Immunology

A rose thorn in a starfish larva

In the winter of 1882, working in Messina, Elie Metchnikoff pushed a rose thorn into the transparent larva of a starfish and looked at it the next morning under his microscope. Motile cells had gathered around the thorn and surrounded it. He had chosen a transparent animal with no blood vessels, no antibodies and no lymphocytes precisely because it stripped the phenomenon down: whatever those cells were doing, they were doing it without any of the machinery that human immunology at the time was built around.

He called them phagocytes and spent the next two decades arguing that cellular ingestion was the basis of immunity, against Paul Ehrlich and the German school who held that immunity was carried by soluble substances in serum. The 1908 Nobel Prize in Physiology or Medicine was awarded to both of them, which was the correct answer, though neither man was pleased about it.

Key idea: Innate immunity is not a primitive precursor to the adaptive response. It is a complete defence system in its own right, and it clears the overwhelming majority of encounters with microorganisms before a lymphocyte is ever recruited.

The surfaces, and how large they actually are

Textbooks have long quoted a gut surface area of 300 to 400 square metres, the size of a tennis court. Herbert Helander and Lars Fandriks went back to the morphometric data in 2014 and recalculated it: the adult human gastrointestinal mucosa is closer to 30 to 32 square metres, roughly half a badminton court. The old figure came from over-generous assumptions about villus and microvillus amplification. It is worth carrying the corrected number, both because it is right and as a reminder that memorable figures propagate through textbooks for decades without anyone rechecking them.

The barriers themselves work in three layers.

• Physical. Stratified keratinised epidermis with tight junctions; mucus over every wet epithelium; the mucociliary escalator moving a sheet of mucus out of the airway at about 1 centimetre per minute; peristalsis; the flushing of urine and tears.
• Chemical. Skin pH near 5 and sebum fatty acids; gastric acid at pH 1 to 2, which is why acid suppression raises the risk of enteric infection and of Clostridioides difficile; lysozyme in tears and saliva cleaving peptidoglycan; lactoferrin sequestering iron. Then the antimicrobial peptides: alpha-defensins from Paneth cells at the base of intestinal crypts and from neutrophil granules, beta-defensins from epithelium, and the cathelicidin LL-37. These are small cationic amphipathic peptides that insert into and disrupt anionic bacterial membranes; the reason they spare host cells is that mammalian outer leaflets are largely zwitterionic and carry cholesterol.
• Microbiological. The commensal flora occupies niches, consumes nutrients and produces inhibitory metabolites. Colonisation resistance is a real and clinically consequential defence: broad-spectrum antibiotics remove it and Clostridioides difficile fills the vacancy, which is why faecal microbiota transplantation cures recurrent disease at rates that no antibiotic matches.

The cells

Two entries deserve expansion because they have changed since older textbooks.

Dendritic cells were identified by Ralph Steinman and Zanvil Cohn in 1973, in a paper that describes an unfamiliar cell in mouse spleen with long branching processes, distinguishable from macrophages by morphology, adherence and surface properties. Their function as the cell type that primes naive T cells, and therefore as the bridge between innate recognition and adaptive response, took another decade to establish. Steinman shared the 2011 Nobel Prize, announced three days after his death.

Tissue macrophages were long assumed to be replenished continuously from blood monocytes. Fate-mapping work, beginning with Florent Ginhoux and colleagues on microglia in 2010, showed that many tissue-resident macrophage populations, including microglia, Kupffer cells and alveolar macrophages, are seeded from yolk sac and fetal liver progenitors before birth and maintain themselves by local proliferation, largely independent of adult haematopoiesis. That changes how you think about them: they are long-lived tissue residents with organ-specific programmes, not a rolling supply of blood cells.

Phagocytosis, step by step

1. Recognition. Non-opsonic receptors bind microbial surfaces directly: the mannose receptor for terminal mannose, dectin-1 for fungal beta-glucan, scavenger receptors for modified lipids. Opsonic receptors bind host molecules already deposited on the microbe: Fc gamma receptors bind IgG, complement receptor 3 binds iC3b. Opsonisation raises the efficiency of uptake by orders of magnitude, which is the mechanistic reason antibody and complement matter to a cell that has no antigen receptor of its own.
2. Engulfment. Actin polymerisation extends pseudopods that zipper around the particle; the membrane closes to form a phagosome.
3. Maturation. Sequential fusion with endosomes and then lysosomes acidifies the compartment to pH 4.5 to 5 and delivers hydrolases.
4. Killing. The NADPH oxidase complex assembles on the phagosome membrane and transfers electrons to oxygen, generating superoxide. This is the respiratory burst. Superoxide is dismutated to hydrogen peroxide, and myeloperoxidase converts that plus chloride into hypochlorous acid. In parallel, inducible nitric oxide synthase produces nitric oxide, and antimicrobial peptides and proteases are delivered from granules.

Chronic granulomatous disease is the experiment nature ran on step four. Mutations in NADPH oxidase subunits, most commonly X-linked CYBB encoding gp91phox, abolish the respiratory burst. Patients suffer recurrent infections with a characteristic set of catalase-positive organisms, notably Staphylococcus aureus, Serratia, Burkholderia cepacia, Nocardia and Aspergillus, and they form granulomas because macrophages that cannot kill an ingested organism keep walling it off. The diagnostic test is functional: the dihydrorhodamine flow cytometry assay measures whether stimulated neutrophils can oxidise a dye, and it has replaced the older nitroblue tetrazolium slide test.

The point: A phagocyte that cannot generate oxidants ingests bacteria perfectly well and simply fails to kill them. Recognition and killing are separable, and each has its own deficiency syndrome.

Neutrophils have a third option. Volker Brinkmann and colleagues described neutrophil extracellular traps in 2004: activated neutrophils extrude decondensed chromatin studded with granule proteins, forming an extracellular mesh that binds and kills bacteria. The mechanism has been implicated in thrombosis and in autoantibody generation in lupus, where the exposed chromatin is a plausible antigen source.

Natural killer cells and the missing-self idea

A cytotoxic T cell needs to see a peptide presented on MHC class I. Viruses and tumours exploit this by downregulating MHC class I, which makes them invisible to T cells. Natural killer cells solve that with an inverted logic: they are inhibited by MHC class I.

Inhibitory receptors, the killer immunoglobulin-like receptors in humans and NKG2A binding HLA-E, deliver a negative signal when they engage self MHC class I. Activating receptors such as NKG2D bind stress-induced ligands, MICA, MICB and the ULBPs, which normal cells do not display. The killing decision is the integration of the two. A healthy cell shows plenty of MHC class I and no stress ligands, and is spared. A virus-infected or transformed cell that has shed MHC class I and put up stress ligands loses inhibition and gains activation, and is killed by the same perforin and granzyme machinery a cytotoxic T cell uses.

Type I interferons tie this together. A cell that detects viral nucleic acid secretes interferon alpha and beta, which induce an antiviral state in neighbouring cells and activate natural killer cells. Deficiencies in this axis, and autoantibodies against type I interferon, were shown during the COVID-19 pandemic to account for a measurable fraction of critical illness in otherwise healthy people, which is a striking demonstration that this arm carries real clinical weight.

Common misconceptions

• "The gut has the surface area of a tennis court." Careful morphometry puts it at roughly 30 square metres, not 300. The larger figure has been repeated for decades without recalculation.
• "Innate immunity is non-specific." It is not antigen-specific, but it is highly specific for conserved microbial structures, and it distinguishes bacteria from fungi from viruses through distinct receptors.
• "Tissue macrophages come from blood monocytes." Many resident populations, including microglia and Kupffer cells, are seeded before birth and self-renew locally.
• "Chronic granulomatous disease impairs phagocytosis." Ingestion is normal; the defect is in oxidative killing after ingestion, which is why the diagnostic test measures oxidative burst rather than uptake.
• "Natural killer cells are activated by MHC." They are inhibited by it. Loss of MHC class I removes the brake, which is the missing-self principle.

Recap

• Metchnikoff's 1882 starfish experiment established cellular phagocytosis as a defence independent of serum factors, and the 1908 Nobel divided the credit between the cellular and humoral schools.
• Barriers are physical, chemical and microbiological; the gastrointestinal mucosa is about 30 square metres, and colonisation resistance is a genuine defence that antibiotics remove.
• Antimicrobial peptides are cationic and amphipathic and spare host membranes because those are largely zwitterionic and cholesterol-rich.
• Neutrophils, macrophages, dendritic cells, natural killer cells, mast cells, eosinophils and innate lymphoid cells each have a defined effector role and a characteristic deficiency.
• Many tissue macrophages are prenatally seeded and self-renewing rather than monocyte-derived.
• Phagocytosis proceeds by recognition, engulfment, maturation and oxidative killing; chronic granulomatous disease removes only the last step.
• Natural killer cells integrate inhibition by self MHC class I with activation by stress ligands, killing cells that have lost the former and gained the latter.

Sources

1. The Nobel Foundation. (1908). The Nobel Prize in Physiology or Medicine 1908: Ilya Mechnikov and Paul Ehrlich. nobelprize.org
2. Helander, H. F., & Fandriks, L. (2014). Surface area of the digestive tract, revisited. Scandinavian Journal of Gastroenterology, 49(6), 681-689. pubmed.ncbi.nlm.nih.gov
3. Steinman, R. M., & Cohn, Z. A. (1973). Identification of a novel cell type in peripheral lymphoid organs of mice, I: Morphology, quantitation, tissue distribution. Journal of Experimental Medicine, 137(5), 1142-1162. pubmed.ncbi.nlm.nih.gov
4. Ginhoux, F., Greter, M., Leboeuf, M., Nandi, S., See, P., Gokhan, S., et al. (2010). Fate mapping analysis reveals that adult microglia derive from primitive macrophages. Science, 330(6005), 841-845. pubmed.ncbi.nlm.nih.gov
5. Brinkmann, V., Reichard, U., Goosmann, C., Fauler, B., Uhlemann, Y., Weiss, D. S., Weinrauch, Y., & Zychlinsky, A. (2004). Neutrophil extracellular traps kill bacteria. Science, 303(5663), 1532-1535. pubmed.ncbi.nlm.nih.gov
6. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The front line of host defense. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
7. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). Cellular defenses. In Microbiology (Section 17.3). OpenStax. openstax.org

A fly covered in mould

The 1996 paper by Bruno Lemaitre, Emmanuelle Nicolas, Lydia Michaut, Jean-Marc Reichhart and Jules Hoffmann contains a photograph that did more work than most figures ever do: an adult Drosophila so densely overgrown with Aspergillus fumigatus hyphae that the animal is barely visible beneath them. The fly carried a loss-of-function mutation in Toll, a gene already famous to developmental biologists for establishing the dorsoventral axis of the embryo. Wild-type flies given the same fungal challenge cleared it.

Toll turned out to control expression of the antifungal peptide drosomycin in the adult. A gene known for patterning an embryo was also a receptor pathway for infection, and flies, which have no lymphocytes and no antibodies, were defending themselves with it.

Charles Janeway had predicted something like this in 1989, from a completely different direction. He pointed out what he called the immunologist's dirty little secret: injecting a purified protein into an animal produces almost no antibody response unless you also inject an adjuvant, which is typically killed bacteria or bacterial extract. If antigen alone were sufficient, adjuvants would be unnecessary. He proposed that the immune system carries germline-encoded receptors for conserved microbial structures, and that signals from those receptors are what license a lymphocyte response.

Why this matters: The prediction came from an unexplained requirement in a routine laboratory protocol. Adjuvants worked and nobody could say why, and taking that seriously produced the framework the rest of this lesson describes.

Two experiments that closed the argument

In 1997 Ruslan Medzhitov, Paula Preston-Hurlburt and Janeway reported a human homologue of Drosophila Toll and showed that a constitutively active version, expressed in a human cell line, activated NF-kB and induced the costimulatory molecule B7.1 along with inflammatory cytokines. The receptor did not merely alarm the cell; it turned on exactly the molecules a T cell needs in order to be activated. That is the mechanistic link Janeway's hypothesis required.

The second experiment came from a mouse strain that had puzzled people for decades. C3H/HeJ mice are almost completely unresponsive to lipopolysaccharide, the endotoxin of Gram-negative bacteria, which kills normal mice at low doses. The locus responsible had been mapped genetically and named Lps. In 1998 Alexander Poltorak, Bruce Beutler and colleagues finished the positional cloning and found the causative mutation in Tlr4: a single proline to histidine substitution in the cytoplasmic signalling domain. A second unresponsive strain, C57BL/10ScCr, carried a deletion of the gene.

Two independent routes, one in flies by forward genetics and one in mice by positional cloning of a spontaneous mutant, converged on the same receptor family. Hoffmann and Beutler shared the 2011 Nobel Prize with Steinman for this work.

What the Toll-like receptors see, and where

Humans have ten functional Toll-like receptors. Their location is not incidental; it is a solution to a hard discrimination problem.

Notice that every nucleic acid sensor is inside an endosome. Host nucleic acid is abundant in the cytosol and nucleus, and a surface receptor for RNA or DNA would be triggered constantly by dying cells. Restricting these receptors to a compartment reached mainly by endocytosed material makes location itself part of the self and non-self discrimination. The trafficking chaperone UNC93B1 is what delivers them there, and its loss removes TLR3, TLR7, TLR8 and TLR9 signalling at once.

That discrimination is imperfect and its failure is instructive. In systemic lupus erythematosus, immune complexes containing self nucleic acid are taken up through Fc receptors into exactly the endosomal compartment where TLR7 and TLR9 sit, which drives type I interferon production by plasmacytoid dendritic cells. Hydroxychloroquine, the oldest drug still in routine lupus use, accumulates in acidic compartments and raises endosomal pH, impairing this signalling. A drug adopted empirically from antimalarial practice turns out to act on the mechanism.

Signalling, and what each branch does

All Toll-like receptors except TLR3 signal through the adaptor MyD88, which recruits the IRAK kinases and leads to activation of NF-kB and the MAP kinases, producing tumour necrosis factor, IL-6, IL-1 and IL-12 and upregulating costimulatory molecules. TLR3, and TLR4 after internalisation, signal instead through TRIF, which activates IRF3 and drives type I interferon.

The human deficiencies map onto these branches with unusual precision.

• MyD88 or IRAK4 deficiency. Recurrent invasive pyogenic infection in childhood, especially Streptococcus pneumoniae and Staphylococcus aureus, with strikingly poor inflammatory signs: children can have meningitis with a normal C-reactive protein. Susceptibility narrows with age, presumably as adaptive memory accumulates, and most patients who survive childhood do well.
• TLR3, UNC93B1 or TRIF pathway deficiency. Herpes simplex virus 1 encephalitis in childhood, with normal resistance to other infections. A defect confined to interferon induction in the central nervous system produces a single, specific, devastating phenotype.

The upshot: Innate sensing defects do not produce global immunodeficiency. They produce narrow, pathogen-specific vulnerabilities, which is strong evidence that the pathways are non-redundant for particular organisms.

The cytosolic sensors

A pathogen that reaches the cytosol has escaped the endosomal sensors, so there is a second layer.

• NOD1 and NOD2 detect peptidoglycan fragments in the cytosol; NOD2 senses muramyl dipeptide. Loss-of-function NOD2 variants are the strongest common genetic risk factors for Crohn disease, which is a reminder that a sensing defect can produce excess inflammation rather than too little, presumably through failure to handle the commensal flora properly.
• RIG-I and MDA5 detect cytosolic viral RNA. RIG-I recognises short double-stranded RNA bearing a 5 prime triphosphate, a feature host capped mRNAs lack; MDA5 recognises long double-stranded RNA. Both signal through MAVS on the mitochondrial outer membrane to induce type I interferon.
• cGAS and STING detect cytosolic double-stranded DNA. cGAS synthesises the cyclic dinucleotide cGAMP, which activates STING and drives interferon production. Gain-of-function STING mutations cause STING-associated vasculopathy with onset in infancy, an interferonopathy with skin vasculitis and interstitial lung disease. The same pathway is a major target in tumour immunology, because DNA from dying tumour cells activates it in dendritic cells.

Inflammasomes and the two-signal rule

IL-1 beta and IL-18 are made as inactive precursors and have no signal peptide, so they are neither processed nor secreted by the conventional route. The inflammasome solves both problems. NLRP3, with the adaptor ASC, oligomerises and recruits pro-caspase-1, which autoactivates and cleaves pro-IL-1 beta and pro-IL-18 into their active forms. Caspase-1 also cleaves gasdermin D, whose N-terminal fragment inserts into the plasma membrane and forms pores. Those pores release the mature cytokines and, if enough form, cause pyroptosis, a lytic inflammatory cell death.

Activation requires two signals, and this is worth remembering because it explains a great deal of clinical immunology. Signal one, typically through a Toll-like receptor, induces transcription of NLRP3 and pro-IL-1 beta. Signal two, a disturbance such as potassium efflux, lysosomal damage from crystals, or mitochondrial stress, triggers assembly. Neither alone suffices. Monosodium urate crystals in gout and calcium pyrophosphate in pseudogout deliver signal two, which is why gout is an inflammasome disease and why IL-1 blockade works in patients who cannot take colchicine or steroids.

The clean genetic proof is the cryopyrin-associated periodic syndromes, caused by gain-of-function mutations in NLRP3 itself: familial cold autoinflammatory syndrome, Muckle-Wells syndrome and neonatal-onset multisystem inflammatory disease, a spectrum of recurrent fever, urticarial rash, arthropathy, sensorineural deafness and amyloidosis. These respond dramatically to IL-1 blockade with anakinra or canakinumab. A single gene, a single cytokine, a targeted drug, and a near-complete response is about as clean as mechanistic medicine gets.

Adjuvants come back here. Stephanie Eisenbarth and colleagues showed in 2008 that aluminium salts, the adjuvant in most human vaccines since the 1920s, activate the NLRP3 inflammasome, and that alum-driven antibody responses were impaired in mice lacking NLRP3 or ASC. The empirical practice Janeway had pointed to was, at least in part, inflammasome activation.

In short: The immune system does not ask whether a molecule is foreign. It asks whether the pattern is one that microbes have and hosts do not, and whether it is in a place it should not be.

Common misconceptions

• "Pattern recognition receptors recognise antigens." They recognise conserved classes of molecule, not individual epitopes, and they are germline encoded rather than somatically generated.
• "TLR4 binds lipopolysaccharide directly on its own." Recognition requires the accessory proteins MD-2 and CD14 and the transfer protein LBP; TLR4 without MD-2 does not respond.
• "Nucleic acid sensors would be dangerous, so the immune system avoids them." It has many, and it manages the risk by putting them in endosomes and the cytosol rather than on the cell surface.
• "An innate sensing defect causes broad immunodeficiency." MyD88 and IRAK4 deficiency give pyogenic infection, TLR3 pathway deficiency gives herpes encephalitis, and each is otherwise narrow.
• "IL-1 beta is secreted like other cytokines." It lacks a signal peptide and requires caspase-1 cleavage and gasdermin D pores to leave the cell.

Putting it together

• Janeway predicted germline-encoded pattern receptors in 1989 from the unexplained necessity of adjuvants.
• Toll mutant flies died of Aspergillus in 1996; a human Toll homologue induced costimulatory molecules in 1997; the lipopolysaccharide-unresponsive mouse locus proved to be Tlr4 in 1998.
• Toll-like receptors sit on the plasma membrane for bacterial surface molecules and in endosomes for nucleic acids, and the compartmentalisation is itself a self and non-self discrimination.
• MyD88 drives NF-kB and inflammatory cytokines; TRIF drives IRF3 and type I interferon, and the corresponding human deficiencies give pyogenic infection and herpes encephalitis respectively.
• Cytosolic sensors NOD1 and NOD2, RIG-I and MDA5, and cGAS with STING cover the compartment that endosomal receptors cannot reach.
• NLRP3 inflammasome activation requires priming plus a disturbance, and produces mature IL-1 beta, IL-18 and gasdermin D pores.
• Gain-of-function NLRP3 causes the cryopyrin-associated periodic syndromes, which respond to IL-1 blockade, and alum adjuvant works partly through the same complex.

Sources

1. Lemaitre, B., Nicolas, E., Michaut, L., Reichhart, J. M., & Hoffmann, J. A. (1996). The dorsoventral regulatory gene cassette spatzle, Toll and cactus controls the potent antifungal response in Drosophila adults. Cell, 86(6), 973-983. pubmed.ncbi.nlm.nih.gov
2. Medzhitov, R., Preston-Hurlburt, P., & Janeway, C. A., Jr. (1997). A human homologue of the Drosophila Toll protein signals activation of adaptive immunity. Nature, 388(6640), 394-397. pubmed.ncbi.nlm.nih.gov
3. Poltorak, A., He, X., Smirnova, I., Liu, M. Y., Van Huffel, C., Du, X., et al. (1998). Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice: Mutations in Tlr4 gene. Science, 282(5396), 2085-2088. pubmed.ncbi.nlm.nih.gov
4. Eisenbarth, S. C., Colegio, O. R., O'Connor, W., Sutterwala, F. S., & Flavell, R. A. (2008). Crucial role for the Nalp3 inflammasome in the immunostimulatory properties of aluminium adjuvants. Nature, 453(7198), 1122-1126. pubmed.ncbi.nlm.nih.gov
5. The Nobel Foundation. (2011). The Nobel Prize in Physiology or Medicine 2011: Bruce A. Beutler, Jules A. Hoffmann and Ralph M. Steinman. nobelprize.org
6. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Receptors of the innate immune system. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov

A third episode of meningococcal meningitis

A 19-year-old student is admitted with meningococcal meningitis. It is her third episode. The first was at age nine, serogroup B; the second at fifteen, serogroup C; this one is serogroup Y. Between episodes she has been entirely well, with no unusual susceptibility to pneumococcus, staphylococcus, fungi or viruses, a normal full blood count and normal immunoglobulins.

That pattern is almost pathognomonic. Recurrent neisserial infection with nothing else wrong points to a deficiency of the terminal complement components, C5 through C9, or of properdin. Neisseria species have a thin peptidoglycan layer and an outer membrane accessible to the membrane attack complex, and they are the one genus for which complement-mediated lysis, as opposed to opsonisation, is the decisive defence. Remove the terminal pathway and everything else still works, except against Neisseria.

Key idea: Complement deficiencies are among the most informative experiments in human immunology, because each level of the cascade produces a different and predictable clinical syndrome.

Three ways in, one place they meet

Complement is about thirty plasma and membrane proteins, most made by the liver, circulating as inactive precursors. Three activation routes converge on the cleavage of C3.

• Classical. C1q, a six-headed molecule, binds the Fc regions of antibody once several are clustered on a surface. One IgM pentamer that has adopted its staple conformation on antigen suffices; IgG requires two or more molecules close together, which is why IgM is far more efficient at fixing complement. Bound C1q activates C1r, which activates C1s, which cleaves C4 and C2 to form the C3 convertase C4b2a. C1q also binds directly to some bacterial surfaces, to C-reactive protein bound to phosphocholine, and to apoptotic cells, so this pathway operates before any antibody exists.
• Lectin. Mannose-binding lectin, ficolins and collectins recognise repeating sugar patterns, particularly terminal mannose and N-acetylglucosamine, that are typical of microbial surfaces and rare on mammalian glycans, which are normally capped with sialic acid and galactose. Associated serine proteases, the MASPs, then cleave C4 and C2 exactly as C1s does. Same convertase, different trigger.
• Alternative. C3 hydrolyses spontaneously in plasma at a low rate, called tickover. The hydrolysed form binds factor B, which factor D cleaves, generating a fluid-phase convertase. If the resulting C3b lands on a surface that lacks host regulators, it forms the surface convertase C3bBb, stabilised by properdin. This pathway needs no recognition molecule at all; it is continuously active and is stopped on host cells by regulators. It also acts as an amplification loop for the other two: C3b generated by any route feeds back into it.

All three make a C3 convertase. Adding another C3b converts it to a C5 convertase, which cleaves C5 into C5a and C5b. C5b then nucleates C6, C7, C8 and multiple C9 molecules into the membrane attack complex, a pore in the target membrane.

What complement actually does, ranked by importance

The membrane attack complex gets the diagrams, but it is the least important effector for most organisms.

1. Opsonisation. C3b and its cleavage product iC3b coat the microbial surface and are bound by complement receptor 1 and complement receptor 3 on phagocytes. This is the dominant antibacterial function and the reason C3 deficiency is severe.
2. Anaphylatoxins. C3a and C5a are small cleavage fragments released into fluid. C5a is a potent neutrophil chemoattractant and activator; both trigger mast cell degranulation and increase vascular permeability. C5a is also a major driver of the pathology in sepsis.
3. Immune complex clearance. C3b-coated immune complexes bind complement receptor 1 on erythrocytes, which ferry them to the liver and spleen where fixed macrophages strip and destroy them. This is why early classical pathway deficiency causes lupus.
4. Enhancement of B cell responses. C3d deposited on an antigen is bound by complement receptor 2, CD21, which is part of the B cell coreceptor complex. Co-ligation of the B cell receptor and CD21 lowers the activation threshold by two to three orders of magnitude. Complement is therefore an adjuvant built into plasma.
5. Lysis by the membrane attack complex. Decisive against Neisseria; of limited importance against most other organisms, which have thick cell walls or capsules.

Remember: If you can only remember one thing about complement effector function, remember opsonisation. Lysis is the exception, not the rule.

The deficiency map

The C1q row deserves emphasis because it inverts the usual expectation. More than 90 percent of reported individuals with complete C1q deficiency develop systemic lupus erythematosus, making it the strongest single-gene risk factor for lupus known. The mechanism is not that complement suppresses autoimmunity directly; it is that apoptotic cells and immune complexes containing nuclear material are not cleared, so nucleic acid persists in tissue, is taken up into endosomes, and drives the interferon response through the nucleic acid sensing Toll-like receptors of the previous lesson. Two lessons, one mechanism.

The screening tests follow the anatomy of the cascade. CH50 measures the classical pathway plus the terminal components, AH50 measures the alternative pathway plus the terminal components, and each returns essentially zero if any single component in its route is absent. So a low CH50 with a normal AH50 localises the defect to C1, C4 or C2; a normal CH50 with low AH50 localises it to factor B, factor D or properdin; and both low points to C3 or the terminal components. Measuring C3 and C4 concentrations separately then distinguishes consumption from deficiency.

Regulation, and what happens when it fails

A continuously active cascade that punches holes in membranes requires tight control. Host cells are protected by several mechanisms.

• C1 inhibitor is a serpin that inactivates C1r and C1s, and also MASP-1 and MASP-2, plasma kallikrein and factor XIIa.
• Factor H binds sialic acid and glycosaminoglycans on host surfaces and, with factor I, degrades C3b there. Microbial surfaces lacking these host markers do not recruit factor H, which is how the alternative pathway distinguishes self from non-self without a recognition molecule.
• Membrane cofactor protein (CD46) and decay accelerating factor (CD55) are membrane regulators that inactivate C3b and dissociate convertases.
• CD59 blocks assembly of C9 into the membrane attack complex.

Three diseases follow directly from failures of these.

Hereditary angioedema is caused by C1 inhibitor deficiency or dysfunction, and it is the most misunderstood entry in this list. Patients present with recurrent episodes of non-pitting subcutaneous or submucosal swelling, often of the face, larynx, extremities or bowel, lasting two to five days. There is no urticaria, no itch, and no response to antihistamines, corticosteroids or adrenaline. The reason is that the mediator is not histamine. C1 inhibitor is the principal brake on factor XIIa and plasma kallikrein, so its loss allows unrestrained cleavage of high molecular weight kininogen and generation of bradykinin, which acts on the B2 receptor to increase vascular permeability. Effective treatments target that pathway: C1 inhibitor concentrate, the B2 receptor antagonist icatibant, and the anti-kallikrein antibody lanadelumab for prophylaxis. C4 is low between attacks and very low during them, which is the useful screening test. A patient given adrenaline for laryngeal swelling that does not respond should prompt this diagnosis.

Paroxysmal nocturnal haemoglobinuria arises from a somatic PIGA mutation in a haematopoietic stem cell. PIGA is required to synthesise the glycosylphosphatidylinositol anchor, so the entire clone loses every GPI-anchored surface protein, including CD55 and CD59. Erythrocytes from that clone are therefore defenceless against the alternative pathway that is continuously ticking over, and they lyse intravascularly, producing haemoglobinuria, iron loss, and a striking thrombotic tendency. Eculizumab, an antibody against C5, blocks terminal pathway activation and transformed the disease. It also produces, iatrogenically, exactly the deficiency described at the start of this lesson: terminal complement blockade with a large increase in meningococcal disease risk. The label carries a boxed warning, and meningococcal vaccination against multiple serogroups, often with antibiotic prophylaxis, is required before treatment. The therapeutic and the genetic experiment are the same experiment.

Atypical haemolytic uraemic syndrome results from loss-of-function mutations in factor H, factor I or CD46, or gain-of-function in C3 or factor B, all of which leave the alternative pathway insufficiently restrained on endothelial surfaces, causing thrombotic microangiopathy with renal failure. It also responds to C5 blockade, which is what distinguishes it therapeutically from Shiga toxin-associated haemolytic uraemic syndrome.

Bottom line: The alternative pathway is always on. Health depends on regulation, and most complement disease is regulatory failure rather than absence of a component.

Common misconceptions

• "Complement kills bacteria by lysis." Opsonisation is the dominant antibacterial function. Lysis by the membrane attack complex matters mainly against Neisseria.
• "The classical pathway requires antibody." C1q also binds directly to some microbial surfaces, to C-reactive protein and to apoptotic cells, so it works in a naive host.
• "Hereditary angioedema is an allergic reaction." The mediator is bradykinin, not histamine, which is why antihistamines and adrenaline do not work and why kallikrein and B2 receptor targeting does.
• "Complement deficiency means infection." Early classical pathway deficiency presents as lupus, through failure to clear apoptotic cells and immune complexes.
• "The alternative pathway needs a trigger." It runs continuously by C3 tickover and is held off host surfaces by factor H, CD55 and CD59.

What to carry forward

• Classical, lectin and alternative pathways converge on a C3 convertase; the alternative pathway also amplifies the other two.
• Effector functions in order of importance: opsonisation, anaphylatoxin release, immune complex clearance, B cell coreceptor engagement through C3d and CD21, and finally lysis.
• Deficiency maps to phenotype: early classical to lupus, C3 to severe pyogenic infection, terminal components and properdin to neisserial disease, mannose-binding lectin to little or nothing.
• More than 90 percent of people with complete C1q deficiency develop lupus, through failure to clear nuclear material that then drives endosomal nucleic acid sensing.
• CH50 and AH50 localise a defect by which pathway is abolished, and C3 and C4 levels distinguish consumption from deficiency.
• C1 inhibitor deficiency causes bradykinin-mediated hereditary angioedema, with a low C4 and no response to antihistamines or adrenaline.
• Loss of CD55 and CD59 in paroxysmal nocturnal haemoglobinuria causes complement-mediated intravascular haemolysis treated by C5 blockade, which then reproduces terminal complement deficiency and its meningococcal risk.

Sources

1. Walport, M. J. (2001). Complement, first of two parts. New England Journal of Medicine, 344(14), 1058-1066. pubmed.ncbi.nlm.nih.gov
2. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The complement system and innate immunity. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
3. Bardhan, M., & Kaushik, R. (2023). Physiology, complement cascade. In StatPearls. StatPearls Publishing. ncbi.nlm.nih.gov
4. Abdulkarim, A., & Craig, T. J. (2023). Hereditary angioedema. In StatPearls. StatPearls Publishing. ncbi.nlm.nih.gov
5. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). Chemical defenses. In Microbiology (Section 17.2). OpenStax. openstax.org

Four words from the first century

In De Medicina, written around AD 30, Aulus Cornelius Celsus set down the signs of inflammation as redness and swelling with heat and pain. Galen later added loss of function. Those five observations survived unchanged for eighteen centuries because they are accurate, and each of them now has a molecular explanation that a modern reader can attach to it.

• Redness and heat come from arteriolar vasodilation and increased blood flow, driven by histamine from mast cells, nitric oxide from endothelium, and prostaglandins E2 and I2.
• Swelling comes from increased venular permeability. Endothelial cells contract at their junctions in response to histamine, bradykinin and leukotrienes, and protein-rich fluid moves into the interstitium.
• Pain comes from bradykinin acting directly on nociceptors, with prostaglandin E2 lowering their threshold so that stimuli which would normally be innocuous become painful. That sensitisation, not the direct activation, is what non-steroidal anti-inflammatory drugs relieve.
• Loss of function is partly protective immobilisation and partly the mechanical consequence of oedema.

The point: Every one of Celsus's signs is a vascular event. Inflammation begins as a change in blood vessels, and the cellular infiltrate follows.

Getting a neutrophil out of the bloodstream

A neutrophil in a postcapillary venule is travelling at roughly a millimetre per second. Extracting it into tissue at a defined location requires a sequence of steps, each with its own molecules, and the sequence is one of the most cleanly worked out in cell biology.

1. Capture and rolling. Activated endothelium displays P-selectin, mobilised within minutes from preformed Weibel-Palade bodies, and E-selectin, transcribed over a few hours in response to IL-1 and tumour necrosis factor. Selectins bind sialylated, fucosylated carbohydrate ligands on the leukocyte, notably sialyl-Lewis x on PSGL-1. These bonds form and break rapidly, so the cell rolls rather than sticks.
2. Activation. Chemokines immobilised on endothelial glycosaminoglycans, CXCL8 for neutrophils and CCL2 for monocytes, engage G protein-coupled receptors on the rolling cell and trigger inside-out signalling that changes leukocyte integrins from a bent, low-affinity conformation to an extended, high-affinity one.
3. Firm adhesion. High-affinity LFA-1 and Mac-1, both containing the beta-2 integrin subunit CD18, bind ICAM-1 on endothelium and arrest the cell.
4. Transmigration. The cell crawls to a junction and squeezes through, using PECAM-1 and JAM family interactions, then crosses the basement membrane and follows the chemokine gradient into tissue.

Two human deficiencies interrupt this at different steps and confirm the model. Leukocyte adhesion deficiency type I is caused by mutations in ITGB2, encoding CD18, so firm adhesion fails. Patients present with delayed separation of the umbilical cord, recurrent bacterial infections without pus, poor wound healing, and a markedly raised neutrophil count, because the neutrophils are made and released but cannot leave the circulation. Absence of pus in a patient with obvious infection is a striking sign and should prompt this diagnosis. Leukocyte adhesion deficiency type II is a defect in a GDP-fucose transporter, so selectin ligands are not fucosylated and rolling fails; it comes with a Bombay blood group phenotype and developmental abnormalities, because the same fucosylation pathway is used elsewhere.

Local becomes systemic

IL-1, tumour necrosis factor and IL-6 are the principal cytokines of acute inflammation. Locally they activate endothelium and recruit cells. In quantity they reach the circulation, and the systemic effects are worth listing because they are what a clinician measures.

• Fever. IL-1 and prostaglandin E2 acting on the vascular organ of the lamina terminalis raise the hypothalamic set point. Antipyretics work by inhibiting cyclooxygenase, and therefore prostaglandin E2, in that structure.
• Acute phase response. IL-6 drives hepatic synthesis of C-reactive protein, serum amyloid A, fibrinogen, haptoglobin and hepcidin. Raised fibrinogen increases erythrocyte aggregation, which is what the erythrocyte sedimentation rate measures; the slower kinetics of fibrinogen explain why the sedimentation rate lags C-reactive protein by days at both onset and resolution.
• Anaemia of inflammation. Hepcidin blocks ferroportin, trapping iron in macrophages and enterocytes. Serum iron falls, ferritin rises, and the patient becomes anaemic with adequate total body iron. It is plausibly a defence, since sequestering iron starves bacteria.
• Shock, in extremis. The same mediators systemically cause vasodilation, capillary leak, myocardial depression and disseminated coagulation.

The eicosanoids, and the drug that reveals the design

Membrane arachidonic acid, released by phospholipase A2, is metabolised down two routes. Cyclooxygenase 1 and 2 produce the prostaglandins and thromboxanes; 5-lipoxygenase produces the leukotrienes, of which leukotriene B4 is a neutrophil chemoattractant and the cysteinyl leukotrienes drive bronchoconstriction and vascular leak.

Aspirin acetylates a serine in the cyclooxygenase active site irreversibly, which is why its effect on platelets, which cannot make new protein, lasts the life of the platelet. Corticosteroids act further upstream and more broadly, inhibiting phospholipase A2 activity and suppressing transcription of the inflammatory programme.

Now the part that changes how you think about inflammation.

Resolution is a programme, not a fading

The intuitive picture is that inflammation ends when the stimulus is cleared and mediators decay. That is wrong, and the evidence is direct: resolution requires new biosynthesis, and blocking that biosynthesis prolongs inflammation even when the pathogen is gone.

Charles Serhan and colleagues showed that during an inflammatory response the same cells switch which lipid mediators they make, a phenomenon called lipid mediator class switching. Prostaglandin E2, which is pro-inflammatory early, induces 15-lipoxygenase and thereby drives the switch to a family of specialised pro-resolving mediators: lipoxins from arachidonic acid, resolvins and protectins from eicosapentaenoic and docosahexaenoic acid, and maresins from macrophages. These are not anti-inflammatory in the sense of blocking; they are agonists at their own receptors and they actively drive an end to the response.

The resolution programme has four measurable components.

1. Stop recruitment. Neutrophil influx ceases while the existing infiltrate is still present.
2. Neutrophil apoptosis. Tissue neutrophils undergo programmed rather than necrotic death, which matters because necrosis would release granule contents and perpetuate injury.
3. Efferocytosis. Macrophages ingest the apoptotic neutrophils. Dying cells release find-me signals such as lysophosphatidylcholine, sphingosine-1-phosphate and nucleotides, and expose phosphatidylserine on the outer leaflet as an eat-me signal. Ingestion of an apoptotic cell actively instructs the macrophage to produce transforming growth factor beta and IL-10 rather than inflammatory cytokines, so clearance and quieting are the same event.
4. Macrophage egress and repair. Macrophages leave through lymphatics, and tissue repair programmes take over.

Aspirin has a second life here. Acetylated cyclooxygenase 2 does not simply stop working; it switches product and generates 15R-hydroxyeicosatetraenoic acid, which is converted to aspirin-triggered 15-epi-lipoxins that are pro-resolving and more resistant to degradation than the native forms. The oldest anti-inflammatory drug in the pharmacopoeia turns out to have a resolution-promoting mechanism nobody suspected for a century.

Why this matters: Failure of resolution, rather than excess initiation, is now the favoured explanation for several chronic inflammatory diseases. That reframes the therapeutic question from how to block inflammation to how to finish it, and blocking and finishing are not the same thing.

Failed efferocytosis is a concrete example. If apoptotic cells are not cleared promptly they progress to secondary necrosis and release nuclear contents. In systemic lupus erythematosus, impaired clearance is one of the mechanisms proposed to supply the nuclear autoantigens, which links this lesson directly to the C1q deficiency phenotype in the previous one.

When it does not resolve: granulomas

If a stimulus cannot be eliminated, as with Mycobacterium tuberculosis, a persistent foreign body or an autoimmune target, the response becomes chronic and macrophage-dominated. A granuloma is an organised structure of epithelioid macrophages, sometimes fusing into multinucleate giant cells, surrounded by lymphocytes, that walls off what cannot be destroyed. Its maintenance depends on tumour necrosis factor and on interferon gamma from TH1 cells.

The clinical proof came from therapeutics. Anti-tumour necrosis factor antibodies used for rheumatoid arthritis and Crohn disease cause reactivation of latent tuberculosis, because existing granulomas lose their structure and contained organisms disseminate. Screening for latent tuberculosis before starting these drugs is mandatory for exactly this reason. A drug's characteristic adverse effect can be the cleanest available evidence about a normal physiological mechanism.

Common misconceptions

• "Inflammation resolves when the stimulus is cleared." Resolution requires active synthesis of specialised pro-resolving mediators, and blocking that synthesis prolongs inflammation even without the stimulus.
• "Prostaglandin E2 causes pain directly." It sensitises nociceptors, lowering their threshold to bradykinin and other stimuli, which is why NSAIDs reduce inflammatory pain more than they reduce acute nociception.
• "Pus indicates severe infection." Pus indicates neutrophils have arrived. Its absence in an obviously infected patient with a high blood neutrophil count suggests an adhesion defect.
• "Anaemia of inflammation is iron deficiency." Total body iron is adequate; hepcidin has trapped it, so ferritin is high while serum iron is low.
• "Efferocytosis is just tidying up." Ingesting an apoptotic cell actively instructs the macrophage to make transforming growth factor beta and IL-10, so clearance and resolution are one process.

Where this leaves us

• Celsus's signs are vascular in origin: vasodilation, permeability, nociceptor sensitisation and consequent loss of function.
• Recruitment proceeds by selectin-mediated rolling, chemokine-triggered integrin activation, integrin-mediated arrest and transmigration.
• Leukocyte adhesion deficiency type I removes CD18 and firm adhesion, giving infections without pus and a high neutrophil count; type II removes fucosylated selectin ligands and rolling.
• IL-1, tumour necrosis factor and IL-6 produce fever, the acute phase response, hepcidin-driven anaemia and, in excess, shock.
• Arachidonic acid feeds cyclooxygenase and lipoxygenase routes; aspirin acetylates cyclooxygenase irreversibly.
• Resolution is an active programme with lipid mediator class switching to lipoxins, resolvins, protectins and maresins, plus neutrophil apoptosis and efferocytosis.
• Granulomas depend on tumour necrosis factor, which is why anti-TNF therapy reactivates latent tuberculosis and why screening beforehand is mandatory.

Sources

1. Serhan, C. N., & Levy, B. D. (2018). Resolvins in inflammation: Emergence of the pro-resolving superfamily of mediators. Journal of Clinical Investigation, 128(7), 2657-2669. pubmed.ncbi.nlm.nih.gov
2. Panigrahy, D., Gilligan, M. M., Serhan, C. N., & Kashfi, K. (2021). Resolution of inflammation: An organizing principle in biology and medicine. Pharmacology and Therapeutics, 227, 107879. pubmed.ncbi.nlm.nih.gov
3. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Induced innate responses to infection. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
4. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). Inflammation and fever. In Microbiology (Section 17.5). OpenStax. openstax.org

The killing that only worked in the right mice

In 1974 Rolf Zinkernagel and Peter Doherty, working in Canberra, infected mice with lymphocytic choriomeningitis virus and took spleen cells from them a week later. They then asked whether those cells could kill virus-infected target cells in culture. The answer was yes, provided the targets came from a mouse of the same strain. Cells from a different strain, infected with exactly the same virus, were not killed.

Sit with how strange that is. The T cells had been primed against a virus. The target cells were displaying that virus. And the killing failed because the two mice had different H-2 types, that is, different major histocompatibility genes, which were known at the time as the genes that determine whether a skin graft is rejected.

Nobody had predicted this. The natural model was that a T cell receptor recognises a viral antigen, as an antibody does. If so, the source of the target cell should have been irrelevant. Instead the receptor was somehow reading virus and host genotype at the same time. The proposals to explain it were unsatisfying: either T cells carry two receptors, one for virus and one for self MHC, or they carry one receptor for a composite of the two, the altered-self model. There was no way to choose between them without a structure.

Key idea: MHC restriction was surprising because it made the T cell receptor's target a property of the presenting cell as well as of the pathogen, which no antibody-based intuition predicts. Zinkernagel and Doherty shared the 1996 Nobel Prize for it.

The structure that explained it

The answer came in 1987 when Pamela Bjorkman, Mark Saper, Don Wiley and colleagues solved the structure of the human class I molecule HLA-A2. The alpha1 and alpha2 domains form a platform of eight beta strands topped by two long alpha helices, and between those helices lies a groove.

The groove was not empty. It contained continuous electron density that did not belong to the protein and that could not be assigned to any single molecule, because the crystallised molecule had been purified from cells and carried a mixture of different peptides in the same site. That unassigned density was the antigen. One receptor, one composite surface made of MHC helices and a bound peptide: the altered-self model was correct, and the two-receptor model was dead.

Two pathways, two audiences

The class I pathway runs as follows. Cytosolic proteins, including misfolded and prematurely terminated products, are ubiquitylated and degraded by the proteasome. Interferon gamma induces alternative catalytic subunits that form the immunoproteasome, which preferentially cuts after hydrophobic and basic residues, generating C-termini that suit MHC class I anchors. Peptides are pumped into the endoplasmic reticulum by the transporter associated with antigen processing, TAP1 and TAP2. There a peptide-loading complex containing tapasin, calreticulin and ERp57 holds the empty class I molecule and edits the bound peptide toward higher affinity. A loaded molecule leaves for the surface; an empty one is unstable and does not.

TAP deficiency, bare lymphocyte syndrome type I, therefore produces very low surface class I, few CD8 T cells, and a distinctive clinical picture of chronic bacterial respiratory infection with granulomatous skin lesions rather than the overwhelming viral disease one might expect. Natural killer cells partly compensate, and this asymmetry between the predicted and actual phenotype is worth noticing.

The class II pathway has a problem to solve that class I does not. Class II molecules are assembled in the endoplasmic reticulum, which is full of peptides destined for class I, and they must not be loaded there. The solution is the invariant chain, which occupies the groove and directs the complex to the endosomal route. Proteolysis trims the invariant chain down to a residual fragment called CLIP that still sits in the groove, and HLA-DM then catalyses release of CLIP and its exchange for peptides derived from endocytosed antigen. HLA-DM also edits, preferentially retaining high-affinity peptides.

Class II expression is controlled by a single master transcriptional coactivator, CIITA, together with the RFX complex. Mutations in these cause bare lymphocyte syndrome type II, in which class II is absent, CD4 T cells fail to be selected in the thymus and fail to be activated in the periphery, and the child presents with severe combined immunodeficiency. One transcription factor, an entire arm of adaptive immunity.

What matters here: The two pathways answer different questions. Class I asks what is being made inside this cell. Class II asks what this cell has eaten. That is why one is on every nucleated cell and the other only on cells whose job is to sample the environment.

Cross-presentation, the necessary exception

The tidy scheme has a hole in it. Many viruses do not infect dendritic cells, and tumours are not dendritic cells at all, yet CD8 responses are raised against both. Priming a naive CD8 T cell requires a professional antigen-presenting cell with costimulation, so exogenous antigen must somehow reach class I.

It does, by cross-presentation, carried out chiefly by the conventional type 1 dendritic cell subset, marked by XCR1 and CLEC9A. These cells take up material from dying cells and route peptides into the class I pathway, either by escape from the phagosome into the cytosol or by loading within the phagosome itself. Without this route, cytotoxic immunity to most viruses and to tumours would be impossible, and it is the reason vaccine and cancer immunotherapy design pays close attention to which dendritic cell subset takes up an antigen.

Why the MHC is the most polymorphic region in the genome

Two features multiply together.

Polygeny. Every person has three classical class I genes, HLA-A, HLA-B and HLA-C, and three classical class II loci, HLA-DR, HLA-DQ and HLA-DP.

Polymorphism. Each of those genes has thousands of known alleles at the population level, and the variable positions are concentrated in the residues that line the peptide-binding groove and contact the T cell receptor. Expression is codominant, so a heterozygous person displays up to six different classical class I molecules, each with a different peptide preference, and a comparable number of class II combinations, including hybrids of alpha and beta chains from different chromosomes.

The evolutionary logic is straightforward. A pathogen can escape a single MHC allele by mutating the anchor residues of the peptides it presents. It cannot escape a population in which no two individuals present the same set. The polymorphism is a population-level defence, not an individual one, which is why it is maintained by balancing selection.

The cost of that strategy is transplantation, which is discussed later in the course, and a set of striking disease associations.

The abacavir example is the one to remember, because it changed practice. Screening for HLA-B*57:01 before prescribing abacavir essentially abolished a hypersensitivity syndrome that was previously seen in about 5 to 8 percent of treated patients, and it is now a required pharmacogenomic test. It is also mechanistically satisfying: a small molecule sitting in a peptide-binding groove alters the repertoire, and the immune system responds to what is effectively a new set of self peptides.

The core of it: A T cell never sees a pathogen. It sees a fragment of one, held by a host molecule whose sequence differs between people, which is why the same infection produces different responses and different diseases in different individuals.

Common misconceptions

• "T cells recognise antigens the way antibodies do." Antibodies bind intact conformational surfaces; T cell receptors bind short linear peptides only when those are held in an MHC groove, and the receptor contacts both.
• "MHC class II presents extracellular antigen and class I intracellular, without exception." Cross-presentation routes exogenous antigen into class I, and it is essential for CD8 responses to most viruses and tumours.
• "MHC polymorphism benefits the individual." Its advantage is at the population level; an individual gains something from heterozygosity, but the strategy is that no pathogen can escape everyone at once.
• "The groove holds one specific peptide." Each MHC molecule binds thousands of different peptides sharing anchor residues, which is why the 1987 structure showed unassignable density rather than a defined ligand.
• "Class I deficiency causes overwhelming viral infection." TAP deficiency presents mainly with chronic bacterial respiratory disease and granulomatous skin lesions, because natural killer cells partly compensate.

Summing up

• Zinkernagel and Doherty found in 1974 that virus-specific killing required target cells sharing the responder's H-2 type, which no antibody-based model predicted.
• The 1987 HLA-A2 structure showed a groove holding unassignable peptide density, settling the argument in favour of a single receptor reading a composite of MHC plus peptide.
• Class I loads 8 to 10 residue peptides from proteasomal degradation of cytosolic proteins, via TAP, in the endoplasmic reticulum, for CD8 T cells, on all nucleated cells.
• Class II loads longer peptides from endosomal proteolysis, after invariant chain removal and HLA-DM catalysed CLIP exchange, for CD4 T cells, on professional presenting cells.
• CIITA controls class II expression; its loss causes bare lymphocyte syndrome type II with absent CD4 T cells.
• Cross-presentation by conventional type 1 dendritic cells routes exogenous antigen into class I and is required for antiviral and antitumour CD8 priming.
• Polygeny plus extreme polymorphism concentrated in groove-lining residues makes the MHC a population-level defence, and produces the associations with ankylosing spondylitis, coeliac disease and abacavir hypersensitivity.

Sources

1. Zinkernagel, R. M., & Doherty, P. C. (1974). Restriction of in vitro T cell-mediated cytotoxicity in lymphocytic choriomeningitis within a syngeneic or semiallogeneic system. Nature, 248(5450), 701-702. pubmed.ncbi.nlm.nih.gov
2. Bjorkman, P. J., Saper, M. A., Samraoui, B., Bennett, W. S., Strominger, J. L., & Wiley, D. C. (1987). Structure of the human class I histocompatibility antigen, HLA-A2. Nature, 329(6139), 506-512. pubmed.ncbi.nlm.nih.gov
3. The Nobel Foundation. (1996). The Nobel Prize in Physiology or Medicine 1996: Peter C. Doherty and Rolf M. Zinkernagel. nobelprize.org
4. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The major histocompatibility complex and its functions. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
5. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The generation of T-cell receptor ligands. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
6. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). Major histocompatibility complexes and antigen-presenting cells. In Microbiology (Section 18.2). OpenStax. openstax.org

A molecule cut into three pieces

In the late 1950s Rodney Porter digested rabbit immunoglobulin G with papain, a protease from the papaya. A molecule of about 150,000 daltons fell into three fragments of roughly 50,000 daltons each. Two of them still bound antigen, and he named them Fab, for fragment antigen binding. The third would not bind antigen at all, but it crystallised readily, which is unusual for a serum protein and told him it was homogeneous. He named it Fc, for fragment crystallisable.

Gerald Edelman attacked the same molecule differently, reducing its disulfide bonds and separating the products by size. He found two chains: a heavy chain of about 50,000 daltons and a light chain of about 25,000. Put the two experiments together and the architecture falls out: two heavy chains and two light chains, held by disulfide bonds, with the antigen-binding sites at the tips of the arms and a stem that is the same in every antibody of a given class. The two men shared the 1972 Nobel Prize.

The point: Porter's Fc fragment crystallised because it is constant. That single observation contains the whole design principle: variable ends for recognition, a constant stem for the effector functions that recognition triggers.

Domains, loops and where variability sits

Both chains are built from a repeating unit, the immunoglobulin domain: about 110 residues folded into two beta sheets packed face to face and pinned by an intrachain disulfide. A light chain has one variable and one constant domain; a heavy chain has one variable and three or four constant domains, with a flexible hinge in IgG, IgA and IgD that lets the two arms move independently and reach epitopes at varying spacing.

Within each variable domain, sequence variability is not spread evenly. It is concentrated in three short loops per chain, the complementarity-determining regions, which sit at one end of the domain and together form the antigen-binding surface. The intervening framework regions are conserved and hold the loops in place. Six loops, three from the heavy chain and three from the light, make the paratope.

Of these, CDR3 of the heavy chain is by far the most variable, because it is formed at the junctions where gene segments are joined and nucleotides are added and removed. The next lesson takes that apart. For now, note that structure and genetics agree: the most variable loop sits at the centre of the binding surface.

Five isotypes, five jobs

Two of these entries hide something worth knowing.

Why IgG lasts three weeks. Most serum proteins are pinocytosed by endothelium and degraded. IgG is rescued: the neonatal Fc receptor, FcRn, binds the IgG Fc region at the acid pH of the endosome, diverts it away from the lysosome, and releases it back at neutral pH outside the cell. The same receptor transports maternal IgG across the placenta and, in rodents, across neonatal gut. It is also the reason therapeutic antibodies have long dosing intervals, why albumin has a similarly long half-life through the same receptor, and why blocking FcRn is now a drug strategy in antibody-mediated autoimmune disease: block the recycling and pathogenic IgG is cleared within days.

Why IgA does not inflame. Dimeric IgA made by plasma cells in the lamina propria binds the polymeric immunoglobulin receptor on the basolateral surface of the epithelium, is transcytosed, and is released at the apical surface with a fragment of the receptor still attached as secretory component, which protects it from proteolysis. Secretory IgA is a poor complement activator and does not engage the activating Fc receptors that IgG does. That is not a shortcoming; a mucosal surface carrying kilograms of commensal bacteria must be able to bind and exclude organisms without triggering inflammation every time it does so.

Fc receptors and the balance that governs them

Antibody does nothing on its own beyond neutralisation. Everything else happens because a cell binds its Fc region.

• Activating Fc gamma receptors (FcGRI, FcGRIIA, FcGRIIIA) signal through immunoreceptor tyrosine-based activation motifs, triggering phagocytosis, degranulation and, on natural killer cells, antibody-dependent cellular cytotoxicity.
• The inhibitory receptor FcGRIIB signals through an inhibition motif and raises the threshold for activation. On B cells it delivers negative feedback when antigen is already coated with IgG, which is a self-limiting mechanism for the antibody response. Polymorphisms reducing its function are associated with lupus.

The therapeutic consequence is that the effect of an antibody drug depends on the ratio of activating to inhibitory engagement, which depends on IgG subclass and on Fc glycosylation. Removing the core fucose from the Fc N-glycan increases FcGRIIIA binding substantially and is engineered into several therapeutic antibodies to enhance cell-mediated killing.

Building a B cell in the bone marrow

Development is a sequence of checkpoints, each testing whether the last step produced something usable.

1. Pro-B cell. Heavy chain D to J joining, then V to DJ joining. If the first allele fails to produce an in-frame product, the second is tried.
2. Pre-B cell and the pre-B cell receptor checkpoint. A successfully rearranged heavy chain pairs with a surrogate light chain made of VpreB and lambda5, together with Ig-alpha and Ig-beta. Signalling from this complex proves that a functional heavy chain exists, drives several rounds of proliferation, shuts off further heavy chain rearrangement, which enforces allelic exclusion, and licenses light chain rearrangement. Bruton tyrosine kinase, BTK, transmits this signal.
3. Immature B cell. Light chain rearrangement, kappa first and lambda if kappa fails on both alleles, produces surface IgM. The cell is now tested against self antigen present in the marrow.
4. Central tolerance. Strong binding to a multivalent self antigen does not usually kill the cell outright. It first triggers receptor editing: rearrangement resumes at the light chain locus and a new light chain replaces the old one, changing the specificity. Only if editing fails is the cell deleted. Weaker binding to soluble self antigen produces anergy, a functionally silenced cell that persists. Estimates from human studies suggest a substantial fraction of newly generated B cells are self-reactive and that these checkpoints remove most but not all of them.
5. Mature naive B cell. Alternative splicing of the heavy chain transcript produces both IgM and IgD from the same VDJ, and the cell leaves for the periphery.

Remember: Receptor editing, not deletion, is the principal mechanism of central B cell tolerance. A B cell that meets self antigen is given the chance to change its mind before it is killed.

The checkpoint made visible: X-linked agammaglobulinaemia

Boys with mutations in BTK cannot signal through the pre-B cell receptor, so development arrests at the pro-B to pre-B transition. The consequences are precise. Circulating B cells are almost absent, all immunoglobulin classes are very low or undetectable, and lymphoid tissue that depends on B cells fails to develop, so tonsils are small or absent and lymph nodes are not palpable, which is a useful examination finding.

The timing is the elegant part. These infants are well for the first four to six months of life, because maternal IgG transferred across the placenta by FcRn protects them and decays with a half-life of about three weeks. As it falls below a protective level, recurrent pyogenic infections begin: Streptococcus pneumoniae, Haemophilus influenzae, otitis, sinusitis, pneumonia. They are also uniquely vulnerable to enteroviruses, which antibody normally neutralises, and chronic enteroviral meningoencephalitis was a characteristic cause of death before immunoglobulin replacement. Treatment is lifelong immunoglobulin replacement, which is antibody without B cells.

Notice that this single disorder demonstrates the half-life of IgG, the role of maternal transfer, the pre-B checkpoint, and the specific pathogens against which antibody is non-redundant.

Making antibodies to order

Georges Kohler and Cesar Milstein reported in 1975 that fusing an antibody-secreting B cell with a myeloma line produced a hybrid that both secreted a single antibody and grew indefinitely. Clone the hybrids and each clone makes one antibody of one specificity, in unlimited quantity. They shared the 1984 Nobel Prize with Niels Jerne.

Every monoclonal antibody drug descends from that method, though modern ones are humanised or fully human to avoid the anti-mouse response that limited the first generation. The naming convention still records the history: the stem for a chimeric antibody is different from that for a humanised or a fully human one.

Worth holding on to: Kohler and Milstein solved a problem of supply, not of recognition. Nature already made antibodies of any specificity; what nobody could do was get a large amount of exactly one of them.

Common misconceptions

• "IgG is the best complement activator." IgM is far better, because a single bound pentamer presents the clustered Fc regions C1q needs, whereas IgG requires two molecules to land close together.
• "IgA is a weak antibody because it does not fix complement well." That is the design. A mucosal surface must exclude organisms without inflaming, and secretory IgA is built to do exactly that.
• "Self-reactive B cells are deleted." Receptor editing is the first and principal response; deletion follows only if editing fails, and anergy handles weaker self-reactivity.
• "X-linked agammaglobulinaemia presents at birth." Maternal IgG protects for four to six months, and symptoms begin as it decays with a three-week half-life.
• "Antibody half-life is determined by proteolysis alone." FcRn-mediated recycling gives IgG its 21-day half-life, which is why blocking FcRn is a therapeutic strategy in antibody-mediated disease.

The takeaway

• Papain digestion gave two Fab fragments and one crystallisable Fc; reduction gave heavy and light chains, and together they define the four-chain structure.
• Six complementarity-determining loops on a conserved framework form the binding site, with heavy chain CDR3 the most variable because it is junctional.
• IgM fixes complement best and is made first; IgG opsonises, crosses the placenta and lasts three weeks through FcRn recycling; dimeric secretory IgA excludes organisms without inflaming; IgE arms mast cells.
• Activating Fc gamma receptors signal through activation motifs and FcGRIIB inhibits, so the activating to inhibitory ratio, tuned by subclass and glycosylation, sets the outcome.
• Development runs pro-B, pre-B receptor checkpoint with surrogate light chain and BTK signalling, immature B, then central tolerance dominated by receptor editing.
• BTK mutation arrests development at the pre-B checkpoint, giving absent B cells, absent tonsils, pyogenic and enteroviral infection beginning at four to six months as maternal IgG decays.
• Hybridoma technology solved the supply problem for antibodies of defined specificity and underlies every monoclonal antibody drug.

Sources

1. The Nobel Foundation. (1972). The Nobel Prize in Physiology or Medicine 1972: Gerald M. Edelman and Rodney R. Porter. nobelprize.org
2. Kohler, G., & Milstein, C. (1975). Continuous cultures of fused cells secreting antibody of predefined specificity. Nature, 256(5517), 495-497. pubmed.ncbi.nlm.nih.gov
3. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The distribution and functions of immunoglobulin isotypes. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
4. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Structural variation in immunoglobulin constant regions. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
5. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). B lymphocytes and humoral immunity. In Microbiology (Section 18.4). OpenStax. openstax.org

Two bands that became one

In 1976 Nobumichi Hozumi and Susumu Tonegawa took DNA from a mouse embryo and DNA from an antibody-secreting myeloma, cut both with the same restriction enzyme, separated the fragments by size, and hybridised them with radiolabelled probes for the variable and the constant region of a light chain.

In the embryonic DNA the two probes hybridised to different fragments: the V and C sequences were far apart in the genome. In the myeloma DNA they hybridised to the same fragment. The DNA between them had been physically deleted, in a somatic cell, during the life of the animal.

That result violated a principle everyone held. Somatic cells were assumed to carry the same genome as the germ line; that is what made genomes stable. Tonegawa had shown that lymphocytes cut and rearrange their own DNA, and that they do so at a defined locus for a defined purpose. He received the 1987 Nobel Prize alone, which is unusual and reflects how decisive the finding was.

Key idea: Antibody diversity is not encoded. It is assembled, from a modest set of parts, in every developing lymphocyte, by a controlled DNA rearrangement.

The problem the rearrangement solves

The human genome contains on the order of 20,000 protein-coding genes. A human makes an antibody repertoire whose diversity is estimated in excess of 10 to the eleventh distinct specificities. William Dreyer and Claude Bennett had proposed in 1965 that one polypeptide might be encoded by two separate genes brought together somatically, precisely to escape this arithmetic, and had been widely disbelieved. Tonegawa proved them right.

The parts list

Antigen receptor loci contain arrays of gene segments rather than complete genes. The human immunoglobulin heavy chain locus carries roughly 40 functional variable segments, about 23 diversity segments and 6 joining segments, upstream of the constant region genes. The kappa light chain locus has about 35 to 40 functional variable segments and 5 joining segments; lambda has roughly 30 variable and 4 to 5 joining segments. Published counts differ between references because pseudogenes and segments of uncertain function are classified differently, so treat these as approximations rather than exact figures.

A heavy chain variable region is assembled by joining one D to one J, and then one V to that DJ. A light chain, having no D segments, joins one V directly to one J.

The combinatorial arithmetic is worth doing.

• Heavy chain: about 40 times 23 times 6, roughly 5,500 different V regions.
• Light chain: about 40 times 5 for kappa, plus about 30 times 4 for lambda, roughly 320 in total.
• Random pairing of one heavy with one light chain: about 5,500 times 320, on the order of 2 million.

Two million is a large number and it is nowhere near 10 to the eleventh. The rest comes from what happens at the joins.

Junctional diversity, which is where most of the repertoire lives

The joining reaction is deliberately imprecise, and three separate processes contribute.

1. Exonucleolytic trimming. A variable number of nucleotides is chewed back from each coding end before joining.
2. P nucleotides. The coding end is sealed as a hairpin during the reaction. When Artemis opens that hairpin it usually nicks off centre, leaving a short single-stranded overhang whose sequence is the palindrome of the adjacent coding sequence. Filling it in adds a few palindromic nucleotides.
3. N nucleotides. Terminal deoxynucleotidyl transferase, a template-independent polymerase expressed in developing lymphocytes, adds nucleotides at random to the exposed ends. Up to about twenty may be added at a heavy chain junction.

All three act at the junctions, which is precisely where the third complementarity-determining loop is encoded. That is why CDR3 of the heavy chain is the most variable part of an antibody and sits at the centre of the binding site: structure and mechanism agree because the mechanism produced the structure.

Including junctional diversity, the potential repertoire is usually estimated well above 10 to the eleventh, which exceeds the roughly 10 to the twelfth lymphocytes a human body contains. The repertoire is therefore limited by cell number, not by genetics.

In short: Combination gives about two million; the imprecision of joining supplies the rest, and it does so exactly at the loop that contacts antigen most directly.

There is a price. Trimming and addition change the reading frame at random, so roughly two-thirds of rearrangements are out of frame and produce nothing usable. This is why development is built as a series of checkpoints: most attempts fail, so each must be tested before the cell invests further.

The machinery

David Schatz, Marjorie Oettinger and David Baltimore identified the recombination activating genes by transfecting genomic DNA fragments into cells carrying an artificial recombination substrate and selecting for cells that had performed the reaction. RAG-1 was reported in 1989 and RAG-2 in 1990, adjacent in the genome and required together.

Their targets are recombination signal sequences flanking each gene segment: a conserved heptamer, CACAGTG, then a spacer of either 12 or 23 base pairs, then a conserved nonamer. The spacer length matters. RAG will only join a segment flanked by a 12 base pair spacer to one flanked by a 23 base pair spacer. This is the 12/23 rule, and it is what enforces correct segment order: in the heavy chain locus the spacer lengths are arranged so that V can join to D and D to J, but V cannot join directly to J.

The chemistry proceeds in defined steps. RAG-1 with RAG-2 binds a signal sequence pair, nicks one strand precisely at the heptamer boundary, and then uses the freed 3 prime hydroxyl to attack the other strand in a transesterification. That single reaction produces a blunt signal end and a covalently sealed hairpin on the coding end. The hairpin is opened by Artemis in complex with DNA-PKcs, and the ends are then joined by the classical non-homologous end joining machinery, Ku70 and Ku80, XRCC4, DNA ligase IV and XLF, the same enzymes covered in the molecular biology course in this catalogue.

The signal joint, between the two signal ends, is precise. The coding joint is imprecise, because of the trimming and additions above. One reaction, two joints, opposite requirements, and both are exactly what the system needs.

RAG is a domesticated transposase. Its catalytic mechanism is that of a DDE transposase, the signal sequences resemble transposon terminal inverted repeats, and purified RAG can perform transposition in vitro. The RAG1 and RAG2 core genes are related to the Transib family of transposons, which were evidently captured in an ancestral jawed vertebrate. Adaptive immunity as we know it appears to date from a transposon insertion into an antigen receptor gene.

What breaks, and how it presents

Omenn syndrome deserves a sentence, because it inverts expectation. A child with an incomplete recombination defect has fewer T cells than normal, yet presents with an aggressive inflammatory disease rather than only with infection. The reason is that the few clones generated escape normal selection and expand enormously, producing an autologous graft-versus-host-like picture. Partial loss of a tolerance-relevant mechanism can be more dangerous than complete loss.

The radiosensitivity entries have a practical consequence: these children must not receive conditioning regimens designed for other forms of severe combined immunodeficiency without dose adjustment, because the same repair defect that blocks recombination also blocks repair of therapy-induced damage.

The same machinery in T cells, with two differences

T cell receptor loci are rearranged by exactly the same RAG-dependent reaction. The beta and delta loci contain D segments; the alpha and gamma loci do not, so they join V directly to J.

Two differences matter.

First, the alpha locus continues to rearrange for as long as the cell fails selection, and because the J segments are numerous, successive attempts on the same allele are possible. Allelic exclusion at the alpha locus is therefore incomplete, and a small percentage of human T cells express two different alpha chains.

Second, and more importantly, T cell receptors do not undergo somatic hypermutation. This is not an accident of biochemistry; it is a requirement. A T cell receptor is selected in the thymus for a defined, weak affinity for self MHC and against strong affinity for self peptide. Mutating that receptor afterwards would produce cells of unknown self-reactivity that had never been tested. Antibodies can be mutated freely because they are selected against foreign antigen in a germinal centre, and because a B cell receiving mutation is still dependent on T cell help. The presence of hypermutation in one lineage and its absence in the other follows directly from how each is selected.

The upshot: The same enzyme builds both receptors, and the differences between the lineages are downstream consequences of how each is selected, not of different machinery.

Common misconceptions

• "Antibody diversity comes mostly from having many V genes." Combination of segments gives about two million; junctional imprecision supplies the great majority of the rest.
• "V(D)J recombination is an accurate process." The signal joint is precise and the coding joint is deliberately imprecise. Two-thirds of rearrangements are out of frame.
• "RAG cuts both strands at once." It nicks one strand, then uses the freed hydroxyl in a transesterification, generating a blunt signal end and a hairpin coding end that Artemis must open.
• "A partial recombination defect is a mild version of a complete one." Hypomorphic RAG mutations cause Omenn syndrome, an aggressive oligoclonal inflammatory disease, not a mild immunodeficiency.
• "T cell receptors improve their affinity during a response as antibodies do." They do not hypermutate, because they were selected on self MHC and mutation would produce untested self-reactivity.

Pulling it together

• Hozumi and Tonegawa showed in 1976 that V and C sequences lie on different restriction fragments in embryonic DNA and the same fragment in an antibody-producing cell, proving somatic rearrangement.
• Loci carry arrays of V, D and J segments; combinatorial joining and heavy-light pairing give on the order of two million specificities.
• Trimming, P nucleotides from off-centre hairpin opening, and template-independent N nucleotide addition by TdT create junctional diversity concentrated in CDR3.
• RAG-1 and RAG-2 cut at recombination signal sequences under the 12/23 rule, generating a precise signal joint and a hairpin coding end opened by Artemis and joined by non-homologous end joining.
• RAG is a domesticated Transib-family transposase, and can transpose in vitro.
• Null RAG gives T minus B minus NK plus severe combined immunodeficiency; hypomorphic RAG gives Omenn syndrome; Artemis and ligase IV defects add radiosensitivity.
• T cell receptors use the same machinery but do not hypermutate, because thymic selection has already fixed their self-reactivity.

Sources

1. Hozumi, N., & Tonegawa, S. (1976). Evidence for somatic rearrangement of immunoglobulin genes coding for variable and constant regions. PNAS, 73(10), 3628-3632. pubmed.ncbi.nlm.nih.gov
2. Schatz, D. G., Oettinger, M. A., & Baltimore, D. (1989). The V(D)J recombination activating gene, RAG-1. Cell, 59(6), 1035-1048. pubmed.ncbi.nlm.nih.gov
3. Oettinger, M. A., Schatz, D. G., Gorka, C., & Baltimore, D. (1990). RAG-1 and RAG-2, adjacent genes that synergistically activate V(D)J recombination. Science, 248(4962), 1517-1523. pubmed.ncbi.nlm.nih.gov
4. The Nobel Foundation. (1987). The Nobel Prize in Physiology or Medicine 1987: Susumu Tonegawa. nobelprize.org
5. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The generation of diversity in immunoglobulins. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
6. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). T-cell receptor gene rearrangement. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov

A child with too much of one antibody and none of the others

A six-year-old has had four episodes of pneumonia and chronic sinusitis. His tonsils and cervical nodes are strikingly large. Immunoglobulins show IgM well above the normal range, with IgG, IgA and IgE undetectable. He has had no Pneumocystis pneumonia, no cryptosporidial diarrhoea and no unusual viral disease.

The pattern of high IgM with absent switched isotypes says that B cells are being activated and making antibody, but cannot change the constant region of what they make. The enlarged lymphoid tissue says the germinal centres are not merely present but hypertrophic. And the absence of opportunistic infection distinguishes this from the commoner X-linked form of hyper-IgM syndrome, in which the same antibody phenotype is accompanied by defective T cell help to macrophages.

The gene responsible in this child is AICDA. In 2000 Masamichi Muramatsu, Kazuo Kinoshita, Tasuku Honjo and colleagues showed that mice lacking the enzyme it encodes, activation-induced cytidine deaminase, could not class switch and could not somatically hypermutate, and that the same enzyme is mutated in a human autosomal recessive hyper-IgM syndrome. One enzyme, two processes that had been assumed to be mechanistically unrelated.

What matters here: Somatic hypermutation and class switch recombination look like completely different operations, one changing the variable region point by point and the other swapping a whole constant region. They begin with the same chemical reaction on DNA.

The structure that makes selection possible

A germinal centre forms in a B cell follicle within days of antigen encounter and organises into two zones.

• The dark zone is packed with rapidly dividing centroblasts, which express CXCR4 and are held there by CXCL12. Division is fast, with cycle times of six to twelve hours, and it is here that somatic hypermutation occurs.
• The light zone contains centrocytes, follicular dendritic cells and follicular helper T cells, organised by CXCR5 and CXCL13. Follicular dendritic cells are not haematopoietic and do not present peptide; they are stromal cells that hold intact antigen on their surface as immune complexes, for long periods, as a display for B cells to test themselves against.

Cells cycle between the zones. A centroblast mutates and divides in the dark zone, moves to the light zone as a centrocyte, is tested, and either dies, returns to the dark zone for another round, or leaves as a memory cell or a plasma cell. The cyclic re-entry model was confirmed by intravital imaging, and it explains how iterative improvement is possible within a single anatomical structure.

How affinity is converted into a survival signal

This is the mechanism worth understanding properly, because it is not obvious how a cell could measure the affinity of its own receptor.

Antigen on follicular dendritic cells is limiting. A centrocyte with a higher-affinity receptor extracts more antigen from that display, internalises it, degrades it and presents more peptide on MHC class II. Follicular helper T cells in the light zone are also limiting, and the help they deliver, through CD40 ligand and cytokines including IL-21, is proportional to the density of peptide-MHC they see. So a receptor property, affinity, is transduced into a quantity of T cell help, and help determines survival and the licence to re-enter the dark zone.

The consequences are worth spelling out. Selection is competitive rather than absolute, so what is selected is the best available clone, not a clone above some threshold. Because T cell help is the currency, class switching and affinity maturation are both dependent on T cells, which is why they largely fail for antigens that cannot recruit T cell help, such as bacterial polysaccharides. That is precisely why plain polysaccharide vaccines work poorly in children under two, and why conjugating the polysaccharide to a carrier protein, which supplies T cell epitopes, transformed Haemophilus influenzae type b and pneumococcal vaccination.

The core of it: A B cell cannot measure its own affinity. It converts affinity into the amount of antigen it can capture and present, and lets a T cell do the measuring.

One enzyme, two outcomes

AID is a single-stranded DNA cytidine deaminase. It converts cytosine to uracil, and it can only act on single-stranded DNA, which in practice means DNA that is being transcribed. Transcription of the target region is therefore a prerequisite, and it is also how targeting is controlled.

In the variable region, the resulting U:G mismatch is handled in three ways, and the mixture is what produces the observed mutation spectrum.

1. Replication over the uracil without repair gives a C to T transition on one strand and G to A on the other.
2. Uracil-DNA glycosylase removes the uracil, leaving an abasic site; translesion polymerases such as REV1 insert any base opposite it, producing transitions and transversions at C:G pairs.
3. The mismatch repair proteins MSH2 and MSH6 recognise the U:G mismatch and recruit the error-prone polymerase eta, which introduces mutations at nearby A:T pairs. This is why mutations appear at A and T positions where no cytosine was deaminated.

The rate is about 10 to the minus 3 per base per division, roughly a million times the ordinary somatic mutation rate, and it is confined to a window of one to two kilobases around the rearranged V region. Most mutations are neutral or deleterious; selection in the light zone keeps the rare improvements.

In switch regions, the same deamination has a different consequence because of where it happens. Upstream of each constant region gene, apart from delta, lies a long repetitive G-rich switch region. Cytokines induce germline transcription through the switch region of the isotype to be used, and because the transcript is G-rich it forms a stable R-loop with the template strand, leaving the non-template strand single-stranded and available to AID. Dense deamination on both strands, followed by uracil-DNA glycosylase and APE1, converts clustered uracils into a double-strand break. Two such breaks, one in the mu switch region and one in the target switch region, are joined by non-homologous end joining, and the intervening DNA is excised as a circle.

The rearranged VDJ exon is untouched. The cell now makes an antibody with exactly the same specificity and a different constant region, which is the entire point: specificity is preserved and effector function is changed.

Which isotype is selected is set by the cytokine environment, that is, by the helper T cell.

The hyper-IgM syndromes, read as a set

Comparing these disorders is the most efficient way to learn what each component does.

• CD40 ligand deficiency (X-linked, the commonest form). The B cell never receives the switching signal, so IgM is normal or high and switched isotypes are absent. But CD40 ligand also mediates T cell help to macrophages and dendritic cells, so these patients additionally have a T cell functional defect: Pneumocystis jirovecii pneumonia, Cryptosporidium infection with sclerosing cholangitis, and neutropenia. Germinal centres are absent or rudimentary.
• CD40 deficiency (autosomal recessive) gives the same phenotype from the receptor side.
• AID deficiency (autosomal recessive). The defect is intrinsic to the B cell and downstream of CD40 signalling, so T cell help to macrophages is intact. These patients have recurrent pyogenic sinopulmonary infection and lymphoid hyperplasia with large germinal centres, but not the opportunistic infections of the CD40 ligand group. Both class switching and somatic hypermutation are absent.
• Uracil-DNA glycosylase deficiency gives a similar but milder picture, with switching severely impaired and hypermutation skewed toward transitions at C:G, exactly as the biochemistry predicts.

Comparing the first and third entries is the diagnostic key. Identical immunoglobulin profiles, opposite germinal centre findings, and opportunistic infections in one group only.

The output, and its cost

Germinal centres produce two long-lived products. Memory B cells circulate and respond rapidly on re-exposure, switching and mutating further. Long-lived plasma cells home to survival niches in bone marrow and secrete antibody continuously for years to decades without further antigen; the durability of antibody titres after some vaccinations and infections rests on them.

There is also a phenomenon worth naming honestly, because it constrains vaccine design. When a person is re-exposed to a variant of a previously encountered antigen, memory cells specific for the original tend to dominate the response over naive cells specific for the new epitopes. This was called original antigenic sin and is now more often called immune imprinting. It is neither purely harmful nor purely helpful; it produces rapid cross-reactive antibody, and it can limit responses to genuinely new epitopes, which is a live issue for influenza and coronavirus vaccine strategy.

The cost of the whole system is that a cell has been licensed to mutate its own DNA and to make double-strand breaks in it. AID has off-target activity at other transcribed loci, and the resulting breaks can be joined to the immunoglobulin locus. The MYC to IGH translocation of Burkitt lymphoma places MYC under the control of immunoglobulin enhancers, and the translocations of follicular lymphoma involving BCL2 arise in the same way. Germinal centre-derived lymphomas are the price of affinity maturation.

Bottom line: Adaptive immunity buys specificity by breaking two rules that every other cell obeys: do not cut your genome, and do not mutate it.

Common misconceptions

• "Follicular dendritic cells present antigen to B cells on MHC." They are stromal, not haematopoietic, and they display intact antigen as immune complexes. They do not process or present peptide.
• "Class switching changes the antibody's specificity." The rearranged VDJ exon is untouched. Only the constant region changes, so specificity is preserved and effector function is altered.
• "Somatic hypermutation and class switching are separate systems." Both are initiated by AID deaminating cytosine in transcribed single-stranded DNA; the different outcomes come from where it acts and how the uracil is processed.
• "All hyper-IgM syndromes look alike." CD40 ligand deficiency adds opportunistic infection and absent germinal centres; AID deficiency has neither and shows lymphoid hyperplasia.
• "Polysaccharide antigens can be affinity matured." Affinity maturation and switching depend on T cell help, which polysaccharides cannot recruit, which is why conjugate vaccines exist.

Looking back

• The germinal centre has a dark zone where mutation and proliferation occur and a light zone where selection occurs, with cells cycling between them.
• Affinity is converted into survival because a higher-affinity B cell captures more antigen, presents more peptide, and receives more help from a limiting number of follicular helper T cells.
• AID deaminates cytosine in transcribed single-stranded DNA; in V regions the resulting uracil is processed into point mutations, and in switch regions into double-strand breaks.
• Class switching preserves the VDJ exon and changes the constant region, with the isotype chosen by cytokines and CD40 ligand engagement required throughout.
• CD40 ligand deficiency gives hyper-IgM plus opportunistic infection and absent germinal centres; AID deficiency gives hyper-IgM with lymphoid hyperplasia and no opportunistic infection.
• Memory B cells and bone marrow long-lived plasma cells provide durable protection, with immune imprinting shaping responses to antigenic variants.
• AID off-target activity generates the translocations of Burkitt and follicular lymphoma, which is the cost of licensed mutation.

Sources

1. Muramatsu, M., Kinoshita, K., Fagarasan, S., Yamada, S., Shinkai, Y., & Honjo, T. (2000). Class switch recombination and hypermutation require activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell, 102(5), 553-563. pubmed.ncbi.nlm.nih.gov
2. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). B-cell activation by armed helper T cells. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
3. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Immunological memory. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
4. Tangye, S. G., Al-Herz, W., Bousfiha, A., Cunningham-Rundles, C., Franco, J. L., Holland, S. M., et al. (2022). Human inborn errors of immunity: 2022 update on the classification from the International Union of Immunological Societies Expert Committee. Journal of Clinical Immunology, 42(7), 1473-1507. pubmed.ncbi.nlm.nih.gov
5. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). B lymphocytes and humoral immunity. In Microbiology (Section 18.4). OpenStax. openstax.org

The same thymocyte, deleted in males and kept in females

In 1988 Pawel Kisielow, Horst Bluthmann, Uwe Staerz, Michael Steinmetz and Harald von Boehmer made a mouse carrying a transgenic T cell receptor specific for the male-specific H-Y antigen presented on a particular MHC class I molecule. Because the receptor was transgenic, essentially every thymocyte carried it, which meant the experiment could ask a question that a normal polyclonal repertoire cannot.

In female mice, the antigen is absent. Thymocytes bearing this receptor developed normally and mature CD8 cells appeared. In male littermates, carrying the identical transgene and differing only in expressing the antigen, the immature CD4 and CD8 double-positive thymocytes were deleted. Same receptor, same genetic background, and the presence or absence of the antigen decided whether the cell lived.

That is clonal deletion demonstrated directly, and it settled a debate that had run since Burnet proposed clonal selection: self-reactive lymphocytes are not merely inactivated somewhere in the periphery, they are physically removed during development.

Key idea: The thymus is not a factory that makes T cells. It is a filter that makes far more than it needs and destroys almost all of them.

The geography, which does the work

Precursors arrive from bone marrow at the cortico-medullary junction and move outward through the cortex, then back into the medulla. Two epithelial populations do most of the selecting, and they are specialised in ways that matter.

• Cortical thymic epithelial cells present the peptides on which positive selection occurs. They are unique in expressing beta5t, a catalytic subunit that forms the thymoproteasome and generates a peptide repertoire distinct from that of any other cell, and they use cathepsin L and thymus-specific serine protease for class II processing. Positive selection therefore occurs on a set of peptides that a mature cell will never see again, which is one proposed reason why the selecting interaction can be weak without being useless.
• Medullary thymic epithelial cells present the self antigens against which negative selection occurs, including tissue-restricted antigens they have no business expressing.

Stages and checkpoints

1. Double negative. Thymocytes lacking CD4 and CD8 progress through stages defined by CD44 and CD25. TCR beta, gamma and delta loci rearrange. A cell that productively rearranges gamma and delta first may commit to the gamma delta lineage.
2. Beta selection. A successful TCR beta chain pairs with an invariant pre-T alpha chain to form the pre-T cell receptor, exactly analogous to the pre-B cell receptor. Signalling proves that a functional beta chain exists, drives proliferation, enforces allelic exclusion at the beta locus, and permits alpha locus rearrangement.
3. Double positive. The cell expresses both CD4 and CD8 and rearranges TCR alpha. Because the alpha locus has many J segments and no D segments, repeated rearrangement on the same allele is possible, so a cell that fails selection can try again with a new alpha chain.
4. Positive selection, in the cortex. A receptor that binds self peptide-MHC weakly to moderately delivers a survival signal. A receptor that binds nothing gives no signal, and the cell dies by neglect within three to four days. Death by neglect is the fate of the great majority.
5. Lineage commitment. A thymocyte selected on MHC class I becomes CD8 single positive; one selected on class II becomes CD4 single positive. The coreceptor that continues to be engaged is the one that is retained.
6. Negative selection, in the medulla. A receptor that binds self peptide-MHC strongly triggers apoptosis, or in some cases diversion into the regulatory T cell lineage.

The whole process is a signal-strength window. Too little and the cell dies by neglect; too much and it is deleted; in between and it survives. More than 95 percent of thymocytes generated never leave the thymus. The wastage is enormous and it is the price of a repertoire that is both self-MHC restricted and self-tolerant.

Why this matters: Positive and negative selection are not two different mechanisms. They are the same recognition event with different affinities, and that is why a single quantitative parameter decides between two opposite fates.

Presenting what the thymus does not contain

There is an obvious problem with negative selection. Insulin is made in the pancreas, thyroglobulin in the thyroid, retinal S antigen in the eye. If a thymocyte can only be tested against what the thymus contains, no tolerance to those proteins can be established there.

Medullary thymic epithelial cells solve it by expressing them anyway. Individually and stochastically, small subsets of these cells transcribe tissue-restricted genes belonging to other organs, so that collectively the medullary epithelium presents a sample of the whole body's protein repertoire. The autoimmune regulator, AIRE, drives a large part of this promiscuous expression, and Mark Anderson, Christophe Benoist, Diane Mathis and colleagues demonstrated in 2002 that Aire-deficient mice fail to express these antigens in the thymus and develop multi-organ autoimmunity.

The human disease is autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy, also called autoimmune polyendocrine syndrome type 1, caused by biallelic AIRE mutations. The classic triad is chronic mucocutaneous candidiasis, hypoparathyroidism and adrenal insufficiency, usually appearing in that order during childhood, with other endocrine and organ-specific autoimmunity accumulating over years.

Two details make it more interesting than a list. The candidiasis is not an infection risk from immunodeficiency in the usual sense; these patients make neutralising autoantibodies against IL-17 and IL-22, the cytokines required for mucosal antifungal defence. And most patients also carry high-titre neutralising autoantibodies against type I interferons, which was of no obvious clinical consequence until it emerged that such autoantibodies predispose to severe viral pneumonia. A tolerance defect produced by failure of thymic antigen expression turns out to cause disease by generating autoantibodies against the immune system's own mediators.

AIRE is not the whole story. Fezf2 controls a partly overlapping, AIRE-independent set of tissue-restricted antigens in the same cells.

The third fate: regulatory T cells

Some thymocytes with relatively high affinity for self peptide-MHC are not deleted but diverted into the regulatory lineage. Shimon Sakaguchi and colleagues showed in 1995 that removing CD4 cells expressing the IL-2 receptor alpha chain, CD25, from a normal mouse and transferring the remainder into a lymphopenic host caused multi-organ autoimmune disease, and that adding the CD25 positive fraction back prevented it. A small, definable subset was actively suppressing autoimmunity.

The lineage-defining transcription factor was identified in 2003 by Shohei Hori, Takashi Nomura and Sakaguchi as Foxp3. Retroviral expression of Foxp3 converted conventional CD4 cells into suppressive cells.

The human disease had in fact been described first. Immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome, IPEX, presents in male infants with severe enteropathy, type 1 diabetes usually in the first months of life, eczema and autoimmune cytopenias, and is often fatal without haematopoietic transplantation. Craig Bennett, Hans Ochs and colleagues showed in 2001 that it is caused by FOXP3 mutations, the same gene mutated in the scurfy mouse.

Set IPEX beside APECED and the contrast is instructive. Both are monogenic failures of self-tolerance. AIRE deficiency fails to present self antigen for deletion, and disease accumulates over years as individual organs are attacked. FOXP3 deficiency removes active suppression, and disease is fulminant from infancy. Deletion and suppression are separate layers, and losing them produces different tempos.

When there is no thymus

The thymus develops from the third pharyngeal pouch. In 22q11.2 deletion syndrome, which includes DiGeorge syndrome, that development is disturbed. Most patients have partial thymic hypoplasia with reduced but adequate T cell numbers; complete athymia occurs in a small minority and is a form of severe combined immunodeficiency treatable by thymus tissue transplantation, which allows the patient's own precursors to be educated in donor stroma. The associated features follow the same embryology: hypocalcaemia from parathyroid hypoplasia, conotruncal cardiac anomalies, and palatal abnormalities.

The thymus involutes from early life, with fatty replacement and falling output. Output can be measured, and this has become a public health tool. When TCR alpha rearrangement excises the intervening DNA, the excised fragment persists as a non-replicating circle, a T cell receptor excision circle. Because it is not copied when the cell divides, its abundance among circulating T cells reports recent thymic emigration. Newborn screening programmes measure TRECs in the dried blood spot, and a low count identifies infants with severe combined immunodeficiency before they present with infection, at a point when transplantation outcomes are far better. A piece of molecular waste from a recombination reaction became a population screening assay.

The point: Newborn TREC screening only works because V(D)J recombination leaves a byproduct that does not replicate. Mechanistic detail turns into clinical capability more often than it looks.

The unconventional lineages, briefly

Not every T cell is a conventional MHC-restricted alpha beta cell. Gamma delta T cells, abundant in epithelia, respond to phosphoantigens and stress ligands with little or no MHC restriction. Invariant natural killer T cells use a near-invariant receptor to recognise lipids presented by CD1d. Mucosal-associated invariant T cells recognise riboflavin biosynthesis intermediates presented by MR1, which is a striking case of a conserved receptor detecting a metabolic pathway that mammals lack and many microbes possess.

Common misconceptions

• "Most thymocytes mature and leave the thymus." More than 95 percent die there, most by neglect after failing to receive a positive selection signal.
• "Positive and negative selection are separate mechanisms." They are the same recognition event at different affinities, which is why one quantitative parameter decides between opposite fates.
• "Tolerance to organ-specific proteins cannot be established in the thymus." Medullary epithelial cells promiscuously express tissue-restricted antigens under AIRE and Fezf2 control.
• "DiGeorge syndrome means no T cells." Most patients have partial hypoplasia with adequate T cell numbers; complete athymia is uncommon and is a distinct, treatable emergency.
• "The candidiasis in APECED reflects general immunodeficiency." It is caused by neutralising autoantibodies against IL-17 and IL-22, which are required for mucosal antifungal defence.

What to remember

• Transgenic receptor experiments in 1988 showed the same thymocyte is deleted when its antigen is present and retained when it is absent, demonstrating clonal deletion directly.
• Cortical epithelium, with its distinctive thymoproteasome, mediates positive selection; medullary epithelium mediates negative selection.
• Development runs double negative, beta selection with the pre-T cell receptor, double positive, positive selection, lineage commitment, then negative selection.
• Selection is a signal-strength window: no signal gives death by neglect, strong signal gives deletion or diversion to the regulatory lineage, and more than 95 percent of thymocytes die.
• AIRE drives promiscuous tissue-restricted antigen expression; its loss causes APECED with candidiasis from anti-IL-17 autoantibodies, hypoparathyroidism and adrenal failure.
• Foxp3 defines regulatory T cells; its loss causes IPEX, which is fulminant from infancy, in contrast to the slower accumulation of organ-specific disease in APECED.
• T cell receptor excision circles do not replicate, so measuring them in the newborn blood spot detects severe combined immunodeficiency before infection occurs.

Sources

1. Kisielow, P., Bluthmann, H., Staerz, U. D., Steinmetz, M., & von Boehmer, H. (1988). Tolerance in T-cell-receptor transgenic mice involves deletion of nonmature CD4 plus 8 plus thymocytes. Nature, 333(6175), 742-746. pubmed.ncbi.nlm.nih.gov
2. Anderson, M. S., Venanzi, E. S., Klein, L., Chen, Z., Berzins, S. P., Turley, S. J., et al. (2002). Projection of an immunological self shadow within the thymus by the aire protein. Science, 298(5597), 1395-1401. pubmed.ncbi.nlm.nih.gov
3. Hori, S., Nomura, T., & Sakaguchi, S. (2003). Control of regulatory T cell development by the transcription factor Foxp3. Science, 299(5609), 1057-1061. pubmed.ncbi.nlm.nih.gov
4. Bennett, C. L., Christie, J., Ramsdell, F., Brunkow, M. E., Ferguson, P. J., Whitesell, L., et al. (2001). The immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome (IPEX) is caused by mutations of FOXP3. Nature Genetics, 27(1), 20-21. pubmed.ncbi.nlm.nih.gov
5. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Generation of lymphocytes in bone marrow and thymus. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
6. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Interaction with self antigens selects some lymphocytes for survival but eliminates others. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov

One bacterium, two diseases

Two patients are infected with Mycobacterium leprae. The first has three well-demarcated hypopigmented plaques with raised borders, each anaesthetic to light touch, and a thickened ulnar nerve. Skin biopsy shows organised granulomas with epithelioid macrophages and very few acid-fast bacilli; a slit-skin smear is negative. The second has diffuse, symmetrical, nodular thickening of the skin of the face and ears, no anaesthesia early on, and biopsy shows sheets of foamy macrophages packed with bacilli, so many that they are visible in almost every high-power field.

The organism is the same. The genome of the bacterium is the same. What differs is the host response. The first patient, at the tuberculoid pole, mounts a strong macrophage-activating cellular response that contains the organism at the cost of nerve damage from the inflammation itself. The second, at the lepromatous pole, mounts a strong antibody response that is useless against an intracellular organism, and the bacterium multiplies unchecked. Leprosy is the clearest clinical demonstration in medicine that choosing the wrong effector programme is worse than choosing a weak one.

In short: An immune response is not a dial that runs from weak to strong. It is a choice among effector programmes, and the wrong choice can be worse than a small right one.

How the subsets were found

In 1986 Tim Mosmann, Holly Cherwinski, Martha Bond, Marilyn Giedlin and Robert Coffman characterised a panel of mouse helper T cell clones by the profile of factors each secreted. The clones did not vary continuously. They fell into two groups: one making interferon gamma and IL-2, the other making what were then called B cell stimulatory factors, now IL-4 and IL-5. They named the groups TH1 and TH2.

The framework has been extended considerably since, and the modern account has more subsets and considerably more plasticity, but the core insight has held: a naive CD4 cell commits to a programme, and the programme determines what kind of immunity results.

Three signals

A naive CD4 T cell needs three inputs to become an effector.

1. Signal 1: the T cell receptor engaging peptide on MHC class II.
2. Signal 2: costimulation, CD28 on the T cell engaging B7-1 or B7-2 on the antigen-presenting cell. Signal 1 without signal 2 produces anergy, which is a peripheral tolerance mechanism and the reason innate recognition matters: B7 is upregulated by pattern recognition receptor signalling.
3. Signal 3: the cytokines present at the moment of priming, which are themselves determined by what the innate system detected. This is what selects the subset.

Note the elegance of the induction logic: interferon gamma reinforces TH1 and inhibits TH2, IL-4 reinforces TH2 and inhibits TH1, and both inhibit TH17. Each programme is self-amplifying and mutually inhibitory, which is what converts a graded cytokine input into a committed choice. That is also why polarised diseases such as leprosy sit at poles rather than in the middle.

What each subset does, and what happens without it

TH1 and macrophage activation. A macrophage that has phagocytosed Mycobacterium tuberculosis cannot kill it unaided, because the organism blocks phagosome-lysosome fusion. Interferon gamma from a TH1 cell, together with CD40 ligand engagement, activates the macrophage: it upregulates the NADPH oxidase and inducible nitric oxide synthase, overcomes the block on phagosome maturation, and increases MHC class II and B7. The granuloma of the previous module is the anatomical form of this interaction.

The human genetics is unusually clean. Mendelian susceptibility to mycobacterial disease is caused by biallelic or dominant-negative defects in the IL-12 and interferon gamma circuit: IL12B, IL12RB1, IFNGR1, IFNGR2, STAT1, and others. Affected people are strikingly susceptible to otherwise harmless mycobacteria, including BCG vaccine and environmental species, and to Salmonella, but are not generally immunodeficient. A single axis, a single class of pathogen.

TH2 and the expulsion programme. IL-5 recruits and activates eosinophils, IL-4 drives IgE and mast cell arming, and IL-13 acts on epithelium and smooth muscle to increase mucus production and gut motility. The combination is sometimes described as weep and sweep, and against a multicellular parasite too large to phagocytose it is a rational strategy. The same programme misdirected at pollen or peanut is allergy, which the hypersensitivity lesson takes up.

TH17 at barriers. IL-17 acts on epithelial and stromal cells to induce G-CSF, CXCL8 and antimicrobial peptides, recruiting neutrophils to mucosal surfaces; IL-22 promotes epithelial repair and defensin production. The corresponding deficiencies are highly specific: loss-of-function STAT3 mutations cause autosomal dominant hyper-IgE syndrome, historically Job syndrome, with recurrent staphylococcal skin abscesses that are strikingly non-inflamed, pneumatoceles after pneumonia, chronic mucocutaneous candidiasis, retained primary teeth and skeletal fragility. Defects in IL-17F, IL-17RA and CARD9 give chronic mucocutaneous candidiasis more narrowly. The pattern to carry is that impaired IL-17 immunity means Candida and Staphylococcus at barrier surfaces.

The upshot: Each helper subset has a matching immunodeficiency, and each of those deficiencies is narrow. That specificity is the strongest evidence that the subsets are functional categories and not just descriptive ones.

Plasticity, honestly

The table above implies terminal, mutually exclusive lineages. That is an approximation. TH17 cells can acquire interferon gamma production and become TH1-like, particularly under IL-12 and IL-23; regulatory T cells can lose Foxp3 expression under inflammatory conditions; and cells co-expressing markers of two programmes are readily found in human tissue. The master transcription factors are better thought of as dominant influences on a network than as switches with two positions. Where this matters practically is in interpreting flow cytometry from human disease, where the neat categories of mouse in vitro polarisation are much less crisp.

Cytokine signalling, and the deficiency that maps onto it

Most of these cytokines signal through receptors coupled to Janus kinases, which phosphorylate STAT proteins that dimerise and enter the nucleus. Which STAT is used largely determines which programme is induced: STAT4 for TH1, STAT6 for TH2, STAT3 for TH17 and Tfh, STAT5 for Treg.

A family of receptors shares a common gamma chain: those for IL-2, IL-4, IL-7, IL-9, IL-15 and IL-21. Mutations in IL2RG, which encodes it, cause X-linked severe combined immunodeficiency, the commonest form. The phenotype is T minus, B plus, NK minus, and each element follows from a specific cytokine: loss of IL-7 signalling abolishes T cell development, loss of IL-15 abolishes natural killer cell development, and B cells are present in number but non-functional because they receive no T cell help. Mutations in JAK3, which associates with the same chain, produce an identical phenotype with autosomal recessive inheritance. Knowing the receptor sharing lets you predict the cell counts.

Blocking cytokines in practice, including one instructive failure

The instructive case is IL-17 in the bowel. Because IL-17 is central to psoriasis, and because psoriasis and Crohn disease share genetic risk loci in the IL-23 pathway, blocking IL-17A in Crohn disease looked promising. The 2012 trial reported by Wolfgang Hueber and colleagues found the opposite: secukinumab was not effective and was associated with higher rates of adverse events, with some patients worsening. The explanation is that IL-17 and IL-22 maintain the intestinal epithelial barrier and its antimicrobial peptide production, so removing them in a disease of barrier failure is counterproductive, whereas blocking IL-23 upstream, which drives pathogenic effector functions without abolishing barrier-protective IL-17 entirely, does work.

Worth holding on to: A cytokine's role differs by tissue. Blocking IL-17 helps in skin and joint and harms in bowel, and that is a fact about anatomy, not about the drug.

One more clinical consequence. Cytokine release syndrome after chimeric antigen receptor T cell therapy is driven substantially by IL-6, largely from macrophages responding to activated T cells, and tocilizumab is the standard treatment. The mechanism-to-therapy chain runs from a subset biology observation straight to a supportive care protocol.

Common misconceptions

• "A stronger immune response is a better one." Lepromatous leprosy features a vigorous antibody response that is useless against an intracellular organism. Programme choice matters more than magnitude.
• "Helper subsets are terminal lineages." Plasticity is substantial, especially between TH17 and TH1, and human tissue routinely contains cells expressing features of more than one programme.
• "Signal 1 alone activates a T cell." Without CD28 costimulation the cell becomes anergic, which is a peripheral tolerance mechanism and the reason innate detection is required to license adaptive responses.
• "Blocking a cytokine that drives a disease in one organ will help wherever that disease appears." Anti-IL-17A helps psoriasis and worsens Crohn disease, because IL-17 maintains the intestinal barrier.
• "Interferon gamma pathway defects cause general immunodeficiency." They cause a narrow susceptibility to mycobacteria and Salmonella, which is why the syndrome is named for that specificity.

The short version

• The leprosy spectrum shows one organism producing opposite diseases according to which helper programme dominates.
• Mosmann and Coffman found in 1986 that helper clones fall into discrete cytokine profiles rather than a continuum.
• Three signals are needed: receptor engagement, CD28 costimulation, and cytokines that select the subset; signal 1 alone gives anergy.
• TH1 with T-bet and interferon gamma activates macrophages; TH2 with GATA3 drives eosinophils, IgE and mucus; TH17 with RORgamma-t recruits neutrophils and induces antimicrobial peptides; Tfh helps germinal centres; Treg suppresses.
• Each programme is self-amplifying and cross-inhibitory, which turns graded input into committed choice.
• IL-12 and interferon gamma pathway defects give mycobacterial and Salmonella susceptibility; STAT3 and IL-17 pathway defects give Candida and Staphylococcus at barriers.
• The common gamma chain is shared by the IL-2, 4, 7, 9, 15 and 21 receptors, which is why IL2RG mutation gives T minus, B plus, NK minus severe combined immunodeficiency.
• Blocking IL-17A helps psoriasis and worsens Crohn disease, because the same cytokine maintains the intestinal barrier.

Sources

1. Mosmann, T. R., Cherwinski, H., Bond, M. W., Giedlin, M. A., & Coffman, R. L. (1986). Two types of murine helper T cell clone, I: Definition according to profiles of lymphokine activities and secreted proteins. Journal of Immunology, 136(7), 2348-2357. pubmed.ncbi.nlm.nih.gov
2. Hueber, W., Sands, B. E., Lewitzky, S., Vandemeulebroecke, M., Reinisch, W., Higgins, P. D. R., et al. (2012). Secukinumab, a human anti-IL-17A monoclonal antibody, for moderate to severe Crohn's disease: Unexpected results of a randomised, double-blind, placebo-controlled trial. Gut, 61(12), 1693-1700. pubmed.ncbi.nlm.nih.gov
3. Casanova, J.-L., & Abel, L. (2002). Genetic dissection of immunity to mycobacteria: The human model. Annual Review of Immunology, 20, 581-620. pubmed.ncbi.nlm.nih.gov
4. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Macrophage activation by armed CD4 TH1 cells. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
5. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). T lymphocytes and cellular immunity. In Microbiology (Section 18.3). OpenStax. openstax.org

An infant who cannot kill, and is dying of inflammation

A five-month-old is admitted with three weeks of unremitting fever. She has a large liver and spleen, a haemoglobin of 68 grams per litre, platelets of 40 times 10 to the ninth per litre, triglycerides well above normal, fibrinogen low, and a ferritin of 12,000 micrograms per litre. Marrow aspirate shows macrophages that have ingested erythrocytes and platelets. She has haemophagocytic lymphohistiocytosis, and sequencing finds biallelic PRF1 mutations: she cannot make perforin.

Now hold the two halves of that together. Perforin is the protein cytotoxic lymphocytes use to kill. This child cannot kill infected or activated cells. And she is dying not of infection but of an overwhelming, self-sustaining inflammatory response.

The explanation is a control loop. A cytotoxic T cell that recognises an infected antigen-presenting cell normally kills it, which removes the source of stimulation and terminates the response. Without perforin, the target survives, continues presenting antigen, and continues stimulating T cells, which continue producing interferon gamma, which activates macrophages, which produce IL-1, IL-6, IL-18 and tumour necrosis factor and consume blood cells. Nothing switches it off. The cytokine storm is the direct consequence of a killing defect.

Key idea: Cytotoxicity is not only an effector function. It is a termination mechanism, and losing it causes inflammation rather than merely permitting infection.

How a cytotoxic T cell is made

A naive CD8 cell needs the same three signals a CD4 cell does, but with an additional constraint: for most antigens, the dendritic cell must first be licensed by a CD4 helper cell. The helper cell recognising peptide on class II engages CD40 on the dendritic cell through CD40 ligand, which raises B7 and produces IL-12 and IL-15, and only then can that dendritic cell fully prime a CD8 cell. The two T cells need not be present simultaneously; the dendritic cell holds the licence.

This is why CD4 depletion in advanced HIV infection impairs CD8 responses as well, and why vaccine design attends to T helper epitopes even when a cytotoxic response is the goal. Antigen from a cell the dendritic cell did not itself infect reaches class I by cross-presentation, as covered earlier.

The synapse

When a cytotoxic T cell finds its target it forms an immunological synapse, which is an organised structure rather than a contact. Integrins, chiefly LFA-1 binding ICAM-1, form an adhesion ring, the peripheral supramolecular activation cluster, and T cell receptors with their signalling machinery cluster centrally. The microtubule organising centre then moves to sit just beneath the synapse, and cytotoxic granules travel along microtubules to be released into the narrow cleft.

Two consequences follow from that geometry. First, killing is directional: the lethal cargo is delivered into a sealed space of a few tens of nanometres, so bystander cells in contact with the same T cell are spared. Second, a single cytotoxic T cell can kill several targets in succession, dissociating after each and reforming a synapse with the next, which is called serial killing.

The granule pathway, and why apoptosis is the right death

Cytotoxic granules are secretory lysosomes. They contain perforin, a family of granzymes, and the proteoglycan serglycin, held at acid pH where perforin is inactive.

On release into the neutral synaptic cleft, perforin binds the target membrane in a calcium-dependent manner and oligomerises. It delivers granzymes into the target cytosol, in part through pores and in part by promoting endosomal escape after uptake. Granzyme B, the best characterised, cleaves after aspartate residues; it directly activates caspase-3 and also cleaves BID, engaging the mitochondrial apoptotic route. Granzyme A acts through a caspase-independent pathway.

The target dies by apoptosis, not by lysis, and that distinction is functional rather than cosmetic. Apoptosis activates caspase-activated DNase, which fragments the cell's DNA, including viral genomes. Necrotic lysis would release intact virions and inflammatory contents into the tissue. Killing an infected cell in a way that destroys the pathogen's genome is a better outcome than bursting it.

The killer protects itself. Cytotoxic lymphocytes express high levels of the lysosomal protease inhibitor and repair machinery, rapidly repair membrane damage, and express cathepsin B at the surface after degranulation. LAMP-1, also called CD107a, is exposed on the surface when granules fuse, and measuring surface CD107a is now a standard clinical assay of degranulation capacity, used in the workup of haemophagocytic lymphohistiocytosis.

Why this matters: The diagnostic tests in this field come directly from the mechanism. CD107a reports degranulation; perforin staining reports the protein; natural killer cell cytotoxicity assays report the end result. Each localises the lesion to a different step.

The second killing route

Fas ligand on the effector engages Fas on the target, trimerising it, recruiting FADD and pro-caspase-8, and activating the extrinsic apoptotic pathway. It is slower than granule delivery and does not require granule release.

Its main role is homeostatic: deleting activated lymphocytes at the end of a response, a process called activation-induced cell death. Failure of that route causes autoimmune lymphoproliferative syndrome, most often from heterozygous FAS mutations. The clinical picture is chronic non-malignant lymphadenopathy and splenomegaly, autoimmune cytopenias, raised vitamin B12 and soluble Fas ligand, and a diagnostic hallmark: an expanded population of alpha beta T cells that express neither CD4 nor CD8, so-called double-negative T cells, which in a healthy person are rare. Lymphoma risk is substantially increased.

Set that beside the perforin story. Loss of granule killing gives a hyperinflammatory syndrome; loss of Fas-mediated killing gives lymphoproliferation and autoimmunity. Two death pathways, two very different failure modes, because they are used for different jobs.

The haemophagocytic syndromes as a family

The pigment entries are worth pausing on. Melanocytes transport melanosomes using the same Rab and lysosome-related organelle machinery that cytotoxic cells use for granules. So a child with silvery-grey hair, partial albinism and an inflammatory syndrome has a trafficking defect, and the hair is the clue that localises the pathway. A light microscopic examination of a hair shaft, showing clumped melanin, is a genuinely useful bedside test.

Haemophagocytic lymphohistiocytosis also occurs without a germline defect. Secondary or acquired forms follow Epstein-Barr virus and other infections, haematological malignancy, and rheumatic disease, where it is called macrophage activation syndrome and complicates systemic juvenile idiopathic arthritis and lupus. It is also a recognised toxicity of chimeric antigen receptor T cell therapy. Treatment aims at the loop rather than at any single cell: etoposide and dexamethasone in the classic protocol, ciclosporin, and more recently emapalumab, an antibody against interferon gamma, which targets the specific cytokine that the failed control loop overproduces. Definitive treatment of familial disease is haematopoietic stem cell transplantation.

Memory, tissue residence and exhaustion

After the response contracts, surviving CD8 cells form memory populations. Central memory cells recirculate through lymph nodes and proliferate strongly on re-challenge; effector memory cells patrol blood and non-lymphoid tissue with faster effector function. A third population, tissue-resident memory cells, does not recirculate at all: they take up permanent residence in skin, lung, gut and other barrier tissues, marked by CD69 and often CD103, and they provide the fastest local response to reinfection. Their existence is a strong argument for mucosal rather than intramuscular vaccination for respiratory pathogens, because intramuscular immunisation generates them poorly in the airway.

Under persistent antigen, as in chronic viral infection or a growing tumour, CD8 cells enter a distinct differentiation state called exhaustion. They progressively lose IL-2, then tumour necrosis factor, then interferon gamma production, and they upregulate inhibitory receptors including PD-1, LAG-3, TIM-3 and CTLA-4 under the control of the transcription factor TOX. Exhaustion is not simply anergy or senescence; it has its own epigenetic programme, which is why it is only partially reversible. That partial reversibility is the entire basis of checkpoint blockade, which the tumour immunology lesson takes up.

The core of it: Exhaustion is a programmed state with its own chromatin landscape, not a passive failure. That is why blocking PD-1 restores some function and not all of it.

Common misconceptions

• "Cytotoxic T cells lyse their targets." They induce apoptosis. Apoptotic death fragments viral genomes through caspase-activated DNase, whereas lysis would release intact virions.
• "Perforin makes a hole through which the cell's contents leak out." Its principal role is delivering granzymes into the target cytosol; the death is caused by granzyme-activated apoptosis, not by osmotic failure.
• "A cytotoxic defect causes infections." The dominant presentation of familial perforin deficiency is hyperinflammation, because killing is how the stimulating cell is removed and the response terminated.
• "The killer cell is at risk from its own granules." Directional secretion into a sealed synapse, rapid membrane repair and surface cathepsin B protect it, and it can kill serially.
• "Exhausted T cells are simply worn out." Exhaustion is a defined differentiation state with a TOX-driven epigenetic programme, which is why checkpoint blockade restores function only partly.

What you now know

• CD8 priming usually requires a dendritic cell licensed by CD4 help through CD40, which is why helper epitopes matter even for cytotoxic vaccines.
• The immunological synapse organises adhesion and receptor clusters and polarises the microtubule organising centre so that granules are released directionally, permitting serial killing.
• Perforin delivers granzymes; granzyme B activates caspase-3 and cleaves BID, and the resulting apoptosis destroys viral genomes rather than releasing them.
• Surface CD107a reports degranulation and is used clinically to localise a cytotoxicity defect.
• Fas-mediated killing terminates lymphocyte responses; its failure causes autoimmune lymphoproliferative syndrome with double-negative T cells.
• Familial haemophagocytic lymphohistiocytosis arises from defects in perforin or granule trafficking, with pigmentary clues in RAB27A and LYST disease, and is treated by interrupting the interferon gamma driven loop.
• Memory CD8 populations include non-recirculating tissue-resident cells, and chronic antigen drives a distinct, partly reversible exhaustion programme.

Sources

1. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). T cell-mediated cytotoxicity. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
2. Zhang, K., Astigarraga, I., Bryceson, Y., Lehmberg, K., Machowicz, R., Marsh, R., et al. (2024). Familial hemophagocytic lymphohistiocytosis. In M. P. Adam et al. (Eds.), GeneReviews. University of Washington, Seattle. ncbi.nlm.nih.gov
3. Bleesing, J. J. H., Nagaraj, C. B., & Zhang, K. (2017). Autoimmune lymphoproliferative syndrome. In M. P. Adam et al. (Eds.), GeneReviews. University of Washington, Seattle. ncbi.nlm.nih.gov
4. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). General properties of armed effector T cells. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
5. Parker, N., Schneegurt, M., Thi Tu, A.-H., Lister, P., & Forster, B. M. (2016). T lymphocytes and cellular immunity. In Microbiology (Section 18.3). OpenStax. openstax.org

Six hundred and forty infants and a peanut

Between 2009 and 2013, George Du Toit and Gideon Lack randomised 640 infants aged 4 to 11 months, every one of them already carrying severe eczema or egg allergy or both, either to eat about 6 grams of peanut protein a week or to avoid peanut completely until their fifth birthday. Standard advice at the time was avoidance. At 60 months, among the 530 children whose entry skin prick test had been negative, peanut allergy was present in 13.7 percent of the avoidance group and 1.9 percent of the consumption group. Among the 98 who had already raised a small weal at entry, the figures were 35.3 and 10.6 percent. Feeding the protein did not sensitise these children. It protected them, and the effect was large enough that national feeding guidance was rewritten around it.

That trial is the clinical face of the mechanism this lesson is about. The gut does not treat a protein that crosses it the way skin treats a protein injected into it. Its default answer to a novel protein arriving through normal traffic is active, antigen-specific suppression, and the name for that is oral tolerance.

So what?: Mucosal immunity is not a thinner version of systemic immunity. It runs an inverted default. The systemic system assumes that a protein which has breached the barrier is dangerous. The mucosal system assumes that a protein arriving through the barrier's ordinary traffic is food, and it takes a specific inflammatory signal to overturn that assumption.

The traffic problem, in numbers

Hold the load in front of you before the mechanism. The gastrointestinal mucosa covers roughly 30 square metres, the corrected figure from the first lesson of this course. Across it sit on the order of 10¹³ to 10¹⁴ bacteria, reaching about 10¹¹ organisms per gram of colonic contents and dominated by Bacteroidetes and Firmicutes. Through it passes something like 60 to 100 grams of dietary protein a day, a small but real fraction of which crosses intact. Around three quarters of the body's antibody-secreting plasma cells live in mucosal tissue rather than in lymph node or spleen, and they push 3 to 5 grams of IgA into the lumen every day, more than the daily output of every other isotype combined. Almost all of it is then excreted.

An immune system built on the systemic template would be at war with breakfast by mid-morning. What the gut runs instead is a set of anatomical and cellular arrangements whose whole purpose is to sample continuously, respond locally, and stop the response escalating.

The architecture: where sampling happens

Two sampling routes bypass the M cell altogether. Goblet cells hand soluble antigen through goblet-cell-associated antigen passages. CX3CR1-positive macrophages extend dendrites between enterocytes, through the tight junctions, into the lumen, then pass what they catch to a CD103-positive dendritic cell that can actually migrate. That division of labour matters: the sampling cell is largely sessile, the presenting cell is mobile, and the handover is where the decision gets made.

How the gut turns a dietary protein into a regulatory T cell

1. Sampling. Antigen crosses by M cell transcytosis, goblet cell passage or macrophage dendrite.
2. Migration. The CD103-positive dendritic cell travels to the mesenteric lymph node under CCR7, carrying the antigen with it.
3. Conditioning. Epithelial cells release TGF beta and thymic stromal lymphopoietin, which keep the local dendritic cells non-inflammatory. Those dendritic cells express retinal dehydrogenase, the ALDH1A2 enzyme that converts dietary vitamin A into retinoic acid. Retinoic acid plus TGF beta drives a naive CD4 cell into a Foxp3-positive peripherally induced regulatory T cell.
4. Imprinting. The same retinoic acid induces the integrin alpha4 beta7 and the chemokine receptor CCR9. Those are addresses: alpha4 beta7 binds MAdCAM-1 on gut endothelium, CCR9 binds CCL25 made in the small intestine. The cell is sent back to the tissue that educated it.
5. Effect. The induced regulatory cells return to the lamina propria and secrete IL-10 and TGF beta, suppressing responses to their own antigen and, through those cytokines, to whatever else is being presented nearby.

Dose changes the outcome. Repeated low doses of antigen favour the regulatory route above. A single very large dose favours deletion and anergy of the responding clones instead. Both end in tolerance, by different routes, which is why experimental protocols that look similar can give different mechanistic answers. Vitamin A deficiency measurably impairs the whole sequence, which is a direct line from nutrition to a specific immunological failure.

Why this matters: Oral tolerance is dominant and transferable. Take CD4 T cells from an animal fed an antigen, put them into a naive animal, and the recipient is suppressed too. That is what separates tolerance from mere ignorance: ignorance cannot be transferred in a syringe.

IgA is a dimer because it has to be shipped

A mucosal plasma cell makes IgA as two monomers joined tail to tail by a small polypeptide, the J chain. That dimeric form is not an accident of assembly; it is a shipping label. The polymeric immunoglobulin receptor, pIgR, sits on the basolateral surface of the epithelial cell and binds only J chain-containing polymers. Monomeric IgA is ignored.

The route runs: pIgR binds dimeric IgA on the basolateral face, the complex is endocytosed, it is transcytosed across the cell in vesicles, and at the apical surface a protease cuts pIgR. The cleaved extracellular piece stays attached to the antibody and is now called secretory component. Secretory IgA is therefore the antibody plus a bodyguard, and the bodyguard does two jobs: it shields the hinge from luminal proteases, and its heavy glycosylation tethers the molecule in the mucus layer where the bacteria are.

What secretory IgA then does is best described as immune exclusion. It agglutinates organisms, blocks adhesins, neutralises toxins and traps everything in mucus for peristalsis to remove. What it mostly does not do is activate complement by the classical pathway or serve as a good opsonin, because its Fc engages neither C1q nor Fc gamma receptors well. Read that as design, not deficiency: 10¹⁴ commensals coated in an antibody that triggered inflammation would be catastrophic. IgA also neutralises virus inside the epithelial cell during transcytosis, and can carry immune complexes out of the lamina propria and into the lumen.

Two production routes feed the lumen. High-affinity IgA comes from Peyer's patch germinal centres with T cell help, the pathway from the germinal centre lesson. Lower-affinity, broadly cross-reactive IgA is induced in the lamina propria without T cells, driven by BAFF and APRIL from epithelium and dendritic cells; this is the fraction that coats commensals indiscriminately. Selective IgA deficiency is the commonest primary immunodeficiency in people of European ancestry, near 1 in 500, and most of those people never find out, largely because pIgR also transports pentameric IgM and covers for it.

Bottom line: Every structural feature of secretory IgA follows from the job. Dimerisation is the transport signal, secretory component is protease armour and a mucus anchor, and the feeble Fc effector function is what makes it safe to deploy by the gram against organisms you have no intention of killing.

The microbiome as an instructing organ

A germ-free mouse is not a clean version of a normal mouse. It has small Peyer's patches, few IgA plasma cells, a thin mucus layer, fewer lamina propria CD4 T cells and a grossly enlarged caecum. Colonise it and most of that reverses. The commensal population is not a passive occupant; it is an instructive input, and three experiments pinned down how specific that instruction is.

Andrew Macpherson and Therese Uhr showed in 2004 that dendritic cells carry live commensal bacteria from the gut to the mesenteric lymph node, induce IgA there, and go no further. Remove the mesenteric nodes and live commensals reach the spleen. The firewall is anatomical, and it is one node deep.

Ivaylo Ivanov and colleagues in 2009 were chasing an irritating discrepancy: the same inbred mouse strain had markedly different numbers of intestinal TH17 cells depending on which vendor supplied it. The difference tracked to a single organism, segmented filamentous bacterium, which adheres tightly to ileal epithelium. Monocolonise a germ-free mouse with it and the TH17 compartment appears. A laboratory artefact turned out to be a mechanism.

Koji Atarashi and colleagues in 2011 ran the complementary experiment: a defined mixture of Clostridium species from clusters IV and XIVa, given to germ-free mice, induced colonic Foxp3-positive regulatory T cells. Part of the signal is chemical rather than cellular. Those organisms ferment dietary fibre to short-chain fatty acids, and butyrate among them inhibits histone deacetylases at the Foxp3 locus, making the gene easier to switch on. Fibre in the diet, butyrate in the colon, an epigenetic change at one promoter, a regulatory T cell.

The epidemiology sits on top of this. David Strachan noticed in 1989, in 17,414 British children born in a single week of March 1958, that hay fever fell steadily as the number of older siblings rose, and proposed what became the hygiene hypothesis. The modern form of the argument is not about cleanliness but about which microbes a young immune system meets. Michelle Stein and colleagues in 2016 compared Amish and Hutterite farm children in the United States: similar ancestry, similar diets, similar family sizes, but Amish farming is traditional single-family dairying with animals close to the house and Hutterite farming is industrialised and distant. Asthma prevalence was 5.2 percent in the Amish children and 21.3 percent in the Hutterite children. Endotoxin in Amish house dust was several times higher. Amish house dust protected mice from experimental airway disease, and that protection was lost in mice lacking MyD88 and Trif, the innate signalling adaptors from the pattern recognition lesson. The exposure is microbial, it is early, and it works through innate receptors.

Common misconceptions

• "Secretory IgA kills bacteria." Mostly it does not. It agglutinates, blocks adhesion and excludes. Weak complement activation and weak opsonisation are the reason it can be secreted by the gram without inflaming the gut.
• "IgA deficiency is a severe immunodeficiency." About 1 in 500 Europeans have it and most are asymptomatic, partly because pIgR also ships pentameric IgM.
• "Oral tolerance means the gut ignores food antigens." It is active suppression by induced Foxp3-positive cells and their cytokines, and it can be transferred to a naive animal by cells alone.
• "Feeding an allergen early sensitises the child." The LEAP trial found the reverse, decisively, in exactly the high-risk infants the old advice was written for.
• "The hygiene hypothesis means we are too clean." Handwashing and clean water are not the exposures in question. The Amish and Hutterite comparison points at farm-animal microbial exposure acting through innate receptors, not at domestic hygiene.
• "M cells are damaged epithelium." They are a distinct differentiated cell type specified by RANKL from stromal cells beneath the follicle.

What to carry forward

• The mucosal default is tolerance, and the LEAP trial showed that reversing avoidance advice cut peanut allergy from 13.7 percent to 1.9 percent in high-risk infants with negative entry skin tests.
• Oral tolerance is produced by CD103-positive dendritic cells that make retinoic acid from dietary vitamin A and, with TGF beta, induce Foxp3-positive regulatory T cells imprinted with alpha4 beta7 and CCR9 to return to the gut.
• Low repeated doses favour regulatory suppression; a single high dose favours deletion and anergy.
• IgA is dimerised by the J chain because pIgR transports only J chain-containing polymers; cleaved pIgR remains bound as secretory component, giving protease resistance and a mucus anchor.
• Secretory IgA works by exclusion rather than by killing, and its poor complement and Fc receptor engagement is what makes that safe.
• Specific commensals give specific instructions: segmented filamentous bacteria induce TH17 cells, Clostridium clusters IV and XIVa induce colonic Tregs, and butyrate from fibre fermentation acts at the Foxp3 locus as a histone deacetylase inhibitor.
• The mesenteric lymph node is the firewall for live commensals, and removing it lets them reach the spleen.

Sources

1. Du Toit, G., Roberts, G., Sayre, P. H., Bahnson, H. T., Radulovic, S., Santos, A. F., et al. (2015). Randomized trial of peanut consumption in infants at risk for peanut allergy. New England Journal of Medicine, 372(9), 803-813. pubmed.ncbi.nlm.nih.gov
2. Johansen, F. E., & Kaetzel, C. S. (2011). Regulation of the polymeric immunoglobulin receptor and IgA transport. Mucosal Immunology, 4(6), 598-602. pubmed.ncbi.nlm.nih.gov
3. Atarashi, K., Tanoue, T., Shima, T., Imaoka, A., Kuwahara, T., Momose, Y., et al. (2011). Induction of colonic regulatory T cells by indigenous Clostridium species. Science, 331(6015), 337-341. pubmed.ncbi.nlm.nih.gov
4. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The mucosal immune system. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov
5. Abbas, A. K., Lichtman, A. H., & Pillai, S. (2021). Specialized immunity at epithelial barriers and in immune privileged tissues. In Cellular and Molecular Immunology (10th ed.). Elsevier.

A thymus that transcribes insulin

In 2002 Mark Anderson, working with Diane Mathis and Christophe Benoist, asked what genes the epithelial cells in the medulla of a mouse thymus actually transcribe. The answer was strange. Insulin was on the list. So were thyroglobulin, salivary protein 1, and proteins of the retina and the exocrine pancreas. None of these belong in a thymus. When the group knocked out the transcriptional regulator AIRE, the list collapsed, and the mice developed lymphocytic infiltration and autoantibodies against exactly the organs whose genes had gone missing from the thymic list.

That is the mechanism of central tolerance stated in one experiment. A developing T cell cannot be shown the whole body, so the body is brought to the thymus in miniature: medullary thymic epithelial cells transcribe several thousand tissue-restricted antigens, each cell expressing a different small subset, so that a thymocyte moving through the medulla meets a mosaic that approximates the self it will later have to leave alone.

The core of it: Tolerance is not the absence of a response. It is a set of active, separable, individually breakable processes. Every one of them has a human disease attached, which is why autoimmunity is the best textbook immunology has.

The natural experiment: APS-1

The human counterpart was cloned in 1997 by two groups working on Finnish, Sardinian and Iranian Jewish families. Autoimmune polyendocrine syndrome type 1, also called APECED, is caused by recessive mutations in AIRE. It is one of the very few monogenic autoimmune diseases, and its clinical triad is chronic mucocutaneous candidiasis, hypoparathyroidism and adrenal insufficiency, usually appearing in that order in childhood.

The candidiasis looked out of place for years. Why would a tolerance defect cause a fungal infection? The answer, found much later, is that these patients make high-titre neutralising autoantibodies against IL-17A, IL-17F and IL-22, the cytokines that maintain antifungal defence at mucosal surfaces. A defect in tolerance produced an immunodeficiency, by way of an autoantibody against a cytokine. Many of the same patients also carry neutralising autoantibodies against type I interferon, which is the same class of antibody that turned out to account for a measurable share of critical COVID-19 in the innate immunity lesson. One gene, three organs, and two acquired cytokine deficiencies.

Central tolerance is nonetheless incomplete by construction. Negative selection removes thymocytes whose receptors bind self peptide-MHC above an affinity threshold; below it, cells leave. Rare antigens may not be represented at all. Antigens that only exist after birth, or after a chemical modification, cannot be shown. So a healthy adult walks around with self-reactive lymphocytes, and something else must hold them.

The four peripheral mechanisms, and what breaks each

B cells have their own layer. A developing B cell whose receptor binds self antigen strongly can rearrange its light chain again, which is receptor editing, or be deleted, or leave in an anergic state. But B cells also class switch and hypermutate after they leave the bone marrow, so a cell can acquire self-reactivity in a germinal centre that it never had at the outset, and the germinal centre must be policed too.

Regulatory T cells and the gene that defines them

Shimon Sakaguchi's 1995 experiment was arithmetic. He removed CD4 cells bearing CD25, then a marker of activation rather than of a lineage, from normal mouse spleen and transferred what was left into a T cell-deficient recipient. The recipients developed gastritis, thyroiditis, oophoritis and insulitis. Adding back the CD25-positive fraction prevented all of it. A minority population, perhaps 5 to 10 percent of CD4 cells, was actively holding the rest.

The transcription factor came in 2003, when Sakaguchi's group and, independently, Fontenot and Rudensky showed that Foxp3 both marked these cells and was sufficient to confer their function when forced into a conventional CD4 cell. The genetics had already pointed there: the scurfy mouse, an X-linked lethal with lymphoproliferation and multi-organ infiltration, carries a frameshift in that gene, and in 2001 Bennett, Ochs and colleagues showed that the human counterpart, IPEX syndrome, is caused by FOXP3 mutations. IPEX boys present in the first months of life with intractable secretory diarrhoea, type 1 diabetes that can begin in the neonatal period, eczema and autoimmune cytopenias. Untreated it is usually fatal, and the definitive treatment is haematopoietic stem cell transplantation, because the defect is in a cell lineage.

Two kinds share the marker: thymically derived cells selected in the medulla at intermediate affinity for self antigen, and peripherally induced cells generated in tissue under TGF beta and retinoic acid, the pathway the mucosal lesson worked through. Their suppressive tools are each a drug target: high CD25 with no IL-2 production of their own, making them an IL-2 sink; CTLA-4-mediated transendocytosis that strips CD80 and CD86 off a dendritic cell; and IL-10, TGF beta and IL-35.

Remember: Regulatory T cells act on the antigen-presenting cell as much as on the responding T cell. Understanding CTLA-4 as a molecule that removes costimulation from the dendritic cell, rather than only as a brake inside the T cell, is what makes both abatacept and the toxicity profile of anti-CTLA-4 antibodies predictable.

How tolerance is actually lost

Molecular mimicry. The clean case is acute rheumatic fever. Two to four weeks after a group A streptococcal pharyngitis, a minority of patients develop carditis, migratory polyarthritis, chorea, subcutaneous nodules or erythema marginatum. The streptococcal M protein is a coiled-coil surface molecule with regions structurally similar to human cardiac myosin, tropomyosin, laminin and the valve endothelial basement membrane. Antibodies and T cells raised against M protein cross-react with these, and Madeleine Cunningham's group has traced cross-reactive human antibody and T cell clones from patients to those targets. Antibody binding to valve endothelium upregulates VCAM-1, T cells enter, and the granulomatous Aschoff bodies of rheumatic carditis follow. The clinically decisive point is that treating the pharyngitis inside about nine days prevents the sequel, and that recurrent infections drive progressive valve damage, which is why secondary prophylaxis runs for years.

Epitope spreading. A response that starts against one epitope does not stay there. In relapsing experimental autoimmune encephalomyelitis induced with the proteolipid protein peptide PLP 139-151, the first relapse is driven not by that peptide but by PLP 178-191, and the next by a myelin basic protein epitope. Tissue destruction releases new antigen, which is taken up and presented by now-activated antigen-presenting cells in an inflamed site. Two consequences follow. Antigen-specific therapy aimed at the initiating epitope stops working once spreading has occurred, and the antigen you can measure in an established disease is often not the one that started it.

Neoantigen formation. Rheumatoid arthritis gives the cleanest example. Peptidylarginine deiminase enzymes convert arginine residues to citrulline in inflamed tissue. Citrullinated fibrinogen, vimentin and alpha-enolase are chemically new, were never present in the thymus, and are bound preferentially by the shared epitope HLA-DRB1 alleles. Antibodies against cyclic citrullinated peptides are more specific for rheumatoid arthritis than rheumatoid factor and appear years before symptoms. Smoking induces citrullination in the lung, and the interaction between smoking and shared epitope alleles is one of the better-documented gene-environment effects in autoimmunity.

Failure of clearance. Systemic lupus erythematosus behaves like a disease of debris. Complete C1q deficiency carries a lupus risk above 90 percent, the highest of any single genetic factor, because C1q opsonises apoptotic cells for silent removal. Neutrophil extracellular traps, from the innate lesson, expose chromatin. Undegraded nuclear material engages TLR7 and TLR9 in B cells and plasmacytoid dendritic cells, drives type I interferon, and the resulting anti-double-stranded-DNA immune complexes deposit in glomeruli. The mechanism explains the serology: complement consumption in active disease means low C3 and C4 with a high anti-dsDNA titre.

Bystander activation and superantigens. Infection can lift the threshold for everything happening nearby, and a superantigen that cross-links MHC class II to the T cell receptor V beta domain activates whole receptor families irrespective of specificity.

Reading a named disease back to its mechanism

Behind all of these sits genetic risk that is mostly not monogenic. HLA associations are the strongest: B27 with ankylosing spondylitis, DQ2 and DQ8 with coeliac disease, DR3-DQ2 and DR4-DQ8 with type 1 diabetes, the DRB1 shared epitope with rheumatoid arthritis. The largest non-HLA effect across several diseases is the PTPN22 variant R620W, in a phosphatase that sets lymphocyte activation thresholds. And most of these diseases fall disproportionately on women, from roughly 2 to 1 in multiple sclerosis to about 9 to 1 in lupus, for reasons that include escape from X inactivation at loci such as TLR7.

What matters here: Naming the mechanism tells you what will help. If the pathogenic agent is a circulating antibody, removing plasma helps within days. If it is a T cell response against a destroyed tissue, nothing you give will bring the tissue back, and the treatment is replacement plus prevention in those not yet affected.

Common misconceptions

• "Autoantibodies prove that antibodies cause the disease." In type 1 diabetes they are markers of an ongoing T cell process. In myasthenia gravis and Graves disease they are the disease. Transfer experiments and placental transfer, not titres, separate the two.
• "Central tolerance deletes every self-reactive cell." It works on an affinity threshold and on whatever antigens the thymus can display. Self-reactive lymphocytes are present in every healthy person.
• "AIRE deficiency causes only autoimmunity." APS-1 patients also acquire neutralising antibodies to IL-17 and IL-22, which is why they get chronic candidiasis, and often to type I interferon.
• "Molecular mimicry means sequence identity." Cross-reactive recognition needs shared surface shape and chemistry, and M protein resembles cardiac myosin in coiled-coil structure rather than in a matching string of residues.
• "Finding the initiating antigen would cure the disease." Epitope spreading means that by the time a patient presents, the response has usually moved on from whatever began it.
• "Regulatory T cells only suppress other T cells." A large part of their effect is on dendritic cells, by CTLA-4-dependent removal of CD80 and CD86.

Where this leaves us

• AIRE drives promiscuous transcription of thousands of tissue-restricted antigens in medullary thymic epithelium, and its loss causes APS-1 with candidiasis, hypoparathyroidism and adrenal failure.
• The candidiasis of APS-1 is caused by neutralising autoantibodies against IL-17 and IL-22: a tolerance defect producing an immunodeficiency.
• Peripheral tolerance rests on anergy, deletion, suppression and ignorance, and each has a named human or experimental failure.
• Foxp3 defines regulatory T cells; its loss causes scurfy in mice and IPEX in humans, and their suppression works through IL-2 consumption, CTLA-4-mediated stripping of B7, and inhibitory cytokines.
• Tolerance is lost by molecular mimicry, epitope spreading, neoantigen formation such as citrullination, and failure to clear apoptotic debris, of which C1q deficiency is the extreme case.
• The mechanism class predicts the therapy: plasma exchange for antibody-mediated disease, replacement for destroyed tissue, B cell or costimulation blockade where the response is still running.

Sources

1. Anderson, M. S., Venanzi, E. S., Klein, L., Chen, Z., Berzins, S. P., Turley, S. J., et al. (2002). Projection of an immunological self shadow within the thymus by the aire protein. Science, 298(5597), 1395-1401. pubmed.ncbi.nlm.nih.gov
2. Sakaguchi, S., Sakaguchi, N., Asano, M., Itoh, M., & Toda, M. (1995). Immunologic self-tolerance maintained by activated T cells expressing IL-2 receptor alpha-chains (CD25). Journal of Immunology, 155(3), 1151-1164. pubmed.ncbi.nlm.nih.gov
3. Bennett, C. L., Christie, J., Ramsdell, F., Brunkow, M. E., Ferguson, P. J., Whitesell, L., et al. (2001). The immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome (IPEX) is caused by mutations of FOXP3. Nature Genetics, 27(1), 20-21. pubmed.ncbi.nlm.nih.gov
4. Cunningham, M. W. (2019). Molecular mimicry, autoimmunity, and infection: The cross-reactive antigens of group A streptococci and their sequelae. Microbiology Spectrum, 7(4). pubmed.ncbi.nlm.nih.gov
5. Abbas, A. K., Lichtman, A. H., & Pillai, S. (2021). Immunologic tolerance and autoimmunity. In Cellular and Molecular Immunology (10th ed.). Elsevier.

A dog called Neptune, and a second dose that should have been safer

In 1901 Paul Portier and Charles Richet sailed with Prince Albert I of Monaco, collecting Portuguese man o' war for a study of its toxin. Back in Paris the following year they switched to extracts of the sea anemone Actinia and set out to do something ordinary: immunise dogs with small doses so that they would tolerate a larger one. The standard account records a dog named Neptune, which took a first injection without difficulty. Three weeks later he was given a second dose, smaller than the first. He vomited, his legs gave way, and he was dead within half an hour.

Richet had been trying to produce prophylaxis, protection. He named what he had produced instead anaphylaxis, against protection, and took the 1913 Nobel Prize for it. A second exposure, at a smaller dose, killed an animal the first had not troubled. Immunological memory, which the rest of this course has treated as the point of the whole system, is here the mechanism of injury.

In 1963 Philip Gell and Robin Coombs proposed a way of sorting such injuries. Their scheme classifies by mechanism, not by clinical appearance, which is why it survives: two diseases that look nothing alike can share a type, and one drug can cause reactions of all four.

Type I: a fourteen-year-old and a biscuit

A fourteen-year-old with known peanut allergy eats a biscuit at a friend's house. Within four minutes her lips tingle and her voice changes. Within eight she is wheezing, covered in urticaria, and her blood pressure is 74 over 40.

Read it mechanistically. At some earlier exposure she made IgE against Ara h 2, and that IgE has been sitting on the high-affinity receptor FcERI on her mast cells and basophils. IgE is unusual among antibodies in being armed on the cell before the antigen arrives, and its serum half-life of about two days becomes weeks once bound. The allergen is multivalent, so it cross-links adjacent IgE molecules; receptor clustering triggers Lyn and Syk, and calcium flux drives granule fusion within seconds. Preformed histamine and tryptase go first, producing vasodilatation, vascular leak and bronchoconstriction. Newly synthesised arachidonic acid products follow within minutes, and cysteinyl leukotrienes are far more potent bronchoconstrictors than histamine, which is a large part of why an antihistamine cannot treat this. Cytokines then recruit eosinophils and TH2 cells over hours, giving the late phase and the possibility of a biphasic reaction after apparent recovery.

Anaphylaxis is treated with intramuscular adrenaline into the anterolateral thigh, and every part of that sentence is mechanism. Alpha1 agonism reverses the vasodilatation and capillary leak dropping her pressure; beta1 agonism supports the heart; beta2 agonism relaxes bronchial smooth muscle and raises mast cell cyclic AMP, suppressing further degranulation. The thigh absorbs faster and more reliably than the deltoid or the subcutaneous route. Antihistamines address itch and urticaria and nothing that will kill her. Serum tryptase, taken at presentation and again as a baseline after recovery, peaks between about 15 minutes and 2 hours and confirms mast cell activation retrospectively.

The point: In type I, the antibody is already loaded onto the effector cell before the antigen arrives. That single structural fact accounts for the speed, for the requirement for prior sensitisation, and for why the first exposure is usually uneventful.

Type II: an antibody against something that cannot move

Haemolytic disease of the fetus and newborn is the cleanest illustration. An RhD-negative woman carries an RhD-positive fetus. Fetal red cells enter her circulation, most substantially at delivery, and she makes anti-D. IgG, alone among isotypes, crosses the placenta by FcRn. In her next RhD-positive pregnancy, maternal anti-D coats fetal red cells, which are then removed by splenic macrophages through Fc gamma receptors. Note that this is extravascular haemolysis: the cells are eaten, not lysed, which is why the picture is anaemia and unconjugated hyperbilirubinaemia rather than haemoglobinuria. Severe cases progress to high-output cardiac failure and hydrops fetalis.

The prevention is one of the most elegant results in clinical immunology. In 1964 Vincent Freda, John Gorman and William Pollack showed that giving anti-Rh gamma globulin to Rh-negative volunteers prevented them from becoming sensitised: giving the antibody suppresses the response to the antigen, most plausibly by clearing the fetal cells before the mother can respond to them. Current practice gives anti-D at around 28 weeks and within 72 hours of delivery, with a Kleihauer-Betke or flow cytometric estimate of the fetomaternal bleed to decide whether more is needed. The test on the baby is the direct antiglobulin test, which detects antibody already on the red cell surface, and which Robin Coombs, of the classification, co-devised.

Goodpasture disease is the same type against a matrix antigen. The autoantibody targets the NC1 domain of the alpha3 chain of type IV collagen, present in glomerular and alveolar basement membrane. Because the antigen is a continuous sheet, immunofluorescence shows a smooth linear ribbon of IgG along the membrane, distinguishing it at a glance from the lumpy granular deposits of type III disease. Richard Lerner, Richard Glassock and Frank Dixon proved causation in 1967 by eluting antibody from diseased human kidneys and transferring nephritis to primates with it. Clinically it is rapidly progressive glomerulonephritis with crescents, plus pulmonary haemorrhage in smokers whose alveolar membrane has been made accessible. Treatment follows from mechanism: plasma exchange to remove the circulating antibody, plus cyclophosphamide and steroids to stop making more.

The type II label also covers pure receptor modulation, where nothing is destroyed at all: the stimulating anti-TSH-receptor antibody of Graves disease, the blocking anti-acetylcholine-receptor antibody of myasthenia gravis. Some authors split those off as a fifth type for that reason.

Type III: the disease that arrives on day nine

Clemens von Pirquet and Bela Schick, in Vienna in the first years of the twentieth century, described what happened to children given horse antiserum against diphtheria. Between seven and twelve days later, reliably, came fever, an urticarial rash, painful joints, swollen nodes and protein in the urine. They called it serum sickness, and the timing is the whole diagnosis: it is how long a primary antibody response takes. The patient becomes ill at the moment her own antibody meets the foreign protein still circulating.

Complex size decides everything. In large antibody excess, complexes are big, bind complement well and are cleared by splenic and hepatic phagocytes. In moderate antigen excess, they are small and soluble, escape clearance, and lodge where hydrostatic pressure and turbulence favour deposition: glomerular capillary walls, small vessel walls, synovium and choroid plexus. There they activate complement, C3a and C5a recruit neutrophils, and the neutrophils, unable to ingest a deposit fixed in a basement membrane, release granule contents outward. Frustrated phagocytosis is the injurious step, not the deposition itself.

The modern versions are drugs, not horses: antithymocyte globulin, chimeric monoclonal antibodies, occasionally beta-lactams. The local version is the Arthus reaction, oedema and haemorrhage at an injection site in an already-immune subject. Lupus nephritis is the endogenous version, the antigen supplied by the patient's own nuclear material. Its immunofluorescence is granular and characteristically full house, staining for IgG, IgA, IgM, C3 and C1q together, and serum C3 and C4 fall during activity because the complexes consume complement.

In short: Type II and type III differ not in the isotype involved but in whether the antigen was fixed in place before the antibody arrived. Linear immunofluorescence means the antibody found the tissue. Granular means the complex formed elsewhere and landed.

Type IV: forty-eight hours, and no antibody anywhere

A student develops an itchy, weeping, sharply demarcated rash under a new watch strap buckle, two days after wearing it. Nickel is the commonest contact allergen in the world, and allergic contact dermatitis is type IV in its purest form: no immunoglobulin is involved at any stage.

The reaction has two phases. In sensitisation, the small molecule penetrates the stratum corneum and acts as a hapten, coupling to skin proteins to make something a T cell can see; Langerhans cells carry it to the draining node and prime T cells over one to two weeks. In elicitation, re-exposure brings memory T cells into the skin within hours, and the visible reaction peaks at 48 to 72 hours because that is how long it takes to recruit and activate the macrophages that do the work. Nickel has an extra twist: Matthias Schmidt and colleagues showed in 2010 that nickel ions bind directly to two histidine residues in human TLR4, which mouse TLR4 lacks, supplying their own innate danger signal, and that is why mice model nickel allergy badly. Urushiol, the catechol of poison ivy, works the conventional hapten way, which is why the rash comes in streaks a day or two after the walk rather than during it.

The tuberculin skin test is the same reaction used deliberately. Robert Koch made tuberculin in 1890 as a failed therapy; Charles Mantoux standardised the intradermal method in 1908. Five tuberculin units of purified protein derivative go into the volar forearm, and at 48 to 72 hours you measure induration, not erythema, transverse to the long axis. Cut-offs move with prior probability and consequence: 5 mm in HIV infection, recent contacts and those with fibrotic chest radiographs, 10 mm in other higher-risk groups, 15 mm in people with no risk factors. Two failure modes matter. BCG vaccination and non-tuberculous mycobacteria cause false positives, which is why interferon gamma release assays use ESAT-6 and CFP-10, antigens present in Mycobacterium tuberculosis but deleted from BCG. And a falsely negative test in disseminated tuberculosis, HIV or severe malnutrition reflects anergy: the test measures the response, so a patient with no response has no reaction, however heavy the burden of organisms.

Werner Pichler's subdivision of type IV is now standard in drug allergy: IVa TH1 and monocyte driven, as in tuberculin and contact dermatitis; IVb TH2 and eosinophil driven, as in DRESS; IVc cytotoxic T cell killing of keratinocytes through perforin, granzyme and granulysin, which is Stevens-Johnson syndrome and toxic epidermal necrolysis; IVd T cell and neutrophil driven, as in acute generalised exanthematous pustulosis.

Where the scheme creaks

Gell and Coombs classifies mechanisms of injury, not diseases, and most real diseases run more than one. Lupus is type II against blood cells and type III in the kidney at once. Chronic asthma has a type I immediate phase and type IVb inflammation with eosinophils and remodelling. A single beta-lactam can produce anaphylaxis, immune haemolysis, serum sickness or a delayed exanthem in four different patients. More awkwardly, several severe reactions that look exactly like hypersensitivity fit none of the four types, because no adaptive receptor is involved.

• Direct mast cell activation. Benjamin McNeil and colleagues showed in 2015 that vancomycin, fluoroquinolones, neuromuscular blockers and opioid analogues activate mast cells through MRGPRX2, a receptor unrelated to IgE. No prior sensitisation is needed, and the reaction is dose and rate dependent, so slowing the infusion helps in a way it never would in true anaphylaxis.
• Complement activation-related pseudoallergy. Liposomal drugs, some contrast media and micelle-forming excipients activate complement directly, producing C3a and C5a and an anaphylaxis-like picture on first exposure.
• Bradykinin-mediated angioedema. Hereditary angioedema from C1 inhibitor deficiency, and ACE inhibitor angioedema, can obstruct an airway. Neither is mast cell mediated: there is no urticaria and no itch, and adrenaline, antihistamines and steroids do not work. The treatments are C1 inhibitor concentrate, icatibant or ecallantide. Getting this wrong is one of the ways a patient dies while being treated correctly for the wrong diagnosis.

Worth holding on to: The clinical question is never only which type. It is whether the swelling in front of you is histamine or bradykinin, because the two look similar for the first five minutes and the drugs that fix them share nothing.

Common misconceptions

• "The first exposure causes the reaction." Type I needs prior sensitisation, which is why an unremarkable first dose is not reassurance. MRGPRX2 reactions are the exception that proves the rule: they can occur on first exposure precisely because they are not immunological.
• "Antihistamines treat anaphylaxis." They treat itch and urticaria. Leukotrienes drive much of the bronchoconstriction, and only adrenaline reverses the vasodilatation and capillary leak.
• "Type II and type III differ by antibody class." Both use IgG. The difference is whether the antigen was already fixed in the tissue, which is exactly what linear versus granular immunofluorescence tells you.
• "A negative tuberculin test excludes tuberculosis, and you read it by the redness." Both wrong. Read induration only, transverse to the arm at 48 to 72 hours, and remember that anergy in disseminated disease, HIV infection or malnutrition gives a negative result with a heavy organism burden.
• "All swelling of the lips and tongue is allergic." Bradykinin-mediated angioedema has no urticaria, no itch and no response to adrenaline, and needs entirely different drugs.

The short version

• Richet and Portier found in 1902 that a second, smaller dose could kill an animal the first had not harmed, and named it anaphylaxis.
• Type I is IgE preloaded on FcERI and cross-linked by multivalent allergen; the alpha1, beta1 and beta2 effects of intramuscular adrenaline each answer a specific part of the pathology.
• Type II is antibody against a fixed antigen: extravascular haemolysis in haemolytic disease of the newborn, prevented by anti-D at 28 weeks and after delivery, and linear immunofluorescence against the alpha3 chain of collagen IV in Goodpasture disease.
• Type III is soluble complexes formed in antigen excess, depositing and recruiting neutrophils that cannot ingest what they attack; serum sickness arrives on day 7 to 12 because that is how long a primary antibody response takes.
• Type IV needs no antibody: haptens for contact dermatitis, with nickel additionally binding human TLR4, and the tuberculin test read as induration at 48 to 72 hours.
• The scheme creaks over receptor-modulating antibodies, over diseases running two types at once, and over MRGPRX2 activation, complement pseudoallergy and bradykinin-mediated angioedema, which are not hypersensitivity at all.

Sources

1. Richet, G. (2003). The discovery of anaphylaxis, a brief but triumphant encounter of two physiologists (1902). Histoire des Sciences Medicales, 37(4), 463-469. pubmed.ncbi.nlm.nih.gov
2. Lerner, R. A., Glassock, R. J., & Dixon, F. J. (1967). The role of anti-glomerular basement membrane antibody in the pathogenesis of human glomerulonephritis. Journal of Experimental Medicine, 126(6), 989-1004. pubmed.ncbi.nlm.nih.gov
3. Freda, V. J., Gorman, J. G., & Pollack, W. (1964). Successful prevention of experimental Rh sensitization in man with an anti-Rh gamma2-globulin antibody preparation. Transfusion, 4, 26-32. pubmed.ncbi.nlm.nih.gov
4. McNeil, B. D., Pundir, P., Meeker, S., Han, L., Undem, B. J., Kulka, M., & Dong, X. (2015). Identification of a mast-cell-specific receptor crucial for pseudo-allergic drug reactions. Nature, 519(7542), 237-241. pubmed.ncbi.nlm.nih.gov
5. Gell, P. G. H., & Coombs, R. R. A. (1963). Clinical Aspects of Immunology (1st ed.). Blackwell Scientific Publications.

A boy with a flat gamma globulin band

In 1952 Ogden Bruton, a paediatrician at Walter Reed, was looking after an eight-year-old boy who had been admitted with pneumococcal sepsis nineteen times in four years. Serum protein electrophoresis had just become available to clinicians, and Bruton ran it. The albumin peak was normal. The alpha and beta bands were normal. Where the gamma globulin should have been there was nothing at all. He began giving the boy subcutaneous gamma globulin, and the episodes of sepsis stopped.

That single paper is the beginning of clinical immunology as a diagnostic discipline, and it establishes the method the rest of the lesson uses. A missing component announces itself as a particular kind of infection, repeated. Learn which organisms go with which defect and you can localise the lesion before any specialised test is ordered.

Key idea: The organism is the assay. Recurrent meningococcal disease is a complement question, not a general immunology question, and a child with delayed cord separation and no pus in an abscess has a leukocyte adhesion problem, whatever else is going on.

SCID: the one that has to be found this week

Severe combined immunodeficiency is a group of disorders in which T cells are absent or non-functional, and because B cells need T cell help, humoral immunity fails too. The genetics sort neatly by which lineages survive.

• T minus, B plus, NK minus. Mutation in IL2RG, the common gamma chain shared by the receptors for IL-2, IL-4, IL-7, IL-9, IL-15 and IL-21. X-linked, and the commonest form. JAK3 deficiency gives the same phenotype autosomally, because JAK3 signals downstream of that chain.
• T minus, B minus, NK plus. RAG1 or RAG2, or Artemis, that is DCLRE1C: the recombination machinery from the V(D)J lesson. No antigen receptor can be assembled, so neither lymphocyte lineage develops, while NK cells, which need no rearranged receptor, are spared.
• T minus, B minus, NK minus. Adenosine deaminase deficiency, in which accumulated deoxyadenosine metabolites are toxic to all lymphocytes.
• T minus, B plus, NK plus. IL7R mutation: IL-7 is the survival and expansion signal for developing T cells specifically.

The presentation is unglamorous and easy to miss: thrush that will not clear, chronic diarrhoea, failure to thrive, Pneumocystis pneumonia, absent thymic shadow on a chest film, and an absolute lymphocyte count below about 2,500 per microlitre in an infant, which is low for an age at which the normal range sits far above the adult one. Two hazards deserve their own line. A live vaccine such as rotavirus or BCG can cause disseminated disease. And unirradiated blood products can cause transfusion-associated graft-versus-host disease, because donor lymphocytes engraft in a host that cannot reject them.

Newborn screening changed the arithmetic. The T cell receptor excision circle assay counts a by-product of receptor rearrangement in dried blood spots; no T cells means no excision circles. Antonia Kwan and colleagues reported the first large series in 2014: across eleven United States programmes and just over 3 million newborns, 52 cases of SCID were found, an incidence near 1 in 58,000. The point of screening is timing. Transplantation before about three and a half months of age, and before an established infection, gives survival above 90 percent; after that the figures fall sharply.

Gene therapy history belongs here because it is honest. Early X-linked SCID trials used gammaretroviral vectors whose strong enhancer sequences integrated near proto-oncogenes, and several children developed T cell leukaemia driven by LMO2 activation. The field responded by moving to self-inactivating lentiviral vectors with the enhancer removed, and results in ADA-SCID and X-linked SCID are now good. Targeted editing at the disease locus rather than random integration is the current direction, and the nuclease chemistry that makes it possible is the subject of the molecular biology course in this catalogue.

Antibody failures: two that are often confused

X-linked agammaglobulinaemia, Bruton's disease, was traced in 1993 by David Vetrie and colleagues to the gene for a cytoplasmic tyrosine kinase, now called BTK. It is required for signalling from the pre-B cell receptor, so development arrests at the pre-B stage. Circulating B cells are under 2 percent of lymphocytes, all isotypes are low, T cells are normal, and tonsils and lymph nodes are strikingly small. The timing of onset is itself a teaching point: these boys are usually well until around six months, because maternal IgG transferred across the placenta protects them until it decays. Treatment is lifelong immunoglobulin replacement, and the residual risk that matters most is chronic enteroviral meningoencephalitis, since clearing enteroviruses depends on antibody.

Common variable immunodeficiency is the opposite kind of entity. B cells are present but fail to differentiate into plasma cells. IgG is low with low IgA or IgM, vaccine responses are poor, secondary causes have been excluded, and presentation is typically in the second to fourth decade. A monogenic cause is found in perhaps a tenth of patients, including TNFRSF13B, NFKB1, CTLA4 haploinsufficiency and LRBA deficiency. What distinguishes CVID clinically is that infection is only half the problem: bronchiectasis, granulomatous and lymphocytic interstitial lung disease, autoimmune cytopenias, enteropathy and a raised lymphoma risk drive much of the morbidity, and immunoglobulin replacement does not touch them.

So what?: CVID is a description, not a diagnosis. Treating it as a single disease is what leads to a patient on immunoglobulin whose lung disease and autoimmunity go unaddressed for a decade.

Two more, briefly, because their mechanisms were worked earlier in this course. Chronic granulomatous disease is the failure of oxidative killing after normal ingestion; management is cotrimoxazole and itraconazole prophylaxis with interferon gamma, and transplantation or gene therapy in severe cases. Complement deficiency splits by position in the cascade: early classical components leave apoptotic debris uncleared and cause lupus-like disease, C3 deficiency causes severe pyogenic infection, and terminal component or properdin deficiency causes recurrent Neisseria because membrane attack complex formation is how neisserial organisms are killed. Eculizumab produces exactly the same terminal defect pharmacologically, which is why patients on it are vaccinated against meningococcus and often given antibiotic prophylaxis.

HIV: a numerical argument about a steady state

HIV enters a cell by binding CD4 with gp120 and then a chemokine coreceptor. Transmitted viruses almost always use CCR5; CXCR4-using variants emerge later in some patients and mark a worse prognosis. The genetics are decisive: people homozygous for the 32 base pair deletion in CCR5 are highly resistant to infection, and the small number of people cured of HIV so far were cured by receiving haematopoietic stem cells from such donors, not by any drug.

For years the clinical latency of untreated infection was read as a slow virus doing little. In 1995 David Ho's group and, independently, Xiping Wei and George Shaw's, disproved that with an experiment that is really an exercise in kinetics. Give a potent protease inhibitor, then measure how fast plasma virus falls. The decay was exponential with a half-life near two days. Since the pre-treatment level had been stable, production must have equalled clearance, and back-calculation gave something on the order of 10⁹ to 10¹⁰ virions produced and cleared every day, with a comparable daily turnover of CD4 cells. The plateau was not quiescence. It was a fast equilibrium, and the gradual CD4 decline of 50 to 100 cells per microlitre per year is the slow loss of a race being run at speed. That reframing is why combination therapy, rather than sequential single drugs, became the design goal: at that replication rate every single-drug resistance mutation already exists in the patient before treatment starts.

Two other features matter clinically. Most of the body's memory CD4 cells live in gut-associated lymphoid tissue, and a large fraction of them are destroyed in the first weeks of infection, long before the blood count moves; the resulting epithelial damage lets microbial products into the circulation and drives the chronic immune activation that predicts progression better than viral load alone. And the CD4 count predicts which infection arrives: Pneumocystis pneumonia below about 200 cells per microlitre, toxoplasmosis and cryptococcal meningitis below 100, disseminated Mycobacterium avium complex and cytomegalovirus retinitis below 50.

Why is there no cure? Diana Finzi and Robert Siliciano answered that in 1997. In patients whose plasma virus had been undetectable on therapy for months, resting memory CD4 cells still carried integrated, transcriptionally silent, replication-competent provirus. That is the latent reservoir, and it is invisible to both arms of the problem. Antiretroviral drugs act on entry, reverse transcription, integration and maturation, and an already-integrated silent provirus does none of these things. Cytotoxic T cells kill cells that display viral peptide, and a silent provirus makes no protein to display. Measured decay half-lives of around 44 months imply well over sixty years of therapy to clear it, and stopping treatment allows rebound within weeks from that reservoir. Hence the strategies now under test: shock and kill, which tries to force expression so the cell can be recognised, and block and lock, which tries to make silencing permanent. Meanwhile the practical result of suppression is large: a person with a durably undetectable viral load does not transmit HIV sexually.

The upshot: Antiretroviral therapy converts a fatal infection into a managed one without ever touching the thing that makes it incurable. Knowing exactly which step of a life cycle a drug blocks tells you precisely what it cannot do.

Common misconceptions

• "Recurrent infection means immunodeficiency." Most recurrent infection in children is anatomical, allergic or environmental. The pattern that should prompt investigation is unusual organisms, unusual sites, unusual severity, or failure to clear.
• "A normal immunoglobulin level excludes an antibody problem." Specific antibody responses can fail with normal totals, which is why vaccine response testing exists.
• "Untreated HIV has a latent period in which little happens." Ho and Wei showed production and clearance on the order of 10 to the 9 virions a day throughout, and gut CD4 depletion occurs in the first weeks.
• "Antiretroviral therapy fails to cure because the drugs are not potent enough." Potency is not the issue. An integrated silent provirus presents no target to any of the drug classes, and no antigen to a T cell.
• "SCID infants look obviously ill from birth." They are typically well for weeks, protected by maternal IgG, which is exactly why newborn TREC screening rather than clinical suspicion is what gets them transplanted in time.
• "Immunoglobulin replacement treats CVID." It treats the infections. Bronchiectasis, interstitial lung disease, autoimmune cytopenias and lymphoma risk need separate attention.

Pulling it together

• Bruton's 1952 electrophoresis, showing an absent gamma globulin band in a boy with nineteen episodes of sepsis, founded the diagnostic method: the organism names the arm.
• Encapsulated bacteria point to antibody, opportunistic viruses and fungi to T cells, catalase-positive organisms and poor wound healing to phagocytes, and recurrent Neisseria to terminal complement.
• SCID subtypes follow the lineages that survive: IL2RG and JAK3 spare B cells, RAG and Artemis spare NK cells, ADA deficiency spares none, IL7R affects T cells alone.
• TREC newborn screening found SCID at about 1 in 58,000 and matters because transplantation before three and a half months and before infection gives survival above 90 percent.
• XLA is a BTK signalling arrest with under 2 percent B cells, presenting at about six months as maternal IgG wanes; CVID has B cells that fail to differentiate, and its autoimmune, granulomatous and malignant complications are not fixed by replacement.
• HIV maintains a high-turnover equilibrium of roughly 10 to the 9 virions a day, and the latent reservoir of integrated silent provirus in resting memory CD4 cells is invisible to both antiretroviral drugs and cytotoxic T cells, which is why suppression is not cure.

Sources

1. Bruton, O. C. (1952). Agammaglobulinemia. Pediatrics, 9(6), 722-728. pubmed.ncbi.nlm.nih.gov
2. Kwan, A., Abraham, R. S., Currier, R., Brower, A., Andruszewski, K., Abbott, J. K., et al. (2014). Newborn screening for severe combined immunodeficiency in 11 screening programs in the United States. JAMA, 312(7), 729-738. pubmed.ncbi.nlm.nih.gov
3. Ho, D. D., Neumann, A. U., Perelson, A. S., Chen, W., Leonard, J. M., & Markowitz, M. (1995). Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 infection. Nature, 373(6510), 123-126. pubmed.ncbi.nlm.nih.gov
4. Finzi, D., Hermankova, M., Pierson, T., Carruth, L. M., Buck, C., Chaisson, R. E., et al. (1997). Identification of a reservoir for HIV-1 in patients on highly active antiretroviral therapy. Science, 278(5341), 1295-1300. pubmed.ncbi.nlm.nih.gov
5. Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Inherited immunodeficiency diseases. In Immunobiology: The immune system in health and disease (5th ed.). Garland Science. ncbi.nlm.nih.gov

Two brothers in Boston, December 1954

On 23 December 1954, at the Peter Bent Brigham Hospital, Joseph Murray, John Merrill and Hartwell Harrison moved a kidney from Ronald Herrick into his twin brother Richard, who was dying of glomerulonephritis. Richard lived another eight years. No immunosuppressive drug was given, because there was nothing to suppress: the twins were genetically identical, and to Richard's immune system the graft was self. Murray shared the 1990 Nobel Prize for it.

That case defines the whole field negatively. Everything that has been built since, the typing laboratories, the crossmatch, the drugs, exists because donors are almost never identical twins. The problem is polymorphism, and specifically polymorphism at the HLA loci, the most variable genes in the human genome.

Why the alloresponse is so violent

Start with a number that ought to be surprising. The frequency of naive T cells responding to any given foreign peptide is around 1 in 10⁵ to 10⁶. The frequency responding to an allogeneic MHC molecule is between 1 and 10 percent of the entire T cell repertoire. Transplant rejection is the strongest T cell response the human immune system mounts, and it is directed at a molecule the recipient has never met.

The explanation follows from thymic selection. Every T cell that left the thymus was positively selected for weak binding to self MHC. An allogeneic MHC molecule is structurally similar enough to engage that same receptor, but it is loaded with thousands of different peptides, and a T cell receptor cross-reacting with allo-MHC plus almost any of those peptides will be triggered. Three recognition routes operate.

• Direct. Recipient T cells recognise intact donor MHC on donor antigen-presenting cells carried in with the graft. This is quantitatively enormous and drives early acute rejection; it fades as donor passenger leukocytes die out.
• Indirect. Recipient antigen-presenting cells process donor MHC into peptides and present them on self MHC, in the ordinary way. Slower, sustained, and the main driver of chronic rejection and of donor-specific antibody.
• Semi-direct. Recipient cells acquire intact donor MHC molecules by membrane transfer and display them.

Bottom line: The two timescales of rejection have two different recognition routes behind them. Direct recognition is a short, intense burst that dies with the passenger leukocytes. Indirect recognition never stops, which is why chronic rejection has been so much harder to solve than acute rejection.

The three, or rather four, ways a graft is lost

The C4d stain deserves a sentence, because it is a clever piece of forensics. Antibody bound to endothelium activates complement, and C4b is covalently deposited on the tissue. C4d is its inactivated fragment, which stays attached long after the antibody has gone. Finding C4d in peritubular capillaries is therefore a footprint: it says an antibody was here.

Keeping the antibody out: typing, PRA and the crossmatch

Hyperacute rejection is entirely preventable, and the paper that made it so was published in 1969. Rajinder Patel and Paul Terasaki tested recipient serum against donor lymphocytes in thirty transplants where the test was positive. Twenty-four of the thirty grafts failed immediately. In the crossmatch-negative group, immediate failure was rare. After that, no kidney was transplanted across a positive crossmatch, and the test became a condition of the operation rather than a research tool.

The modern workflow layers three things. ABO compatibility comes first, because anti-A and anti-B isohaemagglutinins are naturally occurring. HLA typing at A, B, C, DR, DQ and DP is now done by sequencing rather than serology, and mismatches are counted; matching still improves graft survival, though modern immunosuppression has narrowed the gap that existed in the 1980s. Then antibody screening: single-antigen bead assays detect antibodies against individual HLA molecules and generate a calculated panel-reactive antibody value, the percentage of donors in the population against whom the patient has antibody. A patient at 99 percent has almost no compatible donor, which is why paired exchange chains and desensitisation programmes exist. Because the antibody specificities are known, most centres now do a virtual crossmatch, comparing the recipient's antibody list with the donor's type before the physical test is run. The three sensitising events to ask about every time are pregnancy, transfusion and a previous transplant.

What matters here: Antibody and T cells are prevented by completely different means. Antibody is screened out before the operation by typing and crossmatching. T cells are suppressed afterwards with drugs. A programme that does one well and the other badly loses grafts at a characteristic time.

Graft-versus-host disease: the graft rejects the patient

Peter Medawar's group set the conditions. Graft-versus-host disease requires an immunologically competent graft, a host expressing antigens absent from the donor, and a host unable to reject the graft. Haematopoietic stem cell transplantation supplies all three by design. Acute disease targets three organs with unnerving consistency: a maculopapular rash starting on palms and soles, secretory diarrhoea, and a rising bilirubin from bile duct injury. Chronic disease looks like an autoimmune syndrome, with sclerodermatous skin, sicca symptoms and bronchiolitis obliterans.

The complication that makes this hard is that the same alloreactivity is therapeutic. Donor T cells that attack recipient tissue also attack residual leukaemia, which is the graft-versus-leukaemia effect. Deplete T cells from the graft and graft-versus-host disease falls while relapse rises. The clinical art is in separating the two, which is also why matching an HLA-identical sibling does not eliminate the problem: minor histocompatibility antigens, peptides from polymorphic non-HLA proteins such as the male-specific HY antigens and HA-1, are still presented on shared HLA molecules and are still seen.

The drugs, read as blocks on specific signals

Two of these repay a closer look. Mycophenolate is selective for a good reason: lymphocytes depend almost entirely on de novo purine synthesis and have little salvage capacity, so inhibiting IMPDH hits them harder than it hits other dividing cells. And the calcineurin inhibitors carry a genuine irony. The drugs that made kidney transplantation routine are themselves nephrotoxic, producing arteriolar hyalinosis and interstitial fibrosis over years in the organ they were given to protect.

Belatacept was designed around that irony. It is CTLA-4-Ig with two amino acid substitutions that raise its affinity for CD80 and CD86, and it works by the mechanism regulatory T cells use naturally: occupying the ligands so that CD28 cannot engage them, leaving the T cell with signal 1 and no signal 2. In the BENEFIT trial, reported by Flavio Vincenti and colleagues, belatacept-treated kidney recipients at three years had better glomerular filtration rates and less chronic histological damage than those on ciclosporin, but more early acute rejection, some of it severe. The lymphoproliferative risk in EBV-seronegative recipients is real enough that seronegativity is a contraindication, and it must be given as a monthly infusion, which makes adherence visible in a way that tablets do not.

Remember: Every immunosuppressive regimen is a bet on which failure the patient can better survive. Push harder and you buy fewer rejections at the price of cytomegalovirus, BK nephropathy, Pneumocystis and, over decades, squamous skin cancer at many times population rates. Push more lightly and you get donor-specific antibody and chronic rejection.

What tolerance would look like

The goal has been stated since 1953, when Rupert Billingham, Leslie Brent and Peter Medawar injected mouse embryos and newborns with cells from another strain, and found that as adults those animals accepted skin grafts from the donor strain indefinitely while rejecting third-party grafts normally. The tolerance was specific and it was durable, and it earned Medawar a share of the 1960 Nobel Prize with Frank Macfarlane Burnet.

Reproducing it in adults is the unsolved problem. Mixed chimerism protocols, in which donor bone marrow is given alongside the organ so that donor and recipient haematopoiesis coexist, have produced small numbers of patients off all immunosuppression years later. And there is a natural proof that adult tolerance is achievable: a pregnancy is a semi-allogeneic graft carried for nine months without rejection, maintained by trophoblast that expresses no classical class I or class II, by regulatory T cells specific for fetal antigen, and by local tryptophan catabolism.

Common misconceptions

• "A perfect HLA match removes the need for immunosuppression." Only identical twins qualify. HLA-identical siblings still reject, and still develop graft-versus-host disease after marrow transplant, because minor histocompatibility antigens are presented on the shared HLA molecules.
• "The crossmatch and HLA typing are the same test." Typing says what the donor has. The crossmatch asks whether this recipient's serum already contains antibody that binds those cells, which is the question Patel and Terasaki showed was decisive.
• "Rejection means the drugs were too weak." Antibody-mediated rejection often follows non-adherence, and chronic rejection is driven by indirect recognition that current drugs barely touch.
• "Calcineurin inhibitors protect the kidney." They protect the graft from rejection while being directly nephrotoxic, which is a large part of the motivation for belatacept.
• "Removing T cells from a marrow graft is straightforwardly good." It reduces graft-versus-host disease and increases leukaemic relapse, because the two effects share a mechanism.

What you now know

• The 1954 Herrick transplant succeeded without immunosuppression because the donor and recipient were identical twins, which states the problem the whole field addresses.
• Between 1 and 10 percent of T cells respond to a given allogeneic MHC molecule, because thymic selection leaves receptors that cross-react with allo-MHC carrying almost any peptide.
• Direct recognition drives early acute rejection and fades with the passenger leukocytes; indirect recognition persists and drives chronic rejection and donor-specific antibody.
• Hyperacute rejection is preformed antibody and is prevented by ABO matching and crossmatching, which Patel and Terasaki made mandatory in 1969 by showing 24 of 30 crossmatch-positive grafts failed at once.
• C4d in peritubular capillaries is a durable footprint of antibody-mediated injury.
• Calcineurin inhibitors block IL-2 transcription and are nephrotoxic; mTOR inhibitors block the response to IL-2; mycophenolate exploits lymphocyte dependence on de novo purine synthesis; belatacept blocks costimulation and trades early rejection and EBV-related lymphoproliferative risk for better long-term graft function.
• Graft-versus-host disease and the graft-versus-leukaemia effect are the same alloreactivity, which is why T cell depletion trades one for the other.

Sources

1. The Nobel Foundation. (1990). The Nobel Prize in Physiology or Medicine 1990: Joseph E. Murray and E. Donnall Thomas. nobelprize.org
2. Billingham, R. E., Brent, L., & Medawar, P. B. (1953). Actively acquired tolerance of foreign cells. Nature, 172(4379), 603-606. pubmed.ncbi.nlm.nih.gov
3. Patel, R., & Terasaki, P. I. (1969). Significance of the positive crossmatch test in kidney transplantation. New England Journal of Medicine, 280(14), 735-739. pubmed.ncbi.nlm.nih.gov
4. Vincenti, F., Larsen, C. P., Alberu, J., Bresnahan, B., Garcia, V. D., Kothari, J., et al. (2012). Three-year outcomes from BENEFIT, a randomized, active-controlled, parallel-group study in adult kidney transplant recipients. American Journal of Transplantation, 12(1), 210-217. pubmed.ncbi.nlm.nih.gov
5. Abbas, A. K., Lichtman, A. H., & Pillai, S. (2021). Transplantation immunology. In Cellular and Molecular Immunology (10th ed.). Elsevier.

A sarcoma that disappeared after an infection

In 1891 William Coley, a young surgeon at New York Hospital, went through old case notes looking for anything encouraging after a patient of his died of sarcoma. He found the record of Fred Stein, a house painter whose inoperable neck sarcoma had regressed completely in 1885 after two attacks of erysipelas, a streptococcal skin infection. Coley traced Stein to a tenement on the Lower East Side and found him alive, seven years on, with no tumour. He then began deliberately infecting sarcoma patients with streptococci, and later used a heat-killed mixture of Streptococcus pyogenes and Serratia marcescens that became known as Coley's toxins. Some patients had durable remissions. The results were irreproducible, the preparation was not standardised, radiotherapy arrived, and the whole idea was set aside for most of a century.

What Coley had stumbled on is now the core of a therapeutic field: a strong innate stimulus in the right place can license an anti-tumour response that was otherwise being held down. Getting from that observation to a licensed drug took until 2011, and the route runs through an argument about whether the immune system watches for cancer at all.

The immunosurveillance argument, and how it was settled

Frank Macfarlane Burnet and Lewis Thomas proposed in the late 1950s that lymphocytes continually eliminate transformed cells. The idea was widely rejected in the 1970s for an apparently decisive reason: athymic nude mice, which lack a thymus, did not develop more tumours than normal mice. That objection turned out to rest on a bad model. Nude mice retain natural killer cells and gamma delta T cells and are not the clean T cell knockout they were taken to be.

The settling experiment came from Robert Schreiber's laboratory in 2001. Vijay Shankaran and colleagues used mice lacking RAG2, which have no T, B or NKT cells at all, and mice unable to respond to interferon gamma. Both developed more methylcholanthrene-induced sarcomas and more spontaneous epithelial tumours than wild-type animals. Then came the result that gave the field its name. Sarcomas taken from immunodeficient mice were frequently rejected when transplanted into normal mice, while sarcomas taken from normal mice grew in both. The tumours that had grown up under immune pressure had been shaped by it. They were less immunogenic, because the immunogenic variants had already been removed.

That is immunoediting, and it runs in three phases. In elimination, transformed cells are recognised and destroyed. In equilibrium, a heterogeneous population is held in check for what may be years, with variants continually selected. In escape, a variant emerges that the immune system cannot see or cannot reach. Clinically the equilibrium phase is not an abstraction: donor-derived melanoma has appeared in recipients of organs from donors who had been tumour-free for more than a decade, which means occult cells had been held in check in the donor and were released by immunosuppression in the recipient.

Why this matters: The immune system does not merely fail to see a tumour. It has usually already interrogated it and selected for the variants it cannot see. Any therapy that lifts a brake is therefore acting on a population that has been pre-screened for invisibility, which is one reason response rates are what they are.

What is there to see?

Escape works by removing any step in the presentation and killing chain. Loss of beta-2 microglobulin or of the antigen processing transporters removes MHC class I entirely; mutations in JAK1 and JAK2 make the tumour blind to interferon gamma and so unable to upregulate presentation. Around that, the tumour builds a suppressive microenvironment: regulatory T cells, myeloid-derived suppressor cells, tumour-associated macrophages with a repair rather than a killing programme, TGF beta, IDO-mediated tryptophan depletion, and extracellular adenosine. And chronic antigen drives the exhausted T cell state described in the cytotoxicity lesson, with TOX-driven differentiation and rising inhibitory receptor expression.

Two brakes, two places, two Nobel halves

CTLA-4 was identified as a CD28 homologue that binds the same ligands, CD80 and CD86, with much higher affinity. James Allison's reading of it was that it is a brake rather than a second accelerator, and that removing the brake should let an existing anti-tumour response run. In 1996 Dana Leach, Matthew Krummel and Allison showed that an antibody against CTLA-4, given alone, caused rejection of established murine tumours. It is worth noticing what is unusual about that logic. The antibody is not aimed at the tumour at all. It is aimed at a regulatory molecule on the patient's own T cells.

PD-1 arrived from a different direction. Yasumasa Ishida, in Tasuku Honjo's laboratory, cloned it in 1992 from a dying T cell hybridoma and named it programmed death 1 on the assumption that it drove apoptosis. It does not. The function became clear in 1999 when Nishimura and Honjo found that PD-1 knockout mice develop lupus-like glomerulonephritis and, on another background, a dilated cardiomyopathy from anti-troponin antibody. It is an inhibitory receptor, and its cytoplasmic ITSM recruits the phosphatase SHP-2. Enfu Hui and colleagues showed in 2017 that the preferred substrate is not the T cell receptor complex but CD28, so PD-1 works largely by switching off costimulation.

The clinical difference between the two follows from where each acts.

• CTLA-4 acts early, in the lymph node, during priming, and also mediates suppression by regulatory T cells. Blocking it broadens the repertoire of responding clones. Ipilimumab was the first drug of any kind to improve overall survival in metastatic melanoma: in Stephen Hodi's 2010 trial, median survival was 10.1 months with ipilimumab against 6.4 months with a gp100 peptide vaccine alone, and a minority of patients have survived a decade.
• PD-1 acts late, in the tissue, on effector cells that have already arrived and met PD-L1 on tumour and stroma. Blocking it reinvigorates a response that is present but shut down, which is why the tumours that respond best are those with lymphocytes already inside them.

Allison and Honjo shared the 2018 Nobel Prize in Physiology or Medicine for this work. The toxicity profile is the proof of mechanism: releasing tolerance brakes produces autoimmunity, and immune-related adverse events include colitis, hypophysitis, thyroiditis, pneumonitis, hepatitis and, rarely and dangerously, myocarditis. They are more frequent and more severe with CTLA-4 blockade, consistent with acting at the priming stage. They are treated with corticosteroids and, when necessary, with the same biologics used in the corresponding autoimmune disease, which is a fairly direct demonstration that the tolerance mechanisms of the previous module were doing real work all along.

Mismatch repair deficiency: predicting response from a mutation count

Dung Le, Luis Diaz and colleagues at Johns Hopkins tested a specific prediction in 2015. If neoantigens are what checkpoint blockade uncovers, then tumours that cannot repair replication errors should accumulate many more of them and should respond better, whatever organ they arose in. They gave pembrolizumab to 41 patients in three cohorts.

• Mismatch-repair-deficient colorectal cancer: immune-related objective response 40 percent, and 78 percent free of progression at 20 weeks.
• Mismatch-repair-proficient colorectal cancer: response 0 percent, and 11 percent free of progression at 20 weeks.
• Mismatch-repair-deficient cancers that were not colorectal: response 71 percent.

Whole-exome sequencing gave the reason: a mean of 1,782 somatic mutations per tumour in the deficient group against 73 in the proficient group. In 2017 pembrolizumab became the first drug approved by the United States regulator for a molecular feature rather than for a tissue of origin. Mismatch repair deficiency arises either sporadically, usually by MLH1 promoter hypermethylation, or germline in Lynch syndrome; the repair chemistry itself, and the sequencing methods used to detect microsatellite instability, are covered in the molecular biology course in this catalogue.

The core of it: A biomarker works when it measures the thing the drug depends on. Mismatch repair status predicts response because it is a proxy for neoantigen count, and neoantigen count is what a released T cell has to recognise.

Building the receptor instead of finding it

A chimeric antigen receptor is a synthetic protein: a single-chain antibody fragment on the outside, a hinge and transmembrane segment, then a costimulatory domain from CD28 or 4-1BB, then the CD3 zeta signalling chain. The patient's T cells are collected, transduced, expanded and returned. Two features of the design matter. Recognition is by antibody, so it is MHC-independent, and the commonest escape route in solid tumours, loss of class I, does not work against it. And signal 1 and signal 2 are wired into one molecule, so no antigen-presenting cell is required.

Shannon Maude and colleagues reported the pivotal trial of tisagenlecleucel in 2018: 75 children and young adults with relapsed or refractory B cell acute lymphoblastic leukaemia, an overall remission rate of 81 percent within three months with every responder negative for minimal residual disease, and event-free survival of 50 percent at twelve months in a population that had run out of options. The toxicities are specific and severe. Cytokine release syndrome occurred in 77 percent, driven substantially by IL-6 from activated macrophages, and 48 percent needed tocilizumab, an IL-6 receptor antagonist that reverses it without ablating the T cells. Neurological events occurred in 40 percent. Because the target is CD19, a B cell lineage antigen, successful treatment produces prolonged B cell aplasia requiring immunoglobulin replacement: an acceptable trade only because that particular normal tissue is dispensable. Relapse with CD19-negative leukaemia is the characteristic failure, and it is immunoediting happening in months rather than years.

Solid tumours have resisted the approach for reasons worth naming: no comparable dispensable lineage antigen, poor trafficking into tumour, and a microenvironment that shuts the cells down after they arrive.

Common misconceptions

• "Checkpoint inhibitors attack the tumour." They bind receptors on the patient's own lymphocytes. Whether anything happens depends on whether an anti-tumour response already exists to be released.
• "Nude mice disproved immunosurveillance." They retain natural killer and gamma delta T cells. RAG2-deficient and interferon-gamma-insensitive mice gave the opposite answer.
• "PD-1 causes programmed cell death." It was named for a function it does not have. It is an inhibitory receptor whose SHP-2 recruitment acts preferentially on CD28.
• "CTLA-4 and PD-1 blockade are interchangeable." One acts on priming in the lymph node, the other on effector cells in tissue; their response patterns, kinetics and toxicity profiles differ accordingly.
• "A high tumour mutation burden guarantees a response." It shifts the odds. Antigen presentation must also be intact, T cells must reach the tumour, and the microenvironment must not shut them down.
• "CAR-T cells are a general cancer therapy." They transformed the treatment of B cell malignancies specifically, because CD19 is a uniform target on a tissue the body can do without.

The takeaway

• Coley's 1891 observation of sarcoma regression after erysipelas was the first evidence that an innate stimulus can license an anti-tumour response.
• Shankaran's 2001 experiments settled immunosurveillance: RAG2-deficient and interferon-gamma-insensitive mice grow more tumours, and tumours grown without immune pressure are rejected by normal mice.
• Immunoediting proceeds through elimination, equilibrium and escape, and escape usually means losing MHC class I, losing interferon gamma responsiveness, or building a suppressive microenvironment.
• CTLA-4 acts on priming in the lymph node; PD-1 acts on effector cells in tissue and works mainly by dephosphorylating CD28. Allison and Honjo shared the 2018 Nobel Prize.
• Immune-related adverse events are the mechanism made visible: removing tolerance brakes produces autoimmunity, more so with CTLA-4 blockade.
• Mismatch-repair-deficient tumours carried a mean of 1,782 somatic mutations against 73 in proficient tumours, and responded to pembrolizumab at 40 percent against 0 percent, which produced the first tissue-agnostic drug approval.
• CAR-T cells recognise antigen by antibody and are therefore MHC-independent; tisagenlecleucel produced 81 percent remission in refractory B cell leukaemia at the cost of cytokine release syndrome and permanent B cell aplasia.

Sources

1. Shankaran, V., Ikeda, H., Bruce, A. T., White, J. M., Swanson, P. E., Old, L. J., & Schreiber, R. D. (2001). IFNgamma and lymphocytes prevent primary tumour development and shape tumour immunogenicity. Nature, 410(6832), 1107-1111. pubmed.ncbi.nlm.nih.gov
2. Leach, D. R., Krummel, M. F., & Allison, J. P. (1996). Enhancement of antitumor immunity by CTLA-4 blockade. Science, 271(5256), 1734-1736. pubmed.ncbi.nlm.nih.gov
3. Le, D. T., Uram, J. N., Wang, H., Bartlett, B. R., Kemberling, H., Eyring, A. D., et al. (2015). PD-1 blockade in tumors with mismatch-repair deficiency. New England Journal of Medicine, 372(26), 2509-2520. pubmed.ncbi.nlm.nih.gov
4. Maude, S. L., Laetsch, T. W., Buechner, J., Rives, S., Boyer, M., Bittencourt, H., et al. (2018). Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia. New England Journal of Medicine, 378(5), 439-448. pubmed.ncbi.nlm.nih.gov
5. The Nobel Foundation. (2018). The Nobel Prize in Physiology or Medicine 2018: James P. Allison and Tasuku Honjo. nobelprize.org

One child in two hundred

In the early 1980s, roughly 20,000 cases of invasive Haemophilus influenzae type b disease occurred in the United States each year, almost all in children under five. About one child in 200 in that age group developed invasive Hib disease before their fifth birthday, meningitis accounted for around half of the cases, and about two thirds of them occurred before the age of eighteen months. Since the conjugate vaccines were licensed from 1987, incidence has fallen by more than 99 percent. Between 2009 and 2018 the entire United States reported 36 cases in children under five.

The interesting part is the failed attempt in between. A pure polysaccharide Hib vaccine was licensed in 1985 and withdrawn by 1988, because it did not protect the children who needed it. Understanding why is a direct application of the germinal centre lesson, and it is the cleanest illustration in the whole of vaccinology that a vaccine is not a dose of antigen but an instruction to a particular pathway.

Why conjugation works

The Hib capsule is polyribosyl-ribitol-phosphate, a repeating polysaccharide. Repeating polymers are thymus-independent type 2 antigens: they engage many B cell receptors at once, cross-linking them enough to activate the B cell without any T cell involvement. That gives IgM, some IgG2, no germinal centre, little affinity maturation and no memory. Worse, the response depends on marginal zone B cells and on complement receptor CD21, and both mature late, so children under about 18 to 24 months respond poorly or not at all. The vaccine worked in the children who were least likely to get the disease.

Conjugation solves it by changing the pathway rather than the dose. Couple the polysaccharide covalently to a carrier protein, and a B cell whose receptor binds the polysaccharide internalises the entire conjugate. It then processes the protein portion and presents carrier-derived peptides on MHC class II. A helper T cell specific for the carrier, not for the polysaccharide, engages it and supplies CD40 ligand and cytokines. From there the ordinary T-dependent machinery runs: germinal centres, somatic hypermutation, class switching to high-affinity IgG, plasma cells and memory B cells. Rachel Schneerson and John Robbins demonstrated the principle in 1980. The carriers actually used are tetanus toxoid, CRM197, a non-toxic point mutant of diphtheria toxin, and a meningococcal outer membrane protein complex.

The upshot: The antibody produced is still against the polysaccharide. What the carrier protein buys is not specificity but help, and with it every property a polysaccharide alone cannot deliver: affinity maturation, class switching, memory, and a response in a nine-month-old. The same trick underlies the pneumococcal and meningococcal conjugate vaccines.

The platforms, and the trade each one makes

Adjuvants: supplying the signal a purified protein does not have

A purified protein injected on its own is a poor immunogen, and Module 1 explains why. Dendritic cells upregulate costimulation in response to pattern recognition receptor engagement. A clean recombinant antigen engages none, so the T cell gets signal 1 without signal 2, which is the recipe for anergy rather than immunity. An adjuvant supplies the missing danger signal.

Aluminium salts have done that job since Alexander Glenny's work in 1926, and for most of that time nobody knew how. The old depot theory, that alum simply holds antigen at the injection site for slow release, has largely failed experimental test: removing the injection site nodule hours later does not abolish the response. Philippa Marrack, Amy McKee and Michael Munks reviewed the alternatives in 2009. Alum activates the NLRP3 inflammasome; it causes local cell death that releases host DNA and uric acid, which act as endogenous danger signals; and it recruits and activates inflammatory monocytes. It also skews the response towards TH2 and IgG1, which is why it suits vaccines that need antibody and is a poor choice when cytotoxic T cells are wanted.

Modern adjuvants are defined molecular agonists of the receptors from the first module, which makes them a satisfying closing of a loop.

• MF59 and AS03 are squalene oil-in-water emulsions used to boost influenza responses in older adults.
• AS04 combines alum with monophosphoryl lipid A, a chemically detoxified TLR4 agonist derived from Salmonella lipopolysaccharide. It is used in a human papillomavirus vaccine.
• AS01 adds the saponin QS-21 to monophosphoryl lipid A in liposomes. The recombinant zoster vaccine built on it reached about 97 percent efficacy against shingles in adults over 50 in its pivotal trial, against roughly 50 percent for the earlier live vaccine, and it also carries a noticeable reactogenicity.
• CpG 1018 is a synthetic oligodeoxynucleotide TLR9 agonist used in a hepatitis B vaccine that reaches seroprotection in two doses rather than three.

The mRNA platform, and the substitution that made it possible

The idea of injecting mRNA and letting the recipient make the antigen is old. It failed for a specific immunological reason: in vitro transcribed RNA is a potent stimulus for the very receptors that detect viral RNA. TLR3, TLR7 and TLR8 in endosomes and RIG-I in the cytosol all recognise it, and the resulting type I interferon shuts down translation through PKR and activates RNase L. The construct therefore produced a great deal of inflammation and very little protein, which is exactly backwards for a vaccine.

Katalin Kariko and Drew Weissman asked why the body's own RNA does not do this, and answered that natural RNA is chemically modified while in vitro transcripts are not. In 2005 they showed that substituting modified nucleosides found in natural RNA, among them pseudouridine, 5-methylcytidine and N6-methyladenosine, abolished activation of the RNA-sensing Toll-like receptors. In 2008 they showed that pseudouridine-containing mRNA is not merely quieter but better: more stable, and translated to substantially higher levels in vivo. The modification removes the innate alarm and raises protein output at the same time.

Several other pieces were needed: a capping strategy, optimised untranslated regions, a long poly-A tail, chromatographic removal of double-stranded RNA contaminants, and an ionisable lipid nanoparticle that protects the RNA, delivers it into cells and is itself mildly adjuvant. When SARS-CoV-2 was sequenced in January 2020, the platform was ready. Fernando Polack and colleagues reported BNT162b2 in December 2020 in 43,548 participants, with 8 cases of confirmed COVID-19 in the vaccine group against 162 in the placebo group, an efficacy of 95 percent. Kariko and Weissman shared the 2023 Nobel Prize in Physiology or Medicine.

Worth holding on to: The whole platform turned on making a molecule less immunogenic. A single nucleoside substitution converted mRNA from something the innate system attacks into something it ignores long enough to translate. That is a good general lesson about drug design against an immune system that evolved to detect exactly what you are trying to deliver.

The arithmetic of herd immunity

An infected person in a fully susceptible population infects R₀ others on average. To stop transmission you need the effective reproduction number below 1, so you need to remove a fraction of susceptibles large enough that each case produces less than one new case. If the immune fraction is H, transmission continues at R₀(1 minus H), and setting that below 1 gives the herd immunity threshold H greater than 1 minus 1 divided by R₀.

Work the measles row explicitly, because the correction for effectiveness is the step most often skipped. Take R₀ as 15. The threshold is 1 minus 1 over 15, which is 0.933. But a vaccine is not perfect: two doses of measles-containing vaccine confer immunity in about 97 percent of recipients. The coverage you must achieve is therefore 0.933 divided by 0.97, which is 0.962. Ninety-six percent coverage, not ninety-three, and a country sitting at 91 percent is not close to the threshold. It is below it.

One assumption in that derivation is worth naming, because it fails in practice. The formula assumes homogeneous mixing, that every person is equally likely to meet every other. Real populations cluster, and unvaccinated people cluster with each other, by school, by neighbourhood, by community. A country with 95 percent national coverage can contain communities at 60 percent, and an outbreak needs only one of those. Paul Fine's 1993 review remains the clearest account of what the concept does and does not license.

Key idea: Herd immunity is a property of a population's contact structure, not a property of an average. The national figure can be above the threshold while the places that matter are far below it.

What vaccines still cannot do

Three failures are instructive because each fails for a different reason. HIV has extreme sequence diversity, a glycan shield over the vulnerable epitopes on the envelope trimer, and, uniquely among the targets in this list, no example of a person who has cleared it naturally, so there is no protective response to copy. Malaria has a complex multi-stage life cycle and highly polymorphic surface antigens; RTS,S gives partial protection, on the order of a third of clinical episodes over four years, and R21 improves on it, which is real but far from what a conjugate vaccine achieves. And BCG, the most widely given vaccine in the world, protects reliably against disseminated tuberculosis in infancy and unreliably against adult pulmonary disease. A fourth limitation cuts across all of them: intramuscular vaccination generates circulating antibody and memory well but generates tissue-resident memory at mucosal surfaces poorly, which is exactly why respiratory vaccines reduce severe disease far more effectively than they reduce transmission, and why mucosal routes are being revisited.

Common misconceptions

• "A conjugate vaccine produces antibody against the carrier protein." It produces antibody against the polysaccharide. The carrier supplies T cell help by being processed and presented; the specificity of the useful antibody does not change.
• "Adjuvants just make the dose bigger." They supply the innate signal that a purified protein lacks, and the choice of adjuvant steers which kind of immunity you get.
• "mRNA vaccines alter your DNA." The mRNA is translated in the cytoplasm and degraded. There is no reverse transcriptase and no integrase in the construct, and it does not enter the nucleus.
• "Nucleoside modification was about safety." It was about function. Unmodified RNA triggered interferon that shut down translation, so the construct made too little protein to work.
• "If national coverage exceeds the herd immunity threshold, outbreaks cannot happen." The formula assumes homogeneous mixing. Clustered undervaccination produces outbreaks at high national coverage, and the effectiveness correction usually pushes the required figure several points higher than the raw threshold.
• "Natural infection always gives better immunity than vaccination." Sometimes it does and sometimes it does not, and the comparison ignores the cost of acquiring it. Tetanus is the clean counterexample: the disease does not immunise at all, because the lethal dose of toxin is smaller than the immunising dose.

Summing up

• Invasive Hib disease affected about one child in 200 under five and has fallen by more than 99 percent, but the pure polysaccharide vaccine failed because polysaccharides are T-independent antigens that infants respond to poorly.
• Conjugation to tetanus toxoid, CRM197 or a meningococcal outer membrane protein lets a polysaccharide-specific B cell present carrier peptides on MHC class II and recruit T cell help, converting the response into a T-dependent one with affinity maturation and memory.
• Platforms trade against each other: live vaccines give the strongest and most durable immunity but are unsafe in immunodeficiency; subunit vaccines are the safest and the least immunogenic without an adjuvant.
• Alum works through inflammasome activation and release of endogenous danger signals rather than by forming a depot, and it skews towards TH2 and antibody. AS01, AS04 and CpG 1018 are defined agonists of TLR4 and TLR9.
• Kariko and Weissman showed in 2005 that pseudouridine and related modifications stop mRNA triggering the RNA-sensing Toll-like receptors, and in 2008 that the modified message is more stable and better translated; BNT162b2 reached 95 percent efficacy in 2020.
• The herd immunity threshold is 1 minus 1 over R₀, divided by vaccine effectiveness to give required coverage: about 96 percent for measles, and the calculation assumes homogeneous mixing that real populations do not have.

Sources

1. Centers for Disease Control and Prevention. (2025). Haemophilus influenzae. In Epidemiology and Prevention of Vaccine-Preventable Diseases (Chapter 8). cdc.gov
2. Marrack, P., McKee, A. S., & Munks, M. W. (2009). Towards an understanding of the adjuvant action of aluminium. Nature Reviews Immunology, 9(4), 287-293. pubmed.ncbi.nlm.nih.gov
3. Kariko, K., Buckstein, M., Ni, H., & Weissman, D. (2005). Suppression of RNA recognition by Toll-like receptors: The impact of nucleoside modification and the evolutionary origin of RNA. Immunity, 23(2), 165-175. pubmed.ncbi.nlm.nih.gov
4. Polack, F. P., Thomas, S. J., Kitchin, N., Absalon, J., Gurtman, A., Lockhart, S., et al. (2020). Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine. New England Journal of Medicine, 383(27), 2603-2615. pubmed.ncbi.nlm.nih.gov
5. Fine, P. E. M. (1993). Herd immunity: History, theory, practice. Epidemiologic Reviews, 15(2), 265-302. pubmed.ncbi.nlm.nih.gov