The Adaptive Immune Response

The infection you never had#

Chickenpox, once. Measles, once. Most people who catch these get them exactly one time, and the reason is not that the virus disappears from the world afterwards. You are re-exposed constantly — on trains, in classrooms, from your own children. The virus arrives, starts replicating, and is annihilated so quickly that you register nothing at all.

That silence is not the absence of a fight. It is a fight that was over in about thirty-six hours, waged by cells that were manufactured during your first illness and have been idling in your lymph nodes ever since, holding a molecular description of that exact pathogen.

The interesting question is not why you are protected. It is why the first encounter took a week, and what changed such that the second one takes a day.

Two systems, two timescales#

Your immune defences are conventionally split into two arms, and the split is really a split in strategy.

The innate system is fast, generic, and hard-coded. Epithelial barriers, complement proteins, neutrophils, and macrophages recognize broad molecular signatures shared by whole classes of microbe — bacterial cell-wall components, double-stranded RNA, flagellin — using a fixed, genome-encoded set of pattern-recognition receptors. It engages within minutes. It is also, on its own, frequently insufficient: it cannot distinguish one strain of influenza from another, and classically it learns nothing from the encounter.

The adaptive system is slow, exquisitely specific, and it remembers. Its cells are lymphocytes — B cells and T cells — and each individual lymphocyte carries many copies of a single receptor with a single binding specificity. Crucially, that receptor is not inherited. During development, each lymphocyte randomly cuts and rejoins its receptor gene segments (V(D)J recombination), producing a repertoire estimated at more than $10^{11}$ distinct specificities across the body.

This has a strange consequence. Your body builds receptors for pathogens it has never met, including ones that do not exist yet, by generating essentially every shape and keeping the library on file. Somewhere in you right now there are lymphocytes specific for a virus that will not emerge for another decade.

The catch is arithmetic. If specificity is spread across $10^{11}$ different cells, then for any given pathogen you might start with only a few dozen to a few thousand cells that can see it — nowhere near enough to clear an infection. Those few must first be found, and then multiplied. That takes days, and those days are your first illness.

The primary response, and the one that follows#

Below, an infection is introduced into a naive host, allowed to run, and then the same pathogen is delivered again weeks later. Three quantities are tracked: pathogen load, circulating antibody titre, and the memory-cell population.

Press play and watch the first exposure. Notice how long the antibody line stays flat while pathogen load climbs — that flat stretch is the search-and-multiply phase, and it is precisely the interval during which you feel ill. Antibody only becomes detectable around day 6 to 8, peaks a week or two after the pathogen has already been cleared, and then slowly wanes.

Now watch the memory line. It rises during the primary response and then does not come back down — it settles onto a long plateau, and this is the only trace of the infection that persists.

Then let the re-exposure land (or press Re-expose now to deliver it yourself at any point). The difference is the whole subject in one picture: the antibody curve leaps almost immediately and climbs several times higher than the first time, while the pathogen curve barely lifts off the axis. The pathogen was not weaker. The response was earlier.

The arithmetic of clonal expansion#

Once a lymphocyte's receptor binds its antigen and it receives the appropriate co-stimulation, it begins to divide, and it divides fast — an activated lymphocyte can complete a cycle in roughly 6 to 12 hours, among the fastest division rates of any cell in the body. The clone grows exponentially:

$$N(t) = N_0 \, 2^{\,t / T_d}$$

where $N_0$ is the number of antigen-specific cells at the start, $T_d$ is the doubling time, and $t$ is time since activation. With $T_d = 8$ hours, one week of expansion gives

$$\frac{N(7\text{ d})}{N_0} = 2^{\,168/8} = 2^{21} \approx 2 \times 10^{6}$$

a millionfold amplification from a starting handful. Measured expansions of antigen-specific CD8 T cells during acute viral infection are of exactly this order.

Meanwhile the pathogen is doing the same thing. A minimal model of the race writes pathogen load $P$ as growing at intrinsic rate $r$ and being cleared at a rate proportional to the antibody (or effector) level $A$:

$$\frac{dP}{dt} = rP - kAP = (r - kA)\,P$$

The sign of the bracket is the entire story. While $kA < r$, load grows exponentially and you get sicker. The turnover happens the instant antibody crosses

$$A^{*} = \frac{r}{k}$$

so the peak pathogen load — and therefore how ill you actually feel — is governed almost entirely by how long it takes $A$ to reach $A^{*}$, not by how high $A$ eventually gets.

Now put the two equations together. Antibody output tracks the size of the responding clone, so reaching $A^{*}$ requires roughly

$$n = \log_2\!\left(\frac{N^{*}}{N_0}\right)$$

division rounds, taking $n \, T_d$ hours. Because $n$ depends on $N_0$ only through a logarithm, increasing the starting population by a factor of 1000 does not make the response 1000 times faster — it removes about $\log_2 1000 \approx 10$ doublings, or roughly three and a half days at $T_d = 8$ h. That single logarithm is why a week-long primary response becomes a day-and-a-half secondary one, and why immunological memory works by stockpiling cells rather than stockpiling antibody.

One cell, one receptor#

The step that makes all of this possible is the one Frank Macfarlane Burnet proposed in 1957 and called clonal selection: the antigen does not instruct a cell what to make: it selects, from a repertoire that already exists, the rare cells that happen to fit — and those cells alone proliferate.

Watch the antigen drift through a field of lymphocytes, each drawn with a different receptor shape. Only one cell in the field can bind it. The moment it does, every other cell fades to quiescence and the matched cell alone starts dividing, with the live count doubling each round. Follow the counter: by round 10 a single cell has become 1,024, and in a real response the doubling continues past round 20.

Two things are worth noticing. First, the non-matching cells are not destroyed or reprogrammed — they simply stay idle, still holding their own specificities in case something else shows up. Second, the expansion is not clean copying: as the clone grows, some cells commit to short-lived effector fates and a minority differentiate into the long-lived memory population that persists after the pathogen is gone. Those memory cells are the raised plateau you saw on the previous chart.

Antibodies, killers, and the cells that remember#

Clonal selection runs in two parallel lineages that solve different problems.

B cells handle threats in the extracellular space — bacteria in tissue fluid, toxins, free virus between cells. A selected B cell differentiates into a plasma cell, an antibody factory capable of secreting thousands of antibody molecules per second. Antibodies do not kill directly; they mark and block. They neutralize virus by coating the surfaces it uses to enter cells, opsonize bacteria so phagocytes engulf them, and activate complement.

B cells also improve during the response. In germinal centres within lymph nodes, dividing B cells deliberately mutate their receptor genes (somatic hypermutation) and are re-selected on how tightly they bind. Over a couple of weeks, average antibody affinity rises by orders of magnitude — affinity maturation. In parallel, class switching converts early low-affinity IgM into IgG, IgA, or IgE. This is part of why the secondary response is not just faster but qualitatively better: it starts from already-matured, already-class-switched clones.

T cells handle what antibody cannot reach. A virus replicating inside one of your cells is invisible to a circulating antibody, so infected cells display fragments of their internal proteins on MHC class I molecules at the surface. Cytotoxic (CD8) T cells inspect these displays and kill any cell presenting foreign peptide. Helper (CD4) T cells recognize peptide on MHC class II and act as the licensing authority for the whole response — they are required for germinal centre formation, class switching, and effective CD8 memory. This dependency is why losing CD4 T cells, as in untreated HIV infection, collapses immunity far more broadly than the loss of one cell type would suggest.

After clearance, the vast majority of the expanded clone dies by apoptosis over a few weeks. What remains is a memory compartment: memory B cells and memory T cells, plus long-lived plasma cells that take up residence in bone marrow and keep secreting antibody for years or decades. This is why measurable antibody can persist long after an infection, and it is what the plateau in the first animation represents.

What vaccination is actually doing#

Vaccination is best understood not as adding something foreign to the immune system, but as paying the cost of the primary response without paying the cost of the disease.

A vaccine presents antigen — an inactivated pathogen, a purified protein, a polysaccharide conjugate, or an mRNA instruction for the body to make the protein itself — together with signals that convince the innate system this is worth responding to. Clonal selection proceeds exactly as it would in a real infection: the fitting clones expand, germinal centres mature affinity, and a memory population is laid down. What is absent is the replicating pathogen, and therefore the week of climbing pathogen load in which damage accumulates.

Two features of the kinetics explain most of vaccine practice. Boosters exist because each re-exposure to antigen is itself a secondary response — it re-expands the memory pool and pushes affinity maturation further, which is why a second or third dose typically produces much higher and more durable titres than the first. And waning immunity happens because antibody titre decays even while memory cells persist: protection against severe disease (which memory can restore within days) often outlasts protection against any infection at all (which depends on circulating antibody being above threshold at the moment of exposure).

It is also why the timing of a vaccine matters more than its perfection. A vaccine that shortens the response lag from seven days to two does not need to be sterilizing to change the outcome — by the $A^{*}$ argument above, cutting the time to threshold caps the peak pathogen load, and peak load is what correlates with severity.