Pruning Without Dying: Axons That Cut Themselves Back

When a fruit fly turns from larva into adult, each gamma neuron in its mushroom body loses two whole axon branches within about eighteen hours, and the neuron survives.

Later in metamorphosis the same neuron grows a new adult branch. In that interval it took apart a large piece of itself with machinery that could have dismantled the whole cell, and it stopped at the right place. This lesson asks how a neuron cuts back an axon without dying, and how it decides which branch goes.

Progressive events build the “draft” of the nervous system; regressive events sculpt it

Professor Deppmann divides development into two kinds of events. “Progressive events build the ‘draft’ of the nervous system”: neurons are born, migrate, form layers, and send out axons that find their targets. “Regressive events sculpt neural circuits”: whole neurons die, synapses are removed, and axons and dendrites are pruned. Think of a sculptor who first piles up more clay than the figure needs, then carves.

Pruning has a narrow meaning here: removal of an axon branch or a dendrite from a neuron that survives. That separates it from the neuron death of that lesson, where the whole cell goes. The word covers two physical processes. In retraction, a branch pulls back toward the cell body and its material is reabsorbed. In local degeneration, the branch breaks into pieces in place, and other cells engulf the pieces. A good description of any pruning event says which one happened.

Pruning has two broad triggers. In stereotyped pruning the branch to be removed is set by a genetic program, so the same branch goes in every animal. In competitive pruning the outcome depends on which axon gets more trophic support, a growth factor supplied by the target. Both appear below.

Stereotyped pruning of long axon branches requires a program and a repulsive cue

Start with the fly. Liqun Luo’s group used MARCM, a genetic method that labels one or a few neurons in an otherwise unlabelled brain and can make only those neurons mutant. In 2000 Tzumin Lee, Luo and colleagues showed that a gamma neuron needs the steroid receptor EcR-B1, working with its partner USP, inside itself to remodel. The ecdysone receptor is a nuclear receptor, a ligand-activated transcription factor of the kind covered in that lesson. The hormone 20-hydroxyecdysone rises at metamorphosis, and within the mushroom body EcR-B1 is made in the gamma neurons, which need it to prune. The hormone gives the timing; the receptor helps decide which neurons respond.

How is a branch taken apart? In 2003 Ryan Watts, Eric Hoopfer and Luo fixed fly brains at two-hour intervals from the start of metamorphosis to eighteen hours after it. The branches blebbed, then appeared as separate fragments, then were gone, while the main axon shaft stayed continuous. Separated pieces are what local degeneration predicts, since a retracting branch would stay connected to the cell. Blocking ubiquitin tagging, or the proteasome that destroys tagged proteins, kept the larval branches in place. The requirement was inside the pruning neuron. In the same study, blocking the fly’s main apoptosis activators or expressing the caspase inhibitor p35 did not stop this axon pruning, so the fly mushroom body does not appear to rely on the usual apoptosis machinery. The proteasome is covered in that lesson.

Mammals use guidance cues as well. “Repulsive cues are essential in sculpting the CNS (same repulsive cues that participate in the initial assembly).” In the mouse hippocampus, the axons of dentate granule cells, the mossy fibres, form a main bundle and a short infrapyramidal tract. After birth the infrapyramidal tract is cut back. In 2003 Anil Bagri, Marc Tessier-Lavigne and colleagues found that in mice lacking neuropilin-2 or Plexin-A3, parts of the Sema3F receptor, the tract stayed long, and later work showed that Sema3F itself is required. Cultured hippocampal neurons retracted axon branches when exposed to semaphorins, so this example is pruning by retraction. The cue that repels axons during assembly is used again to trim them; that lesson covers semaphorins and their receptors. Hence the professor’s heading: “Stereotyped Pruning of long axon branches require repulsive cues.”

Winners and losers: pruning by competition

Many sympathetic neurons send branches toward targets that make the trophic factor NGF. As that lesson shows, the neurons compete, and the professor calls the result a “zero sum” contest: a branch that gets more support is stabilized, and one that gets less is weakened and removed.

What removes the loser? In 2008 Singh, David Kaplan, Freda Miller and colleagues studied sympathetic neurons in compartmented cultures and in mice. Axons with strong signalling through TrkA, the NGF receptor tyrosine kinase, were protected. Winning neurons made BDNF, a neurotrophin that binds p75, a receptor of the TNF-receptor family. On axons with weak TrkA signalling, BDNF acting through p75 caused degeneration. In the lecture’s version, winning axons release BDNF in an activity-dependent way. A winner therefore does more than take the NGF. It also sends a signal that removes the loser, and its own TrkA signalling makes it resistant to that signal. The pairing of TrkA with p75 is the subject of that lesson.

Using the death machine locally

Caspases are proteases that dismantle a cell during apoptosis, and the Cell Signaling course covers how they are activated in that lesson. How can a neuron run them in one branch and not in its cell body, which shares the same cytoplasm?

Fly sensory neurons that prune their dendrites at metamorphosis gave the first answer. In 2006 Darren Williams, James Truman and colleagues used a genetically encoded reporter to find caspase activity. It appeared only in the dendrites being removed. Suppressing caspases kept the dendrites; activating caspases throughout the neuron killed it. In the same year Chay Kuo, working with Lily Jan and Yuh-Nung Jan, found that the ubiquitin-conjugating enzyme UbcD1 tags DIAP1, an inhibitor that holds the caspase Dronc in check, for destruction by the proteasome. Remove the minder in the branch and the caspase runs there. Keep the minder in the cell body and the cell lives.

Mammalian axons separate the two programs in another way. In 2013 Corey Cusack, Mohanish Deshmukh and colleagues built a microfluidic chamber in which the cell bodies of mouse sympathetic neurons sat in one compartment and their axons grew through narrow grooves into another. When NGF was removed from the axons alone, the axons degenerated and the cell bodies lived. This pruning needed caspase-9, caspase-3 and caspase-6, but it did not need Apaf-1, the protein that builds the apoptosome, the usual activator of caspase-9 in apoptosis. When NGF was removed from the whole dish, the neurons died, death needed Apaf-1, and loss of caspase-6 no longer protected the axons. The two routes share enzymes but are wired differently.

Design problem

You have a two-compartment chamber and sympathetic neurons from wild-type mice and from Apaf-1-null mice. In one set of chambers you remove NGF from the axon compartment only. In another you remove NGF from the whole dish. Predict, for each genotype, which axons degenerate and which neurons die.

Latency to catastrophe: what spheroids do

The professor’s laboratory asked what happens over time in an axon deprived of NGF. Their answer, from Yong, Gamage and colleagues (J Neurosci 2019), divides the process into phases. In a latent phase the axon looks nearly normal. In a transition, late in latency, swellings called axonal spheroids appear. In a catastrophic phase the axon fragments. “Local Versus Global NGF Deprivation have Different Latent Phases”, so how the axon was deprived changes how long it waits.

“So, what’s the deal with these spheroids?” Live imaging showed calcium rising in axons before catastrophic degeneration and gathering in spheroids; calcium signalling is covered in that lesson. Spheroids, but not the rest of the axon, exposed phosphatidylserine, a membrane lipid that phagocytes recognize. Then they ruptured. Dextrans of 3 and 10 kilodaltons entered, but 70-kilodalton dextran did not, and the calcium signal was lost as the dextran entered. Medium collected from deprived axons sped up entry into catastrophe in fresh deprived axons.

Two TNF-receptor family members split the work. “p75NTR but not DR6 is Required for Spheroid Formation.” p75 acted through Rho, a small GTPase that reorganizes actin, covered in that lesson; blocking Rho or actin dynamics prevented spheroids, and forcing Rho on restored spheroids and degeneration in p75-null axons. DR6, in turn, was needed for the catastrophic degeneration that conditioned medium drives, and p75 was not. A later study from the laboratory (Yong and colleagues, Sci Rep 2020) found that, after injury, adding NAD+ or expressing WldS suppresses Rho-dependent spheroid formation. The take-home: “Spheroids gate the transition from latency to catastrophic degeneration.” DR6 has a separate history. A 2009 model proposed that a fragment of APP binds DR6 to trigger pruning; in 2014 Olsen and colleagues revised it, finding that APP and DR6 act in one pathway that does not need the beta-secretase cut.

Development and injury differ. “Caspases required for developmental degeneration but not injury.” After a nerve is cut, the detached part of the axon undergoes Wallerian degeneration. Lunn and colleagues described the WldS mouse, whose cut axons survive far longer, in 1989; Coleman’s group later identified the gene. WldS and loss of SARM1 protect strongly against injury-induced degeneration, WldS does not protect against developmental pruning, and whether SARM1 contributes to degeneration after trophic deprivation is still debated. Gamage and colleagues (Curr Biol 2017) found that DR6 promotes degeneration after injury, which raises the professor’s question, “Are there convergent mechanisms between pathological and developmental regression?” The injury pathway itself belongs to that lesson.