Remade in Place

Keeping by doing what the tree cannot — a salamander that restores a lost limb in the place it occupied, bone, muscle, nerve and skin in their old pattern, rebuilt from a wound-end blastema whose growth depends on nerves

R946 closed with a tree that never undoes an injury: it walls rot into old wood and generates new wood somewhere else. This page takes the turn it pointed to. An axolotl (Ambystoma mexicanum) answers amputation in the opposite way. The missing part is not sealed off and left behind. It is remade, in place, in pattern, and — in the usual account — without a scar. What is kept is not the damaged tissue. What is kept is the form.

First, a cover that is also a signal. Within hours, epidermal cells migrate over the stump and close the wound. That wound epithelium thickens into a signalling layer, the apical epithelial cap (AEC). It is not a passive bandage. Work on the accessory-limb model — a skin wound plus a nerve rerouted to it, with no amputation at all — shows that wound epithelium and nerve together can be enough to start a blastema and, in that model, an extra limb. Remove either partner and the programme does not run in the normal way.

Second, two keepers the stump cannot do without. The regrowth is nerve-dependent, a fact known since the nineteenth century and quantified by Singer in the twentieth: below a threshold density of nerve fibres, an amputated salamander limb heals over instead of regenerating, and the type of nerve matters less than the amount. In newts, part of that nerve signal has a name — the protein nAG, supplied by Schwann cells around regrowing axons, can drive blastema cells to divide, and supplying nAG to a denervated newt limb rescues regeneration (Kumar and colleagues, 2007). Immune cells are the other keeper. When macrophages are depleted early in an axolotl, limb regeneration fails and the stump scars instead; with macrophages present, debris is cleared and the scar-free programme proceeds (Godwin and colleagues, 2013). Regeneration, on this evidence, is not a default the body falls into. It is a guarded outcome that two different systems have to allow.

Third, a blastema that remembers. Under the AEC, local cells near the cut re-enter the cell cycle and gather into the blastema, a bud-like mass at the stump tip; a recognisable bud is present within days to a couple of weeks, varying with the animal and the cut. The older picture — that stump cells dedifferentiate all the way back to pluripotency and start again as blank material — did not survive lineage tracing. Kragl and colleagues (2009), tracking marked tissues in the axolotl, found the blastema is instead a mixed company of restricted progenitors: cells keep a memory of their tissue origin, cartilage-lineage cells making cartilage-lineage descendants, Schwann cells making Schwann cells, dermis contributing chiefly to connective tissue and cartilage rather than to muscle. The blastema is not a reset. It is a muster of specialists who have loosened their state without forgetting their trade.

Fourth, only the missing part. A blastema does not rebuild a whole limb on top of what remains. Cut at the wrist and it makes a hand; cut higher and it makes everything below that higher cut, and no more. Blastema cells therefore carry positional information — where along the limb they sit — and that information can be shifted. Early blastema cells are positionally labile; exposure to retinoic acid can reprogramme them toward a more proximal identity, so a wrist-level blastema can be made to produce a fuller limb. Later work (Duerr and colleagues, 2025) puts part of the mechanism on retinoic-acid breakdown by the enzyme Cyp26b1: levels of the signal, set by how fast it is destroyed in different zones, help assign proximal and distal identities. Pattern, here, is not drawn once and copied. It is renegotiated, in a window, and then stabilised as cells redifferentiate.

Last, a miniature that grows up. From the blastema’s proximal side, cells begin redifferentiating while the tip is still building — cartilage condenses, joints separate, muscle, nerve, vessels and skin return in their relations. Reviews commonly put a complete, patterned replacement at roughly five to seven weeks for the animals and cuts studied, first as a small replica that then enlarges to match the body. That is the contrast with R946 stated plainly: the tree keeps by confining loss and growing elsewhere; the axolotl keeps by remaking the loss itself, in its own place, out of cells that were already there and never stopped knowing what they had been.

A walk from a cut to a limb. Five stages: the wound closed in hours by a wound epithelium; that epithelium thickened into the AEC, with nerve and macrophage as required partners; the blastema, a muster of lineage-restricted progenitors that keep their tissue memory; positional identity renegotiated, so only the missing distal part is specified; and redifferentiation into a miniature limb that grows to size. Schematic, compressed from the published accounts.

Walk it in order.

Schematic only: the stages compress work across axolotls and newts, different ages, cut levels and temperatures, so no single animal runs this exact timetable. Labelled cautions: do not call blastema cells pluripotent — the lineage-tracing result is that they are restricted and keep tissue memory. Do not call nerve dependence absolute in every salamander ever made — limbs that developed without nerves (aneurogenic limbs) are the classic exception, regenerating without the usual requirement. Do not present nAG as the whole nerve signal — it is one identified factor, demonstrated in newts, among several growth signals. Do not present the five-to-seven-week figure as a constant; size, age, species and conditions move it. And do not read regeneration as unlimited: it is patterned and repeatable for limbs, and still has failures and limits — the same axolotl that regrows a limb does not, for instance, bridge every large bone defect placed in it.