Cyclic AMP-Induced Repair of Zebrafish Spinal Circuits

See allHide authors and affiliations

Science  09 Jul 2004:
Vol. 305, Issue 5681, pp. 254-258
DOI: 10.1126/science.1098439


Neurons in the human central nervous system (CNS) are unable to regenerate, as a result of both an inhibitory environment and their inherent inability to regrow. In contrast, the CNS environment in fish is permissive for growth, yet some neurons still cannot regenerate. Fish thus offer an opportunity to study molecules that might surmount the intrinsic limitations they share with mammals, without the complication of an inhibitory environment. We show by in vivo imaging in zebrafish that post-injury application of cyclic adenosine monophosphate can transform severed CNS neurons into ones that regenerate and restore function, thus overcoming intrinsic limitations to regeneration in a vertebrate.

The CNS of humans and other mammals is unable to regenerate, and considerable evidence suggests that this is a consequence of two major impediments. One includes inhibitory molecules found in the CNS environment (14). The other includes intrinsic factors that prohibit neurons from launching a successful regenerative response (411). This latter impediment is evident in the failure of damaged mammalian neurons to regrow even when placed in a permissive environment (1113). The existence of these two confounding variables has complicated efforts to induce functional regeneration after injury.

Fish, unlike mammals, have a CNS that is permissive for regeneration, where some neurons are able to regrow well and others are not (1416). This is analogous to the mammalian situation, in which some neurons grow well in permissive surroundings and others do not (1113). We have analyzed the problem of regeneration in zebrafish (Danio rerio), in which we can examine intrinsic inhibitory factors in the context of an otherwise permissive CNS environment. In zebrafish, we can directly visualize the axons in the CNS of the living animal as well as assess associated motor behavior and recovery of function.

Cyclic adenosine monophosphate (cAMP) can induce axonal sprouting of cultured neurons in vitro (5, 17, 18) and of sensory neurons in vivo (17, 19). However, for a molecule to have therapeutic implications, it must lead to the functional regeneration of neurons within the CNS when applied well after the injury. We tested the effect of cAMP on the regenerative capacity of the Mauthner cell (M cell), a myelinated neuron with well-characterized morphology and function that is part of the escape behavior circuit (2023). The axon of this neuron regenerates poorly after a spinal lesion (1416). We observed the response of the Mauthner axon (M axon) to a lesion and to the subsequent application of cAMP.

We evaluated the ability of the M axon to regenerate by backfilling the axons of 5- to 7-day-old zebrafish with fluorescent dye, lesioning the spinal cord, and visually monitoring the axonal growth response in vivo. Of the lesioned M cells, 15 of 43 (35%), showed some spontaneous regenerative response, which occurred within 48 hours of the lesion. In nearly all of these cases (13 of 15), the regenerating axons did not penetrate the lesion site but grew aberrantly, either growing out through ventral roots into muscle or making a U-turn to grow rostrally in the spinal cord. In two cases, processes from the M axons entered the lesion site, and one was able to cross through it.

In 65% of the animals (28 of 43), the M axon did not regenerate at all, with the axon dying back to form a stump rostral to the lesion site. These nonregenerating fish were followed from day 5 to day 14 post-lesion. A similar poor regenerative response of the M axon is maintained in both adult zebrafish and goldfish (1416).

In contrast to the M axon, some axons in the adult and larval zebrafish spinal cord regenerate well after injury, indicating that the zebrafish CNS is not hostile to growth (14, 15). The poor ability of the M axon to regenerate is evidently related to intrinsic limitations in its ability to regrow.

We then examined whether we could induce M cells to grow across a lesion and restore function. We labeled single M cells with rhodamine dye by electroporation (24) in vivo to allow observation of single cells over time. A day after electroporation of dye, we severed the ventral spinal cord to lesion the M axons. In a few cases, the lesions extended to include the entire cord. The surviving fish were examined 3 to 5 days after the lesion to unambiguously identify M cells with severed, unregenerated proximal axon stumps (Figs. 1 and 2). Such cells never regenerate spontaneously. Membrane-permeable dibutyryl–cyclic adenosine monophosphate (db-cAMP) was then pressure injected, under visual control, onto the somata of the lesioned M cells. The trauma associated with these procedures led to roughly 50% mortality at each step. We started with approximately 100 fish, yielding 26 surviving animals (five of these with whole-cord lesions).

Fig. 1.

Cyclic AMP–induced regeneration of the M axon in vivo. (A) Fluorescently labeled M cell filled by single-cell electroporation viewed in a living zebrafish. Left arrow marks the cell soma; right arrow, the axon in the spinal cord. A full confocal reconstructionof the labeled cell is shown below the fish. (B) Top panel, full-length projection of a confocal z-stack of a severed M axon 3 days after a lesion. Bottom panel, the same cell 1 day after db-cAMP application to the soma. (C) Control results before and after application of vehicle alone. Rostral is to the left in this and subsequent figures. Scale bars, 0.5 mm.

Fig. 2.

Evidence that the M axons are newly regenerated in response to the cAMP. (A) Three pairs of vehicle controls and two cGMP controls. Top panels, the lesion sites 3 days after damage. Bottom panels, the same fish 1 day after vehicle or db-cGMP injection. (B) Responses to cAMP. Top panels of (a) and (b), closeups of lesion sites 3 days after lesioning, immediately before db-cAMP injection. Bottom panels of (a) and (b), the same cell as in the top panels 1 day after db-cAMP. In (a), the distal stump of the severed axon remained; arrows indicate the location of this stump adjacent to the newly regenerated axon. In (b), the M axon, originally labeled with red dye, was labeled with a second green dye at the time of db-cAMP injection. The new axon is yellow; the original, degenerated distal portion of the axon is red (arrows). Top panel of (c), a case of aberrant regeneration prior to cAMP. Effects of db-cAMP on this axon are shown in the bottom panel of (c). Top panel of (d), an axon that failed to regenerate 2 weeks after the lesion. Bottom panel of (d), the same axon 1 day after single-cell electroporation of 8-OH-cAMP. All images are in vivo confocal z-projections in either lateral view (a and b) or dorsal view (c and d). Vertical dashed line in bottom panels of (a to d) marks the middle of the lesion site. Scale bars, 100 μm.

Twenty-four to 32 hours after treatment, we examined the morphology of the cells in vivo. We found long, ventral M axons extending through the lesion site (Fig. 1B, bottom panel; see Fig. 3 for quantification). Application of vehicle-only or dibutyryl–cyclic guanosine monophosphate (db-cGMP) produced no detectable response (Fig. 1C and Fig. 2A, a to e).

Fig. 3.

Summary of distances regenerated. (A) Diagram of the extent of regeneration in all 40 fish that were treated with either db-cAMP or 8-OH-cAMP. Top gray bar shows the length of a normal unlesioned M axon with SD. Gray portions of other bars represent the length of treated axons after lesioning and immediately preceding db-cAMP injection (above dashed line) or 8-OH-cAMP injection (below dashed line). The asterisk indicates the site of lesion. The black bars represent growth of new axon 24 to 32 hours after treatment. The top five axons regenerated after whole-cord lesions. All axons were treated 3 to 5 days after lesioning, with the exception of the two marked by dots, which were treated 14 days after lesioning. Bars are aligned with a schematic of the fish to show the growth in relation to body segments. (B) Histogram of the length of regenerated axon for controls (white bar; includes five vehicle injections, three db-cGMP, two vehicle electroporations, and four extracellular 8-OH-cAMP), db-cAMP (black bars), and 8-OH-cAMP (hatched bars). Scale bar, 0.5 mm.

In all cases, the newly growing axon in the db-cAMP animals eventually ran close to the degenerating debris of the old distal stump (Fig. 2B, a and b). This provided an indication of the new axon's trajectory and confirmed the growth of a truly separate axon. In one case, a substantial portion of the distal stump remained intact 4 days after the lesion, and the new axon could be seen traveling adjacent to it (Fig. 2Ba).

Figure 2Bc shows a case in which we treated an M cell that had spontaneously regenerated after the lesion, but in aberrant fashion, with its axon making a U-turn at the lesion site to grow back toward the head. Treatment with db-cAMP in this animal led unexpectedly to the production of a branch of the aberrant axon that crossed the lesion site and extended appropriately into caudal spinal cord.

In all 26 animals, the db-cAMP injection induced robust, unambiguous regeneration of the M axon through the lesion site into caudal spinal cord (Fig. 3). Regeneration was never observed in response to injection of vehicle alone (n = 5) in blind procedures, which included successfully regenerating, db-cAMP treated animals. The spinal-injured larval fish are fragile, which makes frequent observations of the time course of the response to cAMP in individual fish difficult. We did, however, sample at shorter times after cAMP application in three fish and found sprouting from the axon stump as early as 2 hours after cAMP treatment and penetration of sprouts into the lesion site by 6 hours after cAMP injection. Movie S1 shows the dynamics of a treated regenerating axon over a 36-hour period.

We examined whether the db-cAMP effect was specifically on the M cell by electroporating a membrane-impermeant form of cAMP, 8-OH-cAMP (BioLog, Bremen, Germany) directly into the M cell. Twelve fish in which the M axon failed to regenerate were treated 3 days after the lesion, as in the db-cAMP animals. The M axons in all but one of these fish regenerated through the lesion site into caudal spinal cord (Fig. 3). The axon of the remaining fish sprouted but did not cross the lesion site. In controls, electroporation of vehicle (n = 2), extracellular pressure injection of 8-OH-cAMP (n = 2), or electroporation of 8-OH-cAMP away from the M cell (n = 2) had no effect on regeneration, which supports the conclusion that the cAMP is acting on the M cell itself.

We attempted to maintain one batch of 50 fish for 2 weeks after the lesion to explore the limits of the interval between injury and treatment. At this age (3 weeks old), the animals are near the end of our ability to image the axons in vivo because of increases in muscle mass and pigmentation. Two of these fish survived for 2 weeks after the lesion. M axons in both fish showed robust regeneration after electroporation of 8-OH-cAMP, even after this much longer delay before treatment (Fig. 2Bd and Fig. 3).

Not only did the neurons treated with cAMP launch a regenerative response when previously they could not, but the regenerated axons also adopted normal axonal trajectories. None of the cAMP-induced axons avoided the lesion site but, rather, regenerated directly through it. Furthermore, all of the cAMP-induced regenerating axons eventually assumed a ventral trajectory, extending from 3 to 18 body segments past the lesion site (a total of 297 to 1776 μm in length) (Fig. 3). This corresponds to regeneration of from 12 to 86% of the severed portion of the normal axon.

Knowledge of the circuitry of the M cell and its behavioral role allowed us to measure the functional consequences of the regeneration (21, 22). The M cell initiates an escape behavior in response to a threatening stimulus that is characterized by a very fast, large-amplitude turn away from the stimulus. We used in vivo calcium imaging of circumferential descending (CiD) spinal interneurons normally activated by the M axon during escapes (20, 22) to determine how the activity of individual cells was affected by the lesion and the subsequent regeneration. The calcium responses from three to four identified CiD interneurons caudal to a lesion were imaged before and after the induction of regeneration in each of three fish (a total of 11 CiDs) (Fig. 4). We found that CiD cells were not activated in escapes after the lesion and before the induction of regeneration, as we expected because of the loss of their normal M cell input (Fig. 4A and black lines in 4C). The same CiD cells showed robust calcium responses after regeneration, indicating that the regeneration restored their normal activation in escapes (Fig. 4B and red lines in 4C).

Fig. 4.

Recovery of activity in interneurons caudal to the lesion site. Calcium imaging, in an intact fish, of four CiDs [black dots numbered 1 to 4 in (A) and (B)], which are postsynaptic to the M cell. (A) Top panel, an array of four CiDs caudal to the lesion site filled with a calcium indicator and imaged after lesioning but before regeneration. Bottom panel, pseudocolor imaging of cell 2 (red box) during an escape. Frames were collected every 300 msec; the escape movement occurred in the frame marked with an asterisk. Frames read left to right and top to bottom. (B) Top panel, same cell as in (A), shows a calcium response after induction of regeneration of the M axon (arrows) by db-cAMP. (C) Quantification of the fluorescence changes observed with the calcium indicator for each of the four CiD interneurons in (A) and (B) before (black) and after (red) cAMP-induced regeneration. Arrows mark the time of the escape in each plot. Scale bar, 100 μm.

We examined the behavioral effect of the regeneration by measuring the kinematics of the escape behavior initiated by the M cell in response to a squirt of water directed toward the tail of the fish (21). The lesions substantially impaired all measures of escape performance. In the absence of treatment, these measures exhibited only a small recovery over the time course of our experiment. In contrast, the cAMP-treated fish showed improvement in all measures of escape performance. The characteristics of the escape approached normal after the treatment (Fig. 5).

Fig. 5.

Recovery of behavior. (Top) The escape bend of a fish before regeneration, after regeneration, and in an unlesioned fish. Images are shown every 2 msec after the start of the turn until the maximum of the bend. Bottom panels quantify the escape performance before and after cAMP-induced regeneration (mean + SEM). Performance measures included (A) response latency, (B) peak angular velocity, (C) duration, and (D) maximum angle of the bend. These performance measures are shown for a group of five fish (five trials each) studied before (black bar, 3 days post-lesion) and after (black bar, 5 days post-lesion) cAMP treatment and for a control, untreated group (white bars) over the same time course (P < .0001 in every case for treated versus control). White bars on the right show performance measures from wild-type (w.t.) fish at 9 days.

There is considerable variability in the regenerative capabilities of different axons both within zebrafish and between different species (9, 1416, 25). This disparity is thought to be partially attributable to the degree of up-regulation of growth-associated molecules after the lesion (9, 15). Little to no up-regulation may lead to complete failure of regeneration, as is typically seen with the M cell. Intermediate levels might stimulate sprouting and growth along aberrant paths, whereas higher levels may allow sprouts to penetrate the lesion site and reacquire a normal trajectory. Such differences might explain the natural variation in regeneration and suggest a mechanism by which cAMP promotes regeneration through a lesion.

Cyclic AMP plays numerous roles in intracellular signaling pathways, many of which may have contributed to the effects we observed. It can promote neurite growth (5, 17, 18, 26, 27), act as an axon guidance cue (28, 29), effect growth-associated gene transcription (30), and attenuate the inhibitory interaction of the growth cone and its environment, possibly through the Nogo receptor (3, 8, 26, 27, 31). This broad spectrum of influence may be what is required of a single molecule for it to transform severed neurons into ones that can regenerate and restore function.

If a treatment for damage to the CNS, such as a spinal cord injury, is going to restore function, it must work on neurons that have a prior injury and induce them not only to regenerate but also to reestablish circuits and behavior. Cyclic AMP does this in zebrafish in an environment permissive for growth. Strategies that boost a severed neuron's intrinsic ability to regenerate along with concurrent attenuation of inhibition from the environment may offer the best solution to the problem of functional neuronal regeneration in humans.

Supporting Online Material

Materials and Methods

Fig. S1

Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article