Neuron,

Vol. 9, 1001-1011, December,

1992, Copyright

0 1992 by Cell Press

Precocious Pathfinding: Retinal Axons Can Navigate in an Axonless Brain Elsa Cornel and Christine Holt Department of Biology University of California, San Diego La Jolla, California 92093-0322

Summary The developing axons of retinal ganglion cells follow a stereotyped trajectory through the diencephalon to the optic tectum. In Xenopus, this trajectory closely parallels that of a preexisting fiber tract, the tract of the postoptic commissure (TPOC). This tract comprises part of the early CNS scaffold and has been proposed to play a critical role in guiding the later growing optic axons. We have tested this possibility using heterochronic and xenoplastic transplants of eye primordia to force optic axons to enter the brain before scaffold tracts have arisen in the forebrain. We show that optic axons can navigate appropriately in the absence of the TPOC or any other axons, indicating that axonal pathfinding cues are present in the axonless neuroepithelial sheet. We suggest that molecularly distinct heterogeneities within the neuroepithelium are used for pathfinding by early and late developing axons alike in normal development. Introduction The complex pattern of axon tracts in the adult CNS matures from an initially simple set of “scaffold” tracts that develop early in embryogenesis (Raper et al., 1983; Bastiani et al., 1985; Easter and Taylor, 1989; Chitnis and Kuwada, 1998; Wilson et al., 1990). New axons that develop after the genesis of the scaffold usually add to the preformed tracts rather than forge new ones. Particular axon tracts appear in the brain in a well-defined sequence that is strikingly stereotyped between different individuals and, furthermore, is similar between different species (Thomas et al., 1984; Chitnis and Kuwada, 1990; Wilson et al., 1990; Easter and Taylor, 1989). This conserved pattern of CNS development, with pioneer axons apparently laying down the framework of the mature fiber tracts, suggests the possibility that the axons forming the early scaffold express molecular cues that are key to the pathfinding of later growing axons. Indeed, experimental evidence in several systems has shown that ablations of pioneer axons or tracts cause the later growing axons that normally follow them to halt their growth or to veer off course (Raper et al., 1984; Kuwada, 1986; Klose and Bentley, 1989; Ghosh et al., 1998; Pike et al., 1992). This is not always the case, however; for in some systems ablation of the pioneers does not disrupt subsequent pathfinding (Keshishian and Bentley, 1983; Schubiger and Palka, 1985; Tix et al., 1989; Chitnisand Kuwada, 1991; Pike et al., 1992). Since no single rule clearly applies, even within the same

system (Kuwada, 1986; Chitnis and Kuwada, 1991; Pike et al., 1992), the role that pioneer tracts play in guiding particular follower pathways needs to be tested case by case. Experiments in Xenopus involving 9o” rotations of pieces of the presumptive optic tract before optic axons have extended into it have revealed that local cues within the neuroepithelium, rather than long range cues from the target optic tectum, are responsible for guiding the direction of retinal axon growth through the diencephalon (Harris, 1989; Taylor, 1991). These cues could reside on the surfaces of the cells that comprise the neuroepithelium, or they could be expressed on a scaffold fiber tract that arises earlier than the optic axons. The development of the optic tract of Xenopus and zebrafish is preceded by an earlier forming longitudinal tract, the tract of the postoptic commissure (TPOC), which pioneers a dorsoventral trajectory through the diencephalon (Easter and Taylor, 1989; Chitnis and Kuwada, 1998; Taylor, 1991). Indeed, recent experimental evidence reveals that the TPOC becomes rotated when pieces of the presumptive optic tract neuroepithelium are rotated (Taylor, 1991). These observations have led to the proposal that the TPOC provides essential guidance information to optic axons in normal development (Easter and Taylor, 1989; Taylor, 1991). To test directly whether scaffold tracts, particularly the TPOC, provide guidance cues essential for optic axon navigation, we have used heterochronic and xenoplastic transplants to force optic axons to grow precociously into the embryonic brain, before the development of any other axon tracts. Our results show that these axons can navigate to the optic tectum in the absence of the TPOC or any other pioneering tracts in the brain and thus suggest that guidance information is available in the early neuroepithelial sheet. Results A Scaffold Tract, the TPOC, As a Possible Substrate for Optic Axons The initial time of appearance of the TPOC in normal embryos was determined using an a-acetylated tubulin antibody (a-AT; Piperno and Fuller, 1985) that selectively labels the axons of differentiating neurons (Chitnis and Kuwada, 1990; Wilson et al., 1990). Comparison of the staining obtained with a-AT versus Dil (l,l’-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate) and HNK-1 (Chitnis and Kuwada, 1998; Wilson and Easter, 1991; Taylor, 1991; C. H., unpublished data) and from immunolabeling with a-AT of horseradish peroxidase (HRP)-filled retinal axons shows that acetylated tubulin is present in most axons at the early stages of axonogenesis and that a-AT labeling extends close to the base of the growth cone at

NWKT 1002

Figure 1. Development mal and Hydroxyurea

Tel

\POC

the distal tip of the axon. The TPOC begins to form at stage 25, approximately 25 hr after fertilization and approximately 7 hr after neural tube closure. At this time, it can be seen as a thin tract comprising 1-4 axons running dorsoventrally in the outer wall of the diencephalic vesicle (Figure 1; see also Taylor, 1991). It courses dorsally along the caudal edge of the optic stalk, then veers caudally into the tegmentum where it joins the previously formed ventrolateral longitudinal tract in the mesencephalon. At this stage of development, the TPOC is the only fiber tract in the lateral wallof thediencephalicvesicle.Thecell bodieswhose axonscontributetothistract stain positivelywith a-AT at stage 24/25 (Taylor, 1991; this staining subsequently disappears from the cell bodies) and can be seen dispersed along the length of the tract. W ith time more axons are added, and the tract elongates and broadens. Optic axons first begin to enter the presumptive optic tract at stage 33/34, 8-10 hr after the TPOC has begun to form. Their developmental timetable, pathway, and behavior have been previously characterized (Holt and Harris, 1983; Holt, 1984,1989; Harris et al., 1987; Harris, 1989; Taylor, 1991). During the period of initial optic tract development, from stage 33134to 37138, the optic fibers can be visualized in relation to the other axon tracts in the brain with double immunofluorescence (see Figures4Aand 4B). The optic tract forms along the anterior edge of the TPOC, following a course that parallels it until the two diverge in the dorsal diencephalon. Confocal microscopy on the leading axon tips in the mid optic tract shows that the growth cones of retinal axons commonly extend in TPOC-free territories (see Figure

of theTPOC (HU)-Treated

in NorBrains

Anti-acetylated tubulin staining in normal (A and B) and HU-treated (C and D) Xenopus brains. (A) Stage 25 brain showing the TPOC, which begins to form in the diencephalon (Di) at this stage. (B) By stage 28 (approximately3 hr later), theTPOC is comprised of more fibers and can be seen coursing from the postoptic commissure (POC) ventrally through the diencephalon to join the ventrolateral tract (VLlI in the tegmentum. New tracts and commissures, such as thedorsoventral diencephalic tract (DVDT) and the ventral tegmental commissure (VTC), begin to form by this stage. (C) Stage 25 HU-treated brain shows delayed fiber tract development. The TPOC is not present, and the scant few fibers in the rhombencephalon indicate the beginning of the VLT. (D) Stage 28 HU-treated brain, theTPOC begins development at this time andcomprisestwoorthreefibers.Embryos were incubated in 20 m M HU from stage IO-12 onward. An HRP-conjugated secondary antibody was used to visualize the a-AT staining of whole-mount brains, and cameralucidadrawingsweremadeusinga16x objective. Bar, 100 pm.

4B). Stereoimaging (data not shown) provides information about axon fasciculation and reveals that retinal axons fasciculate with TPOC axons along only short stretches of their length and that growth cones together with leading portions of the axon are often not in direct contact with theTPOC or any other a-ATlabeled axons (see Figure 4B). Old to Young Transplants: Precocious Pathfinding To discover whether optic fibers can pathfind at times earlier than normal, optic axons were made to invade the brain several hours earlier than normal through heterochronic eye transplantation (class 1 experiment illustrated in Figure 2). Surgical ablation of the cells giving rise to the TPOC was not a possible strategy for eliminating this fiber tract because the TPOC fibers originate from cells that are widely dispersed through the diencephalon and surgery in this area would severely disrupt later growing axons. Eyes were excised from donor embryos at the beginning of retinal axonogenesis at stage 28 and transplanted to younger host embryos in place of host eyes at early stages of optic vesicle formation, stages 18-21, giving rise to a donor-host time disparity of 6-9 hr. At varying time intervals following transplantation, when host embryos ranged from stages 24 to 35/36, the projections from the grafted eyes were labeled with HRP and visualized in whole-mount brain preparations. Of 58 transplant cases in the class 1 experiment, 47 sent axon projections into the brain (Table 1). In the other 11, either the transplanted eyes did not extend projections into the CNS or the projections were less than 50 urn in length and were consequently too short

Precocious

Pathfinding

1003

Figure 2. Schematic Summary Classes of Experiments

HOSTS

Axolotl

to assay for directional orientation. The lack of a projection could be explained by several possibilities; for example, the HRP filling step may have been unsuccessful, or the grafted eye may not have healed sufficiently to support axon growth across the host-donor junction. The positive cases extended projections that ranged from 50 urn to 358 urn in length (from the ventral optic tract to the optic tectum), and 94% of these were appropriately oriented within the diencephalon and exhibited directed growth toward and into the optic tectum (Figures 38 and 3C). Their trajectories through the diencephalon were similar to those of normal and control optic tracts. The terminals of precocious projections in the tectum appeared dispersed and disorganized, suggestive of a lack of topography within the target (compare Figure 3A and 3B). When the host brain was at stage 32, the stage at which optic axons in normal embryos have just reached the midline chiasm (Holt, 1989), precociously growing optic axons from heterochronically transplanted eyes commonly formed a complete pathway to the optic tectum (Figure 3C). Complete fiber projections to the tectum were seen as early as stages 28 and 29130 (n = 3). This illustrates that optic axons are able to navigate appropriately in the CNS and can recognize the target tectum at least 6-8 hr earlier than normal. Thus, the necessary pathway guidance cues and target “stop” cues must be available at these earlier times. A small fraction of these projections showed abnormal fiber tracts besides the normal projection (Table

Table 1. Number

of Appropriately

Oriented

Trajectories

in the Three

In all experiments, stage 28 Xenopus embryos were used as donor embryos, and host eye rudiments were excised before being replaced by grafted eye primordia. (Class 1) Stage 28 eye primordia were transplanted into young Xenopus embryos ranging in age from stages 18-21. (Class 2) Host Xenopus embryos were raised in 20 m M HU from stages IO-12 onward. At stage 21, these embryos received grafts of normal stage 28 eye primordia. (Class 3) Stage 28 Xenopus were grafted xenoplastically into stage 22-28 axolotl embryos. Bar, 1 mm.

1). One extended a projection dorsally in the diencephalon, and 2 sent projections into the telencephaIon from the mid optic tract. In 11 cases, HRP-labeled projections were detected in the ventral and rostra1 telencephalon, which seemed to join the optic fibers ventrally at the chiasm. These were judged to be not of optic origin but rather due to inadvertent application of HRP to the nasal epithelium during the eye filling procedure. (Control experiments demonstrated that these telencephalic projections could be labeled simply by applying HRP to an intact embryo’s nose.) Sometimes the optic nerve entered the brain at a location that was shifted from normal due to imperfect positioning of the grafted eye stalk. Projections from these eyes made normally oriented projections in accordance with data from previous studies on ectopically transplanted eyes (Harris, 1986). To determine whether the precocious axons entered the CNS before the TPOC had arisen, projections were examined at short time intervals following transplantation. The earliest time that optic axons from transplanted eyes were found to enter the CNS was stage 25, the very stage at which the TPOC begins to form. This finding is consistent with the possibility that the pioneering tract is, indeed, important for optic axon pathfinding. There are, however, two alternative possibilities. First, it could be that the entire neuroepithelium is not permissive to axon growth before this stage of development. Second, it could be that the period of time it takes for the donor eye to heal and send out its first axons into these young host

Classes of Transplant

Experiments

Type of Experiment

Number of Transplants

Projections >50 Pm

Entire Projection Appropriately Oriented

Projection Appropriately with Minor Projections to Telencephalon or Diencephalon

Class 1 Class 2 Class 3

58 110 98

47

44 (94%) 60 (86%) 30 (79%)

3 (6%) 10 (14%) 8 (21%)

70

38

of theThree

Oriented

. Figure

3. Normal

and Precocious

Optic

Projections

in Whole-Mount

Xenopus

Brains

(A) Stage 37/38 brain showing the normal HRP-filled optic tract COT). Optic fibers enter the contralateral diencephalon (Di) ventrally, just caudal to the optic stalk-diencephalic junction, and course dorsally. In the dorsal diencephalon, they veer caudally toward the optic tectum (Tee), which they begin to innervate at this time. (B) Precocious innervation of the tectum from a class 1 experiment. The host brain is at the same stage of development as in (A) (stage 37/38) but received a heterochronic eye transplant at stage 21. The optic fibers follow a normal trajectory through the diencephalon and halt their growth at the tectum, but note that the projection pattern is abnormally diffuse within the tectum. (C) Stage 32 host brain, the embryo received an older eye transplant at stage 18 (class 1 experiment), showing a heterochronic optic projection, which appears normal in its course through the diencephalon. One fiber (arrowhead) has reached the tectum. There are 8-11 hr between stages 32 and 37138. Normally optic axons have not grown beyond the midline chiasm at stage 32. This brain also showed HRP-labeled axons in the telencephalon. Telencephalic projections of this sort occurred in several cases but probably arise from inadvertent application of HRP to the nasal epithelium rather than misrouted retinal axons. (D) Stage 29/30 HU-treated host brain (class 2 experiment) showing the heterochronic optic pathway following a normal course though approximately two-thirds of the presumptive pathway. P, pineal; Tel, telencephalon; Hy, hypothalamus; D, dorsal; R, rostral. Bars, 188 urn (A and B) and 75 urn (C and D).

brains is about 7 hr, by which time the hosts have reached stage 25. To distinguish between these possibilities, it was necessary to increase the temporal disparity further between the host and donor. Heterochronic Elimination of the TPOC with HU-Treated Hosts One of two strategies we used to increase the time disparity between host and donor embryos was to

retard the development of the host brain by rearing host embryos in the drug HU, a DNA synthesis inhibitor, which slows the rate of embryogenesis without affecting the formation of the basic scaffolding tracts (Harris and Hartenstein, 1991). HU-treated embryos progress through developmental stages slower than normal, and theTPOC does not begin to develop until stage28, instead of stage25 (Figure 1). This experimental procedure, illustrated in Figure 2 (class 2 experi-

Precocious 1005

Pathfinding

Table 2. Number

of Retinal

Projections

Formed

in the Absence

of the TPOC

Type of Experiment

Number of Transplants

TPOC Absent Optic Absent

TPOC Absent Optic Present

TPOC Present Optic Absent

TPOC Present Optic Present

Class 2

167

89 (53%)

32 (20%)

29 (17%)

17 (10%)

ment), introduced at least 6 additional hours between thetimeofgraftingtheeyeandthetimeofappearance of the TPOC, amounting to a total IO-14 hr disparity between the donor and host. Eye primordia were transplanted from normal stage 28 embryos to HUtreated stage 21 hosts, and the projections were subsequently assayed with HRP when the hosts ranged from stage 27 to stage 35136. Seventy of these transplanted eyes made projections into the CNS that were long enough to assay, and 86% of these were found to project along the presumptive optic tract in the direction of the tectum (Table 1). A few cases (14%) extended a fiber bundle or single fibers inappropriately into the diencephalon and telencephalon in addition to the normally oriented projection. To determine whether these precociously growing retinal axons were, indeed, growing in an otherwise axonless environment, a second series of class 2 experiments was done, in which double immunostaining was performed on stage 27-29 HU-treated host brains so that optic axons and the earliest fiber tracts could be visualized simultaneously (Table 2). Twenty percent of the double-labeled cases (n = 32) developed an optic projection that was long enough to score in the complete absence of a-AT-stained fibers (Figures 4C and 4D), and all of these exhibited trajectories that were appropriately oriented. Indeed, sometimes a single optic axon could be seen extending toward the tectum through an otherwise axonless diencephalon. In the other cases, either TPOC fibers were present or the optic axons had failed to enter the brain and make a substantial projection. In those cases that had TPOC fibers present, it was evident that, for the most part, the optic axons were navigating in areas free of a-AT-stained fibers. Elimination of Preexisting Tracts with Xenoplastic Grafts A second strategy we used to delay the development of the host brain and thus the formation of the TPOC was to perform xenoplastic grafts using axolotl embryos as hosts (see Figure 2, class 3 experiment). Axolotls are a slow developing species of amphibian that undergoes embryogenesis at about a third the rate of Xenopus. Previously, it has been shown that optic axons from grafted Xenopus eyes can project to the tectum of axolotls, demonstrating that pathfinding cues are conserved across species (Harris and Cole, 1984). Eyes were transplanted from stage 28 Xenopus embryos to stage 22-28 axolotl embryos, and the projections were assayed when the hosts reached stages 28-35/36. A developmental series of anti-acetylated tubulin-stained brains revealed that the TPOC does

not form until stages 35-36 in axolotls (Figures 5A-5C), some 18-24 hr after eye grafting. Table 1 shows the data from class 3 experiments, in which the host axolotl brains were assayed at stage 34 and younger, i.e., before the formation of the TPOC. Thirty-eight cases made projections into the host brain that were long enough to assay, and 79% of these exhibited a single tract of optic fibers growing in the appropriate direction toward the tectum (Figure 5). An additional 21% showed a pattern of growth in which a major component of the projection formed a distinct tract of fibers that was directed toward the tectum, while a minor component formed a diffuse array of shorter axons that radiated invariousdirectionsawayfrom the point of brain entry (see Figure 5D). In no case did the transplanted optic axons form distinct tracts that were misdirected away from the tectum. Some optic projections developed as early as stage 28, which is approximately 18 hr in advance of the TPOC. Positional Cues Exist Deep within the Neuroepithelium The growth cones of optic fibers usually grow within 3-15 t.trn of the pial surface in the marginal zone among the apical end feet of the neuroepithelial cells and are not found close to the ventricular surface (Holt, 1989). By contrast, axons from heterochronically transplanted eyes frequently grow away from the pial surface and penetrate deep into the neuroepithelium (Figure 6). Indeed, axons and growth cones were observed close to and at the ventricular surface (Figures 6A and 6C). This occurred at a higher frequency when younger hosts were used. With stage 18-19 hosts, 59% developed deep projections, whereas with stage 2021 hosts, 23% developed deep projections. Deeply projecting axons were not seen in control experiments following eye grafts between stage 28 embryos. Despite their location, most deeply growing axons were nevertheless able to navigate appropriately, as evidenced by the formation of deep tracts extending toward the tectum (Figures 6A and 6B). This strongly indicates that guidance information must be expressed on both the basal and apical surfaces of the neuroepithelial cells early in neurogenesis and raises the possibility that the restriction of axons to the marginal zone in normal development is caused by the expression of inhibitoryor nonpermissivecueswithin the depth of the neuroepithelium. Discussion The results presented here show that retinal ganglion cell axons from heterochronically transplanted eyes

Figure 4. Optic

Projections

with

and without

the TPOC

Double-label immunocytochemistry showing the spatial relationship between optic fibers (red/orange) and nonoptic scaffolding tracts (green) in normal (A and B) and heterochronic (C and D) brains. Optic fibers were filled with HRP and immunolabeled with a-HRP and rhodamine-conjugated secondary antibodies; a-AT and a fluorescein-conjugated secondary antibody were used to label the early scaffold tracts. (A) Confocal image of a stage 3304 normal brain. The optic fibers (OT), which have reached the mid optic tract in the diencephalon (Di), just dorsal to the dorsal edge of the optic stalk (OS), can be seen coursing along the anterior edge of the TPOC. Note that there are many a-AT fibers present at the time of normal optic tract development. (6) High magnification confocal image of leading edge of optic projection shown in (A). The stretch of axon behind the lead growth cone (small arrowheads) does not fasciculate with the TPOC axons but rather crosses over them, and the growth cone fgc, arrow) is located dorsal to the TPOC. Stereoimaging shows clearly that the lead growth cone and the growth cone at the bottom of the figure are not growing on the TPOC and that the other axons only intermittently contact the TPOC fibers. (C) and (D) show a double-labeled stage 25 HU-treated host brain (class 2 experiment) with a clearly formed optic projection stained with a-HRP in (C) (arrow) and a-AT in (D) (arrow). In (D), note the absence of nonretinal a-AT-labeled fibers in the diencephalon where the optic fibers are growing. a-AT staining is also absent in the telencephalon and the midbrain, indicating that axonogenesis has not begun in theseareas. Arrowheads indicate a-AT-positive fibers in the rhombencephalon, where they are beginning to form the ventrolateral tract. This optic projection entered the brain slightly anterior to normal but nevertheless orients correctly. Tee, tectum; Tel, telencephalon. Bars, 100 rrrn (A, C, and D) and 20 urn (B).

are able to extend in a specific and directed manner in the younger host CNS. Furthermore, when these axons are challenged to grow into a host brain before any other fiber tracts have formed, including the TPOC, they navigate correctly. Observations of selective fasciculation in the grass-

hopper CNS led Goodman and colleagues to propose the “labeled pathways” hypothesis (Goodman et al., 1982; Raper et al., 1983), which posits that a small number of neurons establish a simple set of fiber tracts in the early CNS, forming a primary scaffold on which are deposited guidance cues essential for later grow-

Precocious

Pathfinding

1007

Stage 34

Stage 30 Figure 5. Optic

Projections

following

Xenoplastic

Stage 35

Transplantation

(A)-(C) illustrate several stages in the development of a-AT staining in axolotl brains showing the protracted development of the TPOC. At stage 30 (A), the TPOC is absent. At stage 34 (B), TPOC fibers begin to appear and are dispersed at wide intervals along the presumptive tract. At stage 35 (C), more fibers are present but the tract is discontinuous and does not become continuous until after this stage. (D) and (E) showxenoplastic optic projections resulting from class 3 experiments.(D) Optic projection with a dense bundle of axons heading in the correct direction toward the tectum plus a scattered set of shorter axons radiating anteriorly. (E and F) Xenopus optic projections in stage 34 and 35 axolotl brains exhibiting normal trajectories. In (F), the optic fibers terminate in the tectum. All of these projections are ipsilateral to the transplant. Bars, 300 km.

ing axons to navigate to their targets. Early embryonic axons have been shown to express unique surface cues along subregions of their lengths that are developmentally regulated (Kotrla and Goodman, 1984; Dodd et al., 1988). Indeed, there is much experimental evidence to support this hypothesis in several invertebrate and vertebrate systems. Easter and Kuwada and colleagues (Chitnis and Kuwada, 1990; Wilson et al., 1990) have shown that in the vertebrate brain, like the insect brain, a simple set of fiber pathways is laid down initially and that later growing axons subsequently add to this defined set of tracts, thus building on the existing pattern of fiber tracts rather than pioneering new ones. Our finding that a later developing pathway, the optic tract, is able to form without a scaffold tract shows that the molecular information guiding these later growing axons is not confined exclusively to the scaffold axons. Our results do not preclude the possibility that during normal development optic fibers use guidance cues located on the surfaces of other axons, but they do show that this sourceof a cue is not essential for correct pathfinding. In normal development, optic axons may use cues that are located either exclusively on the surfaces of the neuroepithelial cells or on both the neuroepithelial cells and the scaffold axons. It is interesting that precocious optic axons still travel in a close-knit bundle along the same route through the diencephalon even in the absence of other tracts. This suggests that the patterning information for establishing the stereo-

typed setsof initial tracts is a property of the neuroepithelium. The question of whether pioneer axons or tracts are necessary for the correct development of later growing axons has been investigated in several systems with varying results. Ablations of pioneer neurons or tracts have resulted in varying degrees of disrupted growth by the follower axons (Lo Presti et al., 1973; Raper et al., 1984; Kuwada, 1986; Chitnis and Kuwada, 1991; Chosh et al., 1990; Pike et al., 1992) or, conversely, normal growth (Keshishian and Bentley, 1983; Chitnis and Kuwada, 1991; Pike et al., 1992). The types of disrupted behavior exhibited by pioneerless follower axons include the following: halting of gro&h (Raper et al., 1984; Kuwada, 1986), retardation of growth with otherwise correct pathfinding (Pike et al., 1992), selection of inappropriate pathways (Chitnis and Kuwada, 1991), and failure to invade appropriate targets (Ghosh et al., 1990). One hypothesis to account for these diverse behaviors relates to the extent of fasciculation of the followers with the pioneers. In the grasshopper CNS, for example, the follower G axons fasciculate with axons in the AIP fascicle (Bastiani et al., 1984), and when the A and P axons are ablated, G axon growth is stunted (Raper et al., 1984). By contrast, in the zebrafish brain, where the fasciculation between the nucleus of the posterior commissure neurons and the pioneer TPOC fibers may not be extensive, the nucleus of posterior commissure neurons is capable of pathfinding with-

Neuron 1008

Figure 6. Deep Axonal Transplantation

Projections

following

Heterochronic

Eye

(A) Transverse section through the diencephalon of a stage 31 host brain following a class 1 experiment. Soon after the optic nerve enters the brain (EP, entry point) it bifurcates into two main fiber bundles. One (OT2) penetrates deep into the neuroepithelium (n) and runs along the ventricular surface (vs.), and the other (OTI) remains superficial and runs along the pial surface (ps). Several other fibers can be seen traveling separately within the depth of the neuroepithelium (small arrows). (6) Stage 32 wholemount HU-treated brain from a class 2 experiment showing a bifurcated projection from the heterochronically transplanted eye. As in (A), one fiber bundle (OT2) travels deep in the neuroepithelium, but it also navigates appropriately. OTl is traveling superficially and terminates in the tectum. (C) Shows a single growth cone of a retinal ganglion cell traveling close to the ventricular surface (vs) in the ventral diencephalon (stage 32 brain from a class 1 experiment). Tel, telencephalon; V, ventricle. Bars, 50 km (A and B) and 20 pm (C).

out pioneers (Chitnis and Kuwada, 1991). Our confocal observations reveal that pioneering retinal ganglion cell axons in Xenopus do not fasciculate extensively with TPOC axons, which is consistent with previous electron microscopy data showing only intermittent contacts between these two populations of axons (Easter and Taylor, 1989; Holt, 1989). Similarly in zebrafish embryos, recent electron microscopy analysis has shown that leading optic axons only occasionally contact TPOC axons (Burrill and Easter, 1991, Sot. Neurosci., abstract). Thus, the degree of fasciculation between pioneers and their followers might predict

the type of behavior of follower axons in the absence of pioneers. Previous electron microscopy studies have identified the neuroepithelial cells and axons of nonretinal origin as being the two sorts of cellular substrate that pioneer optic axons use in the diencephalon (Bovolenta and Mason, 1987; Easter and Taylor, 1989; Holt, 1989). The basal lamina located on the pial surface of neuroepithelial cell end feet is rarely contacted by retinal growth cones and so probably does not play a major role in axon guidance (Holt, 1989). Since the retinal axons are able to pathfind in the absence of

Precocious 1009

Pathfinding

nonretinal fibers, they must use guidance cues that are associated with the neuroepithelial cells. Neuroepithelial cells form the basic framework of the early CNS and are destined to give rise to neurons and glia in the mature brain. They are elongated cells with processes that often span the entire width of the neuroepithelium. Their marginal processes swell into sinuous end feet, and adjacent end feet interdigitate to delimit the surface of the CNS. Axons grow between the marginal processes of these cells, and optic axons sometimes make insertions into them and vice versa (Bovolenta and Mason, 1987; Holt, 1989). The absence of dye coupling between retinal axon growth cones and adjacent neuroepithelial cells in the optic tract eliminates the possibility that the mode of transmission of neuroepithelium-axon guidance is gap junctionally mediated (Holt, 1989) and, instead, indicates that it is cell surface mediated. What molecular cues could nascent axons be following in the neuroepithelium?The main class of molecules known to be expressed at very early stages in neurogenesis is the cell adhesion molecules, including N-cadherin and neural cell adhesion molecule. These molecules begin to be expressed at neurulation, and their levels of expression are, for the most part, similar in different regions of the neural tube (Hatta et al., 1987; Detrick et al., 1990; Levi et al., 1987). In contrast, there is mounting evidence that different homeobox genes are expressed in regionally distinct domains in the developing mouse, zebrafish, and Xenopus brain. engrailed, for example, first appears in a band of neuroepithelial cell nuclei at the isthmus of the midbrain and hindbrain and later extends anteriorly in a graded fashion into the optic tectum (Pate1 et al., 1989; Hemmati-Brivanlou et al., 1991); distaless and pax are expressed in the presumptive forebrain in distinct patterns before axonogenesis (Walter and Gruss, 1991; Krauss et al., 1991; N. Papalopulu and C. Kintner, personal communication). This group of genes may act to turn on downstream genes encoding cell surface or secreted proteins that would lead to the emergence of distinct regional identities within the neuroepithelium that comprises the developing brain. In sampling the molecular differences between the various domains (a single domain may extend across as few as three to five cell diameters), growth cones could continuously monitor their positions by comparing local differences in their environment. By referring to an inherent “idea” of their destination, growth cones could make adjustments to their trajectories and, thereby, selectively advance toward their target. This sort of model requires that intrinsic properties of different sets of axons define how they behave in response to the positional cues they encounter and could account for the high degree of regulation that is characteristic of vertebrate CNS development. In the present study, precociously growing retinal axons were not confined to the marginal zone but often grew deep in the presumptive diencephalon,

wending their way through the cell bodies close to the ventricular surface. Retinal ganglion cell axons normally grow in the marginal zone close to the pia in the optic tract, and they never invade the ventricular zone (Silver and Rutishauser, 1984; Bovolenta and Mason, 1987; Holt, 1989; Easter and Taylor, 1989). This spatially restricted pattern of axon growth might reflect the differential expression of specific permissive or repulsive molecules along the apico-basal dimension of the surfaces of mature neuroepithelia. Normally, permissive factors may be localized to the marginal surfaces of the neuroepithelial cells, making this a favorable environment in which to grow. Indeed, deposits of the neurite-promoting extracellular matrix molecule laminin are localized on the marginal end feet of the neuroepithelial cells in the optic pathway (chick, Cohen et al., 1987; mouse, Liesi and Silver, 1988). Alternatively, repulsive factors could be expressed along the ventricular two-thirds of the cells, making this a hostile environment for axon growth. The finding that precocious retinal axons often dive deep into the neuroepithelium yet still grow correctly, indicates that permissive cues are present along the entire apico-basal dimension of the early neuroepithelium and suggests that differential expression of inhibitory factors or the marginal confinement of permissive factors may occur later to restrict axon growth to the marginal zone. That they nevertheless navigate correctly further suggests that the guidance cues are not restricted to the marginal zone of the neuroepithelium. Mostofthecellsthatcomprisethe neuroepithelium are mitotically active at stage 25. In showing that optic axons respond to cues on the early neuroepithelium, our results support the idea that cells express specific guidance information on their surfaces before their terminal divisions. Our HU experiments corroborate the results of Harris and Hartenstein (1991) and demonstrate that the neuroepithelial cells do not need to divide in order to express these cues, since mitosis can be blocked from stage 10 on without apparently affecting axonal guidance. The finding that optic axons navigate appropriately without the TPOC is in accord with previous experiments by Harris showing that optic axons can navigate to the tectum from a variety of brain entry points and along trajectories distant to the TPOC (Harris, 1986). Furthermore, he showed that the guidance information was local, since 90” rotations of pieces of the presumptive optic tract neuroepithelium gave rise to optic projections that became correspondingly rotated within the graft (Harris, 1989). Taylor recently repeated this study with similar findings, and in addition, he showed that the TPOC became rotated when the presumptive tract was rotated (Taylor, 1991) and concluded that the local cues were, in fact, provided bytheTPOC. By contrast, our experimental data show that both the TPOC axons and the optic axons can independently follow cues associated with the neuro-

Neuron 1010

epithelial cells. We suggest that these local cues are used byearlyandlategrowingaxonsofdiverseorigins and are part of a global map of positional identity that arises in the early neuroepithelium. Experimental

Procedures

Embryos and Surgery Xenopus laevis eggs were obtained by hormone-induced matings of adult frogs from our breeding colony at the University of California, San Diego. Six hundred IUs of human chorionic gonadotropin (Calbiochem) was injected into the female and 250 IUs into the male. Axolotl eggs were obtained from the Axolotl Colony at Indiana University. Embryos of both species were raised in 10% Holtfreter’s solution at three temperatures, 14OC, 21°C, and 27OC, to maximize the range of developmental stages available. Xenopus embryos were staged according to the criteria of Nieuwkoop and Faber (1967J, and axolotl embryos were staged using the table of Bordzilovskaya and Dettlaff (1979). The ages given in hours in the text refer to embryos raised at 21°C. For the Xenopus to Xenopus eye transplant experiments, embryos were removed from their jelly coats with 2% cysteine and transferred to a clay-coated dish containing anesthetic (0.01% ethyl-m-aminobenzoate; Sandoz) in 100% modified Ringer’s solution (Gimlich and Gerhart, 1984). Smooth-sided depressions were made in the clay with a glass ball pipette to fit snugly the size and shape of the embryos. The embryos were positioned on their left sides in the depressions and held in place with an L-shaped minutien pin. Fine minutien pins (0.10 m m or 0.15 m m diameter) were used to dissect out the right eye primordia from host and donor embryos, and care was taken to excise the entire eye vesicle, including the optic stalk. When neural plate stage embryos were used as hosts, they were placed in round wells, rostra1 side up, and the right eye vesicle was removed from the presumptive eye-forming region in the anterior lateral neural plate (Brun, 1981; Eagleson and Harris, 1990). The donor eye primordium was positioned in thevacated orbit of the younger host with the cut surfaces of the optic stalks aligned and held in place with a chip of coverslip glass. The width of the optic stalk of the older donors was about half that of the host stalk so thetwo were apposed together at the ventral edge, since it is the vetltral edge of the optic stalk along which the optic fibers grow. After sufficient healing had occurred, about20 min postoperative, the glass chip was removed, and the embryo was transferred to a recovery dish containing 100% modified Ringer’s solution for about 1 hr, then to 10% Holtfreter’s solution. Operated embryos were usually placed in the 14OC incubator overnight to slow their overall rate of development. Similar operative procedures were used in the xenoplastic series of experiments in which Xenopus eye primordia were transplanted to axolotl host embryos. HRP labeling of Optic Fibers To label the optic fibers from the transplanted eyes, the lens primordium was removed, and a plug of semidried HRP (type VI; Sigma; approximately 30% HRP in 1% lysolecithin) was placed in the lens cavity. Embryos were fixed 20-30 min later in 4% paraformaldehyde for 4-24 hr for immunohistochemistry, or 2% glutaraldehydefor 1 hrfor diaminobenzidine(Sigma) histochemistry. The fixatives were diluted in 0.1 M phosphate buffer at pH 7.4, and fixation occurred at room temperature. After fixation, the brains and spinal cords were dissected out in one piece and processed as whole-mount preparations. a-HRP and a-Acetylated Tubulin labeling To visualize the pioneer tracts in the brain, in particular the TPOC, an antibodyto a-acetylated tubulin (a-AT, kindly provided by Dr. C. Piperno) was used. Whole-mount brains were incubated in a-AT (I:20 dilution) at 4OC overnight, followed by fluorescein- (for double labeling) or HRP-conjugated secondary antibody staining. To double label the brains so that both the optic fibers and the TPOC could be visualized in the same specimen,

theopticfiberswerefilledwith HRP,thenthebrainwasimmunolabeled first with an a-HRP antibody (Sigma), using a rhodamineconjugated secondary antibody, and second with a-AT, using a fluorescein secondary antibody. The brains were examined briefly on both sides with epifluorescence to determine whether an ipsilateral or contralateral optic projection existed, then mounted in glycerol with an antibleaching agent (0.1% paraphenylenediamine) under a coverslip supported by two clear reinforcement rings (Avery). Diaminobenzindinereacted brains were dehydrated, cleared in benzylbenzoate, and mounted in Coverbond (BSP) under a coverslip supported by one or two reinforcement rings. It was desirable to flatten the brains slightly with the coverslip so that as much as possible of the surface of the brain could be viewed in a single plane of focus. Sectioning and Microscopy To investigate the growth and behavior of precociously growing axons within the depth of the neuroepithelium, diaminobenzidine-reacted brains with optic projections were embedded in albumin and sectioned transversely at 40 urn on a vibratome (Oxford). Serial sections were cleared in xylene and mounted in Coverbond. Sections and whole mounts of diaminobenzidinereacted material were examined with Nomarski optics on a Zeiss photomicroscope and photographed with T-MAX 100 film using a green filter or with Kodakolor 400 using a blue filter. Drawings of the projections were made either directly from the sample using a camera lucida attachment or from color slides using acetate sheets to overlay the image projected onto a small (12” x 8’1 screen. Confocal microscopy (Bio-Rad MRC 600) was performed on normal stage 33/34 and 35/36 Xenopus embryos that were double stained with a-AT and a-HRP to visualize the relationship be tween the TPOC and the optic axons. A focus-through Z series was collected at 0.5 pm intervals on separate fluorescein and rhodamine channels, and the images were recombined to produce double-labeled photomicrographs that could be viewed stereoscopically. Acknowledgments We thank Bill Harris for his insightful input and David Rapaport, Chi-Bin Chien, and Ajay Chitnis for their critical reading of the manuscript. We are grateful to John Kuwada and Cianni Piperno for providing uswith anti-acetylated tubulin antibody. We thank Mark Ellisman for the use of the confocal microscope facility at the Neurosciences Department at the University of California, San Diego and Tom Derrink for excellent technical assistance. This work was supported by grants from the National Institutes of Health (NS23780) and the March of Dimes (#I-FY91-0348) and a Pew Scholars Award (C. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC Section 1734 solely to indicate this fact. Received

August

14, 1992; revised

September

15, 1992.

References Bastiani, M. J., Raper, J. A., and Goodman, C. S. (1984). Pathfinding by growth cones in grasshopper embryos. Ill. Selective affinity of the G growth cone for the P cells within the A/P fascicle. J. Neurosci. 4,2311-2328. Bastiani, M. J., Doe, C. Q., Helfand, S. L., and Goodman, C. S. (1985). Neuronal specificity and growth cone guidance in grasshopper and Drosophila embryos. Trends Neurosci. 8, 257-266. Bordzilovskaya, N. D., and Detlaff, T. A. (1979). Tables of stages of axolotl embryos and the prognostication of timing of successive developmental stages at various temperatures. Axolotl Newslett. 7, 2-22. Bovolenta, P., and Mason, C. (1987). Growth cone morphology varies with position in the developing mouse visual pathway from retina to first targets. J. Neurosci. 7, 1447-1460.

Precocious 1011

Pathfinding

Brun, R. B. (1981). The movement of the prospective eye vesicles from the neural plate into the neural fold in Ambystoma mexicanum and Xenopus laevis. Dev. Biol. 88, 192-199.

Kotrla, K. J., and Goodman, C. S. (1984). Transient expression of a surface antigen on a small subset of neurons during embryonic development. Nature 377, 151-153.

Chitnis, A. J., and Kuwada, J. Y. (1998). Axonogenesis of zebrafish embryos. J. Neurosci. 70, 1892-1905.

Krauss, S., Johanson, T., Korzh, V., and Flose, A. (1991). Expression pattern of zebrafish pax genes suggests a role in early brain regionalization. Nature 353, 267-270.

in the brain

Chitnis, A. J., and Kuwada, J. Y. (1991). Elimination of a brain tract increases errors in pathfinding by follower growth cones in the zebrafish embryo. Neuron 7, 277-285. Cohen, J., Burne, J. F., McKinlay, C., and Winter, J. (1987). The role of laminin and the laminin/fibronectin receptor complex in the outgrowth of retinal ganglion cell axons. Devl. Biol. 722,487418. Detrick, J. R., Dickey, D., and Kintner, C. R. (1990). The effects of N-cadherin misexpression on morphogenesis in Xenopus embryos. Neuron 4,493-586. Dodd, J., Morton, S. B., Karagogeos, D., Yamamoto, M., and Jessell,T. M. (1988). Spatial regulation of axonal glycoprotein expression on subsets of embryonic spinal neurons. Neuron 7, 105116. Eagleson, G. W., and Harris, W. A. (1998). Mapping of the presumptive brain regions in the neural plate of Xenopus laevis. J. Neurobiol. 27, 427-448. Easter, S. S., and Taylor, J. S. H. (1989). The development of the Xenopus retinofugal pathway: optic fibers join a preexisting tract. Development 7, 553-573. Ghosh, A., Antonini, A., McConnell, S. K., and Shatz, C. J. (1998). Requirement for subplate neurons in the formation of thalamocortical connections. Nature 347, 179-181. Cimlich, promote 130.

R. L., and Gerhart, J. C. (1984). Early cellular interactions axis formation in Xenopus laevis. Dev. Biol. 704, 117-

Goodman, C. S., Raper, J. A., Ho, R., and Chang, S. (1982). Pathfinding by neuronal growth cones in grasshopper embryos. Symp. Sot. Dev. Biol. 40, 275-316. Harris, W. A. (1986). Homing behaviour of axons in theembryonic vertebrate brain. Nature 320, 266-269. Harris, W. A. (1989). Local positional cues in the neuroepithelium guide retinal axons in the embryonic Xenopus brain. Nature 339, 218-221. Harris, W. A., and Cole, J. (1984). Common mechanisms brate axonal navigation: retinal transplants between related amphibia. J. Neurogenet. 7, 127-140.

in vertedistantly

Harris, W. A., and Hartenstein,V. (1991). Neuronal determination without cell division in Xenopus embryos. Neuron 6499-515. Harris, W. A., Holt, C. E., and Bonhoeffer, F. (1987). Retinal axons with and without their somata, growing to and arborizing in the tectum of Xenopus embryos: a time-lapse video study of single fibres in vivo. Development 707, 123-133. Hatta, K., Takagi, S., Fujisawa, H., and Takeichi, M. (1987). Spatial and temporal expression pattern of Ncadherin molecules correlated with morphogenetic processes of chick embryos. Dev. Biol. 720, 215-227. Hemmati-Brivanlou, A., de la Torre, J. R., Holt, C., and Harland, R. M. (1991). Cephalic expression and molecular characterization of Xenopus En-2. Development 777, 715-724. Holt, C. E. (1984). Does timing of axon outgrowth influence initial retinotectal topography in Xenopus? J. Neurosci. 4, 1130-1152. Holt, C. E. (1989). A single cell analysis of early retinal ganglion cell differentiation in Xenopus: from soma to axon tip. J. Neurosci. 9, 3125-3145. Holt, C. E., and Harris, W. A. (1983). Order in the initial retinotectal map in Xenopus: a new technique for labellinggrowing nerve fibres. Nature 307, 150-152. Keshishian, H., and Bentley, D. (1983). Embryogenesis of peripheral nerve pathways in grasshopper legs. III. Development without pioneer neurons. Dev. Biol. 96, 116-124. Klose, M., and Bentley, essential for formation ence 245,982-984.

D. (1989). Transient pioneer neurons of an embryonic peripheral nerve.

are Sci-

Kuwada, J. Y. (1986). Cell recognition by neuronal in a simple embryo. Science 233, 748746.

growth

cones

Kuwada, J. Y., Bernhardt, R. R., and Chitnis, A. B. (1998). Pathfinding by identified growth cones in the spinal cord of zebrafish embryos. J. Neurosci. 70, 1299-1388. Letourneau, P. C. (1982). Nervefibre growth and extrinsic factors. In Neuronal Development, N. C. Spitzer, ed. (New York: Plenum Publishing Corp.), pp. 213-254. Levi, G., Crossin, K. L., and Edelman, G. M. (1987). Expression sequences and distribution of two primary cell adhesion molecules during embryonic development of Xenopus laevis. J. Cell Biol. 705, 2359-2372. Liesi, P., and Silver, J. (1988). lsastrocyte laminin involved in axon guidance in the mammalian CNS? Dev. Biol. 730, 774-785. Lo Presti, V., Macagno, E., and Levinthal, C. (1973). Structure and development of neuronal connections in isogeneic organisms: cellular interactions in the development of the optic lamina of Daphnia. Proc. Natl. Acad. Sci. USA 70, 433-437. Nieuwkoop, P. D., and Faber, J. (1967). Normal Tables of Xenopus laevis (Daudin) (Amsterdam: North-Holland Publishing Co.). Patel, N. H., Martin-Blanco, E., Coleman, K. G., Poole, S. J., Ellis, M. C., Kornberg, T. B., and Goodman, C. S. (1989). Expression of engrailed proteins in arthropods, annelids and chordates. Cell 58, 955-968. Pike, S. H., Melancon, E. F., and Eisen, J. S. (1992). Pathfinding by zebrafish motoneurons in the absence of normal pioneer axons. Development 774, 825-831. Piperno, C., and Fuller, M. T. (1985). Monoclonal antibodies specific for an acetylated form of a-tubulin recognize the antigen in cilia and flagella from a variety of organisms. J. Cell Biol. 707, 2085-2894. Raper, J. A., Bastiani, M. J., and Goodman, C. 5. (1983). Pathfinding by neuronal growth cones in grasshopper embryos. II. Selective fasciculation onto specific axonal pathways. J. Neurosci. 3, 31-41. Raper, J. A., Bastiani, M. J., and Goodman, C. S. (1984). Pathfinding by neuronal growth cones in grasshopper embryos. III. The effects of ablating the A and P axons upon the behavior of the C growth cone. J. Neurosci. 4, 2329-2345. Schubiger, M., and Palka, J. (1985). Genetic suppression of putative guidepost cells: effect on establishment of nerve pathways in Drosophila wings. Dev. Biol. 708, 399-410. Silver, J., and Rutishauser, U. (1984). Guidance of optic axons in vivo by a preformed adhesive pathway on neuroepithelial endfeet. Dev. Biol. 706, 485-499. Taylor, J. S. H. (1991). Theearlydeveiopment of the frog retinotectal projection. Development (Suppl.) 2, 95-184. Thomas, J. B., Bastiani, M. J., Bate, M., and Goodman, C. S. (1984). From grasshopper to Drosophila: a common plan for neuronal development. Nature 370, 203-207. Tix, S., Bate, M., and Technau, C. M. (1989). Preexisting pathways in the leg imaginal discsof Drosophilawings. ment 707, 855-862.

neuronal Develop

Walter, C., and Gruss, P. (1991). fax-6, a murine paired box gene, is expressed in the developing CNS. Development 773, 14351449. Wilson, S. W., and Easter, S. S. (1991). Stereotyped pathway selection by growth cones of early epiphysial neurons in th\e embryonic zebrafish. Development 772, 723-746. Wilson, S. W., Ross, L. S., Parrett, T., and Easter, S. S. (1990). The development of a simple scaffold of axon tracts in the brain of the embryonic zebrafish, Brachydanio rerio. Development 708, 121-145.