Weapons Generated at Random

Your immune system can recognise a molecule that has never existed before. A synthetic compound made last week, or a virus that mutated this morning. It does this with a genome that contains no information about either.

How is the interesting question, and the answer is one of the few places in biology where deliberate randomness is the design.

Take the two halves of the system first. The innate immune system responds within minutes to broad molecular patterns shared by whole classes of pathogens: the components of bacterial cell walls, the structures of viral RNA. It is fast, general, and has no memory. Inflammation, fever and phagocytes belong to it.

The adaptive system takes days to respond and is exquisitely specific: it produces antibodies matching a particular antigen, and it remembers. That memory is why you get chickenpox once and why vaccination works at all.

The recognition problem belongs to the adaptive side. A receptor that binds a specific molecule has a specific shape, which must be encoded somewhere, and the number of possible antigens is unbounded while the genome is finite.

Susumu Tonegawa's solution, which won a Nobel in 1987, is that the receptor genes are not inherited whole. Each developing lymphocyte cuts and rejoins a selection of gene segments (V, D and J) in a combination chosen essentially at random, with additional nucleotides inserted at the joins. A few hundred segments generate a receptor repertoire estimated in the hundreds of millions to billions.

The body builds an enormous random library of shapes before knowing what it will need, and then selects from it when something turns up. Each lymphocyte carries one receptor; when a pathogen arrives, the cells whose receptors happen to bind it are stimulated to proliferate, and the ones that match best proliferate most.

This is natural selection, running inside you, over days.

It also creates an obvious problem. If receptors are generated at random, a large fraction will recognise your own tissues.

The answer is negative selection. Developing lymphocytes are tested against self-antigens in the thymus and bone marrow, and those that bind strongly are killed or inactivated. The process discards a substantial majority of what is produced, which is expensive and is the price of generating the repertoire randomly in the first place.

It is also imperfect, and its failures are autoimmune disease. The immune system's central computational problem is distinguishing self from non-self, and it solves that statistically rather than perfectly, which means some false positives are structural.

Vaccination follows directly from all this and is conceptually simple whatever the delivery method. Present the antigen without the disease, so memory cells are generated in advance. A killed pathogen, a weakened one, a purified surface protein, or mRNA instructing your own cells to make one viral protein; the immunological mechanism is identical in each case.

The mRNA vaccines were novel in delivery and entirely conventional in principle, which is why they could be designed within days of the virus's sequence being published. The remaining time went on manufacturing and trials.

Herd immunity is worth getting precise, because it is used loosely. When enough of a population is immune, chains of transmission break and the pathogen cannot sustain itself, protecting people who are not immune: newborns, the immunocompromised, those for whom the vaccine failed.

The threshold depends on how contagious the disease is, roughly one minus one over the reproduction number. A disease where each case infects two others needs around 50% immunity. Measles, where each case infects twelve to eighteen, needs above 90%, which is why measles returns first and fastest when coverage slips, and why the margin for error is so much narrower than for other diseases.

One last thing, because it is counterintuitive and clinically important.

In severe infections, much of the damage is frequently done by the immune response rather than by the pathogen. Inflammation is destructive by design (it is supposed to be, since the point is to destroy something) and a poorly regulated response floods tissue with enzymes and reactive compounds that do not distinguish carefully.

Sepsis and severe acute respiratory distress are substantially this. Which is why some treatments for severe infection are immunosuppressive, a strategy that sounds backwards until you recognise that what is being suppressed is the response and what is being protected is the patient's own lungs.