Developmental Neuroscience · Unit 10: Glia

Feeding the Axon: Glia as Metabolic Partners

An axon can reach a metre from its cell body, spends energy without pause to pump ions, and is mostly covered by myelin that blocks its access to the fluid around it. This lesson asks how such an axon is fed, weighs the evidence for and against the idea that astrocytes hand lactate to neurons, and follows the evidence that oligodendrocytes and Schwann cells supply the axons they wrap. Afterwards you can design and read the experiments that test whether a glial cell feeds an axon.

Watch first Glia Feed the Axons They Wrap · 3:08

Your brain is about 2 percent of your body weight and uses about 20 percent of the energy your body spends at rest. Much of that energy is spent in axons and synapses, and the axon is the part of a neuron farthest from where most of its supplies are made. A motor axon running from your spinal cord to your foot is about a metre long. Along most of that length it is wrapped in myelin, a membrane coat made by glial cells that blocks access to the fluid outside. So how is the far end of an axon fed?

This lesson’s answer is that glial cells help. The evidence is good for the glia that wrap axons and contested for the astrocytes around synapses, and you will see why.

The idea: an axon has a supply problem

Start with where the energy goes. Every action potential lets sodium in and potassium out, and every excitatory synapse does the same at a smaller scale. The Na+/K+-ATPase restores the gradients, spending one ATP to move three sodium ions out and two potassium ions in. Energy budgets worked out for rodent grey matter, beginning with Attwell and Laughlin in 2001 and revised since, put most of the energy used for signalling into this pumping, after action potentials and synaptic currents. The pumps never get a day off.

A quick refresher on cellular metabolism, since the rest of the lesson depends on it. Glycolysis splits one glucose into two pyruvate in the cytosol, with a net gain of two ATP, and turns NAD+ into NADH along the way. If the cell is to keep going, it must turn that NADH back into NAD+. Mitochondria can do it by taking pyruvate into the citric acid cycle and oxidative phosphorylation, which yields far more ATP. Or lactate dehydrogenase can reduce pyruvate to lactate, regenerating NAD+ in the cytosol. That reaction runs in both directions. A cell that takes up lactate can turn it back into pyruvate and burn it, so lactate works as a portable fuel. It crosses membranes on monocarboxylate transporters (MCTs), which also carry pyruvate. In rat liver the free NAD+-to-NADH ratio in the cytosol has been estimated at several hundred to one, which keeps glycolysis running in the forward direction; values in neurons depend on the cell and the measurement.

Now place that chemistry in a neuron. Consider the problem of metabolic compartmentalization in neurons:

  • The soma has dense mitochondria, makes nearly all of the cell’s proteins and has every pathway close at hand.
  • The axon is long and thin. Its ATP has to be made locally, by axonal mitochondria and axonal glycolysis, because ATP cannot diffuse a metre. The enzymes and mitochondria themselves are made in the soma and carried down by motor proteins that also spend ATP.
  • The synapse has a high, fluctuating demand from recycling vesicles and pumping ions after each release.

Myelin adds a twist. In a myelinated axon, compact myelin covers the long stretches between nodes of Ranvier, and the axon membrane meets the extracellular fluid mainly at those narrow nodes. Insulation that speeds conduction also stands between the axon and its glucose. The glial cell that makes the myelin, though, sits right against the axon along its whole length. That position makes it a candidate supplier.

How many glial cells are there to do this work? Older textbooks said glia outnumber neurons by ten to one. Counts of whole human brains by the isotropic fractionator method (Azevedo and colleagues, 2009) found roughly as many non-neuronal cells, most of them glia, as neurons overall, with glia outnumbering neurons in the cerebral cortex and far outnumbered by neurons in the cerebellum.

Controversies surrounding lactate and astrocytes

Astrocytes are the other candidate. They wrap synapses, take glutamate back up after release, and put end-feet on blood vessels, so they sit between the blood’s glucose and the neuron.

In 1994 Luc Pellerin and Pierre Magistretti gave glutamate to cultured mouse astrocytes. The cells took up more 2-deoxyglucose, a glucose analogue, and released more lactate. D-aspartate, an analogue taken up by the same transporters, had the same effect, and the authors concluded that receptor activation was not responsible. Blocking the sodium pump with ouabain blocked the response. Glutamate transporters bring sodium in with each glutamate, the pump has to remove it, and the pump’s demand for ATP drives glycolysis. From this they proposed the astrocyte-neuron lactate shuttle: an active synapse makes its astrocyte burn glucose to lactate, and the lactate goes to the neuron as fuel. Astrocytes were reported to export lactate mainly through MCT1 and MCT4, and neurons to take it up through MCT2.

The model is attractive because it couples supply to use, and because it would explain why glucose uptake rises where synapses are active. Some later in vivo work supports it. In 2016 Philipp Mächler, Bruno Weber and colleagues used a fluorescent lactate sensor in mouse cortex and reported a lactate gradient running from astrocytes to neurons.

The objections are serious. Critics, including Gerald Dienel, have argued that culture results do not transfer cleanly to the brain and that lactate’s near-equilibrium chemistry makes net flow hard to infer. In 2017 Gary Yellen’s laboratory imaged sensors of NADH and other metabolites in single neurons of mouse brain slices and live mice. Stimulation made neurons raise their own glycolysis, and the group saw no sign of lactate import during the stimulus. Both sides now agree that neurons can take up and use glucose directly. They disagree about how much of a neuron’s fuel comes through astrocytes, when, and in which brain region.

So treat the shuttle as a hypothesis with partial support. The answer may differ between resting and active states, between cortex and other regions, and between preparations.

Myelinating glia feed the axons they wrap

The case is firmer for oligodendrocytes, the myelinating glia of the CNS. It began with a puzzle from myelination research. Mice lacking the myelin protein PLP (1998) or the enzyme CNP (2003) made myelin that looked close to normal, yet their axons swelled and degenerated as the mice aged. The insulation was there; something else was missing.

Design problem

You want to test whether oligodendrocytes must respire to keep axons alive. You delete Cox10, a gene needed to assemble complex IV of the respiratory chain, only in myelinating cells, using a Cre line active in those cells. The oligodendrocytes can now make ATP only by glycolysis. What do you expect if oligodendrocytes must respire to support axons, and what result would instead support glycolytic oligodendrocytes feeding axons with lactate?

One way to do it

If oligodendrocytes must respire, they lose energy, and either they fail or the myelin and axons they serve fail. Over months you would expect thinning or loss of myelin, swollen and degenerating axons, and perhaps death of neurons whose axons run through the white matter.

If instead oligodendrocytes can live by glycolysis and pass its products on, the picture is the reverse. The oligodendrocytes survive, the myelin stays, the axons stay intact, and, because the glia now make more lactate than they can burn, lactate may build up in the tissue unless the axons use it as fast as it is made.

This is roughly what Ursula Fünfschilling, Klaus-Armin Nave and colleagues reported in 2012. In the adult CNS, once myelination was complete, myelin and axons were largely preserved over the months the mice were followed. Magnetic resonance spectroscopy showed more lactate in mutant than in control brains, but only in mice under isoflurane anaesthesia; it fell below detection within an hour after the anaesthetic ended. The peripheral nerves told a different story. Schwann cells lacking Cox10 failed to myelinate properly, and the mice developed a severe neuropathy. In these mice, then, oligodendrocytes could do without respiration and Schwann cells could not.

The Cox10 result shows that oligodendrocytes can live by glycolysis, and it fits with them feeding axons. It does not by itself show the feeding. That came from the transporter. In 2012 Youngjin Lee, Jeffrey Rothstein and colleagues reported that MCT1 is concentrated in oligodendrocytes. Lowering MCT1 in cultured slices of spinal cord, or in the spinal cord of mice, killed motor neurons, and in mice it also caused axon swellings. In the cultures, adding lactate protected neurons stressed by an MCT1 inhibitor and a short period without glucose. The group also found less MCT1 in tissue from people with ALS and from an ALS mouse model. Put together, the picture is of oligodendrocytes that take up glucose, run glycolysis, and export lactate or pyruvate through MCT1 to the axon, which takes it in and burns it in its mitochondria. The exact route from glial cytoplasm to axon is still being worked out, and so is how much of an axon’s ATP depends on it in a living animal.

After injury: when glial metabolism changes

In the peripheral nervous system, Schwann cells wrap axons the way oligodendrocytes do in the CNS. In 2011 the laboratory of Jeffrey Milbrandt reported that mice whose Schwann cells cannot maintain their mitochondrial DNA develop a neuropathy in which axons degenerate. Schwann cells also respond to injury with their own metabolic programs. After a nerve is cut, the pieces of axon cut off from the cell body degenerate, a process covered in injury and regeneration. In 2020 Elisabetta Babetto, Keit Men Wong and Bogdan Beirowski reported that mouse Schwann cells shift toward glycolysis after injury, and that this shift slows the degeneration of the injured axons.

The signals that steer such shifts are ones you know from mTOR and AMPK. AMPK senses a falling energy charge and turns on catabolism and autophagy. mTOR drives growth and lipid synthesis when nutrients and growth factors are plentiful. HIF1-alpha, a transcription factor stabilized at low oxygen, turns on glycolytic genes. In injured mouse nerve, mTORC1 and HIF1-alpha were reported to drive the glycolytic shift in Schwann cells; the role of AMPK there is less clear, and how they divide the work in a repairing nerve is not settled.

Satellite glia, which surround neuron cell bodies in sensory ganglia, change too. In 2020 the laboratory of Valeria Cavalli reported that after peripheral nerve injury in mice, satellite glia turn on genes for fatty acid synthesis, and that this program helps the sensory axons regrow.

Here is the big picture question: how might these metabolic changes in glia be beneficial or detrimental to neuronal survival and function? On the benefit side, a glial cell that makes more lactate may tide a stressed axon over, and a glial cell that makes lipids may supply membrane for regrowth. On the cost side, lactate comes with protons, so a large glycolytic output can acidify the tissue, and a glial cell that diverts its resources to its own repair program may give less to the axon. Which outcome wins probably depends on the cell, the injury and the timing, and that is still being measured.

How we measure it

Glucose analogue uptake in cultured cells

Cells are given 2-deoxyglucose, which is taken up and phosphorylated like glucose but goes no further and so stays trapped inside. The amount trapped measures glucose uptake. Pellerin and Magistretti used it to show that glutamate raises glucose uptake in cultured astrocytes.

Conditional knockout with Cre and lox

A gene is flanked by short lox sequences, and the enzyme Cre, made only in chosen cells, cuts it out there. The rest of the animal keeps the gene. Deleting Cox10 only in myelinating cells asked whether those cells need respiration, without harming the neurons.

Genetically encoded metabolite sensors

A fluorescent protein is engineered so that its brightness or colour changes when it binds a metabolite, such as lactate, or reports a ratio, such as NADH to NAD+. Expressed in one cell type, it reports that cell's metabolism in living tissue, which bulk chemistry cannot.

Organotypic slice culture with gene knockdown

A thin slice of tissue, here spinal cord, is kept alive in a dish with its cells in place. A virus delivers a short RNA, or an antisense oligonucleotide is added to the medium, to lower one gene, and neurons and axons are counted afterwards. Adding a metabolite to the medium tests whether a missing supply caused the loss.

Finish line

Test yourself

8 questions

Practice and MCAT-style, with explained answers.

Next · Lesson 46

Injury and Regeneration: Why the Brain Does Not Regrow

Cut a mouse nerve and the stump falls apart in about a day. In one strain of mouse it lasts for weeks. Why?

Continue →

← Lesson 44 · Myelination: Wrapping the Axon

Lactic acid — the fuel in question
Lactic acid — the fuel in question ● LIVE · PubChem CID 612
open on PubChem ↗
Loading live structure 612 from PubChem…

Three carbons, a hydroxyl and a carboxyl. At the pH of the body it is almost all lactate, the anion, which needs a transporter such as MCT1 to cross a membrane.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
NAD+ — the cofactor that lactate dehydrogenase recycles
NAD+ — the cofactor that lactate dehydrogenase recycles ● LIVE · PubChem CID 5892
open on PubChem ↗
Loading live structure 5892 from PubChem…

Look for the nicotinamide ring at one end: it accepts a hydride to become NADH. Glycolysis stalls without NAD+, and making lactate from pyruvate is one way to return it.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Human lactate dehydrogenase A — pyruvate to lactate and back
Human lactate dehydrogenase A — pyruvate to lactate and back ● LIVE · RCSB PDB 1I10
open on RCSB PDB ↗
Loading live structure 1I10 from RCSB PDB…

Four identical subunits, each with a bound NADH beside the substrate analogue oxamate. The same active site runs the reaction in either direction; which way it goes depends on the concentrations around it.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Human GLUT1 — glucose entry into glia and across the blood-brain barrier
Human GLUT1 — glucose entry into glia and across the blood-brain barrier ● LIVE · RCSB PDB 4PYP
open on RCSB PDB ↗
Loading live structure 4PYP from RCSB PDB…

Twelve membrane-spanning helices in two halves, with a cavity open to the inside of the cell between them. The transporter rocks between inward- and outward-facing forms to move glucose down its gradient.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Na+/K+-ATPase — where much of the brain's ATP goes
Na+/K+-ATPase — where much of the brain's ATP goes ● LIVE · RCSB PDB 3B8E
open on RCSB PDB ↗
Loading live structure 3B8E from RCSB PDB…

A large alpha subunit carries the ion sites and the site of phosphorylation, with a beta subunit and a small regulatory subunit beside it. Each cycle spends one ATP to move three sodium ions out and two potassium ions in.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Luc Pellerin

Luc Pellerin

With Pierre Magistretti, showed in cultured mouse astrocytes that taking up glutamate makes astrocytes take up more glucose and release lactate, and proposed that this lactate fuels active neurons: the astrocyte-neuron lactate shuttle.

In the early 1990s, brain imaging was built on a puzzle. Positron emission tomography tracked glucose use, and glucose use rose where neurons were active, but nobody could say which cells were burning the extra glucose or what linked the two. Pellerin, working in Pierre Magistretti's laboratory at the University of Lausanne, tested one candidate link in astrocytes grown in culture from mouse cortex. When the cells were given glutamate, the transmitter that excitatory synapses release, they took up more of a labelled glucose analogue and released more lactate. The effect followed glutamate transport, which carries sodium into the cell, and not glutamate receptors, and it needed the sodium pump. The pair proposed that every glutamate cleared from a synapse sets off glycolysis in the astrocyte, whose lactate then goes to the neuron as fuel. The proposal gave imaging a cellular mechanism to test, and it set off a debate about lactate that is still running.

✓ verbatim Pellerin & Magistretti, Proc. Natl. Acad. Sci. USA 91:10625 (1994).
Klaus-Armin Nave

Klaus-Armin Nave

His laboratory showed that mice whose oligodendrocytes cannot carry out mitochondrial respiration keep their CNS myelin and axons, evidence that oligodendrocytes can live by glycolysis and may pass its products to axons.

Nave's laboratory at the Max Planck Institute of Experimental Medicine in Göttingen came to metabolism through myelin genes. Mice lacking the myelin protein PLP (studied by Ian Griffiths, with Nave) or the myelin enzyme CNP made myelin that looked close to normal, yet their axons swelled and degenerated as the animals aged. If the axon loss was not caused by bad insulation, the glial cell had to be giving the axon something else. In 2012 Ursula Fünfschilling and colleagues in the laboratory deleted Cox10, a gene needed to assemble complex IV of the respiratory chain, in myelinating cells. In the adult CNS the oligodendrocytes survived, kept their myelin, and the axons they wrapped stayed intact over the months the mice were followed. Under anaesthesia, lactate became detectable in the mutant brains. The result showed that oligodendrocytes, once they have made myelin, can live on glycolysis alone, and it made glycolytic glia that pass lactate to axons a model that other laboratories could test.

✓ verbatim Fünfschilling et al., Nature 485:517 (2012); Lappe-Siefke et al., Nature Genetics 33:366 (2003).
Jeffrey Rothstein

Jeffrey Rothstein

His laboratory showed that the lactate transporter MCT1 is concentrated in oligodendrocytes and that reducing it damages axons and kills neurons in mouse and culture models.

Rothstein's laboratory at Johns Hopkins University studied amyotrophic lateral sclerosis (ALS), the disease in which motor neurons die, and for years its focus was glutamate: in 1992, with Lee Martin and Ralph Kuncl, he reported that tissue from people with ALS took up less glutamate than control tissue. That work was about glial cells clearing a transmitter. The 2012 paper led by Youngjin Lee turned to glial cells delivering fuel. Using reporter mice, the group found that MCT1, a transporter that carries lactate and pyruvate across membranes, is concentrated in oligodendrocytes. When MCT1 was reduced in cultured slices of spinal cord, or in the mouse spinal cord, axons degenerated and neurons were lost, and in the slice cultures lactate added to the medium prevented neuron loss when MCT1 was blocked with an inhibitor and glucose was withheld. The group also reported less MCT1 in tissue from people with ALS and from an ALS mouse model. The work tied a glial transporter to axon survival and to a human disease.

✓ verbatim Lee et al., Nature 487:443 (2012); Rothstein, Martin & Kuncl, New England Journal of Medicine 326:1464 (1992).
Gary Yellen

Gary Yellen

His laboratory built fluorescent sensors of cytosolic NADH and reported that stimulated neurons in mouse brain raise their own glycolysis instead of relying on imported lactate.

Yellen's laboratory at Harvard Medical School, long known for work on ion channels, came to brain energy through a measurement problem. Arguments about the lactate shuttle rested largely on cultured cells and on bulk measurements of tissue, which cannot say which cell is doing what. In 2011 the laboratory reported Peredox, a fluorescent protein whose signal reports the ratio of NADH to NAD+ in the cytosol of a single cell. Glycolysis makes NADH, and so does converting imported lactate to pyruvate, so the sensor can watch a neuron's fuel use as it happens. In 2017 Carlos Manlio Díaz-García and colleagues imaged such sensors in neurons of mouse brain slices and live mice. Stimulating the neurons raised their own glycolysis, and the group found no sign that the neurons were taking up lactate during the stimulus. The work gave the debate a single-cell measurement in intact tissue and a counter-model to set beside the shuttle.

✓ verbatim Díaz-García et al., Cell Metabolism 26:361 (2017); Hung et al., Cell Metabolism 14:545 (2011).
Test yourself
Question 1 of 8 Score 0 / 0

Why does a myelinated axon in the CNS have limited direct access to glucose in the extracellular fluid?
Sources & Integrity

Every structure, slide, simulation, and clip on this page is listed below with the source it came from. None of the 3D structures is drawn by hand or by AI: each one loads by accession ID from RCSB PDB or PubChem. This panel is generated from the lesson's own data file, so what it lists is what the page loads.

Molecular structures live data

  • Human lactate dehydrogenase A — pyruvate to lactate and back — RCSB PDB 1I10 · coordinates CC0 · Human muscle (M) lactate dehydrogenase with NADH and oxamate, X-ray crystal structure (Read et al., Proteins, 2001)
  • Human GLUT1 — glucose entry into glia and across the blood-brain barrier — RCSB PDB 4PYP · coordinates CC0 · Human glucose transporter GLUT1, X-ray crystal structure (Deng et al., Nature, 2014)
  • Na+/K+-ATPase — where much of the brain's ATP goes — RCSB PDB 3B8E · coordinates CC0 · Pig kidney Na+,K+-ATPase, X-ray crystal structure (Morth et al., Nature, 2007)
  • Lactic acid — the fuel in question — PubChem CID 612 · PubChem compound record for lactic acid (2-hydroxypropanoic acid)
  • NAD+ — the cofactor that lactate dehydrogenase recycles — PubChem CID 5892 · PubChem compound record for nicotinamide adenine dinucleotide (NAD+)

Scientist profiles & quotes

  • Luc Pellerin — Pellerin & Magistretti, Proc. Natl. Acad. Sci. USA 91:10625 (1994).
  • Klaus-Armin Nave — Fünfschilling et al., Nature 485:517 (2012); Lappe-Siefke et al., Nature Genetics 33:366 (2003).
  • Jeffrey Rothstein — Lee et al., Nature 487:443 (2012); Rothstein, Martin & Kuncl, New England Journal of Medicine 326:1464 (1992).
  • Gary Yellen — Díaz-García et al., Cell Metabolism 26:361 (2017); Hung et al., Cell Metabolism 14:545 (2011).

Quotes are verbatim from the cited source. Profile text is original to this site and was checked against the sources cited.

Animated explainer

Glia Feed the Axons They Wrap · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation: AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures, plots computed from their equations · captions from the narration script.

Images & figures

This lesson carries no raster figures. Every structure is rendered live from RCSB PDB or PubChem, and every plot is drawn from its equation in the browser.

integrity Machines check the mechanics. The build confirms that every local file exists, that every question's answer key points at a real choice, and that any AI-generated scientific figure names the person who checked it; a weekly job re-resolves every accession ID against its database. Questions are original; MCAT-style items are written for this site and are not real AAMC questions. Banned sources (the Tymoczko textbook, publisher decks, copyrighted figures) are not used.

Like I'm five

A nerve cell has a very long arm, and the far end of the arm is always working and always hungry. Most of the arm is wrapped in a coat made by helper cells, so it cannot grab food from around it. The helper cells that make the coat are thought to pass small bits of fuel through the coat to the arm. If they stop, the arm slowly breaks down.
High school

Neurons spend most of their energy pumping sodium and potassium back across their membranes after each signal. An axon can be very long, so fuel and proteins made in the cell body take a long time to reach its end. In many axons, myelin made by glial cells covers almost the whole surface. Glial cells can break glucose down to lactate, a small molecule that can be moved across membranes by transporters. Several lines of evidence suggest that myelinating glial cells pass lactate or pyruvate to the axons they wrap. Whether astrocytes also feed neurons lactate during activity is still argued about.
College

The Na+/K+-ATPase restores ion gradients after action potentials and synaptic currents and is a major sink for brain ATP. Axons are far from the soma and, in myelinated fibres, exposed to extracellular fluid mainly at nodes, so they depend on local mitochondria, local glycolysis and their glia. The astrocyte-neuron lactate shuttle proposes that Na+-coupled glutamate uptake drives astrocytic glycolysis and that the lactate is exported (via MCT1 and MCT4) and oxidized by neurons after entry through MCT2. Supporting data come from astrocyte cultures and some in vivo sensor work; opposing data show neurons raising their own glycolysis during activity. For myelinating glia the evidence is firmer: oligodendrocytes lacking complex IV assembly (Cox10 deletion) keep CNS axons intact, and loss of oligodendroglial MCT1 damages axons, in mice.
Graduate

The shuttle debate turns on where glucose is phosphorylated and whose NADH is reoxidized by lactate dehydrogenase. Lactate is in near-equilibrium with pyruvate through LDH and moves through MCTs down gradients, so net flux depends on the cytosolic NADH-to-NAD+ ratio and lactate concentration in each cell, not on LDH isoform alone. Single-cell sensors (Peredox, Laconic) have produced evidence both for an astrocyte-to-neuron lactate gradient in mouse cortex and for activity-driven neuronal glycolysis without lactate import, under different preparations and stimuli. For white matter, the Cox10 and MCT1 results fit a model in which oligodendrocytes export glycolytic products into the periaxonal space, possibly through cytoplasmic channels in non-compact myelin, though the route has not been established.

An open question is how much of an axon's ATP, in a given tract and at a given level of activity, comes from glia rather than from glucose the axon takes up itself, and whether that fraction changes after injury, when Schwann cells shift toward glycolysis.
Myelin that looked fine, axons that did not

Mice lacking the myelin protein PLP, and later mice lacking the myelin enzyme CNP, made myelin that looked close to normal under the electron microscope. Yet as they aged, their axons developed swellings and degenerated. A disease of myelin genes was turning out to be a disease of axons, with the insulation still in place. The simplest reading was that oligodendrocytes do more for an axon than insulate it, and the search for what else they provide led toward metabolism.

— Griffiths et al., Science 280:1610 (1998); Lappe-Siefke et al., Nature Genetics 33:366 (2003).
A glucose puzzle under the brain scanner

Brain imaging with labelled glucose analogues shows glucose use rising where neurons are active. That raises the question of which cells take up the glucose. Pellerin and Magistretti proposed in 1994 that much of it goes into astrocytes, which clear glutamate from synapses, burn the glucose to lactate, and pass the lactate on. If so, part of the signal in such scans would come from glial cells responding to neurons. The proposal is still tested and argued about.

— Pellerin & Magistretti, Proc. Natl. Acad. Sci. USA 91:10625 (1994).

 ↑ Top