Receptor Tyrosine Kinases

Fifty-eight of your 518 protein kinases are built into the cell surface, with a growth-factor-binding half outside the cell and the kinase inside. Only one helix crosses the membrane between the two halves. So how does a growth factor that never enters the cell switch on an enzyme on the far side? The cell’s answer is to bring two receptors together.

The idea: one plan, built 58 times

A receptor tyrosine kinase (RTK) is a membrane receptor whose inner half is an enzyme that adds phosphate to tyrosine side chains. You met the architecture in the receptor field guide. Every RTK has three parts.

• The ectodomain is the part outside the cell. It binds the ligand.
• A single helix crosses the membrane. It anchors the receptor and joins its two halves.
• The kinase and its tail sit inside. The kinase is the two-lobed fold from the phosphorylation lesson, with ATP held in the cleft between the lobes. Past the kinase runs a flexible tail that carries tyrosines of its own.

The human genome encodes 518 protein kinases. About 90 are tyrosine kinases, and 58 of those are receptors, sorted into about 20 families that include the receptors for EGF, insulin, NGF, PDGF and VEGF. The kinase is recognizably the same fold in each one. What changes is the ectodomain and the tail.

The working unit is a pair. A lone receptor’s kinase stays off, or close to it, and the whole plan fits on one line:

two receptors side by side → kinases switch each other on → phosphotyrosines on the tails → SH2 proteins dock → several pathways

The dimerization question: three ways to make a pair

Growth factors are often dimeric, meaning built of two chains, and most receptors are monomeric, one chain each. How might ligands activate receptors? Cells use at least three answers.

A ligand that is itself a pair. NGF is two identical chains, and the pair binds two copies of its receptor, TrkA, one on each side. PDGF and VEGF are pairs too and work the same way.

A ligand that makes each receptor expose an arm. EGF is a single small chain, and one EGF binds one receptor, so EGF cannot be the bridge. At rest, the EGF receptor’s ectodomain is mostly folded over on itself, with one loop, the dimerization arm, tucked out of reach. EGF binds between two regions of the ectodomain and holds it extended with the arm out. Two extended receptors clasp each other by their arms, and in the crystal structure of the pair the two EGF molecules sit on the outer faces, far from the contact. Whatever holds the pair together belongs to the receptors themselves.

A receptor that is already a pair. The insulin receptor is built as two halves joined by disulfide bonds. Each half has an alpha chain outside and a beta chain that crosses the membrane and carries the kinase. Pairing cannot be its signal, because the pair is permanent. Insulin instead changes the shape of the pair, so that the two kinase-bearing legs, held apart at rest, come together.

All three routes end in the same place: two kinases held side by side. If contact between two kinases is the switch, contact supplied any other way should throw it too.

”Ligand activates receptor…now what?” The two kinases switch each other on

Recall the activation loop, the mobile strand beside a kinase’s active site. In the insulin receptor’s resting kinase the loop lies folded into the active site, and one of its own tyrosines sits where a substrate tyrosine should go. The kinase is plugged by a piece of itself.

When two kinases are held together, each phosphorylates the loop of the other. This is called autophosphorylation in trans: the receptor phosphorylates its own kind, but across the pair. With three phosphates on the insulin receptor’s loop, the loop stays clear of the active site, the two lobes close on ATP, and the kinase runs.

Most RTKs use this scheme, in order: in an FGF receptor followed step by step, a loop tyrosine was phosphorylated first, which raised the kinase’s rate, and the docking tyrosines followed.

The EGF receptor is the exception, although the picture borrowed from other receptors gave it the same mechanism. In 1987 Yosef Yarden and Joseph Schlessinger had shown that EGF switches on the receptor’s kinase through contact between receptor molecules, and by 1989 there was evidence that one receptor adds phosphate to another, so a phosphate on the activation loop was the expected switch. Experiments from 1992 on had already found that this loop phosphate is not needed, but they did not say what does the switching. Two results in 2006 pointed to an answer.

First, changing one amino acid in the loop, a leucine to an arginine, makes the isolated kinase about 20 times more active than normal, and it does not need a phosphate on the loop. Clinicians call this change L858R, and it is found in many lung cancers whose EGF receptor is always on. Second, in John Kuriyan’s laboratory, Xuewu Zhang took the normal kinase alone, with no ectodomain and no growth factor, and crowded it by gathering it on the surface of lipid vesicles. It switched on, again without needing the loop phosphate.

So the resting kinase holds itself off, and contact with a second copy releases the hold. Crystal structures from Kuriyan’s group, published in 2006, showed the contact, and the pair is lopsided. The large lobe of one kinase presses against the small lobe of the other and pushes that kinase’s active site into its working shape. A cyclin, the protein that switches on the kinases that drive cell division, works on its kinase in much the same way. The pressed copy is the one that switches on. The pressing copy only has to supply a surface.

The EGF receptor has three relatives, HER2, HER3 and HER4, and all four use the same contact, so they can work in mixed pairs. HER3’s own kinase has almost no activity, yet a pair of HER2 and HER3 signals strongly, because HER3 can do the pressing.

”RTK gets activated…what next?” The tail becomes a set of docking sites

Once on, the kinases phosphorylate tyrosines outside the kinase core, most of them on the two tails. Each of these phosphates is a docking site.

You know the reader of these marks from the lesson on interaction modules: an SH2 domain binds a phosphotyrosine and checks the few residues that follow it. So the docking-site sequence decides which target binds. Take the EGF receptor family.

• Phosphotyrosine 1068 on the EGF receptor’s tail has an asparagine two residues later. That suits the SH2 domain of Grb2, which leads to Ras and the MAP kinase cascade.
• Phosphotyrosine 992 on the same tail binds PLC-gamma, the enzyme that makes IP3 and DAG.
• Six tyrosines on HER3’s tail each have a methionine three residues later. That suits the SH2 domains of PI 3-kinase, which leads to Akt.

One activated pair therefore starts several pathways at once, and the choice of partner changes the mix. The PDGF receptor carries sites for all three proteins on a single chain, and there the sites were tested one at a time: change the tyrosine that serves one protein to phenylalanine, which cannot be phosphorylated, and that protein fails to dock while the others still do. PI 3-kinase uses two such tyrosines. The next three lessons follow the branches.

The same marks arrange the receptor’s removal. A phosphotyrosine recruits Cbl, an E3 ubiquitin ligase, which tags the receptor with ubiquitin. The receptor is taken into the cell by endocytosis, the pinching-in of a patch of membrane as a vesicle, and the ubiquitin tag routes it to the lysosome, the compartment that digests proteins, and not to the proteasome. Meanwhile protein tyrosine phosphatases (PTPs) strip the phosphates, and there are plenty of them: the human genome has 107 PTP genes, of which 38 are classical enzymes that act only on phosphotyrosine, set against about 90 protein tyrosine kinases (PTKs).

RTKs in cancer: three ways to stay on, two kinds of drug

“Nearly Every Kinase Has Been Associated with Disease”, and RTKs come up again and again in human tumours: the EGF receptor and its relatives, c-KIT, the VEGF receptors, c-MET and RET. If contact between kinases is the switch, a tumour can throw it without the ligand in at least three ways.

Make too much receptor. Gene amplification gives a cell extra copies of a gene. In 1987 Dennis Slamon and colleagues found the HER2 gene amplified in 30 percent of 189 breast tumours, and those patients relapsed sooner and survived for less time. HER2’s arm is exposed without any ligand, so crowding alone makes pairs.

Change the kinase. L858R sits in the kinase itself. The outside of the receptor is normal, but the kinase no longer waits for it.

Lose or replace the outside. In some thyroid cancers the RET gene breaks, and the half that encodes the kinase is joined to part of another gene whose protein forms pairs without help. The fused protein has no ectodomain at all.

Two kinds of drug answer these faults. An antibody is a large protein that binds one target tightly. It cannot cross the membrane, so it works on the ectodomain: trastuzumab (approved in 1998) binds HER2, and cetuximab (2004) binds the EGF receptor partly where EGF would and holds it folded. A small molecule can enter the cell and sit in the kinase’s ATP pocket. Gefitinib (2003) and erlotinib (2004) do this to the EGF receptor, and imatinib (2001), the Abl drug you have met, also blocks c-KIT. In 2004 three papers, by Lynch, Paez, Pao and their colleagues, reported that most lung cancers that shrank on gefitinib carried mutations in the EGF receptor’s kinase, L858R among them.

Why so many kinase drugs? “Highly Conserved— target for lots of drugs.” The ATP pocket is much the same across the kinase family, so one chemical strategy suits many kinases, and a molecule made for one pocket tends to fit others.

Design problem

A tumour cell must keep an RTK switched on with no growth factor present. Design three different ways to do it, each one changing a different part of the receptor or the amount of it. For each, predict whether an antibody against the outside of the receptor would still work, and whether a small molecule in the ATP pocket would.