally occurs. Thus, either the primary vulval cells themselves or one or more of their precursor cells could be required for defasciculation. Similarly, we do not know if defasciculation occurs at the time of HSN axonal ougrowth or later, after the axon has passed the region of the vulva. As illustrated in Fig. 8C, vulval cells could directly affect HSN axonal outgrowth in this region by defining the paths along which the HSN axons grow. Alternatively, vulval cells could separate the HSN axons from most of the other VNC axons after the HSN axons have extended beyond the vulval cells. During the L4 stage, the vulval cells undergo a complex set of rearrangements to generate the vulva (Sul-ston and Horvitz, 1977; J. White, personal communication) and vulval morphogenesis might well be responsible for this defasciculation.

Either the primary vulval cells or their precursors are required for the HSN axons to branch. Since we have been unable to use light microscopy to study the HSN axons as they are growing after they enter the VNC, we do not know when or how the HSN axons form branches. In other organ-isms, branches can be formed by bifurcation of an extend-ing growth cone (Fujisawa, 1987; Raper et al., 1983a) or by sprouting either from just behind the growth cone (Harris et al., 1987) or from the side of a fully extended axon (O’Leary and Terashima, 1988). Thus, the HSN branch might form either as the HSN axon begins to grow anteri-orly in the VNC or after the HSN axon has grown beyond this region. The vulval cells or their precursors could induce branching, provide a permissive substrate for growth of the branch or stabilize the branch once it has formed (Fig. 8D).

Studies of neurite branching in other organisms have often found that growing branches lack obvious growth cones (Fujisawa, 1987; Harris et al., 1987; O’Leary and Terashima, 1988), which suggests a mechanistic difference between most forms of axonal elongation and branching. In this context, it is interesting to note that mutations in many of the genes (unc-6, unc-14, unc-33, unc-34, unc-40, unc-44, unc-51, unc-71, unc-73 and unc-76) required in general for axonal outgrowth in C. elegans (Desai et al., 1988; Hedgecock et al., 1985, 1990; McIntire et al., 1992) do not lead to obvious defects in HSN branching (G. G., unpublished observations).

Our observations suggest that branching is not an absolute requirement for HSN function. Assuming our immunocy-tochemical methods can detect all HSN branches, some of the HSN migration mutants that are missing HSN branches lay eggs normally. For example, approximately 25% of mig-1 animals are missing branches of both HSNs, but only 5% of mig-1 animals are egg-laying defective (Desai et al., 1988). Similarly, of the 45 egl-43 animals that we could score for HSN branching, only one contained an HSN branch; however, 10% of egl-43 animals are egg-laying competent. These results might explain why we have not found any mutants that are specifically affected in HSN branching in our collection of egg-laying defective mutants (Trent et al., 1983; Desai and Horvitz, 1989; C.D. and G.G., unpublished results).

The active role played by C. elegans muscles in synap-togenesis could explain why the HSN branch might not be required for HSN function. C. elegans muscles send processes known as muscle arms to the axons of motor neu-rons in both the ventral and dorsal nerve cords (White et al., 1976, 1986). These muscle arms apparently can regu-late their growth to find misplaced axons. For example, a defect in circumferential axonal outgrowth in unc-5 mutants results in the loss of the dorsal nerve cord because the axons of the motor neurons normally found in the dorsal nerve cord are displaced ventrally. Muscle arms of the dorsal body wall muscles that are normally innervated by axons in the dorsal cord regulate their growth to interact with these dis-placed axons in unc-5 animals (Hedgecock et al., 1990). It is possible that although most of the synapses between the egg-laying muscles and the HSNs are normally made on the HSN branches, when the HSN axons do not branch the egg-laying muscle arms regulate their growth and nonethe-less synapse with the HSN axons.

As illustrated in Fig. 8B, both vulval cells and the axons of the PVP and PVQ neurons are required for the left HSN axon to fasciculate with the axonal bundle of the left VNC. The requirement for vulval cells is interesting, since in the absence of these cells the left HSN axon grows over axons that are normally chosen as substrates for fasciculation. One explanation for how vulval cells might influence the path-way choice of the left HSN axon is that the growth cone of the left HSN prefers the right VNC bundle but the vulval precursor cells, which constitute the hypodermal ridge over which the left HSN axon must pass to enter the right side of the VNC, is not a permissible substrate for HSN axonal outgrowth (Fig. 8B). In other words, the vulval cells could actively inhibit the HSN growth cones. In vertebrates, sev-eral examples of substances that are inhibitory to growth cone extension have recently been described (for reviews, see Schwab, 1990; Walter et al., 1990). An alternative hypothesis is that vulval cells play an active role in HSN axonal pathfinding. For example, vulval cells might actively cause the left HSN growth cone to interact with axons in the left VNC bundle, by altering the properties of either the HSN growth cones or the axons on which they extend.

Our observations of HSN axonal morphology in animals missing PVP and PVQ axons indicate that even in an axon bundle as simple as the left VNC, which has only four axons in the vulval region, there are subgroups of axons that selec-tively fasciculate with one another. Although the left HSN axon fasciculates with the PVP and PVQ axons, it does not appear to interact with the AVK axon. That the left HSN axon must pass over additional axons within the right VNC bundle before reaching the right HSN axon (White et al., 1986) further suggests that the VNC is a complex structure with subdomains created by selective fasciculation.

White and his colleagues (1985, 1986) proposed that synaptic connectivity in C. elegans is determined largely by the ‘neighborhoods’ of axons. Ultrastructural analysis of the nematode nervous system revealed that synapses are made between adjacent axons en passant and that axons maintain their neighbors (adjacent axons) for long distances within axonal bundles. Since within a nerve bundle C. elegans neurons synapse with 23% of their neighboring axons (for highly adjacent axons; Durbin, 1987), the synaptic con-nectivity of a neuron is dictated to a great extent by the neighborhood its axon inhabits. How are these neighbor-hoods established? Our results support the hypothesis (White, 1985; White et al., 1986) that neighborhoods are generated by selective fasciculation. Analogous to the ‘labeled pathway hypothesis’ proposed to account for selec-tive fasciculation in grasshoppers (Raper et al., 1983b, 1984), selective fasciculation in C. elegans could be spec-ified by neighborhood-specific surface molecules that act as recognition markers for individual growth cones.

We are grateful to Scott Clark, Ed Hedgecock, Chris Li, James Manser and Michael Stern for kindly providing strains and mate-rials used in this work and to Laird Bloom, Scott Clark, Erika Hartwieg, Chris Li and Michael Stern for sharing unpublished data. We also thank Michael Basson, Martha Constantine-Paton, Erik Jorgensen and members of the Horvitz lab for comments on the manuscript. This work was supported by US Public Health Service research grant GM24663. G. G. was supported by post-doctoral fellowships from the Damon Runyon-Walter Winchell Cancer Research Fund and the Charles A. King Trust and US Public Health Service grant MH09926. C. D. was supported by US Public Health Service predoctoral training grant GM02787. H.R. H. is an Investigator of the Howard Hughes Medical Institute.