Autoimmunity

Why does a defence system built to tell self from foreign ever turn on the body it protects?

The usual framing is that the immune system makes a mistake — a glitch, friendly fire in an otherwise reliable machine. That invites the wrong question, because self-reactivity is not rare. Lymphocytes capable of binding your own tissues circulate in you right now, in everyone, and low levels of antibodies against self are ordinary in healthy people and commoner with age.

The error is not exceptional, then; it is continuous, and almost always silent. So the question is harder than "what went wrong?" It is: why is a system built so that self-reactivity is guaranteed, and what normally keeps it quiet?

Start with how a receptor is made. The system must recognise pathogens that do not yet exist — microbes that will mutate next year, or first infect a human next decade — and no inherited catalogue covers that. So the body builds its receptors by randomly cutting and rejoining gene segments in each developing lymphocyte, generating a repertoire vast enough to bind almost any shape. Notice the trap. A process random with respect to shape cannot be told to avoid your own shapes: coverage of the unknown and blindness to self are the same property, so self-reactive receptors are not a failure of manufacture but its by-product, produced continuously, in everyone, by design.

It goes further, and this overturns the folk picture. Developing T cells in the thymus are tested against the body's own molecules, and those binding too weakly die of neglect; only cells with a weak but real recognition of self mature, while strong self-binders are killed. The system does not sort into "self" and "foreign" at all. It sorts on a continuous scale of binding strength, with a threshold — and every threshold has two ways of being wrong. Set it strict and you delete useful cells, leaving holes a pathogen can walk through; set it loose and self-reactive cells survive. No setting gives zero autoimmunity except at the price of worse infection.

Then ask what the thymus can test against: only antigens present there. Proteins made solely in the pancreas or the myelin sheath would never appear — except that a transcription factor called AIRE drives thymic cells to express a sample of tissue-restricted proteins, insulin among them, so clones against them can be caught. When AIRE is defective, people develop autoimmunity across several organs at once: about as direct a demonstration as biology offers that the layer is load-bearing.

Since deletion is necessarily incomplete, further layers follow. Regulatory T cells suppress self-reactive responses, and when the gene FOXP3 that specifies them is broken the result is overwhelming multi-organ autoimmunity in infancy. And a T cell meeting its antigen without signals of damage is switched off rather than activated. That is the crucial principle: activation requires antigen plus context. The system never asks "is this self?", a question it cannot answer. It asks whether this shape appears alongside evidence that something is wrong.

Which tells us where disease should come from, and it does: an infection supplies the context. If a pathogen protein resembles a self protein — molecular mimicry — a correct response to the microbe becomes a response to you, as in rheumatic heart disease after streptococcal infection. Epstein-Barr virus and multiple sclerosis look like the same story: a study of more than ten million US military personnel found MS risk rose roughly thirty-two-fold after EBV infection, with the disease essentially absent in the never-infected — strong evidence for a necessary trigger, not a sufficient one, since nearly everyone carries EBV and few develop MS.

Two honest limits. For most autoimmune diseases the specific self-antigen and trigger remain unknown, and "autoimmunity" names a family of conditions with different mechanisms, not one disease. Susceptibility is heavily genetic — particular HLA variants dominate the risk — yet identical twins are concordant well under half the time for type 1 diabetes, so environment does much of the work. The sex bias, roughly four in five patients being women, is real and unexplained; X-chromosome gene dosage is the leading line of inquiry, not a settled answer.

Three corrections follow.

First, "the body attacking itself" describes the outcome, not the decision. Nothing identifies a target as self and proceeds anyway; the system applies a binding threshold and a context requirement, and autoimmunity is what those rules produce when a shape sits near the threshold and the context arrives.

Second, tolerance of your own tissue is not passive — not the system simply ignoring what it recognises. It is maintained by continuous work: deletion, receptor editing, suppression, withholding of the second signal. Being a process, it can be lost, which is why it fails suddenly in an adult who was fine the day before.

Third, this is not allergy, treated separately in this collection. There the target genuinely is foreign — pollen, peanut — and the failure is one of threat assessment, an expulsion programme run against something harmless. Here the target is your own tissue and the failure is one of identity: self against non-self, not harmless against dangerous. The two share a lesson — the damage is done by the response, not the antigen — but not a mechanism.

• Why the eye, brain and testis are partly walled off from immune surveillance, and what happens when the wall is breached.
• How treatments that suppress autoimmunity trade the disease against the infections immunity was preventing.

Every term the collection defines is gathered in the glossary.