What your immune system is actually doing
Before anything else, the boring but necessary bit: this is a description of how a biological system works. It isn't medical advice, it isn't about any particular illness, and it can't tell you anything about your own body. Anything to do with your health belongs with a doctor, not an article.
With that out of the way — the immune system is one of the strangest things you're made of, and almost nothing about it works the way the everyday language suggests.
It isn't an organ. It's tens of billions of individual cells with no central command, patrolling everywhere, making decisions locally on partial information. It doesn't have a list of things to attack. It generates its recognition equipment at random, before it knows what's coming, then edits it under fire. And a good deal of what feels like being ill isn't the microbe doing anything to you. It's your own response, and you're feeling the collateral.
Two systems, running at different speeds
The usual split is innate and adaptive, and the honest way to describe the difference is speed against precision.
The innate side is ancient, shared with insects and plants in various forms, and it goes off in minutes. It doesn't identify specific invaders. It recognises broad categories — molecular signatures that lots of microbes have and your own cells don't. Bits of bacterial cell wall. The protein flagella are built from. Double-stranded RNA, which almost always means a virus. Even DNA in the wrong compartment.
The receptors that do this are hardcoded in your genome, the same set in everyone, refined over hundreds of millions of years. Their discovery earned Jules Hoffmann and Bruce Beutler a share of the 2011 Nobel Prize, alongside Ralph Steinman for dendritic cells. The innate system doesn't learn and doesn't remember. It's a smoke alarm that only knows about smoke.
The adaptive side is slow and specific. It exists in essentially all jawed vertebrates and nowhere else, which puts it at maybe 500 million years old rather than a billion. It can recognise something it has never encountered, including molecules that have never existed in nature. It takes days to get going the first time. And it keeps a record.
Those two aren't independent. The innate response is what tells the adaptive one there's a problem and what kind, and the adaptive one calls the innate cells back in to do the actual demolition.
The first line is mostly not cells at all
Most of what keeps microbes out isn't immunological in any interesting sense. It's plumbing.
Skin is a barrier of dead, keratin-filled cells with an acidic surface. Mucus traps things and cilia sweep it steadily upwards and out. Tears and saliva contain lysozyme, an enzyme that chews open bacterial cell walls, which Alexander Fleming found in 1922 six years before he noticed the mould. Stomach acid runs around pH 1.5 to 3.5 and destroys most of what you swallow. Urine flushes. Coughing and sneezing are expulsion mechanisms, however unpleasant they are to be near.
Add competition. The bacteria already living on you occupy the space and consume the resources a newcomer would need. That's not the immune system helping, but the effect is the same.
Nothing here is intelligent. It's all just harder to get through than it looks.
What inflammation actually is
The four signs were written down by Celsus in the first century AD and haven't needed revising: redness, heat, swelling and pain. What's changed is that we know why each one happens.
When tissue is damaged or a sentinel cell detects something it recognises, chemical signals are released that make the local blood vessels widen and go leaky. Widening brings more blood through, which is the redness and the heat. Leaking lets fluid and proteins escape into the tissue, which is the swelling. Some of the signalling molecules act directly on nerve endings and lower their threshold, which is the pain.
All three effects serve the same purpose: getting cells and molecules out of the bloodstream and into the tissue where the trouble is. Neutrophils arrive first, in enormous numbers, following a chemical gradient to the exact spot. They engulf bacteria and destroy them internally with reactive oxygen and enzymes, and some of them fling out webs of their own DNA to trap microbes, dying in the process. Neutrophils are short-lived and you make them on an industrial scale — the usual estimate is around a hundred billion a day.
Pus is dead neutrophils. That's all it is.
Macrophages come in behind, eat the debris including the spent neutrophils, and then switch roles to help shut the whole thing down and start the repair. Inflammation that doesn't get switched off properly causes real damage, which is why the off-switches are as elaborate as the on-switches.
Complement, which nobody has heard of
There's a set of around thirty proteins circulating in your blood in an inactive form, and they're doing a surprising amount of the work.
Complement was named in the 1890s because it was thought to complement the action of antibodies, and the name stuck even though it turned out to do plenty on its own. The proteins sit there dormant until one gets triggered — by a microbial surface, by an antibody that's already stuck to something, or by certain sugars found on bacteria but not on your cells. That first activation cleaves the next protein, which cleaves the next, and the signal amplifies enormously at each step.
The output is three things. Fragments that coat the microbe and act as handles for phagocytes to grab. Fragments that drift away and summon more cells to the site. And a set of proteins that assemble into a ring, punch it through the microbe's membrane, and let its contents leak out.
A cascade like that is fast and violent, and the obvious danger is setting it off on your own tissue. Your cells carry surface proteins whose only job is to shut complement down before it gets going. Microbes don't have those, which is a large part of how the system tells the difference.
The adaptive system builds its receptors at random
Here's the genuinely counter-intuitive part, and it's worth slowing down for.
Your adaptive immune system can recognise essentially any molecular shape, including synthetic compounds that never existed until a chemist made one. It clearly can't have a gene for each of those. So how?
It doesn't inherit finished receptors. It inherits a set of interchangeable gene segments — several dozen of one kind, a handful of another, a few of a third — and when a lymphocyte matures it physically cuts its own DNA and splices one segment of each type together, more or less at random. Then, at the joins, it adds and deletes a few extra bases sloppily, which multiplies the variety again. Susumu Tonegawa demonstrated this in the 1970s and it won him the 1987 Nobel Prize, because until then the idea that a cell would deliberately rearrange its own genome was close to heresy.
The arithmetic gets silly quickly. The number of distinct receptors the process could in principle produce runs past a trillion. You don't carry that many lymphocytes, so at any moment you're holding a large random sample of the possible space rather than all of it.
Each lymphocyte makes exactly one receptor and sticks with it for life.
Which sets up the second half of the idea, proposed by Frank Macfarlane Burnet in 1957 and called clonal selection. The repertoire is generated blind, before any threat appears. When something does appear, it finds the handful of cells whose receptors happen to fit, and those cells — and only those — are triggered to divide repeatedly. The response isn't designed. It's selected, from a pre-existing random library, by the thing it needs to fight.
That's the same logic as evolution, running inside your body, on a timescale of days.
Antibodies, and how they get better mid-fight
An antibody is a Y-shaped protein made of four chains — two heavy, two light. The stem is constant and is what other immune components grab hold of. The two tips are the variable regions, produced by the shuffling described above, and they're the part that binds.
Antibodies don't kill anything by themselves. What they do is mark and obstruct. Coating a virus can physically block the bit it uses to enter a cell. Coating a bacterium makes it far easier for a phagocyte to get a grip. Clumping several microbes together stops them spreading. And an antibody stuck to a surface is one of the triggers that sets complement going.
There are five classes, distinguished by the stem and by where they operate. IgM is made first in a response and is a bulky pentamer. IgG is the most abundant in blood, and it crosses the placenta, which is how a newborn starts life carrying its mother's antibody repertoire for a few months. IgA is secreted into mucus, saliva, tears and breast milk. IgE is present in tiny amounts and is central to allergic reactions and responses to parasites. IgD's job is still not fully pinned down, which is a nice reminder that this is an active field rather than a settled one.
Now the part that's easy to miss. During a response, activated B cells pile into structures in the lymph nodes called germinal centres and start deliberately mutating the DNA coding for their own antibody's variable region, at a rate around a million times the normal background. Most of those mutations make the antibody worse or useless, and those cells die. The rare ones that improve the fit get selected to survive and divide.
So the antibodies you have at the end of a response bind much more tightly than the ones you started with. The system runs a miniature round of mutation and selection on itself, on purpose, over about a week.
T cells and the inspection system
Antibodies work on things floating about outside cells. That leaves an obvious gap: a virus that's already got inside is invisible to them.
The solution is that every nucleated cell you have continuously chops up samples of the proteins it's currently making and displays the fragments on its surface, on molecules called MHC. It's an open declaration of internal business, published constantly by every cell, whether anything's wrong or not.
Cytotoxic T cells patrol and read those displays. If a cell is showing fragments of something that shouldn't be there, the T cell instructs it to self-destruct, and the cell complies — dying tidily and taking the virus factory with it. This is why the display system exists, and why viruses go to considerable lengths to interfere with it. Natural killer cells cover that loophole: they're suspicious of any cell showing an unusually low level of display, on the reasonable assumption that something has been tampered with.
Helper T cells do something different and arguably more important. They don't kill. They coordinate — licensing B cells to switch antibody class and enter germinal centres, activating macrophages, and shaping which kind of response gets mounted. Without helper T cells the rest of the adaptive system barely functions, which is why they're such an effective target for anything that wants to disable immunity.
MHC genes are the most variable genes in the human population by a wide margin, with thousands of known variants. That diversity is a population-level defence — no single pathogen can evade everyone's display system at once — and it's also the reason transplanted tissue has to be matched.
Memory, and why some infections only happen once
A first encounter takes time. The right lymphocyte has to be found among millions, activated, and expanded into a large clone, and that's roughly a week's work, sometimes more. You feel ill during it.
Afterwards, most of that expanded army dies off — it's expensive to maintain. But a subset persists as memory cells, and they change the arithmetic completely. They're already present in far greater numbers than the original one-in-a-million. They're pre-selected for a good fit, after affinity maturation. They activate faster and need less prompting. Separately, long-lived plasma cells settle in the bone marrow and keep secreting antibody for years or decades without being asked.
The result is that a second encounter with the same thing gets dealt with in a day or two, often before you notice anything. The pathogen never reaches the density that makes you feel unwell. It isn't that you can't be infected again; it's that the infection gets extinguished before it amounts to anything.
This is the plain mechanism that vaccination is built on — presenting the recognition features without the disease so the memory exists in advance. That's the whole of the principle, and everything specific about it belongs with a clinician.
Memory isn't uniform, though, and the reason some infections come round again and again is worth understanding. Memory is specific to particular molecular shapes. If a pathogen changes those shapes, existing memory no longer matches well. Some viruses mutate their surface proteins constantly, so last year's memory is a poor fit for this year's version. Others simply come in dozens of distinct varieties, and immunity to one buys you nothing against the rest. Nothing has failed in those cases. The lock changed.
Fever is a decision, not a malfunction
Your body temperature isn't fixed at a value that illness disturbs. It's regulated to a set point in the hypothalamus, and during infection that set point gets deliberately moved upwards.
Signalling molecules released during the immune response — including interleukin-1, interleukin-6 and tumour necrosis factor — act on the brain and raise the target. Everything after that is your body working hard to reach the new target: shivering to generate heat, blood vessels in the skin narrowing to conserve it, and the sensation of being cold in a warm room, which is the odd part. You feel cold because relative to the new set point, you are.
The evidence that this is a defence rather than a breakdown is fairly persuasive. Many bacteria replicate less efficiently above normal body temperature. Several immune functions run faster when warm. And there's a neat observation from cold-blooded animals: infected lizards will deliberately move to a hotter spot and hold themselves at a raised temperature, which they can only be doing behaviourally, since they have no way to generate the heat internally. Michael Kluger's work on desert iguanas in the 1970s is the classic demonstration.
The same logic applies to most of the misery of being ill. Aching, tiredness, loss of appetite and wanting to lie still are driven by the same signalling molecules, and they look a lot like a coordinated response rather than a set of unrelated damages. What that means for any individual person who's ill is a question for a doctor, and this article isn't going anywhere near it.
Why the system attacks the wrong thing sometimes
If receptors are generated at random, a large fraction of them will fit something in your own body. That isn't a bug in the design; it's an unavoidable consequence of generating variety blindly. So the filtering has to be severe.
Developing T cells go through the thymus, where they get tested twice. First, can the receptor engage the display molecules at all — if not, the cell is useless and dies. Second, does it bind self-material too strongly — if so, it's dangerous and gets deleted. The great majority of cells entering the thymus fail one test or the other and never leave. The thymus even switches on genes belonging to other tissues entirely, specifically so that developing cells can be shown proteins they'd otherwise never meet until it was too late.
It's a good filter. It can't be a perfect one.
Not every self-protein can be shown during training. Some tissues are physically walled off from immune traffic, so nothing there was ever presented for inspection — and if that barrier is later breached, the material inside registers as foreign. Some microbial molecules resemble human ones closely enough that a response raised against the microbe cross-reacts with tissue, an effect called molecular mimicry. And inflammation itself changes the context, because material released from dying cells gets presented at exactly the moment the system is primed to treat everything it sees as hostile.
There's a second layer of protection outside the thymus. Regulatory T cells actively suppress responses against self, and there are mechanisms that shut down a lymphocyte that sees its target without the accompanying danger signals. Autoimmunity is what happens when the randomly generated repertoire, the imperfect filter and the suppression system all fail on the same target at the same time. Genetic background, particularly the variants of the display molecules a person carries, changes the odds considerably. So does sex — a large majority of people affected by autoimmune conditions are women, and why that is remains an open question rather than a solved one.
Allergy is a related kind of misfire aimed outwards instead of inwards: a vigorous response, IgE and histamine and all, mounted against something that posed no threat in the first place.
Which is the honest summary of the whole system. It's a randomly generated, self-editing, distributed recognition network with no supervisor, running continuously, and the same properties that let it handle threats it has never seen are the properties that let it occasionally get things badly wrong. Nobody has worked out how to have one without the other.
