Every day your kidneys strain about a hundred and eighty litres of water out of your blood. You hold about five litres of it, so that is your whole blood volume through the sieve some thirty-odd times between one morning and the next. Then more than ninety-nine per cent goes back. What leaves is a litre and a half, wrung out at the last against the standing hill of salt No. 28 was about.
It is worse than that. Out with the water goes something like a hundred and eighty grams of glucose — a day’s sugar, discarded — and better than twenty-five thousand millimoles of sodium, the sodium in roughly a kilogram and a half of table salt. Nearly all of both is hauled back in: under a hundred millimoles of sodium is let go, and in a healthy person, none of the sugar.
Read as a design, this is deranged. No engineer throws out the inventory each morning to keep it.
Start with the cost, because the first surprise is that the throwing-out is free: no cell spends anything at the filter. Blood arrives in the glomerulus at about 55 mmHg; back-pressure and the inward pull of the plasma proteins subtract about 45 of it. The ten left over are enough. The heart does the filtering; the kidney only holds the sieve.
And the sieve knows nothing. Three layers — capillary wall, basement membrane, the interlocking feet of the podocytes — discriminating on two properties only: size and charge. Molecules under about twenty ångströms cross with no measurable restriction; past about forty-two, almost nothing does. Albumin is turned back partly for being large and partly for being negative. Nothing there could tell glucose from urea, or a vitamin from a poison. It is a hole with a diameter and an opinion about charge, and that is the whole of its judgement.
Which is the point. Because it cannot tell, the body’s standing posture toward every small molecule in your blood is get rid of it. Discard is the default. Keeping anything requires a specific, evolved transporter in the tubule wall, reaching into the stream and pulling that molecule back.
That inverts the obvious architecture. You would expect a body to need a mechanism for removing each waste. Instead it needs one for keeping each valuable — and that list is short, ancient and nearly unchanging: sodium, glucose, amino acids, bicarbonate, phosphate, water. The list of things worth removing is unbounded and changes with every meal. A molecule your ancestors never met — a drug, a plant alkaloid, some novel metabolite — is excreted by default, with nothing having evolved to do it. The kidney copes with the future by refusing to recognise the present.
One honest exception. Filtration takes only the unbound fraction, so a drug riding on albumin mostly does not get through — and for those the proximal tubule had to evolve active secretion, class by class. Everything the sieve can reach comes free.
Now the bill. Filtering is free; taking it back is ruinous. Nearly all the reclaiming is bought from one pump, the sodium–potassium ATPase on the blood-facing side of every tubule cell — the same pump that holds the voltage across every cell you own. It drags sodium out into the blood, and sodium falls in from the tubule to replace it. Glucose, amino acids and phosphate never get pumps of their own. They ride in alongside the sodium, passengers on a gradient paid for by something else. Moving sodium is where most of the kidney’s oxygen goes.
Passengers on a fixed number of carriers can be crowded out. Raise blood glucose far enough, the carriers saturate, and sugar that would otherwise have been reclaimed appears in the urine — the sign that named diabetes mellitus, honeyed, long before anyone knew what a transporter was. The spill is not the kidney failing; it is a ceiling doing what ceilings do.
Last, the rate. If the filter is indiscriminate, how fast it runs is the regulation — and two local mechanisms hold it, with nothing central involved. Stretch the feed arteriole and its smooth muscle contracts: the crudest available pressure defence. Then the elegant one. Each nephron’s tubule loops back and touches its own glomerulus, where a patch of cells — the macula densa — tastes the salt going past. Salt still arriving this far along means the filter is outrunning the tubule, so those cells release ATP and adenosine and the arteriole pinches shut. The load has a wire back to the tap, and the wire is one cell long. Between them, filtration holds nearly steady from a mean arterial pressure of eighty to one of a hundred and eighty — and when it does wobble, the proximal tubule takes a constant fraction, about two-thirds, so the load handed downstream scales rather than swamping.
Three issues now on one organ, and no two of them the same machine. Last time the kidney was a sensor, and a sensor must be exquisitely particular: it read one variable through a proxy and moved one dial. Here it is an effector, and this kind must be perfectly indifferent — acting on everything at once, regulating a thousand substances by declining, one at a time, to take them back. Specificity is the whole virtue in the reading and a fatal defect in the doing.
≈
One loop I’m watching
Next: what happens when that wire from the load back to the tap is long instead of one cell short. A steady trickle of retail demand turns into wild swings in factory orders as it travels upstream, each tier ordering against its forecast of the tier below, each correction arriving late and landing past the mark. Forrester described it in 1961; Sterman sat students down in a simulated beer supply chain in 1989 and watched them produce it reliably while behaving perfectly sensibly. The amplification is nobody’s mistake. It is the shape of the chain. Next time.