Limits of regeneration

Why can a starfish regrow a whole arm when a person cannot regrow a lost finger?

A starfish loses an arm and grows another. A salamander loses a whole forelimb — bone, muscle, nerve, skin, the lot — and eighteen months later has a working forelimb with the right number of digits in the right order. A person loses a finger and gets a stump. It is tempting to read this as a ladder, with simple animals near the bottom and complex ones near the top having outgrown the trick. But that is suspicious. Why would evolution discard something so plainly useful?

Begin with what regrowing a limb actually demands. You need cells willing to divide in large numbers. You need those cells to be uncommitted enough to become several different tissues. And you need them to know where they are — that this stump is a mid-humerus and what is owed is an elbow, a forearm, a wrist and five digits, not simply more of whatever was cut. Three requirements: a proliferating population, flexible cell identity, and positional information.

Now look at what a salamander does. The wound is covered quickly by a thin layer of epidermis rather than a clot and a scab. Beneath it, cells near the cut re-enter the cell cycle and lose some of their specialised character, piling up into a mound called a blastema — and this mound is the thing mammals do not form. The blastema grows, then rebuilds the missing part in order from the stump outwards, because its cells carry a memory of position: transplant a blastema from a wrist to an upper arm and it makes wrist-and-hand, not the whole intervening limb.

A precision worth keeping: the blastema is not a pool of do-anything cells. Careful lineage tracing in axolotls showed that its cells are largely restricted — old muscle mostly rebuilds muscle, dermal cells rebuild skeleton and dermis. So the trick is not that salamanders keep embryonic cells lying around. It is that adult cells can be talked back into a proliferating, semi-flexible state, under a nerve-dependent signal, within a wound that permits it.

So what happens in a mammal instead? Speed happens. The wound clots, inflammation arrives fast, fibroblasts fill the gap with collagen, and a scar seals the breach in days. That is a different objective — close the hole before infection takes hold — and it is pursued so quickly and so thoroughly that the slow, orderly business of building a blastema never gets a chance to start. A scar is not a failure to regenerate. It is a competing programme that wins.

Which invites the obvious question: why prefer the fast, crude answer? The usual proposal is a trade-off with cancer. A large, long-lived, warm-blooded animal runs an enormous number of cell divisions over a lifetime, and any cell that can be persuaded to dedifferentiate and proliferate is precisely the cell most likely to become a tumour. Mammals carry heavy proliferation brakes, and those brakes plausibly cost them regeneration. It is a good hypothesis and probably part of the answer. It is not proven, and there is an awkward fact sitting in the middle of it: salamanders, which regenerate freely, appear to be strikingly cancer-resistant rather than cancer-prone. Whatever the trade-off is, it is not as simple as regeneration buying tumours.

Two corrections keep this honest.

First, mammals are not simply incapable. Human liver is the standard example, and it is usually described imprecisely: remove a large fraction and the mass and function return, but by the surviving cells enlarging and dividing throughout the remaining tissue — compensatory growth. The removed lobe itself does not grow back, and the organ ends up a different shape. That is restoration of quantity, not of form. Fingertips are the better example: in young children, and sometimes adults, a fingertip amputated beyond the base of the nail can regrow with nail and sensation, and it depends on cells associated with the nail bed. Cut below that level and it does not. And deer regrow antlers — genuine bone, cartilage, skin and nerve, annually. So the mammalian ceiling is not absolute; it is patchy and positional.

Second, the ladder framing is wrong in both directions. Regenerative ability is scattered across the tree of life rather than concentrated at its base, and closely related species differ. Some starfish regrow an entire animal from one arm carrying part of the central disc; others cannot. Zebrafish regenerate heart muscle; adult mice mostly scar, though newborn mice can do it for a few days after birth. That last observation is the most telling of all — the capacity is not absent from mammals, it is switched off early. Which suggests the limit we are trying to explain is less a lost machine than an actively suppressed one.

• What actually holds positional information in a limb, and how a blastema reads it.
• Why newborn mice can regenerate heart muscle and adults cannot — what closes in those few days.

Every term the collection defines is gathered in the glossary.