Partner or extortionist is not a property of the bacterium — it is settled by how the bacterium gets to the next generation
2026-09-15 · BigCat
"Symbiosis" sounds like two parties who struck a good deal. It is closer to a landlord and a tenant: how well they get on rarely depends on anyone's character, and mostly on how the lease is written and what it costs to move out.
Those little green insects on a rose stem, aphids, spend their whole lives drinking nothing but plant sap.
By rights that should not keep one alive. Plant sap is cloyingly sugary but short of a set of amino acids that animals cannot make for themselves and have to eat — amino acids being the parts that proteins are assembled from, and missing a few means the body cannot be built. An aphid ought to starve slowly.
Cut one open and you find a few dozen unusually large cells in its abdomen, packed with bacteria. The missing amino acids are made by those bacteria.
Which is exactly where the problem starts. These bacteria live inside someone else's body eating someone else's food. What stops them from taking a little more and handing over a little less?
// 01
It does not take advantage of you because it burned its own way out long ago
The bacteria have a name: Buchnera. They live inside those few dozen oversized cells the aphid sets aside for them and never come out. They have exactly one route to the next generation: getting into the mother's eggs. There is no second one.
The cost shows. Their entire DNA is down to about 640,000 base pairs — a base pair being one paired letter of DNA, which you can think of as a single character — and roughly 580 genes, about a seventh of what their relative E. coli carries. What was lost is exactly what "making a living outside" requires: coping with a changing environment, making the amino acids the aphid can already make. What was kept is exactly the machinery for the ones the aphid lacks. And on the aphid's side, the ability to make those few amino acids is long gone too.
Neither can now survive without the other. This is not a recent arrangement: the aphid family tree and the bacterial family tree have almost the same shape — every time aphids split into two lineages, their bacteria split too. They did not each pick up a partner separately; they branched together, for at least 160 million years.
The thing that matters is not to tell this as a morality tale. Nothing "learned loyalty". The accounting is dull: copies of the bacterium can only leave inside an egg, so every extra offspring the host raises means another batch of copies, and if the host dies that whole line ends. So a "take a bit more" variant may win on the spot, but it shrinks its own only exit at the same time — such variants do not persist. Nothing chose not to cheat; the cheating strains left no descendants. The transmission channel welds both sides' reproductive success into a single number.
Mitochondria, from the piece on endosymbiosis and the eukaryotic revolution, are what this road looks like at the very end: genes reduced to barely a dozen, and nobody counts them as a separate organism any more.
// 02
The same bacterium is a necessity in one host and an extortionist in another
A second one is called Wolbachia. Estimates of how many arthropod species (insects, spiders, crustaceans — that whole huge group) carry it range from forty to seventy percent depending on the survey; the methods differ too much to pin it down, but either way the answer is an enormous number.
It too travels only in eggs and cannot squeeze into sperm. So males are a dead end for it. It has accordingly acquired a whole repertoire for tilting host sex ratios towards females: killing the male embryos of infected mothers, developing genetically male individuals into females, letting females lay without mating at all.
The commonest trick is more roundabout. An infected male crossed with an uninfected female produces embryos that fail to develop; an infected female is fine with any partner. The infected are not more fertile — the uninfected females are relatively penalised — so the bacterium spreads right across a population while doing nothing good for the insects themselves. This is the same logic as meiotic drive in the piece on conflict inside the genome, except that this time the cheat is not a stretch of the host's own chromosome but another species.
Yet the same genus of Wolbachia is something else entirely inside filarial worms, the parasitic nematodes behind elephantiasis and river blindness. There the bacterium is indispensable: clear it with antibiotics and the worms become sterile, their lifespan dropping from about ten years to two or three. That is now a clinical approach in actual use — hit the bacterium to treat the worm infection indirectly. And the close relative Loa loa simply has no such bacterium at all.
Mutualism and parasitism are not two boxes; they are positions on this axis. Note the middle point: Wolbachia in insects also travels only in eggs, yet it can ride only the female line — a narrow exit is not the same as aligned interests; what matters is where the narrowing is. (Schematic, not a measured scale.)
There is one more counter-case on the right-hand side of the axis. The Hawaiian bobtail squid uses Vibrio fischeri in its belly to glow, erasing its own shadow on moonlit nights. These bacteria are not passed in eggs; every young squid has to recruit them from seawater afresh. With no egg to ride, what stops the recruits from skimming? Filtering: the entrance to the light organ admits only this one species, and each dawn the squid vents ninety to ninety-five percent of them, leaving the rest to regrow. Binding and filtering are two roads to the same thing — leaving no exit for the lineage that takes advantage.
// 03
"You and your microbes are one whole" is a fine line, but it cannot serve as the unit of account
Start by granting the real part. Mice raised in a completely germ-free environment genuinely are abnormal: the lymphoid tissue near the gut is underdeveloped, the thymus is small, and they are leaner on the same food, needing to eat more to hold the same weight. Microbes are not optional lodgers.
So a very natural idea appeared: treat a host and all the microbes on and in it as one thing (some people call it a holobiont), add both sets of genes into a single genome, and let selection act on that whole.
The objection lands on something hard: for anything to be settled as a unit, the whole package has to reach the next generation reliably and as a set. Aphids and their bacteria qualify — the bacteria are inside the egg and the set travels together. People do not: most of the bacteria in your gut are picked up piecemeal from the environment, from food, from the people around you, and one course of antibiotics, one trip, one change of staple food swaps out a large chunk. If the composition is not transmitted as a set, the "whole" has no continuous lineage for selection to sift; what does get sifted stably is still the host's genes and each microbe's own genes. As one researcher put it, treating host plus microbes as an evolutionary unit should be the hypothesis under test, not the default assumption.
The thing to watch here is not being carried along by numbers. The most-repeated line — that bacteria in the body outnumber human cells ten to one — is wrong; a 2016 recalculation put it at roughly one to one (about thirty-eight trillion bacteria against thirty trillion human cells). That a tenfold-inflated figure circulated for decades unchecked tells you how easily this field overheats.
How far has the causal evidence got? Solid for recurrent Clostridioides difficile infection: transplanting a healthy person's gut community across as a set cures well over eighty percent of cases, far better than continuing antibiotics. Claims that reshaping the community drives weight loss look beautiful in mice but have never reproduced at the same strength in humans, and recent reviews have visibly narrowed their language; that one is still contested. The difference between the two lands back on this piece's lever: the transplant works because a position emptied out gets refilled as a set, whereas permanently remaking a combination that had no transmission channel in the first place will drift back on its own.
🎒 In Context · BigCat
① Health and energy—You finish a course of antibiotics and for a week or two your digestion and appetite are off; you pick up a box of probiotics, take them for a fortnight, cannot tell whether they did anything, and go round again next time. The mechanism at work is the one from "'You and your microbes are one whole' is a fine line": whether a microbe stays depends on whether there is a channel and an opening for it, not on how much you swallow — most commercial strains are undetectable a few weeks after you stop, while the gut is drifting back on its own anyway, and the two recovery curves look identical. What to change: keep the same single measure for two weeks before and two weeks after the antibiotics (bowel regularity, or that one specific symptom); only a before-and-after comparison separates natural recovery from the supplement.
② Parenting and education—While a child is small, a household argues in circles about how clean to be: disinfectant wipes over every surface on one side, letting them roll in the dirt on the other. The mechanism at work is the one from "The same bacterium is a necessity in one host": whether a microbe is a partner or a problem is not a property of the microbe but a position given to it by the manner and timing of contact — so sorting actions by "clean versus dirty" sorts them into a category that does not exist. What to change: treat "washing hands" and "disinfecting the house" as two separate things, the first with hard evidence behind it and the second with far thinner; and instead of asking "is this thing clean", ask "if this contact turns out to be a mistake, can it be undone".
🌀 Crossing Over · Connections Beyond the Field
History and war · An armed force takes a town — does it loot and leave, or start repairing the bridges: the economic historian Mancur Olson compared two kinds of predator. A roving band strips one place and moves to the next; the one that settles and stays often turns around and maintains the irrigation, sets rules, keeps other raiders out, and converts plunder into regular taxation. The mechanism at work is the one from "It does not take advantage of you because it burned its own way out long ago": once a party's take can only be realised through the long-run output of the same host, what it extracts and what its own exit is worth become a single number, and the dearer it is to switch hosts the tighter that weld. From which follows something you could not otherwise see: to judge whether an occupier will turn "moderate", do not look at its declarations or its promises, look at what switching to another host would cost it. The cheaper the switch, the further behaviour slides towards extraction whatever is said; and raising the cost of exit — assets, family, reputation all locked in the same place — changes behaviour more than any moral constraint. The steadiest rule usually does not start from goodwill, it starts from having nowhere to go.Where this breaks down: a ruler who has settled can change their mind within a year, and revert to roving when conditions shift — they are doing the sums within a single lifetime. Buchnera has no such thing as "deciding to be moderate"; it merely lost every strain that was not, and it burned the way out permanently by discarding the genes. The human version of that welding can be undone and the biological one cannot — which is why "moderate" periods in history are always so much shorter than symbioses.
Eastern thought · The Buddhist "non-self" can be turned into a question you can actually compute: non-self does not say there is no you. It says that when you take it apart you cannot find a self that stands on its own with a clean boundary — the five aggregates (roughly the body plus four kinds of mental activity) each hold together only by depending on other conditions, and none can be pulled out to serve as the owner. The mechanism at work is the criterion from "'You and your microbes are one whole' is a fine line": whether a collection can be settled as a unit does not depend on how unified it looks, but on whether it reaches the next generation whole and reliably. From which follows something you could not otherwise see: this turns a philosophical assertion into a question answerable scale by scale, and the answer is not "yes" or "no". Aphids and their bacteria meet the criterion, so that composite really is a unit; a person and their gut community do not, so the "me" that is person-plus-microbes fails while the "person" layer stands perfectly firm. That the same object lacks own-being at one scale and holds at another is no contradiction — the criterion never asked how much it resembles a whole, only how it gets passed on.Where this breaks down: the Buddhist point of non-self lands on how to live and how to loosen one's grip on self; the criterion here answers only "at which layer does selection settle accounts", and says nothing whatever about how to live. Going from "no independent self can be found" to "so it does not matter what happens to me" is a step this mechanism cannot take — that is a jump from what is to what ought to be, and there is no bridge across it.
Going Deeper
How much evidence is there really behind "being too clean gives you allergies"?
One strand is relatively hard. A 2016 study compared Amish and Hutterite children in the United States: two groups with similar ancestry and similar ways of life, differing mainly in that the Amish still farm traditionally right next to their livestock barns. Asthma prevalence among the Amish children was about a quarter of the other group's and allergic sensitisation about a sixth; endotoxin (a component of bacterial outer walls) in the dust of Amish homes was nearly seven times higher, and giving mice that dust to breathe suppressed an experimental asthma response. But what it shows is early exposure to a certain kind of microbial environment, which is not the same as "the dirtier the home the better".
If Wolbachia manipulates its hosts, can people turn that around and use it?
They already do. Projects have introduced the bacterium into Aedes aegypti mosquitoes, whose ability to transmit dengue drops markedly once they carry it; and the trick where infected males crossed with uninfected females leave embryos that fail to develop is exactly what helps it spread through wild mosquito populations by itself, with no need to re-release every year. In a randomised controlled trial in Yogyakarta, Indonesia, dengue incidence in the release areas was about seventy-seven percent lower than in the control areas. Whether mosquitoes or the virus eventually get around it is a question for the coming decades.
Do humans and their gut bacteria count as "coevolving"?
Partly. Breast milk contains a class of sugars humans cannot digest themselves and that certain bifidobacteria are specialised to break down, and that match does not look like coincidence. But most gut bacteria have no dedicated lineage relationship with your ancestors — change your diet and you get a different set. The more accurate statement is that the human side evolved a whole apparatus for managing microbes (the mucus layer, immune filtering, that targeted ration in breast milk), while exactly who gets managed is replaceable.