Innate Immunity: Toll-like Receptors and NF-κB

How does a cell recognize an enemy it has never met? Antibodies take days to appear, and a fruit fly makes none at all. Yet a fly survives on rotting fruit, and your macrophages, the white blood cells that swallow microbes, begin to change their gene expression within minutes of touching a bacterium. This lesson follows that fast answer from a receptor at the membrane to a transcription factor, a protein that binds DNA and switches genes on, that was waiting in the cytoplasm the whole time.

The idea: detect what a microbe cannot give up

In 1989 the Yale immunologist Charles Janeway pointed at a puzzle in his own field. A purified protein injected alone provokes little immunity unless it is mixed with an adjuvant, an additive that provokes the immune system, and nobody knew why. His answer was a prediction: animals must carry receptors for “certain characteristics or patterns common on infectious agents but absent from the host”. The microbial features became known as pathogen-associated molecular patterns, and he called the proteins that bind them pattern-recognition receptors. The receptors he had in mind had not yet been identified as a class.

A receptor for adrenaline fits one molecule. A pattern receptor fits a whole class, chosen because the microbe cannot do without it: the lipid coat of a bacterium, or the double-stranded RNA of a virus. A bacterium that mutated its outer membrane away to escape detection would have no outer membrane. That is why a short list of receptors can do work for which antibodies need billions of variants.

A fly gene for back and belly

The evidence came from an animal with no antibodies. Toll was known in the fruit fly as a gene the mother uses to set the back-to-belly axis of her embryo. In 1996, in Jules Hoffmann’s laboratory in Strasbourg, Bruno Lemaitre tested adult flies with a disabled Toll pathway. They could not switch on the gene for drosomycin, a peptide that kills fungi, and they died after infection with the mold Aspergillus fumigatus. The relay inside the cell was mostly the one the embryo uses, and it ends at fly relatives of the mammalian transcription factor NF-κB.

A year later Ruslan Medzhitov and Janeway reported a human relative of Toll, which they had cloned, locked in the “on” state and expressed in human cells. NF-κB switched on, and inflammatory genes followed. What the human receptor detects was still unknown.

The receptor for endotoxin

That answer came from mice. Since the 1960s geneticists had known strains that barely react to lipopolysaccharide (LPS, or endotoxin), the lipid-and-sugar molecule that coats Gram-negative bacteria, the ones with an outer membrane. A dose that sends a normal mouse into shock leaves these strains nearly unaffected, and one gene makes the difference. Bruce Beutler’s laboratory found it by positional cloning, which means reading the genes in the stretch of chromosome that always travels with the trait. In 1998, with Alexander Poltorak as first author, they reported that the gene was Tlr4, a mouse relative of Toll. One resistant strain had a single amino acid changed in the part of the receptor inside the cell, and a second had no Tlr4 at all. Danielle Malo’s group reported the same gene early in 1999, and that year Shizuo Akira’s laboratory showed that mice with Tlr4 deleted barely respond to LPS. Akira’s laboratory and others then matched most of the rest of the family, ten receptors in humans, to their targets, largely by deleting one receptor gene at a time.

Receptor	What it detects	Where it sits
TLR4, with its partner MD-2	lipopolysaccharide	cell surface
TLR2, paired with TLR1 or TLR6	bacterial lipoproteins	cell surface
TLR5	flagellin, the protein of bacterial flagella	cell surface
TLR3	double-stranded RNA	endosome
TLR7 and TLR8	single-stranded RNA	endosome
TLR9	DNA with unmethylated CG sequences	endosome

Molecules from the outside of a microbe are read at the cell surface. Nucleic acids are exposed only after a microbe has been swallowed and broken open, so their receptors sit in endosomes, the compartments that hold what a cell has swallowed.

From receptor to gene: an answer held ready

A Toll-like receptor has a horseshoe of leucine-rich repeats outside the cell and a TIR domain inside, named for Toll and the interleukin-1 receptor, which share it. Ligand binding holds two receptors side by side and so pairs their TIR domains. The paired domains are a docking site for the adaptor MyD88. Every Toll-like receptor uses MyD88 except TLR3, which uses a different adaptor, TRIF, on a route that leads to interferon genes.

MyD88 has a TIR domain at one end and, at the other, a death domain, a small module that binds others of its kind. Those of MyD88 stack with those of the kinases IRAK4 and IRAK1 or IRAK2 into a helical tower. Inside it the IRAK4 kinases are close enough to phosphorylate one another and then their IRAK partners, and an activated IRAK binds the next protein, TRAF6.

TRAF6 is an E3 ubiquitin ligase. You know the rule from ubiquitin and proteolysis: a chain joined through lysine 48 sends a protein to the proteasome, and a chain joined through lysine 63 is a scaffold. In 2000 Zhijian Chen’s laboratory worked out this step in a test tube, purifying a missing factor from cell extracts. TRAF6, with a two-part enzyme called Ubc13/Uev1A, makes lysine-63 chains, and those chains switch on the IκB kinase complex, IKK. The chain is a platform that gathers a kinase called TAK1 next to IKK, and TAK1 phosphorylates the activation loop of IKK.

TLR pair → MyD88 → IRAK4 and IRAK1 → TRAF6 → lysine-63 chain → TAK1 → IKK → IκB destroyed → NF-κB enters the nucleus

NF-κB and its inhibitor

The target of IKK had been found years earlier in David Baltimore’s laboratory. In 1986 Ranjan Sen found a nuclear protein that bound a ten-base sequence in a control region of an antibody gene, the κ light chain, and the two named it NF-κB. In a cell line that lacked the activity, LPS made it appear, even with protein synthesis blocked. The protein must already have been there, inhibited. In 1988 Patrick Baeuerle found the inactive form in the cytoplasm, released it with a detergent, and named the inhibitor IκB.

NF-κB is a pair of Rel-family proteins. The family: Dorsal, Dif and Relish in the fly; RelA, RelB and c-Rel in mammals, along with p50 and p52. IκB proteins are built of ankyrin repeats, a common protein-binding fold, and they cover the signal that would send NF-κB into the nucleus. IKK adds two phosphates to IκB, a second E3 ligase adds a lysine-48 chain, and the proteasome destroys it. NF-κB, already folded, enters the nucleus. Nothing has to be transcribed or translated first, so it reaches its genes within minutes.

Ending the response

One gene that NF-κB switches on encodes IκBα, its own inhibitor. New IκBα enters the nucleus, takes NF-κB off the DNA and carries it back out. This is the negative feedback loop of signaling circuits, and under a lasting stimulus NF-κB can enter and leave the nucleus in waves.

”The anti-death, death pathway”: one receptor says die and stay alive

NF-κB earns this name from what the pathway does next. One of its target genes encodes tumor necrosis factor (TNF), a cytokine, which is a signaling protein that immune cells secrete. TNF binds TNF receptor 1 on the same cell and on its neighbors, and that receptor also signals to IKK and NF-κB, so the alarm spreads.

TNF receptor 1 is a death receptor: it can instruct a cell to kill itself. Usually it does not, because NF-κB switches on genes whose products block the death machinery. The receptor says “die” and “stay alive” at once, and the second message normally wins. In cultured cells whose NF-κB arm is blocked, for example by an IκB that cannot be phosphorylated, the same TNF kills. The death arm belongs to a later lesson.

A second alarm: nucleic acid in the wrong place

A virus has no cell wall to detect. It does have nucleic acid, and so does the cell.

Design problem

A virus is built from the same four nucleotides as the cell that it infects. Design a way for a cell to detect the nucleic acid of a virus and ignore its own, using only where you place the receptor and one or two chemical differences. Then say what you would expect to go wrong in a person whose own DNA leaks into the wrong compartment.

One way to do it

Start with location. Put receptors inside endosomes, facing inward. A cell keeps its own DNA in the nucleus and mitochondria, while a swallowed virus is taken apart in the endosome and its nucleic acid is exposed there. TLR3, TLR7, TLR8 and TLR9 sit here.

Then add chemistry. TLR9 binds DNA in which the sequence CG is unmethylated. That is common in bacterial DNA and in many viral DNAs and rare in yours, where CG is scarce and usually methylated. TLR3 binds long double-stranded RNA, which a healthy cell makes little of and many viruses make as they copy their genomes.

In the cytosol, location alone is enough, because DNA should never be there. The sensor described below binds DNA of any sequence.

The weak point is the assumption that your own DNA stays put. If it leaks into the cytosol, the sensor fires with no infection and the cell makes interferon without stopping. People who lack TREX1, a nuclease that clears DNA from the cytosol, develop an inflammatory disease, Aicardi-Goutières syndrome, that resembles a viral infection present from birth.

The sensor that links cytosolic DNA to interferon was purified in Zhijian Chen’s laboratory and reported in 2013. The enzyme, cGAS, binds double-stranded DNA and then joins one GTP and one ATP into a small ring, cyclic GMP-AMP, a second messenger of the kind you met with cAMP. The ring binds STING, a protein in the membrane of the endoplasmic reticulum. STING recruits a kinase that activates the transcription factor IRF3, which switches on interferon genes. Interferon then warns neighboring cells through the JAK-STAT pathway.

Too little and too much

Children born without working MyD88 or IRAK4 get severe, repeated infections with pus-forming bacteria such as Streptococcus pneumoniae and Staphylococcus aureus, often with little fever. They resist most other microbes, and the infections become rarer as they grow up.

Septic shock is the opposite failure. When Gram-negative bacteria reach the blood, macrophages throughout the body meet LPS at once. The TNF they release makes blood vessels widen and leak, so blood pressure falls and organs fail. Much of the damage is done by the host’s own response. The LPS-blind mice show both sides: endotoxin does not shock them, and infection with Gram-negative bacteria kills them far more easily.

In small, controlled amounts the alarm is useful. Monophosphoryl lipid A, a modified fragment of LPS that activates TLR4 with far less toxicity, has been used as part of the adjuvant in licensed vaccines against human papillomavirus and shingles. It is one answer to the question Janeway asked in 1989.