What a cell does all day

Put a drop of pond water under a cheap microscope. The first surprise isn't the shapes. It's how much everything is shaking.

That shaking is the thing most descriptions of cells leave out. The diagram in the textbook is calm — neat outlines, labelled blobs, an arrow pointing at the nucleus, everything holding still to be identified. Real cells aren't like that at all. They're warm, wet, absurdly crowded and shoved about constantly by the molecules around them, and nearly everything a cell does comes down to pulling order out of that mess and paying for it. If you take one idea away from this, take that one.

There are somewhere around thirty trillion of them in you. Here's roughly what they're doing.

Being small changes the rules

A typical human cell is ten to thirty micrometres across. A red blood cell is about seven. You can't see any of them without help, which is why nobody knew they existed until Robert Hooke put a slice of cork under a lens in 1665 and named the little compartments after the rooms monks slept in.

At that size the physics you're used to stops applying. Gravity is irrelevant — a protein doesn't sink, because thermal jostling swamps its weight completely. Water behaves less like water and more like treacle, because at those scales viscosity dominates over momentum. Nothing coasts. If a bacterium stops swimming, it stops dead within a fraction of its own body length.

The useful part of being tiny is diffusion. Molecules move about randomly, and the time a random walk takes goes up with the square of the distance. Cross a micrometre and you're done in a small fraction of a second. Cross a millimetre and it's several hours. Cross a metre and you're looking at centuries.

So the cell doesn't need a delivery system for most things. It just needs to be small enough that random bumping does the job for free, which it very nearly always is. That square law is also why you can't simply scale a cell up. Double the width and you've got eight times the volume to supply through four times the surface, and diffusion has to travel twice as far while taking four times as long. Big cells solve it with tricks — long thin shapes, internal churning, dedicated transport — but they don't get to ignore it.

The inside isn't a solution, it's a crowd

Cytoplasm gets drawn as a pale wash. It's actually something like 200 to 300 grams of protein per litre, which is roughly the consistency of a thick soup, and a fifth to a third of the available space is already occupied by something large.

This matters more than it sounds. In a crowded space, big molecules get pushed together simply because there's nowhere else to go — the excluded volume nudges them towards each other, which speeds up reactions between them and makes weak associations stick. Enzymes behave differently in a test tube of dilute buffer than they do at home. A good deal of twentieth-century biochemistry was done in dilute buffer, and biologists have spent decades since working out which of those results survive contact with the real thing.

Nothing in there is far from anything else. A protein in a bacterium bumps into the cell wall thousands of times a second.

Everything runs on one currency

Cells don't burn sugar directly to do work. They convert it, in stages, into a single small molecule called ATP — adenosine triphosphate — and then spend that.

ATP carries energy in the repulsion between its three phosphate groups, which are all negatively charged and don't much like being lined up in a row. Snap the last one off and you release enough energy to drive a reaction that wouldn't happen on its own. That's the whole idea. Muscle contraction, pumping ions, building proteins, moving cargo along a fibre — all of it, in every organism anyone has looked at, is paid for in the same coin.

The turnover figures are the part people don't believe. You carry maybe 250 grams of ATP at any given moment. You get through something in the region of your own body weight of it every day, which means each molecule is charged and spent thousands of times over the course of a day. There's no reservoir. It's built and consumed continuously, and if the manufacturing stops the cell has seconds of supply, not minutes.

The trick isn't chemistry, it's a gradient

Here's the part that took biochemistry an embarrassingly long time to accept.

For decades people assumed there had to be some high-energy chemical intermediate linking food to ATP, and they went looking for it, and they couldn't find it. In 1961 Peter Mitchell proposed there wasn't one. Instead, he argued, the energy from food is used to pump protons across a membrane, building up a difference in concentration and charge from one side to the other. That difference is the intermediate. It's not a molecule at all, it's a state of the membrane.

The idea was received badly. It sounded like a dodge. Mitchell ended up funding a good deal of his own research from a converted manor house in Cornwall, and it took most of the 1960s and 70s before the evidence became impossible to argue with. He got the Nobel Prize in 1978.

What sits in the membrane collecting the payment is one of the strangest objects in biology. ATP synthase is a rotary motor — a genuine one, with a spinning rotor and a fixed stator. Protons flowing back down the gradient turn the rotor, the rotor twists a set of catalytic sites through a cycle of shapes, and each turn squeezes out ATP. It spins fast enough that its speed is quoted in revolutions per second. Paul Boyer worked out the mechanism, John Walker got the structure, and they shared the 1997 chemistry prize with Jens Christian Skou.

Skou was there for a different pump, which brings us to the bill.

Pumping ions is the biggest expense you have

Your cells hold far more potassium inside than out, and far more sodium outside than in. Neither of these arrangements is stable. Both leak constantly, and both have to be restored constantly, by a pump that throws three sodium ions out and hauls two potassium ions in for each ATP it spends.

Estimates vary by tissue, but a fifth or more of your resting energy budget goes on this. In nerve cells it's considerably more. You are, to a surprising extent, an animal whose main job is keeping salt in the right places.

The gradient isn't waste, though. It's stored work, and the cell spends it on all sorts of things — hauling glucose and amino acids in against their own concentration gradients by coupling them to sodium flowing back down its gradient, and running the electrical signalling that nerves and muscles depend on.

That electrical part deserves a number. The voltage across a resting cell membrane is about 70 millivolts, which sounds trivial. The membrane it's across is about five nanometres thick. Divide one by the other and you get a field strength of roughly ten million volts per metre, which is in the same territory as the air just before lightning. Every cell you have is sitting under that, all the time, and mostly not doing anything dramatic about it.

Membranes hold themselves together

Nothing glues a membrane. There's no structural framework holding the lipids in place. They stay put because of water.

A phospholipid has a phosphate head that mixes happily with water and two greasy tails that don't. Drop a pile of them in water and they'll arrange themselves, without any instruction, into a double sheet with the tails tucked inwards and the heads facing out. This isn't the lipids being attracted to each other so much as the water refusing to accommodate them and forcing the issue. Puncture the sheet and it seals itself, because the alternative exposes tails to water, and the water won't have it.

The result is about five nanometres thick, roughly as fluid as olive oil, and constantly moving. Individual lipids swap places with their neighbours millions of times a second and drift right across the cell in under a minute. The proteins embedded in it float about like buoys. That picture — a fluid sheet with things drifting in it, rather than a rigid wall with things bolted on — was set out by Singer and Nicolson in 1972 and it's held up remarkably well.

What the membrane really is, functionally, is a barrier to charge. Ions can't cross a greasy interior on their own, so anything charged has to go through a protein channel or pump, and that means the cell controls it. Being able to say no to a sodium ion is the foundation of everything else in this article.

Proteins have to fold, and folding is genuinely hard

A protein comes off the production line as a floppy chain of amino acids in a specific order. On its own, that chain is useless. It has to collapse into one precise three-dimensional shape before it does anything, and the shape is what determines the function.

Christian Anfinsen showed in the 1950s and 60s that the information for the shape is in the sequence and nowhere else. He unfolded an enzyme completely, removed the unfolding agents, and it re-formed itself into working enzyme. No template, no external help. Nobel Prize in 1972.

Which raises a problem that Cyrus Levinthal pointed out in 1969. A chain of a couple of hundred amino acids has an astronomical number of possible shapes, and if it tried them at random it would take longer than the universe has existed to stumble on the right one. Real proteins fold in milliseconds to seconds. They aren't searching randomly — the folding funnels downhill, with partially correct structures forming early and locking in — but the detail took decades to work out, and computationally predicting a structure from a sequence stayed unsolved until very recently.

Inside a crowded cell it's worse, because a half-folded protein has sticky bits exposed and there are neighbours everywhere for it to stick to. So cells run a whole class of proteins whose job is babysitting other proteins: chaperones, which hold new chains apart from each other, and chaperonins, which are barrels that a chain can be shoved inside to fold in private. A lot of them were first identified because cells make more of them when they overheat, which is why they're called heat shock proteins.

When this fails, misfolded proteins clump. Aggregates are hard to clear and tend to seed more aggregation. It's a failure mode the cell spends real resources trying to avoid.

Copying the instructions, and getting them slightly wrong

The DNA in one of your cells is about two metres long if you stretch it out, packed into a nucleus a few micrometres across. It codes for something like twenty thousand proteins.

To make one, a stretch gets copied into RNA, the RNA goes out to a ribosome, and the ribosome reads it three letters at a time, adding an amino acid per triplet. Human ribosomes manage roughly five or six amino acids a second. Bacterial ones run three times faster. A ribosome is itself mostly RNA rather than protein, and the actual bond-forming chemistry is done by the RNA — which is a strong hint that RNA was doing this job before proteins were around to help.

Mistakes happen at every stage, and the cell tolerates very different rates depending on the stakes. Translation errors run around one in ten thousand amino acids, which is fine, because a single duff protein gets thrown away. DNA replication is held to something like one error in a billion bases or better, because a mistake there is permanent and inherited. The difference comes from proofreading — the copying enzyme checks its own work and backs up to fix errors, and a separate repair system sweeps up afterwards.

Not perfect, though. Nothing is. The residual errors are mutations, and they're the reason there's anything to write about.

Mitochondria used to be somebody else

The compartments doing the ATP work have a peculiar set of features, and once you notice them the explanation is hard to avoid.

Mitochondria have two membranes rather than one. They carry their own DNA — a small circular genome, the same shape bacteria use, holding 37 genes in humans. They make their own ribosomes, and those ribosomes look bacterial rather than human, which is why some antibiotics that target bacterial protein synthesis have effects on them too. They don't get built from scratch by the cell; they divide in two, and you only get new ones from old ones. And you inherit yours from your mother, because the ones in sperm don't make it through.

Put that together and you get endosymbiosis: a long time ago, an ancestral cell took in a free-living bacterium and didn't digest it. The lodger stayed, and over an enormous stretch of time most of its genes migrated into the host nucleus or were lost, leaving the stub of a genome that mitochondria still keep.

Lynn Margulis published the case for this in 1967, under her then name Lynn Sagan, after the paper had been turned down by about fifteen journals. It's now standard. The best current evidence puts the mitochondrial ancestor among the alphaproteobacteria. Chloroplasts went through the same thing separately, starting from a cyanobacterium, which is why plants can do photosynthesis and you can't.

Margulis went on to argue for endosymbiotic origins of several other things, and most of those claims didn't survive. The mitochondrial one did, comprehensively.

Nothing in there is permanent

Proteins don't get built and left alone. They get built, used, damaged and destroyed, on timescales from minutes to weeks, and replaced.

There are two main disposal routes. Individual proteins get tagged with a small marker called ubiquitin and fed into a proteasome, a barrel-shaped shredder that chops them back into amino acids for reuse — the work that won Ciechanover, Hershko and Rose the 2004 chemistry prize. Bulk material, including whole knackered mitochondria, gets wrapped in a membrane and delivered to a lysosome for digestion. Christian de Duve found lysosomes more or less by accident in the 1950s; Yoshinori Ohsumi worked out the genetics of the wrapping process in yeast and took the 2016 medicine prize for it.

The point of all this churn is quality control. There's no way to repair a badly damaged protein, so the cell doesn't try. It recycles it and builds another. You are, at the molecular level, mostly not made of the same material you were made of last month, even though you're unmistakably the same shape.

Getting things across the room

Diffusion covers most distances inside a cell, but not all of them, and the exceptions need machinery.

Cells are threaded with fibres. Actin filaments handle shape and movement at the edges; microtubules are stiffer hollow tubes that run longer distances and act as tracks. Walking along those tracks are motor proteins — kinesin heading one way, dynein the other — which take eight-nanometre steps, one foot swinging past the other, burning ATP per step, dragging cargo behind them.

This looks like an extravagance until you remember the square law. A nerve cell running from your spine to your foot has an axon up to a metre long. Diffusion would take centuries to move anything down it. Motors do it in days, and if that transport fails, the far end of the cell starves.

The same fibres tear the cell in half when it divides, and haul the chromosomes apart. They're also what a lot of cell shape is: remove the scaffolding and cells slump.

Knowing when to stop

Dividing is easy. Not dividing is the hard part, and most of the regulation is devoted to it.

The cell cycle runs through checkpoints — is the DNA fully copied, is it damaged, are the chromosomes properly attached before we pull? At each one the cycle can be halted while the problem is dealt with. If the damage can't be fixed, there's a further option: the cell can kill itself, in an orderly way that packages the remains up neatly for the neighbours to eat rather than spilling its contents everywhere.

That process, apoptosis, isn't a failure. It's a normal developmental tool. The nematode worm Caenorhabditis elegans builds exactly 1,090 cells and then deliberately kills 131 of them, the same 131 every time, which is how the genetics of the whole thing got worked out. Your hands were webbed once. The webbing didn't wear away; the cells in it were told to go, and went.

And ordinary cells don't divide indefinitely anyway. Leonard Hayflick showed in the 1960s that normal human cells stop after a few dozen divisions, largely because the protective caps on the ends of chromosomes get shorter each time and eventually run out.

Which is a lot of infrastructure devoted to restraint, in a system that spends its whole existence a few seconds away from running out of ATP.