Neurons and Glia at Work

When a small patch of brain fires hard, potassium leaking from its neurons can pile up outside them fast enough to shift E_K by tens of millivolts. Something has to put it back.

Astrocytes, a kind of glial cell, do a large share of that work. This lesson takes the neuron first, compartment by compartment, then the glial cells that keep it working, and ends by separating what glia are known to do from what is still argued. How these cells were discovered and built belongs to Developmental Neuroscience (the neuron doctrine, origins of glia); here you meet them at work in the mature brain.

One neuron, five jobs

Think of a neuron as a relay station with five rooms, each doing one job with the signal. The scheme is the canonical one, drawn from cortical and hippocampal pyramidal neurons and cerebellar Purkinje cells; there are exceptions.

Dendrites receive input. A synapse onto a dendrite produces a postsynaptic potential, a local change in membrane voltage whose size depends on how much transmitter arrived. It is graded: bigger input, bigger potential, with no threshold. An excitatory one is an EPSP, a depolarization that moves the cell toward firing. In a passive dendrite it shrinks as it spreads, the way a ripple fades from where a stone fell. The next lesson, on the resting membrane and passive signals, puts numbers on how far it travels.

The soma, or cell body, holds the nucleus and most of the machinery that makes proteins. It is also a place where inputs meet. Spatial summation adds potentials arriving at different synapses at about the same time; temporal summation adds potentials arriving at one synapse in quick succession. Integration is not the soma’s alone, since dendrites sum and transform inputs before they arrive.

The axon initial segment (AIS) is the first stretch of axon, packed with voltage-gated sodium channels. That density makes it the easiest place to reach threshold, so it is where the action potential usually starts. The next section shows where, exactly, in one cell type.

The axon carries the spike away, regenerating it as it goes. In mature mouse layer 5 pyramidal neurons, voltage imaging showed the spike propagating saltatorily, from node to node, toward the axon’s branches. The axon also carries cargo both ways on microtubule tracks, described in molecular motors.

The terminal turns the spike back into chemistry. Depolarization opens voltage-gated calcium channels, and calcium entry triggers fusion of transmitter-filled vesicles with the membrane. At fast synapses this release can follow the spike in less than a millisecond, though the delay is not the same at every synapse. Synaptic transmission takes this apart.

Now the exceptions: some mammalian neurons fire spikes in their dendrites, and the trigger zone is not always where textbooks put it.

Where the spike starts

Many textbooks say the action potential begins at the axon hillock, the cone where the axon leaves the soma. Lucy Palmer and Greg Stuart tested this in acute slices of rat somatosensory cortex, in layer 5 pyramidal neurons from rats usually three to five weeks old. A somatic recording cannot answer the question, since a spike starting tens of micrometres away arrives almost at once, so they imaged the axon at several points simultaneously with a voltage-sensitive dye.

The earliest signal appeared about 35 µm from the hillock, in the distal part of an initial segment about 40 µm long. Experiments targeting the first node suggested it is not involved in initiation. The trigger zone is a measurement in one cell type, not a rule for every neuron. How the AIS is assembled is covered in neuronal polarity; what it does, in the action potential.

Astrocytes keep the space clean: K+ and glutamate

Every action potential lets potassium out of the neuron. The space between brain cells is narrow, so during intense firing potassium can build up in it, and because a neuron’s resting potential depends largely on potassium, the build-up depolarizes it.

An equilibrium potential is the voltage at which an ion’s concentration gradient and the electrical force on it balance; the Nernst equation gives it, and is derived in membrane transport. At 37 °C, for potassium:

E_K = 61.5 mV × log10([K+]o / [K+]i)

Worked problem. Take [K+]i = 140 mM, steady, and treat potassium on its own. The values are illustrative.

Step 1: rest. With [K+]o = 3 mM, the ratio is 3/140 = 0.0214, and log10(0.0214) = −1.669. So E_K = 61.5 × (−1.669) = −102.6 mV.

Step 2: after intense firing. Suppose a small extracellular space reaches 12 mM. The ratio is 12/140 = 0.0857, log10(0.0857) = −1.067, and E_K = −65.6 mV. That is 37.0 mV more positive. A fourfold rise always shifts E_K by 61.5 × log10(4) = 37.0 mV, whatever [K+]i is. The resting potential of a real neuron moves less than E_K, because other ions contribute to it, but it moves the same way, toward threshold.

Step 3: clearance. Suppose astrocytes remove the excess potassium with a single-exponential time constant of 2 s (an illustrative value, not a measured one). The excess falls to 10 percent when e^(−t/2 s) = 0.1, so t = 2 s × ln(10) = 2 × 2.303 = 4.6 s.

Step 4: read a real number correctly. In mouse hippocampal slices, Tyurikova and colleagues (2025) measured the decay of the potassium current recorded in astrocytes after a train of five stimuli at 50 Hz: 1.85 s in control astrocytes and 1.13 s in astrocytes overexpressing Kir4.1. The decay is (1.85 − 1.13)/1.85 = 38.9 percent shorter. That is a property of a current, not a time constant for extracellular potassium, so do not drop it into Step 3.

How do astrocytes do it? In spatial buffering, potassium enters an astrocyte where the extracellular concentration is high, travels through it and through the network of astrocytes joined to it by gap junctions, and leaves where the concentration is low. Think of a sponge that soaks up a spill at one end and wrings out at the other. The entry channel is largely Kir4.1, an inwardly rectifying potassium channel. Separately, the sodium-potassium ATPase brings potassium into glia for net uptake. Redistribution and net uptake are distinct mechanisms.

The genetic evidence comes from the laboratory of Ken McCarthy. Djukic and colleagues deleted Kir4.1 from glia and found that CA1 astrocytes in hippocampal slices from mice 15 to 20 days old took up potassium poorly. They also took up glutamate poorly. That does not mean Kir4.1 carries glutamate: glutamate transport is driven by ion gradients and depends on the astrocyte’s voltage, which Kir4.1 helps hold negative. The deletion began in development and also reached oligodendrocytes, so the result cannot be pinned on mature astrocytes alone.

Glutamate, the main fast excitatory transmitter, must be cleared from outside synapses too. The transporter GLT-1 (also called EAAT2) is mostly astrocytic, with a smaller pool in excitatory terminals. In mice, Petr and colleagues deleted it from astrocytes with an inducible Cre, and the animals had seizures and early death. Deleting it from neurons cut uptake into crude forebrain synaptosomes (pinched-off nerve terminals) by about 40 percent, which shows how much one assay can favour a minor pool.

Myelinating cells and microglia at work

Myelin is a membrane wrapped many times around an axon. It gives the axon a high-resistance, low-capacitance sheath, so current spreads farther between the gaps called nodes of Ranvier, and the spike jumps from node to node. One myelinated stretch between two nodes is an internode. In the peripheral nervous system a myelinating Schwann cell makes exactly one internode. In the CNS one oligodendrocyte can make several, often on different axons; there is no fixed count to memorize. Some Schwann cells never myelinate: Remak cells hold small unmyelinated axons. How myelin is made is in myelination.

Myelin is not frozen in adults. Hughes and colleagues imaged adult mouse somatosensory cortex for weeks (the sheath-length figure is from mice 10 to 14 months old) and saw newly generated oligodendrocytes add new sheaths, while only about 1 percent of existing sheaths changed length over seven weeks.

Oligodendrocytes also feed axons. Lee and colleagues knocked down the lactate transporter MCT1 in oligodendrocytes of the optic nerve, and the axons were damaged; in slice culture, added lactate prevented injury when MCT1 was blocked and glucose withheld. The fuller story is in glia as metabolic partners.

Microglia are the brain’s resident immune cells. Using two-photon imaging in healthy adult mouse neocortex, Nimmerjahn, Kirchhoff and Helmchen saw them in their presumed resting state with cell bodies in place but processes that extended and retracted continually, surveying the tissue. Their role in removing synapses during development is told in microglia and complement pruning.

What is still argued: gliotransmission

Gliotransmission is the proposal that astrocytes release transmitters, such as glutamate, that act quickly on synapses. Astrocytes show calcium signals (see calcium signalling), and a calcium rise could in principle trigger vesicle release, as it does in a terminal.

The evidence conflicts. Agulhon and colleagues raised or disrupted astrocyte calcium signalling in mouse hippocampus, including by deleting the IP3 receptor type 2, and found no change in excitatory transmission or plasticity. But Srinivasan and colleagues later showed that adult mouse astrocytes lacking that receptor still have calcium fluctuations, mostly in their processes, in slices and in vivo. So the null result does not exclude effects driven by the remaining calcium.

On the other side, de Ceglia, Volterra and colleagues found a specialized subgroup of adult mouse hippocampal astrocytes expressing synaptic-like release machinery, including VGLUT1, and showed that deleting VGLUT1 from astrocytes suppressed the glutamate hotspots linked to that subgroup. That is positive evidence, restricted to one subgroup, in one species. Neither study refutes the other: they used different manipulations and asked different questions.

Design problem

Do adult hippocampal astrocytes release glutamate that changes synaptic transmission? You know that IP3R2 knockout found no effect on transmission or plasticity, and that a VGLUT1-expressing astrocyte subgroup shows glutamate hotspots that astrocytic VGLUT1 deletion suppresses. Design an experiment in mice that could show whether release from those astrocytes changes transmission, and say what your result could and could not establish.