Remodeling of the leg sensory system during metamorphosis of the hawkmoth, Manduca sexta

Animals

M. sexta (L) were obtained from a laboratory culture reared on an artificial diet (Bell and Joachim, 1976) under a long-day photoperiod regimen (17 hours light and 7 hours dark) at 26°C and at approximately 60% relative humidity. Both chronologic and morphologic criteria were used for the staging of animals (Nijhout and Williams, 1974; Bell and Joachim, 1976; Reinecke et al., 1980; Tolbert et al., 1983; Consoulas et al., 1996). In summary, larval day 0 (L0), L1, L2, etc., represent the days of the last (5th) larval instar. Similarly, wandering day 0 (W0) signifies the first day of wandering, and W1–W4 represent the remaining days before pupation. Stages W2a and W2b are defined as the periods on day W2 before and after ocellar retraction occurs. Stage W4 is subdivided into four substages according to Truman et al. (1980). After pupation, stage pupal 0 (P0) indicates the day of the pupal molt, and P1–P18 are the next stages of pupal development.

Nerve staining techniques

1,1′-Dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate staining

Animals (n = 72) from the second day of the last larval instar (L2) were anesthetized on ice and placed ventral side up on plasticine. A small incision along the ventral midline was made without damage of any internal tissue. The prothoracic ganglion was exposed with the aid of a glass hook made from a Pasteur pipette. Small crystals of 1,1′-dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate (DiI; Molecular Probes, Eugene, OR) were deposited with the aid of a glass electrode within central areas of the ganglion, avoiding spread of the dye in the hemolymph. After the DiI deposition, the incision was sutured with the aid of surgical thread, and hemolymph leakage was stopped with a drop of adhesive tissue glue (Consoulas and Levine, 1997; Vetbond, St. Paul, MN). Oxidation of hemolymph was prevented by adding a few crystals of phenylthiocarbamide (PTU) to the wound. The rate of animal survival was approximately 80%. After the operation, the animals were allowed to develop under normal conditions and then were dissected in later larval and pupal stages.

Sectioning

The positions of the sensory neurons in the developing leg (stages W3, W4, and P2) were examined in serial longitudinal sections (n = 12). The legs were removed from the animals, fixed in alcoholic Bouin's solution for 2–3 days, embedded in paraffin (paraplast), and serially sectioned (10–12 μm). After deparaffinization and rehydration, the sections were stained with hematoxylin and eosin.

5-Bromodeoxyuridine labeling

To reveal the time course during which the sensory neurons of the adult legs appear as the result of the division of epidermal cells, 50 μg/g body weight of 5-bromodeoxyuridine (BrdU; Sigma) dissolved in distilled water was injected into the animals at specific developmental stages (from W0 to P5). The animals were then allowed to reach pupal stages P10 or P13, when all the axons of adult leg sensory neurons have reached the CNS but the epidermal tissue is still soft and easily sectioned. The prothoracic legs were then dissected out of the animals and fixed for 2 days in alcoholic Bouin's solution, embedded in paraffin, and sectioned. After deparaffinization, rehydration, and extensive washing in PBS and PBSX (0.1%) the DNA of the cells was denatured by treatment with 2 N HCl in PBS for 15 minutes. Nonspecific activity was blocked with 10% NGS in PBS for 30 minutes. The sections were then incubated for 2 hours in the primary antibody against BrdU (Becton Dickinson Immunocytometry Systems, San Jose, CA) diluted 1:100 in PBS with 5% NGS. After washing the sections in PBSX and PBS for 1 hour, they were incubated in goat anti-mouse secondary antibody diluted 1:200 (Cy3-conjugated; Sigma) in PBS for 1 hour (Consoulas et al., 1997; Consoulas and Levine, 1997).

The stained preparations were viewed with a confocal microscope (MRC-600 with a Nikon Optiphot-2 microscope [Nikon, Tokyo, Japan] and a Krypton/Argon laser light source [BioRad, Cambridge, MA]). By using the two shambling channels and dichromatic cubes (BioRad K1 and K2; excitation wave lengths, 488 nm and 568 nm, respectively), optical sections were recorded simultaneously through the depth of wholemount preparations for the one or two dyes used. In cases in which two dyes were used, the images were merged by using different pseudocolors (red for Cy3-conjugated streptavidin or DiI and green for streptavidin-dichlorotriazinylamino fluorescein). Images were prepared by using Confocal Assistant (Bio-Rad) and Corel 8 (Corel Corp., Ottawa, Ontario, Canada).

Electrophysiology

The animals were dissected as described above. The spontaneous activity of leg sensory neurons was recorded with a saline-filled glass suction electrode from the main nerve, IN2a. The extracellular signals were amplified with a differential alternating current amplifier (A-M systems, Everett, WA), stored on a Vetter PCM 8-channel recorder (A.R. Vetter Co. Inc., Rebersburg, PA), and subsequently transferred to a computer (acquisition sample rate, 10 KHz) for analysis using Data-Pac II software (Run Technologies, Laguna Hills, CA).

Larval leg sensory system

The larval prothoracic legs in Manduca sexta are small, segmented, but poorly articulated structures. The trochanter is fused with the femur, and the two coxae are fused with one another. The overall organization of the sensory system of the distal leg segments, as revealed after filling the nerves with biocytin, is shown in Figures 1D and 2A. Two main nerves (IN2a and IN1b; Eaton, 1974) enter the fused trochanter-femur segment. Nerve IN1b is a motor nerve innervating the tibial extensor muscle (Consoulas et al., 1996). The other leg nerve, IN2a, gives off five secondary nerve branches: Branches 2a5, 2a6, 2a8, and 2a9 are purely sensory nerves. Branch 2a7 is a motor nerve carrying the axons of the motoneurons that innervate all of the femoral and tibial muscles (Consoulas et al., 1996). Sensory organs located at the anterior region of the leg are innervated by neurons with axons in nerve branches 2a5 (trochanter and femur) and 2a8 (tibia, tarsus, and pretarsus), whereas sensory organs of the posterior region of the leg are innervated by branches 2a6 (trochanter and femur) and 2a9 (tibia and tarsus; Figs. 1D, 2A).

Femoral, tibial, and tarsal leg segments are covered by a number of proprioceptive sensilla of different size (Kent et al., 1996). Hair-like sensilla and spines are innervated singly by sensory neurons with cell bodies measuring 10–20 μm (Figs. 1D, 2B,C). Single sensory neurons supply the eight campaniform sensilla that are located at the trochanter-femur boundary (n = 3), at the tibia (n = 1), the tarsus (n = 3), and at the pretarsus (n = 1; Kent et al., 1996). Multipolar joint receptors (MJRs; Coillot and Boistel, 1968) were found in the femur (n = 5) and the tibia (n = 3). The fine dendrites of the MJRs arborize extensively not only close to the joints (Fig. 2E) but also in almost every part of the leg segments, thus forming an epidermal plexus (Fig. 2B,C, arrows). A structure resembling a putative tension receptor (Theophilidis and Burns, 1979) attaches to the posteromedial bundle of the tibial flexor muscle (Fig. 1D). Its axon enters nerve 2a8, and its dendrites, which differ characteristically from motor terminals, project to the most proximal part of the muscle fibers (Fig. 2D). In the femur and tibia, clusters of sensory neurons are associated with sensory structures resembling chordotonal organs (Fig. 1D). In the femur, groups of sensory neurons are associated with two connective strands (Fig. 1D, 2F). One of the strands (associated with six to nine neurons) is attached to the apodeme of the tibial extensor muscle. The other strand (associated with two to four neurons) spans the femorotibial joint and is attached laterally to the tibia (Fig. 1D). In the tibia, five neurons are associated with a strand that spans the tibiotarsal joint to be attached to the lateral tarsus (Figs. 1D, 2A, see inset).

Remodeling of the larval leg sensory system

The adult leg develops within the larval leg. In M. sexta, like in other lepidoptera, the epidermis of the adult leg develops from cells of the larval leg epidermis that constitute the differentiation center (Kim, 1959; Miles and Booker, 1993). These cells begin to proliferate during the last larval instar and radiate out under the cuticle of the larval legs to form a new, highly folded epidermis. By early stage W2, the imaginal epidermis encloses the larval neuromuscular and tracheal systems and begins to retract from the larval cuticle (Fig. 3A). Biocytin staining of nerve 2a revealed a larval-like sensory pattern of innervation. The cell bodies of the sensory neurons innervating cuticular proprioceptors were detached from the larval cuticle and became associated with the retracted imaginal epidermis (Fig. 3A–D). The dendrites of these neurons elongated considerably to innervate their sensory organs (Fig. 3D, see arrows). Extracellular recordings from the main leg nerve (2a) during this stage of development revealed trains of action potentials in response to mechanical displacement of hair sensilla or spines (Fig. 6A), which suggests that dendritic elongation does not prevent transduction. In addition, tactile stimuli applied to a single hair or to a group of hairs elicit reflex responses similar to those recorded in the intact larval leg (stage L2, not shown; see also Kent and Levine, 1988), suggesting that the sensory neurons maintain proper connections with their targets within the CNS and that the underlying circuits remain functional. During this period, the number and position of the sensory neurons supplying the femoral and tibial chordodonal organs remained unchanged (Fig. 3B,C), but their apodemes were detached from the larval leg cuticle (not shown).

The imaginal leg completes its retraction from the larval cuticle and begins to grow and elongate considerably during the last 2 days of the fifth larval instar. Biocytin staining of the peripheral nerves revealed that the four primary sensory pathways (2a5, 2a6, 2a8, and 2a9) persist within the imaginal leg. However, the number of the sensory neurons decreases substantially (Table 1). For example, of the 21 neurons supplying external sensory organs in the femur of an early fifth instar larval leg (stage L2), only 13 ± 3 sensory neurons remain by stage W2b, and only 10 ± 2 neurons remain by late stage W4 (Table 1; also compare Figs. 1D–4C). During the last 2 larval days, the sensory neurons are not associated with external proprioceptors, and their elongated dendrites degenerate (Figs. 4A–C, 5A– C). The sensory neurons are concentrated in groups in the medial region of the imaginal leg segments (Figs. 4A–C, 5A–C). However, two of the femoral chordotonal organ sensory (FCO) neurons migrate toward the femorotibial border (see Figs. 4A–C, asterisk; Fig. 5B). Similarly, the axons of the tibial chordotonal organ (TiCO) neurons, together with the tarsal/pretarsal group of neurons, elongate considerably to occupy the distal-most region of the developing leg (Fig. 4A–C, Fig. 5A, inset). Cross sections from late larval (Fig. 5D) and early pupal (not shown) preparations revealed that the cell bodies of the sensory neurons that supply exteroceptors in the larva are located at the outer cell layer of the imaginal epidermis. Although, during this time of development, the remaining larval sensory neurons are not associated with sensory organs and do not serve any function, they can be active spontaneously (Fig. 6B). It was not possible to stimulate specific neurons, but spontaneous activity and bursts of activity were apparent in response to tactile stimuli applied to the epidermis in the vicinity of the cell bodies (Fig. 6C).

After pupation, the imaginal leg begins to unfold, and, by stage P1, it reaches its final length. Although fewer larval sensory cells were present in retrograde biocytin stainings of the peripheral nerves of stage P1 and P2-early preparations (Table 1), all of the major sensory branches were apparent by this stage of development. Nerve branch 2a6 had shifted to the imaginal trochanter. It carried the axons of two larval sensory neurons that are designated here as larval sensory neuron 1 and 2 (LS1 and LS2; Figs. 4D, 5E). Six larval FCO neurons were present in the proximal imaginal femur (Fig. 5E). Their axons, together with the axons of the other two chordotonal organ neurons that migrate close to the femur/tibia border (designated here as larval sensory neurons 3; LS3; Figs. 4D, 5E), run through two collateral branches (2a5a and 2a5b) of nerve 2a5. Three larval sensory neurons survive larval-to-pupal transition in the imaginal tibia (Fig. 4D, circled area; Fig. 5F,G). Two groups of three larval neurons each, together with the larval TiCO neurons, are located at the tip of the developing adult leg (Figs. 4D, 5H,I). Their axons project to the CNS through branches 2a8 and 2a9. Similar to what occurs in late larval stages, several of these neurons are active spontaneously and can respond to tactile stimulation of the nearby epidermis (Fig. 6C).

Birth of adult sensory neurons

To determine the birth dates of new-adult sensory neurons, BrdU was injected into late larval stage (W0–W4) and early pupal stage (P0–P6) animals. The injected animals were left to develop normally to stages P10 or P14, then the legs were dissected, sectioned, and processed for BrdU immunocytochemistry. BrdU incorporated by the nuclei of dividing epidermal cells should be diluted due to many rounds of division. In contrast, BrdU still should be detectable in the nuclei of adult sensory neurons if the latter derive from precursors undergoing relatively few divisions. Although BrdU incorporation by the nuclei of epidermal cells and by myonouclei still was detectable in late pupal stages, the strongly labeled, large nuclei of the adult sensory neurons could be distinguished easily (Fig. 7). Neurogenesis within the legs occurred in a proximal-to-distal fashion. BrdU injections into animals (n = 15) at early pupal stages (from P0 to P2-early) and dissection 8–10 days later (stage P10) revealed labeled nuclei of sensory neurons in the trochanter, femur, and tibia and, occasionally, in the tarsus (Fig. 7A,B). BrdU injections into animals (n = 17) at stages P2–P5 revealed BrdU incorporation by the nuclei of sensory neurons in the tibia and tarsus (Fig. 7C). BrdU injections into late fifth instar larvae or into pupae later than stage P5 did not result in any labeling (data not shown). Because BrdU in the late larva or pupa remains available for approximately 2 days (unpublished data; L.A. Oland, personal communication), the above findings suggest that adult sensory neurons derive from epidermal cells that undergo a single or a few divisions between stages P2 and P5.

Development of the adult pattern of sensory innervation

The different classes of sensory organs of the adult prothoracic legs and their underlying sensory neurons have been described previously (Kent and Griffin, 1990). The establishment of the innervation pattern is described here. During the first 3 days of the pupal stage, the adult leg epidermis is separated gradually from the pupal cuticle, and the first signs of segmentation appear by stage P3. The imaginal leg becomes fully segmented during the next 3 days. The main sensory pathways are established within 1 day (stages P2-late and P3; Fig. 8A,B), and during the next 3 days (until stage P6), most of the adult sensory neurons grow axons into the CNS (Fig. 8C). During these early pupal stages, the sensory organs have not yet differentiated, making it difficult to identify specific neurons, but hair-plate and chordotonal organ neurons can be distinguished easily due to their unique appearance and arrangement. Another caveat is that the data presented here are based solely on biocytin fills. During the critical time of early pupal development, in which most of the larval neurons degenerate and new-adult neurons differentiate, there is always the possibility that the absence or relocation of neuron(s) may be due to the failure of the dye to incorporate into all of the neurons. This problem could be resolved by using, in addition to diffusable markers, other molecular tools, such as antibodies against neuronal proteins, that are not available currently for Manduca.

Biocytin backfills of nerve IN2a revealed that the axon of the first adult sensory neuron in the developing leg reaches the CNS by middle stage P2. Its cell body is situated at the distal femur (not shown) adjacent to the LS3 neurons, and its axon runs together with the LS3 axons (nerve 2a5b). By late stage P2, a single neuron in trochanter and three groups of neurons appear almost simultaneously along the medial axis of the femur (Figs. 8A, 9A). The most proximal group (group I) consists of five to seven neurons. Together with the persistent larval FCO neurons, group I neurons participate in the formation of the FCO in the adult leg (Consoulas, Rose, and Levine, in preparation). The second group of neurons (group II) is comprised of four cells with axons joining the preexisting nerve branch 2a5 (Figs. 8A, 9A). The most distal group (group III) is comprised of three neurons. Their axons travel toward and fasciculate with the axons of group II neurons (Figs. 8A, 9A). At this stage of development, LS3 neurons migrate distally to the proximal tibia (Figs. 8A, 9A), leaving behind the new-adult sensory neuron at the distal part of the femur (Fig. 8A; Fig. 9A, arrow). This axon then becomes the precursor of the new-adult sensory pathway 2a6a.

By stage P3, newly emerged sensory neurons could be stained in the trochanter, femur, and tibia of the imaginal leg with biocytin introduced into the CNS (Fig. 8B; Fig. 9B, inset). In the trochanter, the axons of the posterolateral hair plate and anterolateral 1 hair plate neurons project toward and fasciculate with the axons of the preexisting LS1 and LS2 neurons to form two adult sensory nerve branches (designated here as 2ab and 2ac). The medial and anterolateral 2 hair plate neurons and other sensory neurons appear with a short delay, and their axons join 2ab and 2ac branches (Figs. 8B, 9B). New-adult sensory neurons that are located at the posterolateral, lateral, and anterolateral parts of the femur project axons through nerve 2a6a. Nerve 2a5a carries the axons of the persistent larval and new sensory cells that supply the FCO, together with the axons of the sensory neurons located in the medial and anteromedial regions. The axons of the sensory neurons located in the posteromedial part of the femur join nerve branch 2a5b, which was established previously by the axons of LS3 neurons (Figs. 8B, 9B). In the tibia, two larval sensory neurons persist until this time of development (Figs. 8B, 9B, encircled areas). Two pairs of sensory neurons, together with the larval tibial chordotonal neurons, are apparent at the distalmost part of the leg (Fig. 9B, inset).

During the next 4 days of pupal development, more sensory neurons differentiated within the developing leg (Fig. 10). The full compliment of adult sensory neurons and most of their target-sensory organs became apparent by stage P8. A total of seven sensory nerves carried the axons of all of the sensory neurons (Fig. 10A–C). In the trochanter, hair plate, campaniform, and hair sensilla, neurons projected their axons through nerves 2a6b and 2a6c (Fig. 10B). It is interesting to note that the larval sensory neurons LS1 and LS2 that were apparent in the previous stages were not stained in any biocytin fill of the peripheral nerves of later stages (Fig. 10A,B,H). In the femur, sensory axons projected through three nerves; the preexisting larval nerve (2a5b), which was established by the LS3 axons, and the two new sensory nerves (2a5a and 2a6a), which were established by the axons of early adult sensory neurons (Fig. 10A). In the tibia, tarsus, and pretarsus, adult sensory axons projected through one of the two preexisting larval nerves, 2a8 and 2a9 (Fig. 10E–G). This pattern of sensory innervation remained unchanged throughout the pupal stage and in the adult.

Biocytin fills in late pupal stages and in the adult revealed that, in contrast to LS1 and LS2 larval neurons, LS3 neurons survive metamorphosis to innervate the tibiotarsal chordotonal organ (Fig. 10D,I). Six larval FCO neurons survived metamorphosis and, together with new neurons, comprised the new organ in the adult leg (Fig. 10A; Consoulas, Rose, and Levine, unpublished observations). The larval TiCO neurons migrate to the last tarsal segment to become the tarsal chordotonal organ. Finally, the six larval sensory neurons (two in the tibia and four in the tarsus) persist into the adult stage to innervate hair sensilla.

Assessment of persistent sensory neurons

Biocytin stainings of the peripheral nerves in late larval preparations revealed that, despite the degeneration of the larval leg, many of the larval sensory neurons become associated with the imaginal leg and survive until the end of the last larval instar. After pupation, most of the larval sensory neurons degenerate but, few of them survive to the adult. Although biocytin fills of these neurons at every stage during the larval-to-pupal transition strongly support the conclusion that these larval neurons persist to the adult, there is always the possibility that these larval neurons are replaced by similarly located adult neurons that were not detected in BrdU preparations. To ensure that these neurons are larval and not newly born adult sensory neurons, the lipophilic vital dye, DiI, was injected into a central area within the prothoracic ganglion in early fifth instar larvae (stage L2). After the injection, the animals were left to develop normally to late larval or early pupal stages. In early stage W3 preparations, DiI labeling was detected in some of the sensory nerve branches and in almost all of the sensory cell bodies (Fig. 11A). In late larval or early pupal preparations (W4 and P3), DiI had almost disappeared from nerve branches but remained in the sensory cell bodies (Fig. 11E,F). For these stages, to ensure that DiI staining was localized in the cell bodies of neurons, animals that were injected with the dye at stage L2 were dissected at late larval (stage W4) and early pupal (stage P3) stages, and the peripheral nerves of the same preparations were then backfilled with biocytin (Fig. 11B,C). Colocalization of DiI and biocytin in most of the sensory cell bodies was apparent (Fig. 11D). Although definite proof that some larval neurons persist to the adult could be gained by detecting DiI staining in the adult leg, unfortunately, DiI disappears completely from preparations older than pupal stage P3. Nevertheless, the above observations, in combination with the biocytin fills and the BrdU labeling experiments, strongly support the notion that these larval sensory neurons persist to the adult stage (see below).

Degeneration of the larval sensory neurons

Most of the larval sensory neurons degenerate within a period of 5 days (stage from W2b to stage P3; see Table 1) during the larval-to-pupal transition. The neurons do not degenerate immediately after the loss of contact with their associated sensory organs but degenerate a few days later. LS1, LS2, and some tibial and tarsal neurons degenerate 5 days later. These neurons may persist for this additional time to facilitate the establishment of new sensory branches (see below), but most do not appear to serve this role in the developing leg. At present, there is not a suitable explanation for the delayed degeneration of these larval sensory neurons.

Larval sensory neurons that supply exteroceptors (hair and campaniform sensilla and spines) remain functional for an additional day after the formation of the imaginal leg. They retain their mechanotransduction capability by growing elongated dendrites that span the distance between the cell bodies, which are localized in the retracted imaginal epidermis, and their target-sensory organs, which are located at the larval cuticle. Dendritic elongation of sensory neurons supplying mechanosensory hairs of the prolegs during the molt also has been shown (Jacobs and Weeks, 1990). Not only the sensory neurons but also circuits within the CNS remain functional, because the legs can exhibit larval-like reflexes in response to sensory inputs. It remains unclear why limited leg motor function is preserved for 2 additional days after the completion of the puparium cell (Dominick and Truman, 1986).

During the prepupal stages, the axons of the larval sensory neurons retract, which results in the grouping of the cell bodies within central regions of the imaginal segments. Subsequently, the axons of particular groups of neurons (LS3, TiCO, and tarsal/pretarsal neurons) undergo a phase of expansion as the imaginal leg grows and elongates considerably during the last larval and first pupal days. These initial phases of axonal retraction and expansion may represent passive responses to the shrinkage and later expansion of the developing imaginal epidermis. Alternatively, active migration may account for the large-scale relocation of the LS3 neurons. Similarly, axonal processes of leg motoneurons undergo similar phases of retraction and expansion as they lose their larval target muscles and become associated with imaginal tissue during elongation of the imaginal leg (Consoulas et al., 1996).

Persistence of larval sensory neurons and their role in the adult

Small numbers of larval sensory neurons that occupy strategic positions in the imaginal segments survive the larval-to-pupal transition. They participate in the establishment of the adult sensory pathways by providing a route to the CNS for the growing axons of the adult neurons. Evidence that larval sensory neurons persist in the pupa was based on 1) biocytin fills of peripheral nerves at frequent intervals; 2) DiI labeling of the sensory system of intact, early last instar larvae and examination for stained neurons of the same animals at later developmental stages; and 3) electrophysiological recordings of spontaneous sensory activity. These approaches, together with the fact that no adult neurons are born before stage 2 in the pupa, leave little doubt that the neurons identified in this study are persistent larval sensory neurons. Other examples of larval sensory neurons that persist into the pupal stage have been described in Manduca. Specific larval sensory neurons in abdominal segments survive to innervate hairs within the gin traps (Bate, 1973; Levine et al., 1985; Levine, 1989). Similarly, the abdominal stretch receptors (Tamarkin and Levine, 1996) as well as other larval thoracic and abdominal sensory neurons survive metamorphosis (Consoulas, unpublished observations). Larval sensory neurons associated with leg and eye imaginal discs also persist through metamorphosis in diptera (Jan et al., 1985; Tix et al., 1989a,b; Lakes-Harlan et al., 1991b). Also, an array of larval sensory neurons persist in each segment of Drosophila to become functional adult neurons (Shepherd and Smith, 1996). Thus, the survival of sensory neurons, like many motor and interneurons (see above), may be a general feature among holometabolous insects.

Do larval sensory neurons survive metamorphosis because they are required for adult behavior? In Manduca, gin trap sensory neurons become capable of mediating a pupal-specific defense reflex (Levine et al., 1985, Waldrop and Levine, 1989, 1992), and the abdominal stretch receptor organ survives metamorphosis to provide information about the length of the abdomen to persistent motoneurons and interneurons (Tamarkin and Levine, 1996). In the dipteran, Phormia regina, two sensory neurons that are associated with the Keilin's organ (mechanoreceptor) in the larva survive metamorphosis to innervate campaniform sensilla in the adult (Lakes-Harlan et al., 1991a,b). Similarly, in Drosophila, multidendritic and stretch receptor neurons survive metamorphosis to perform the same function in the adult (Shepherd and Smith, 1996; Smith and Shepherd, 1996). Larval FCO, TiCO, LS3 neurons and six tibial and tarsal neurons (that supply campaniform sensilla) survive to adulthood (this study; Consoulas, Rose, Levine in preparation) to serve the same function and participate in the locomotor behavior of the adult (Johnston and Levine, 1996).

Role of persistent larval sensory neurons in the establishment of the adult sensory pathways

A common feature of the developing nervous system in a broad range of organisms is that later developing neurons fasciculate with certain nerve pathways that are established by earlier developing neurons, known as pioneers, to reach their target area within the CNS or in the periphery (LoPresti et al., 1973; Bate, 1976; Raper et al., 1983; Goodman et al., 1984; Eisen et al., 1986; Kuwada, 1986; Dodd et al., 1988; Ghosh et al., 1990; Peinado et al., 1990; Stanier and Gilbert, 1990; McConnell et al., 1994; Gan and Macagno, 1995; Jellies et al., 1995). During metamorphosis, newly emerged sensory neurons arise far from the CNS and, unlike during embryonic development, must span distances that, in the Manduca leg, are in the range of centimeters.

The data presented here suggest that persistent sensory neurons may provide guidance cues for adult sensory neuron growth and, thus, may act as pioneers. Adult sensory axons always fasciculate with preexisting larval sensory nerves. In other holometabolous insects, persistent neurons may guide adult sensory axons to the CNS. In dipterans, for example, larval neurons that are associated with the leg (Tix et al., 1989a; Lakes-Harlan et al., 1991b) and eye (Tix et al., 1989b) imaginal discs, and neurons that innervate internal proprioceptors in thoracic and abdominal segments (Shepherd and Smith, 1996) survive metamorphosis and provide an axonal pathway to the CNS for the newly generated sensory neurons. Persistent neurons may be necessary for guiding the new axons, or new axons simply may prefer to use a preexisting larval nerve for navigation.

Flies homozygous for mutant alleles of the disco gene have defects as the larval Bolwing's nerve, which may explain the failure of the adult retinular neurons to innervate the optic lobe (Steller et al., 1987). Similarly, in Tenebrio, ablation of larval sensory neurons disrupts the growth of adult sensory neurons (Breidbach, 1990). In contrast, the pattern of nerve pathways formed in ectopic leg discs of Antennapedia flies seems to be unaffected by the absence of larval neurons (Tix et al., 1989b), and, in the absence of pioneer neurons, peripheral nerves in the embryonic grasshopper limb can be established successfully by neurons that appear later (Keshishian and Bentley, 1983b).

Newly emerged neurons have several choices and must make specific decisions. New adult axons in the tibia and tarsal segments, for example, grow along larval nerves 2a8 and 2a9 (Figs. 4, 9). Similarly, adult axons in the tibia and femur fasciculate with the axons of LS3-persistent neurons (nerve 2a5b; see Figs. 4, 9). In these cases, new sensory neurons that appear in the vicinity of larval sensory nerve branches make the decision to use these larval pathways as a scaffold for reaching the CNS. In contrast, nerve branches 2a5a and 2a6a (Fig. 9) are newly generated branches. They are established not by larval axons but by the axons of a few adult sensory neurons that appear earlier than the follower sensory neurons. These early adult sensory neurons, therefore, may play the same guiding role as the persistent larval sensory neurons. The formation of branches 2a6b and 2a6c is of particular interest. New sensory axons in the trochanter grow toward and fasciculate with the axons of LS1 and LS2 neurons, which disappear immediately after the formation of the adult pathways. These larval neurons have many of the characteristics of the embryonic pioneer neurons (Bate, 1976; Keshishian and Bentley, 1983a). They are present in the imaginal trochanter earlier than the adult neurons, their axons serve as a substrate for follower adult neurons, and they eventually degenerate. The only feature that differentiates these neurons from pioneer neurons is that, prior to metamorphosis they supply functional mechanoreceptors.

In conclusion, this study describes the establishment of the sensory pathways in the adult leg. It can be used as a tool in designing experimental perturbations that may allow a better understanding of the mechanisms underlying axon outgrowth and neural circuit remodeling during postembryonic development.