In 2005, Masahiro Nakajima and colleagues dissolved three proteins and some ATP into a test tube, left it on the bench, and measured what happened. The three proteins were KaiA, KaiB, and KaiC — named after the Japanese word for "rotation." What happened was a circadian rhythm. The phosphorylation state of KaiC rose and fell with a period of almost exactly 24 hours, for days, in solution, with no cells, no genes, no membranes, no transcription happening at all.
This was not what anyone expected. The prevailing model for how biological clocks work — built over decades of genetics in fruit flies and mice — was an elaborate gene-expression feedback loop. In mammals, two proteins (CLOCK and BMAL1) bind to DNA and turn on the genes that produce two other proteins (PER and CRY). As PER and CRY accumulate, they eventually reach concentrations high enough to shut off CLOCK and BMAL1. Then PER and CRY are degraded, CLOCK and BMAL1 are free again, the genes turn back on, and the whole cycle repeats. The loop takes roughly 24 hours because transcription, translation, protein folding, accumulation, and degradation are all slow steps with their own delays. It requires a nucleus. It requires ribosomes. It requires cells doing active biology.
KaiA, KaiB, and KaiC do not. The core of what KaiC does is: it slowly phosphorylates itself. Very slowly. It hydrolyzes about 15 molecules of ATP per day. That is not a typo — 15. For comparison, a typical ATPase in your cells hydrolyzes millions of ATP molecules per second. KaiC is, by some measure, the slowest known ATPase. And that sluggishness is apparently the point. The phosphorylation cycle produces the period.
The mechanism goes like this: KaiA promotes KaiC phosphorylation. As KaiC becomes more phosphorylated, it changes shape and begins to sequester KaiB. KaiB then sequesters KaiA, removing the thing that was driving phosphorylation forward. Without KaiA, KaiC dephosphorylates itself. As phosphorylation falls, KaiB releases KaiA again. And the cycle restarts. No nucleus. No gene expression. Three proteins and a phosphate donor, doing their slow chemistry in the dark.
The cyanobacterial clock and the mammalian clock evolved independently — they share no homologous proteins, no conserved mechanism. They arrived at the same period through different molecular routes. Which raises a question about what the period is tracking. The answer seems to be Earth. The circadian resonance hypothesis says organisms whose internal period matches the environmental cycle do better than those whose periods are mismatched. Competition experiments with cyanobacterial strains bearing different-period mutations bear this out: the strain that matches the light-dark cycle wins, regardless of absolute fitness in constant conditions. The 24-hour period is not a property of any particular clock mechanism. It is a shape that chemistry gets selected toward, because the planet rotates every 24 hours.
There is one more strange thing. Most chemical reactions follow Arrhenius kinetics: higher temperature means more thermal energy, higher collision rates, faster reactions. A useful rule of thumb is that a 10°C increase roughly doubles the rate of most biochemical processes. But circadian clocks run at nearly the same period across a wide temperature range — from around 20°C to 37°C in many organisms, the period stays close to 24 hours. This is called temperature compensation, and it means there is something self-correcting built into the system. The most likely explanation is that the network contains opposing reactions with different temperature sensitivities that cancel each other out — one reaction speeds up with heat, another speeds up more, and their interaction stays roughly constant. But the exact mechanism is not fully resolved even now.
I find myself sitting with the test-tube result more than the others. The gene-expression model felt like a reasonable kind of clock — a slow relay with multiple stages, like a Rube Goldberg machine timed by the number of steps. The protein-oscillation model feels different. Three molecules in a solution, keeping time for days, because one of them is very slow. There's a minimum in there — a question of what a clock needs at the bottom. Not cells. Not genes. Just chemistry with the right period, happening slowly enough that Earth can keep up with it.