Ectoderm to Mesoderm Lineage Switching During Axolotl Tail Regeneration

See allHide authors and affiliations

Science  06 Dec 2002:
Vol. 298, Issue 5600, pp. 1993-1996
DOI: 10.1126/science.1077804


Foreign environments may induce adult stem cells to switch lineages and populate multiple tissue types, but whether this mechanism is used for tissue repair remains uncertain. Urodele amphibians can regenerate fully functional, multitissue structures including the limb and tail. To determine whether lineage switching is an integral feature of this regeneration, we followed individual spinal cord cells live during tail regeneration in the axolotl. Spinal cord cells frequently migrate into surrounding tissue to form regenerating muscle and cartilage. Thus, in axolotls, cells switch lineage during a real example of regeneration.

Lineage restriction into ectodermal, mesodermal, and endodermal germ layers that occurs during development has been thought to be a process that is not reversed. However, recent data indicate that adult cells from various sources, including brain, skin, and bone marrow, can form cell types of other lineages when exposed to novel or foreign environments (1–5). Whether such examples represent true cases of cell-type switching and whether lineage switching represents a rare or frequent event are still being debated (6, 7). The issue of neural cell plasticity has been particularly controversial. Cells derived from the ventricles of the adult brain have been documented to form muscle in transplantation and coculture experiments and bone marrow when infused into irradiated mice (1, 4, 8, 9). Furthermore, when transplanted into chicken embryos, these neural progenitors populated tissues arising from all three embryonic germ layers (3). The following questions have been raised: (i) whether the cell that gave rise to other cell types was truly of neural origin rather than a contaminating cell type, (ii) whether long-term culture may have selected cell characteristics not present in the original neural cell itself (7), and (iii) whether the cells merely fused with other cells and their fate was thereby determined by fusion to a predetermined cell type (10, 11). Such ambiguities are largely due to the cell assays that involve placing populations of cultured cells into unnatural environments and examining the outcome after numerous cell doublings have occurred. Understanding cell plasticity is important because of its potential for creating progenitor pools for tissue regeneration, but because most mammalian tissues do not naturally regenerate, it has been difficult to assess the relevance of this phenomenon for tissue repair.

To determine whether lineage switching is an integral part of tissue regeneration when it occurs in a natural context, we developed a method to follow individual neural precursor cells live during axolotl tail regeneration. In amphibian axolotls, tail amputation results in perfect regeneration of all tail structures, including the spinal cord, muscle, cartilage, dermis, and skin, in several weeks (12–14). The entire process of tail regeneration can be visualized with a light microscope with single-cell resolution in 2-cm-long animals (15). Immediately after amputation, epidermal cells migrate over the cut end to form the wound epidermis. Over the next few days, a zone of proliferating progenitor cells called the blastema forms. Blastema cells largely derive from muscle and dermis dedifferentiation and in the late stages of regeneration redifferentiate into muscle, cartilage, and dermis (15,16). The regenerating spinal cord constitutes a tube of neuroepithelial cells separate from the blastema (see fig. S1). As regeneration proceeds the tube elongates, and spinal cord neurons differentiate.

Histological studies indicated that lineage restrictions were likely maintained during regeneration; the wound epidermis derived from mature epidermis, the regenerating spinal cord and neural crest formed from the spinal cord, and mesodermal tissues such as muscle and cartilage formed from dedifferentiation of tissues originating from mesoderm (17–20). However, these observations depended on reconstructing events from static images of cell populations within regenerating tissue. Therefore, whether cells crossed lineage boundaries during regeneration remained an open question.

To follow spinal cord cells during regeneration, we used single-cell electroporation to express the green fluorescent protein (GFP) in the neural precursor cells of the spinal cord. These precursor cells are radial glial cells similar to radial glial cells that act as neural precursors in the mammalian brain (21). We targeted GFP expression to radial glial cells by inserting the DNA-filled microelectrode into the lumen of the spinal cord through the cut end of the tail (22) and by driving GFP expression with the glial acidic fibrillary protein (GFAP) promoter (23) via the GAL4-VP16 enhancer (24, 25). We chose the GFAP promoter because GFAP is expressed in the radial glial cells of the mature axolotl spinal cord and expression is maintained in the regenerating spinal cord cells (26). Expression of GFP was reliably visible 48 hours postelectroporation, the starting point of our tracking experiments. By this time one to four cells were labeled in the spinal cord, depending on the animal. Whether the profile of cell labeling is due to division of a single cell or to electroporation of multiple cells is not known. We confirmed the specificity of the labeling method in cryosections of the transfected tails (n = 10) where the labeled cell was always found in the cell layer lining the neural canal (Fig. 1). Figure 1A shows a transfected cell in a cross section of the spinal cord. The cell has the characteristic fine radial processes of a GFAP-positive spinal cord cell, and the overlay with a differential interference contrast (DIC) microscopy image of the cross section confirms that the cell is in the layer lining the spinal cord lumen (27). Figure 1C illustrates the overall layout of the cross section and the positions of the GFAP-positive radial glial cells.

Figure 1

Transverse cryosection of the axolotl tail containing a transfected radial glial cell. (A) Fluorescent image of a GFP-expressing radial cell. (B) The DIC image overlaid with the fluorescence image shows that the cell lies in the spinal cord and extends a process toward the pial layer of the spinal cord. (C) Spinal cord morphology. Radial glial cells of the ependymal layer lie closest to the lumen (L) and extend processes to the pial edge; a labeled cell is illustrated in green. Scale bar, 50 μm.

When we followed individual radial glial cells during regeneration, we found that their descendants were not all restricted to an ectodermal cell fate. Although many labeled cells formed the expected neural cell types (see fig. S2), at high frequency the descendants of labeled spinal cord cells migrated out of the regenerating spinal cord and contributed to the regenerating mesodermal structures: muscle and cartilage (Table 1). Two of the major classes of cell fates are illustrated in Figs. 2 and3.

Figure 2

Radial glial cells exit the spinal cord and contribute to muscle. (A to C) Day 2; (D to F) day 6; (G and H) day 13. Two days postamputation, four cells labeled by GFAP-GFP are visible in the spinal cord (B and C). In (B), the brightness of the fluorescent image has been increased to allow visualization of cells 1 and 3. Arrows in (A), (D), and (G), indicate the plane of amputation; the regenerating spinal cord (Reg. SC) is highlighted in blue in (D). By day 6, the cells have begun to proliferate (E and F). By day 13, the cells of cluster 4 have also exited the spinal cord, while cells from clusters 1 and 2 have migrated into the regenerate but remain within the spinal cord (G). Bar = 100 μm. At higher magnification (H), it is clear that the exited cells display an elongated shape resembling nascent muscle fibers. (I) The exited cells (green) are positive for the differentiated muscle marker, myosin heavy chain (red). Scale bar, 50 μm.

Figure 3

GFAP-positive radial glial cells can also form cartilage. (A to C) Day 2. Two cells are initially labeled in the spinal cord (B and C). (D toF) Day 8. The cell closest to the plane of amputation, cell 2, migrates out into the regenerating spinal cord and begins to proliferate (E and F). Cell 1 is not visible in this image. It remained close to the plane of amputation and gave rise to new radial glial cells. (G to I) By day 9, cells from cell cluster 2 begin to exit the spinal cord (H and I). (J toL) By day 11, all labeled cells have left the spinal cord, rapidly changing shape to become morphologically distinct cartilage cells; see arrowheads (K and L). Scale bar, 100 μm.

Table 1

Different cell types deriving from GFAP-positive cells during tail regeneration. Frequency is calculated based on the number of animals followed (25). Animals contained multiple labeled cells, and some labeled cells gave rise to descendants that had different fates, so the total frequency is greater than 100%.

View this table:

The most frequent cases of cell type switching were to muscle (Fig. 2). Two days posttransfection, four cells were labeled in the mature spinal cord (Fig. 2, A to C). By day 6 the cells had divided and spread into four clusters in the regenerating spinal cord (Fig. 2, D to F). By day 13, cells of clusters 3 and 4 (three cells) had moved outside the spinal cord and displayed an elongated shape resembling nascent muscle fibers (Fig. 2, G and H). In contrast, the cells from clusters 1 and 2 remained in the spinal cord (Fig. 2, G and H). To confirm that the cells of clusters 3 and 4 had transdifferentiated into muscle, we fixed the regenerating tail and processed it for whole-mount immunohistochemistry with muscle-specific myosin heavy chain.Figure 2I shows that the cells were positive for the myosin marker and were lying within the area of the tail undergoing extensive muscle differentiation. This illustrates that radial glial cells originally derived from the ectoderm can switch lineage during tail regeneration and contribute to muscle, a cell type originally made from mesoderm.

Whether these radial glial cells were predetermined to form muscle before exiting the spinal cord and fusing with nearby cells is not yet known. However, we also observed labeled cells contributing to the cartilage rod that forms ventral to the regenerating spinal cord (Fig. 3). Differentiation into cartilage does not involve cell fusion. In this example, two cells were initially transfected in the mature spinal cord (Fig. 3, A to C). By day 8, cell 2 had divided within the regenerating spinal cord (Fig. 3, D to F), but then these cells exited out of the spinal cord, rapidly changed shape (day 9) (Fig. 3, G to I), and became morphologically distinct cartilage cells by day 11 (Fig. 3, J to L, arrowheads). Cartilage cells are mononucleate, and their fate could not be determined by fusion with other predetermined cells.

In 24% of the animals, we observed glial cells that exited the spinal cord to form muscle; in 12% of the animals, cells exited to form cartilage (Table 1). Because each animal contained multiple labeled cells at the start point of our experiments, this represents 20% and 8%, respectively, of the starting number of labeled cells. Other radial glial cells formed the expected neural cell types, including glia, neurons, and neural crest derivatives such as fin mesenchyme and melanocytes. During formation of the neural crest, cells took two exit routes from the regenerating spinal cord: they exited through the side walls of the spinal cord, as expected from previous histological results, and also out of the end of the growing tube (Table 1).

Neural stem cells have been described as having the potential to differentiate into neurons, astrocytes, and oligodendrocytes and to undergo self-renewal through the process of asymmetric cell division. Although individual cells may harbor such potential, our observations suggest that, in practice, a single neural progenitor may generate a limited number of cell types and may not even undergo self-renewal. We have observed that some cells proliferate to replenish the pool of radial glial cells, whereas others produce progeny that all differentiate. All the visible progeny of a cell apparently exit the spinal cord, leaving no self-renewing cell behind in the spinal cord (Figs. 2 and 3). It is not known whether this observation reflects differences in innate potential or in the extracellular cues encountered.

In the axolotl, we have observed that GFAP-positive radial glial cells have the ability not only to re-form a functional spinal cord but also to contribute to regenerating tissues outside the spinal cord such as muscle and cartilage. These experiments establish the relevance of neural cell plasticity to regenerating functional tissue in amphibia. The question remains whether mammalian neural stem cells also have the inherent ability to switch lineage. Is the contrast in regeneration ability between amphibians and mammals due to intrinsic differences in neural cell plasticity or to the environment the cells encounter in the injured tissue? In axolotls, tail regeneration occurs through the formation of the blastema—a zone of undifferentiated cells surrounding the regenerating spinal cord. It is likely that mammalian and axolotl neural stem cells are similar, but mammals lack the ability to form an inductive environment like the blastema that induces cells to switch lineage.

Supporting Online Material

Figs. S1 and S2

  • * To whom correspondence should be addressed. E-mail: tanaka{at}


View Abstract

Stay Connected to Science

Navigate This Article