Protein Modules: SH2, SH3, PTB, PDZ and PH

A receptor tyrosine kinase has just put phosphate on its own tyrosines. What reads those marks, and how does each reader pick its mark?

Ligand activates receptor… now what? A phosphotyrosine is only a mark, and a mark does nothing until some protein recognises it. This lesson is about those readers: small folded domains that grip a motif, which is a short stretch of amino acids a few residues long.

The problem: a phosphotyrosine is a mark, and the cell needs a reader

A phosphate on a tyrosine adds two negative charges to a small patch of a protein’s surface. By itself that changes little. What the cell needs is a coordinated message, many proteins arriving at the receptor together, and the phosphate is how the message is written.

It works as a handhold for a practical reason. Kinases put it on and phosphatases, the enzymes that take phosphate off, remove it within seconds. A docking site can therefore be switched on and off without the cell making or destroying any protein.

A reader has two jobs. It must grip tyrosine only when the tyrosine carries a phosphate, and it must tell one phosphotyrosine from the many others in the same cell. So where did we get such readers?

The idea: a module is a small fold that keeps working when you move it

A modular interaction domain is a small stretch of protein, from a few dozen to about 120 amino acids, that folds by itself and recognises a motif or a lipid. It keeps doing so when it is moved into a different protein. Think of a snap-on tool head: the handle can be anything.

One of the first was found through an experiment that looked like a failure. In 1986 Ivan Sadowski, a graduate student in Tony Pawson’s laboratory in Toronto, mutated a stretch of about a hundred amino acids just ahead of the kinase domain of a sarcoma-virus protein. The same stretch sits in the same place in Src, Abl and Fps, so it probably mattered, but nobody knew for what. The mutant kinase, made in bacteria, put phosphate on tyrosine in a test tube about as well as the original. In rat cells, it no longer made them cancerous, and its kinase activity there was low. An enzyme assay reports how fast phosphate moves, and a region whose whole contribution is to hold on to something else converts nothing. It had looked like packaging.

The lab named the region Src homology 2, or SH2, after the second conserved block in Src. By 1990 they had shown that SH2 domains from three proteins with little else in common, a phospholipase, a GTPase-activating protein and Src itself, each bind activated growth-factor receptors in vitro. Crystal structures in 1992 showed how. Arginine and lysine side chains, which carry positive charge, close around the phosphate on one tyrosine of the other protein. The grip holds until a phosphatase removes the phosphate, and then the two proteins part.

A virus then showed that a reader is a complete part. Bruce Mayer, in Hidesaburo Hanafusa’s laboratory at Rockefeller University, cloned the cancer gene of an avian sarcoma virus collected in 1931. Its protein, Crk, was an SH2 domain joined to a second module, SH3, with no enzyme anywhere in it. Crk does not itself put phosphate on tyrosine, yet cells it transforms carry more phosphotyrosine than normal cells. It reads phosphotyrosines and gathers the cell’s own enzymes next to them, outside the cell’s usual control.

How does the binding happen? Different modules are shaped for different targets:

Module	What it binds
SH2 (about 100 residues)	a phosphotyrosine and the residues just after it
PTB (phosphotyrosine-binding)	an NPXY motif, which some PTB domains need phosphorylated and others bind unphosphorylated; the Shc type reads the residues before the tyrosine
SH3 (about 60 residues)	a proline-rich motif with a PxxP core
WW	a proline-rich motif such as PPxY
PDZ	the last few residues at a protein’s C-terminal end, including its free carboxyl group
PH (about 120 residues)	phosphoinositides, the phosphorylated lipids of the membrane, for some PH domains
14-3-3	phosphoserine or phosphothreonine motifs; a whole small protein that works as a dimer

Many PH domains bind lipid weakly or not at all, and only a minority bind a particular phosphoinositide with real specificity. The one that matters for the design problem below is the PH domain of Akt, which binds PIP3.

Specificity: the residues after the phosphotyrosine choose the reader

Many tyrosines are phosphorylated in a stimulated cell, and around a hundred human proteins carry SH2 domains, so many kinds of SH2 domain are loose among them. A phosphate pocket alone cannot sort that crowd.

In 1992 Gabriel Waksman, a postdoctoral researcher in John Kuriyan’s laboratory at Rockefeller, crystallized the SH2 domain of v-Src with a short phosphopeptide. That structure and a 1993 follow-up with a tighter-binding peptide showed two grips on one small fold. One was a pocket clamped on the phosphotyrosine, with lysine and arginine side chains stacked against the ring and bonded to the phosphate. Beside it lay a second groove that holds the residues just after the tyrosine.

One domain holding one peptide cannot tell you whether that groove matters in a cell. If the pocket did most of the work, every SH2 domain would pick up almost any phosphoprotein, and sorting would have to happen elsewhere, in how much of each protein is made and where it sits. In 1993 Zhou Songyang, in Lewis Cantley’s laboratory in Boston, tested it. He made a mixture of many different short peptides, each with a phosphotyrosine at a fixed position and random residues just after it, and let purified SH2 domains from at least ten signaling proteins fish in it one at a time.

The Src-family domains (Src, Fyn, Lck and Fgr), screened separately, all came back with the same four residues: phosphotyrosine, glutamate, glutamate, isoleucine. Domains from p85, the regulatory subunit of PI 3-kinase, from phospholipase C-gamma and from the phosphatase SHPTP2 preferred phosphotyrosine followed by a hydrophobic residue, any residue, then another hydrophobic residue. Most other domains picked sequences of their own.

So the groove sets the specificity, and the groove travels with the domain. The PTB domain of Shc shows the same logic with the layout reversed: it reads an NPXY motif, and the residues before the phosphotyrosine do the choosing.

How tightly a domain holds its target follows the ordinary binding equation:

fraction of domain bound = [ligand] / (Kd + [ligand])

Kd is the ligand concentration at which half the domains are occupied. A modest Kd is a feature here, because a grip that is not too tight lets go soon after a phosphatase acts. Sequence is also only one filter. Where a protein sits, how much of it there is and when it appears all narrow the field further.

Src: a protein that reads itself

Bishop and Varmus won the 1989 Nobel Prize for showing that the cancer gene of a chicken sarcoma virus, v-Src, is a captured copy of a normal cell gene, c-Src. The two proteins differ in how well they stay quiet.

Reading from the amino end, c-Src has a fatty acid, myristate, attached to its second residue, which anchors it to the inner face of the plasma membrane. Then come an SH3 domain, an SH2 domain, a linker, the kinase domain and a short tail ending in tyrosine 527 (chicken numbering). The enzyme Csk phosphorylates that tyrosine. The tail folds back, the protein’s own SH2 domain grips the phosphotyrosine, and its own SH3 domain grips the linker. Both domains settle on the back of the kinase and hold it in an inactive shape.

The clamp is deliberately loose. The tail sequence after tyrosine 527 is a poor fit for the SH2 pocket compared with a sequence like pTyr-Glu-Glu-Ile, so a better ligand can pull the SH2 domain away. A phosphorylated PDGF receptor does this, and so does the kinase FAK once integrins have switched it on. A proline-rich partner such as beta-arrestin can take the SH3 domain, which is how an activated beta-adrenergic receptor reaches Src. Once the clamp opens, the kinase phosphorylates tyrosine 416 in its own activation loop and stays active.

Now v-Src makes sense. The viral protein has lost the tail and tyrosine 527 with it, along with some point mutations, so the SH2 domain has nothing to grip and the clamp never forms. Changing tyrosine 527 to phenylalanine in c-Src has the same effect and makes it transform cells. The viral form is always on because it has lost its off-switch.

The virtue of modularity: new pathways from old parts

How did we get such rapid expansion of receptor tyrosine kinases and signaling pathways? The human genome encodes 518 protein kinases, and 58 of them are receptor tyrosine kinases. A new pathway needs new connections, not new chemistry, and modules make a connection cheap. A reader spliced onto an enzyme that already works lets a receptor control it.

Phospholipase C-gamma shows how. It is a lipid-cutting enzyme that also carries two SH2 domains, an SH3 domain, a PH domain and a C2 domain. The SH2 domains are what let an activated receptor recruit it. Grb2, which the next lesson takes up, is little more than an SH3, an SH2 and a second SH3 in a row. Humans have around a hundred SH2-containing proteins; yeast, with no dedicated tyrosine kinases, has almost none. Shuffling readers onto old enzymes is a cheap way to wire a new pathway.

Design problem

Design a protein that binds PIP3 and an activated receptor tyrosine kinase. Then make it respond to calcium as well. PIP3 is a lipid made on the inner face of the membrane when PI 3-kinase phosphorylates PIP2. Say which modules you use, how you join them, and one experiment that shows each part is doing its job.

Try your design in the mixer before you open the answer.

Add a part

Your protein

clear

Or start from

The cell

receptor phosphorylated on tyrosine PI 3-kinase has made PIP3 calcium is high every grip is weak (coincidence detector)

A module keeps its job when you move it. Each reader grips only when its target is there, and an enzyme with no reader is not brought anywhere. This is a sketch of the logic, not a prediction for a real construct.

One way to do it

Fuse a PH domain that binds PIP3, such as the one from Akt, to an SH2 domain whose preferred sequence matches the phosphotyrosine motif on your receptor. Join them with a flexible linker long enough that both can reach the membrane at once. Each grip alone is weak, so the two support each other, and the protein stays only where a receptor tail and PIP3 are both within reach. It is a coincidence detector.

For calcium, add a Ca2+-dependent C2 domain, which binds acidic membrane lipids only when it has Ca2+ bound. Now all three inputs are needed before the protein stays at the membrane. A second route is to hide one of the grips behind an inhibitory segment that Ca2+ and calmodulin pull away, as a better ligand pulls the SH2 domain off the Src tail.

Test each part. A mutation in the arginine of the SH2 pocket, or an unstimulated receptor, should abolish recruitment. A PI 3-kinase inhibitor, or a mutation in the basic pocket of the PH domain, should abolish it too. Chelating Ca2+ with EGTA should abolish it as well. If removing any one input still leaves the protein at the membrane, the grips are not weak enough and the design needs a poorer SH2 or PH domain.