Cut off an axolotl's arm. Within hours, the wound closes — epidermal cells slide over the cut surface and form a cap. Over the next few days, cells in the stump begin to change. Fibroblasts, Schwann cells, satellite cells loosen their differentiated identities and accumulate beneath the cap as a proliferating mass called the blastema. The blastema looks, under a microscope, like nothing in particular — a cluster of small, pale, morphologically undifferentiated cells. It looks like an early embryo. It looks like something that hasn't decided what it is yet.
It has already decided. It just doesn't look like it.
In 2018 the Tanaka lab ran single-cell RNA sequencing on axolotl blastema cells from multiple limb levels — wrist, elbow, shoulder — and found what they called a molecular funnel. The diverse connective tissue subtypes that enter the blastema (fibroblasts, osteoblasts, pericytes, tenocytes, all with distinct transcriptional identities) converge, in the early blastema, on a single shared gene expression state. They look, transcriptionally, like undifferentiated progenitors. The funnel takes them in as distinct and outputs them as seemingly equivalent. This is what was once called dedifferentiation, and it's real.
But something doesn't funnel. A 2024 paper in Developmental Cell tracked chromatin accessibility and histone modification patterns across blastema cells from different limb segments and found that the positional identity — the information encoding whether this cell came from the upper arm or the hand — is carried not in gene expression but in histone marks. Specifically H3K27me3, a repressive modification at Hox and MEIS gene loci. Wrist-derived cells have different H3K27me3 patterns at these loci than shoulder-derived cells. The marks persist through the transcriptional convergence. The cells look like the same blank state, but their chromatin is differently annotated. Each cell is carrying a quiet record of where it came from.
This is the answer to a question that confused limb regeneration research for decades. If you transplant a wrist-level blastema to a shoulder amputation site, it regenerates a wrist, not a shoulder. The cells remember. You can override this with retinoic acid — flooding the cells with RA proximalizes them, pushes their identity toward the upper limb — but the default behavior is to regenerate what was lost, no more and no less. For years it wasn't clear where that information lived, since the cells looked undifferentiated. Now it looks like the information never left. It was just moved out of transcription and into the chromatin layer. The cell gene expression was quieted; the annotation was kept.
The other finding that interests me: the same cells, in a different immune environment, produce entirely different outcomes. James Godwin's 2013 PNAS paper depleted macrophages from axolotls before amputation by having the animals phagocytose clodronate-filled liposomes, which killed macrophages as they engulfed them. All depleted animals failed to regenerate. Not impaired regeneration — complete failure. The stumps formed fibrotic scar tissue, dense and disorganized, like a mammalian wound response. Then Godwin re-amputated these animals after their macrophage populations had recovered. The same stumps — the ones that had scarred — regenerated normally. The regenerative capacity wasn't destroyed. It had just gone unused.
What macrophages do in early regeneration is still being worked out. They clear debris, resolve inflammation rapidly, and probably signal the structural cells in ways not fully characterized. What the experiment establishes is that the fibroblasts and Schwann cells and satellite cells are not, themselves, the limiting factor. They have the regenerative program. What determines whether it runs is the context those cells are in during the first few days after injury. The decision between scar and regeneration is not made by the cells that will do the regenerating. It's made earlier, by cells whose job is to read the wound and set the conditions.
Put these two findings together and you have a picture of regeneration as something that requires both preserved information and the right context to read it. The blastema cells are carrying encoded positional identity through the apparent dedifferentiation — the histone marks persist even when the transcriptional program goes quiet. But whether that information gets used — whether the cells ever reconstitute a limb at all — depends on an earlier decision by the immune system. The annotation is there; the macrophages determine whether anyone acts on it.
I keep returning to the asymmetry: what axolotls can do that mammals can't is not exotic. The axolotl's cells are not special in their composition. They have the same basic toolkit. What seems to differ is the early inflammatory response — the axolotl resolves it faster, runs pro- and anti-inflammatory programs simultaneously rather than sequentially — and the capacity to hold positional identity in a form that survives dedifferentiation. Two adjustments, one to timing and one to where certain information is stored, and the outcome switches from scar to arm.
There's no general lesson here I want to force out of this. It's just a striking mechanism: cells that look blank but aren't, and a capability that exists but depends on the right environment to run. The axolotl's arm is already annotated. It just needs the macrophages to say go.