Injury and Regeneration: Why the Brain Does Not Regrow

Cut the sciatic nerve of a mouse and the part of each axon beyond the cut looks normal for about a day, then falls to pieces. In one strain of mouse the same cut axons stay intact for two weeks or more.

A cut axon was long assumed to die the way a severed branch withers: cut off from its supply, it runs down. If one mutation can stop the process, something is running it. This lesson follows that idea from a frog’s tongue to an enzyme that destroys NAD+. Then it asks a second question with a less happy answer: once an axon is lost, why does it grow back in your arm and not in your spinal cord?

Augustus Waller and the case of the disappearing axon

The axon is the long process that carries a neuron’s output; the cell body holds the nucleus and makes most of the cell’s proteins. In a paper published in the Philosophical Transactions of the Royal Society in 1850, the physiologist Augustus Waller described cutting two nerves that serve the frog’s tongue, the glossopharyngeal and the hypoglossal, and following the fibres under the microscope. Over the following days the fibres beyond the cut broke up. The parts still joined to the brain stayed normal. Breakdown of the distal stump, the part cut off from its cell body, is still called Wallerian degeneration.

Waller read his result as a matter of nutrition: a fibre depends on a centre for its upkeep. That is true, and it suggests a passive death. Starve the stump and it fails.

Design problem

After a nerve is cut, is the degeneration of the distal stump passive starvation, because it has lost its supply from the cell body, or an active program of self-destruction? Design a genetic test, and say what result would decide it.

The slow mouse, and “Sarm null phenocopies wlds”

The first such mouse came by accident. In 1989 Lunn, Perry, Brown, Rosen and Gordon described a substrain of C57BL mice in which cut sciatic nerve axons degenerated very slowly. It became known as Wlds, for slow Wallerian degeneration, and crosses reported in 1990 showed that a single dominant gene was responsible. In 1998 Michael Coleman, Hugh Perry and colleagues found that the gene’s region on mouse chromosome 4 holds an 85-kilobase stretch of DNA in three tandem copies. In 2001 Coleman’s group showed that the rearrangement fuses the start of a ubiquitination factor, Ube4b, to NMNAT1, an enzyme that makes NAD+ from nicotinamide mononucleotide (NMN). Adding that fusion gene to normal mice protected cut axons, and more of it protected them more.

Then came a loss-of-function result of the kind the design problem asks for. In 2012 Marc Freeman’s laboratory and its collaborators reported that flies lacking a protein called dSarm keep severed axons for weeks, and that mice lacking its counterpart, Sarm1, keep cut axons intact for a long time, in the nerve and in culture. Deleting one ordinary gene gave protection like that of the odd fusion. The stump is not starving. It is being dismantled by a program it carries.

The program: from latency to catastrophe

The current model, built by several laboratories, has four steps.

1. A short-lived survival factor. NMNAT2 is the NMNAT isoform that works in axons. In 2010 Jonathan Gilley and Coleman showed that it is the least stable NMNAT isoform and must be replenished constantly by transport from the cell body (long-distance signaling). After a cut the supply stops, and NMNAT2 in the stump falls below a threshold before the axon breaks up. Blocking the proteasome slows its loss (ubiquitin and proteolysis).
2. A sensor. With NMNAT2 gone, its substrate NMN builds up and NAD+ begins to fall. SARM1 waits in the axon, held shut by its ARM domain. In the leading model the ARM domain binds either NAD+ or NMN, and a rise in NMN relative to NAD+ switches SARM1 on.
3. An executioner. The TIR domain of SARM1 is an enzyme. In 2015 the laboratories of Aaron DiAntonio and Jeffrey Milbrandt showed that forcing SARM1 TIR domains together is enough to destroy axonal NAD+ and trigger degeneration, and in 2017 that the TIR domain itself cleaves NAD+. ATP falls after NAD+.
4. Collapse. As energy fails, calcium rises inside the axon. Calpains, proteases switched on by calcium, cut the cytoskeleton, and the axon fragments.

Picture a building whose sprinkler tank must be topped up every hour by truck. Cut the road and the tank runs low. A sensor reads the low level and, instead of sounding an alarm, starts the fire. The Wlds protein works because it is a longer-lived source of NMNAT activity that keeps the tank full after the trucks stop.

This is not apoptosis, the death program of a whole cell (see Bcl-2 and caspases). Injury-induced degeneration of the stump proceeds when caspases are blocked or missing, while the developmental pruning of axons deprived of trophic factor needs them. The program also runs locally, in a stump with no nucleus. Stress kinases such as DLK and JNK take part after injury (stress signals), and where they fit in the sequence is still being worked out.

Why are axons so vulnerable to injury compared to the cell body? Part of the answer is scale. In an adult human, the axon of a motor neuron that moves a toe runs roughly a metre from the spinal cord, and its far end depends on proteins made in the cell body and carried all that way. A survival factor as short-lived as NMNAT2 makes the far end only as healthy as the transport that reaches it.

Go deeper: spheroids [Deppmann lab]. Before an axon breaks up, it swells into beads called spheroids. Working with sympathetic neurons from newborn mice, the Deppmann laboratory found that after withdrawal of nerve growth factor, axons stayed intact for about 18 hours (the latent phase) and then came apart over about 3 hours (the catastrophic phase). Toward the end of latency, calcium rose inside the axons, and calcium-rich spheroids grew and ruptured, releasing their contents. Medium from such axons hastened the catastrophic phase in others (Yong et al., J Neurosci 2019). After injury, the same group reported that spheroid growth depends on Rho and actin, that Wlds neurons and added NAD+ suppress it, and that Sarm1-null and DR6-null neurons still form spheroids that rupture (Yong et al., Sci Rep 2020). DR6 is a member of the TNF receptor family. The group had found that cut sciatic nerve axons in mice lacking DR6 were preserved up to four weeks after injury (Gamage et al., Curr Biol 2017). They propose that spheroid rupture marks the transition from latency to catastrophic degeneration. Are there convergent mechanisms between pathological and developmental regression? Spheroids and calcium appear in both; caspases do not.

Why the peripheral nerve regrows and the brain does not

In a peripheral nerve, losing the stump is not the end. The proximal stump, still attached to its cell body, forms a growth cone and regrows, often at around a millimetre a day in humans, depending on the nerve and the injury. Schwann cells, the glia of peripheral nerve, convert into repair cells under the control of the transcription factor c-Jun. They help break down myelin, recruit macrophages from the blood, and line up into tubes that guide regrowing axons back toward their targets. In the central nervous system, myelin debris is cleared far more slowly and incompletely.

Is the difference in the neurons or in their surroundings? In 1981 Samuel David and Albert Aguayo grafted segments of peripheral nerve into adult rats as bridges between the medulla and the spinal cord. Axons from central neurons at both levels grew about 30 millimetres along the grafts. Where they re-entered central tissue, they stopped within a short distance. Central neurons can regrow; their surroundings stop them. Three barriers have been studied most.

Myelin inhibitors. In 1988 Pico Caroni and Martin Schwab reported membrane proteins from central myelin that block neurite growth, and an antibody, IN-1, that neutralized them. The larger antigen was later cloned as Nogo-A. Nogo-A’s Nogo-66 domain, myelin-associated glycoprotein (MAG) and oligodendrocyte myelin glycoprotein (OMgp) all bind the Nogo receptor, NgR, which has no cytoplasmic domain and signals through co-receptors including p75 (p75) to activate RhoA and collapse the growth cone. Mice lacking these inhibitors regrow only modestly, so they are one barrier among several.

The glial scar. Reactive astrocytes wall off the lesion, and the scar is rich in chondroitin sulfate proteoglycans. Digesting their sugar chains with the bacterial enzyme chondroitinase ABC improved axon growth and recovery in rats with spinal cord injury. The scar also limits the spread of damage, so the aim is to change it, not remove it.

Low intrinsic growth. Adult central neurons turn their growth program down. In 2008 Zhigang He’s laboratory reported that deleting PTEN, a brake on the mTOR pathway, from retinal ganglion cells in mice let many of their axons regenerate through a crushed optic nerve. In sensory neurons, cutting the peripheral branch first raises the regrowth of the central branch, an effect called the conditioning lesion.

Fixing it

There are two problems here, and they need different fixes. The first is to keep axons from degenerating. Small molecules that inhibit SARM1 have protected axons in cultured neurons and in animal models, and raising NAD+ with precursors such as nicotinamide riboside or NMN is being explored. The model contains a catch: NMN is itself an activator of SARM1, so which precursor reaches the axon, and in what amount, matters. A fully severed stump cannot rejoin its cell body however long it survives, so the strongest case for these drugs is slower damage, such as some peripheral neuropathies, where failing axons are still connected.

The second problem is regrowth after the axon is lost, and no single fix is enough. Blocking Nogo signalling, digesting the scar, raising intrinsic growth and supplying bridges each help in rodents, usually partly, and none yet restores function reliably in people. Acute care for spinal cord injury today centres on surgery to relieve pressure on the cord and stabilize the spine, supportive care, and rehabilitation. High-dose methylprednisolone, a steroid, was once widely given in the first hours, but many current guidelines no longer recommend it routinely, because its benefit was small and uncertain and it carries risks.