Neuron · Grey Matter

A neuron is the cell that carries signals in the nervous system: it gathers inputs on its dendrites, adds them up across its membrane and, when the sum is large enough, sends an electrical pulse down its axon to the cells it contacts.

Neuron. A neuron is the cell that carries signals in the nervous system: it gathers inputs on its dendrites, adds them up across its membrane and, when the sum is large enough, sends an electrical pulse down its axon to the cells it contacts.

The adult human brain holds about 86 billion of them, and roughly as many glial cells beside them. Each neuron is a small electrochemical device. Pumps in its membrane keep sodium high outside and potassium high inside, which leaves the inside about 70 mV negative at rest; channels that open and close with voltage or with a chemical turn that stored gradient into signals. Inputs arrive as small graded changes of voltage at synapses, and the output is a single kind of event, the action potential, which either happens in full or does not happen.

In the cortex, neurons come in two broad families that the rest of this theme keeps returning to.

Pyramidal cells are the excitatory majority (most estimates put them near 75 to 80 % of cortical neurons in rodents, fewer in humans). They release glutamate, send long axons to other areas and are aligned side by side with their apical dendrites toward the surface, which is why their currents add up into what an EEG records.

Interneurons are the inhibitory minority. They release GABA, keep their axons local and set the timing and the gain of the pyramidal cells around them; several types exist, distinguished by shape, by the proteins they carry and by which part of the pyramidal cell they target.

The ratio is a regional and species property. Electron-microscopy counts in human cortex find a larger share of inhibitory neurons than in mouse, so the 80 to 20 split is a rule of thumb for rodent cortex rather than a constant.

A neuron's output is binary, its input is analogue.

Everything graded (the strength of each synapse, the balance of excitation and inhibition, the neuromodulators in the background) acts by moving the membrane toward or away from the threshold of a single all-or-none event.

The spike itself is the subject of the action potential card, the junction between two neurons of the exocytosis card, and the inhibitory side of the circuit of the GABA and excitation-inhibition balance cards.

Questions: Does the adult human brain make new neurons? Possibly, in one place, and how much is disputed. Rodents keep making new granule neurons in the dentate gyrus of the hippocampus throughout life, and carbon-14 dating of human hippocampal DNA suggested about 700 new neurons per day in each hippocampus in adults. In 2018 two studies looking for immature neurons in post-mortem human tissue reached opposite conclusions, one finding them into old age and the other almost none after childhood, and differences in tissue handling and patient groups explain part of the conflict. Either way, the number is tiny next to the brain's 86 billion neurons, and almost all adult change happens in synapses. What makes grey matter grey and white matter white at the level of cells? Grey matter is where neurons do their local work: it holds the cell bodies, dendrites, synapses and short local axons, together with astrocytes, microglia and a dense web of capillaries, and it looks pinkish grey in living tissue. White matter is made of bundles of long axons that connect distant regions, most of them wrapped in myelin by oligodendrocytes, and the high lipid content of the myelin makes it look white. In the cerebrum the grey matter forms the outer cortex and the deep nuclei, with the white matter between them; in the spinal cord the arrangement is reversed, with grey matter in a butterfly-shaped core. The division is functional as much as visual: grey matter computes, white matter carries. How many synapses does a brain have, and how many does one neuron receive? Estimates for the adult human brain range from about 100 to 500 trillion synapses, with about 150 trillion in the neocortex alone, roughly a billion in every cubic millimetre of cortex. A cortical or hippocampal pyramidal cell receives thousands to tens of thousands of synapses, most of its excitatory ones on dendritic spines, and the number varies widely by cell type and region. These counts mean that each neuron hears from thousands of others at once, and that the brain's storage lies in the pattern of strengths across these synapses more than in the neurons themselves. What would it take to record every neuron in a human brain? The human brain has about 86 billion neurons. Sampling each one's voltage at 1 kHz with 10 bits (a modest choice, since Neuropixels samples its spike band at 30 kHz with 10-bit converters) gives about 8.6×10148.6 \times 10^{14}8.6×1014 bits per second, close to 101510^{15}1015, against 384 channels for one Neuropixels shank. Electronics in the head also make heat, and implant designs keep tissue warming within about 1 °C, which caps how much amplification, digitisation and transmission can sit inside the skull. Sorting spikes on the device and sending only events helps, but reaching every cell would also take a way of placing sensors through the whole volume without damaging it, which no current technology has. What is a cortical column, and how large is it? A cortical column is a vertical module running through the layers of the cortex, in which neurons share inputs and respond to related features, first described by Vernon Mountcastle in the 1950s. The word covers two scales: minicolumns, chains of cells about 50 micrometres across, and larger columns (macrocolumns) about 300 to 600 micrometres across, built from many minicolumns; the often-quoted figure of a millimetre is at the generous end. Each column contains pyramidal cells and inhibitory interneurons wired in a recurring microcircuit, so it is a natural unit in which to study the balance of excitation and inhibition. How sharp their boundaries are varies by area and species, and whether columns are a universal computational unit is debated. What happens to a neuron's membrane between −70 mV and +30 mV? Inputs nudge the resting voltage of about −70 mV upward until it reaches threshold, near −55 mV, where voltage-gated sodium channels open together. Sodium floods in and the inside swings positive to about +30 mV in a fraction of a millisecond. The sodium channels then inactivate and potassium channels open, so potassium leaves and the voltage drops back, briefly below rest. The whole cycle takes one to two milliseconds, and the spike is always the same size: a stronger input changes how often the neuron fires, never how high. Which parts of the neuron make up grey matter, and why does it use most of the brain's energy? Grey matter holds the neuron's cell bodies, its branching dendrites with their synapses, and the local unmyelinated stretches of axons, packed among glial cells and capillaries. Without much myelin it lacks the pale fat of white matter and looks grey in a preserved brain. Synaptic transmission and the ion pumping that follows every signal happen mostly here, so grey matter uses several times more energy per gram than white matter and is the first to suffer when blood flow falls. Where on a neuron does the action potential start, and why there? In most neurons it starts at the axon initial segment, the first few tens of micrometres of axon after the cell body. Sodium channels there are packed several times more densely than on the soma, and the low-threshold Nav1.6 subtype concentrates at its far end, so this is where inputs summed from the dendrites first reach threshold. From there the spike runs forward down the axon and also backward into the soma and dendrites, where it tells the synapses that the cell has fired. Because so much depends on this short stretch, inhibitory chandelier cells that target it, and changes in its length or position, have a strong grip on a neuron's output. Why does EEG record synaptic currents of pyramidal cells and miss their action potentials? To reach the scalp, currents from thousands of cells have to add up. Pyramidal cells stand in parallel with their dendrites toward the surface, and their synaptic currents last tens of milliseconds, so inputs that arrive roughly together overlap in time and sum into one large dipole. An action potential lasts about a millisecond and its currents flow in opposite directions over a short stretch of axon, so spikes from many cells rarely overlap and largely cancel at a distance. Even so, a visible deflection needs something like 6 to 10 square centimetres of cortex active in synchrony. Why are neurons especially vulnerable to misfolded proteins? Most cells dilute damaged proteins by dividing, but neurons do not divide and must keep the same cell working for decades, so whatever their chaperones, proteasomes and autophagy fail to clear accumulates. Their shape adds strain: synapses lie up to a metre from the cell body where most proteins are made, so repair and disposal depend on long-distance transport along the axon. Their energy budget is tight, and protein quality control itself consumes ATP. Aggregates of tau, amyloid beta, alpha-synuclein and other proteins that build up this way are the hallmarks of Alzheimer's, Parkinson's and related diseases. Why does the brain probably not learn by backpropagation as artificial networks do? Backpropagation sends an exact error backwards through the same connections used in the forward pass, which neurons cannot do directly: synapses transmit one way, so the backward path would need separate connections mirroring every forward weight. It also needs graded error values and the derivative of each unit's response, separate phases for computing and for learning, and a controller alternating them, while neurons send discrete spikes and learn continuously as they act. Most researchers conclude that the brain uses local rules that approximate some of its effect, helped by feedback connections and neuromodulators, and the finding that error sent through random feedback weights still trains networks shows the symmetry requirement can be relaxed. Why does the brain use about a fifth of the body's oxygen at about 2 % of its mass? Signalling is paid for by ion gradients, and the gradients are restored by the sodium-potassium pump, which burns ATP made almost entirely by oxidative metabolism. In an adult at rest the brain accounts for about 20 % of the body's oxygen use, around 20 W of a resting total near 100 W. Budgets for rodent cortex attribute most of the signalling cost to pumping ions back after action potentials and after the currents opened by glutamate at synapses, with the resting potential and transmitter recycling costing much less. That is why the brain has almost no energy reserve and fails within seconds to minutes when its blood supply stops.