TGF-β and Smads

A protein was purified as the thing that makes cells grow like cancer cells, and it turned out to be one of the body’s main orders to stop dividing. Its name, transforming growth factor β (TGF-β), records the first job and hides the second. To see how one protein does both, follow the short relay that carries its message from the membrane to a gene.

A factor named for the wrong job

Around 1980 the laboratory test for a cancer-like cell was a dish of soft agar. A normal fibroblast (a connective-tissue cell) divides only when it is anchored to a surface. A transformed cell, one with the habits of a tumour cell, needs no anchor and grows into a colony while suspended in the gel. De Larco and Todaro had reported that cells transformed by a sarcoma virus release something that makes normal cells pass this test. The activity was called a transforming growth factor, and the name carried an assumption: this is what cancer cells make.

At the National Cancer Institute, Anita Roberts and Michael Sporn set out to purify it. In 1980 Roberts extracted the activity from cells transformed by a virus or by chemicals. In 1981 the same kind of extraction pulled an EGF-potentiated activity, which the authors called a new class of transforming growth factors, out of tissues of healthy adult mice, among them kidney, liver, heart and brain. Normal organs were making the “cancer” factor.

Then the protein was caught doing the opposite job. In 1984 Tucker, Shipley, Moses and Holley reported that a growth inhibitor from kidney epithelial cells (the cells that line surfaces such as skin and gut) was closely related to TGF-β. In 1985 Roberts, Sporn and colleagues called their purified protein “a bifunctional regulator of cellular growth”: it raised colonies in one assay and held cells back in another, depending on the cell and the conditions.

So the name describes one assay. What does TGF-β do to a cell once it lands, and where does “grow” or “stop” get chosen?

The idea: the ligand only introduces two kinases

The insulin and EGF receptors from the receptor field guide are tyrosine kinases. TGF-β receptors are receptor serine/threonine kinases: each crosses the membrane once, and its inner part is a kinase that puts phosphate on serine or threonine. Humans have five type II receptors and seven type I receptors, shared among more than thirty related ligands: the TGF-βs, the activins, the bone morphogenetic proteins (BMPs) and others.

TGF-β itself is a dimer, two chains joined by a disulfide bond. Here is what it does:

1. The dimer binds two type II receptors. The type II kinase is active before any ligand arrives, and it stays active.
2. The ligand and type II together recruit two type I receptors, so four receptor chains surround one ligand.
3. Type II phosphorylates type I on a stretch just inside the membrane called the GS domain, named for its glycines and serines.
4. The type I kinase, now active, phosphorylates the next protein in line, a Smad.

TGF-β + type II + type I → type II phosphorylates the GS domain of type I → type I phosphorylates a Smad

Think of type II as a lit match and type I as a fuse. The ligand’s only job is to hold the match to the fuse.

Jeffrey Wrana and Liliana Attisano worked this out in Joan Massagué’s laboratory between 1992 and 1994. First they showed that the type II receptor binds TGF-β but cannot signal without type I, and that type I cannot bind TGF-β without type II. Then they showed that the phosphate travels one way, from the kinase that never turns off to the kinase that cannot start alone.

What keeps the fuse from lighting by accident? In a resting type I receptor the unphosphorylated GS domain holds the kinase in an inactive shape, and a small protein called FKBP12 sits on top of it. Phosphorylation by type II ends that arrangement and turns the GS domain into a docking site for Smads.

The pairing also decides which ligand a cell can hear. Attisano’s 1993 paper found that different pairings of a type I with a type II receptor gave complexes with different signaling capacities, so a cell answers only the ligands for which it displays a working pair. (Some relatives, such as BMP2, grip type I more tightly than type II, but the phosphate still passes from type II to type I.)

The messenger is the transcription factor

In the cAMP and Ras pathways several proteins stand between the receptor and a gene. Here there is one. The protein that the type I receptor phosphorylates is itself a transcription factor, a protein that binds DNA, on its own or in a complex, and changes how often a gene is read.

These proteins are the Smads. The name joins two genes that geneticists found in animals: sma, from the worm C. elegans, whose mutants are small, and Mad, from the fruit fly. In 1996 Macías-Silva, Wrana, Attisano and colleagues, by then in Toronto, reported that the TGF-β receptor phosphorylates a human relative they called MADR2, now Smad2. Other laboratories, Massagué’s among them, reported related Smads the same year.

Humans have eight Smads in three groups:

• Receptor-activated Smads (R-Smads) are the receptor’s substrates. TGF-β and activin receptors phosphorylate Smad2 and Smad3. BMP receptors phosphorylate Smad1, Smad5 and Smad8.
• Smad4 is the partner that both branches share.
• Smad6 and Smad7 are inhibitory Smads, covered below with the brakes.

An R-Smad has two folded domains. The front one, MH1, binds DNA in most Smads. The back one, MH2, binds other proteins and ends in a short tail. The type I receptor phosphorylates two serines at the very end of that tail. A phosphorylated tail fits a pocket on a neighbouring MH2 domain, so R-Smads leave the receptor and assemble with Smad4, typically two R-Smads to one Smad4. The complex collects in the nucleus and binds DNA.

The STATs of the last lesson take the same shortcut, except that a STAT is phosphorylated on tyrosine by a separate kinase, a JAK, while a Smad is phosphorylated on serine by the receptor itself.

The BMP branch does much of its work in shaping the embryo, which belongs to the next unit.

Same signal, different answers

The MH1 domains of Smad3 and Smad4 read a DNA sequence only four base pairs long, and they hold it weakly. A sequence that short turns up by chance about once every 256 base pairs (4 × 4 × 4 × 4), which means millions of sites in a human genome. A Smad complex alone cannot pick out a gene.

It picks a gene with help. Smads bind firmly only next to partner transcription factors that read their own sequences nearby. Each cell type makes its own set of partners and keeps its own set of genes open for reading. The Smad complex carries one message, “TGF-β is here.” The partners decide what that means in this cell.

In many healthy epithelial cells it means “stop dividing.” Smad3 and Smad4, with their partners, switch on the genes for p15 and p21, two proteins that inhibit the cyclin-dependent kinases (the enzymes that push a cell through its division cycle). They also switch off the gene for Myc, a transcription factor that drives growth. In a fibroblast the same pathway switches on collagen genes, which is why TGF-β drives wound repair and, in excess, scarring.

This is also why the pathway has two faces in cancer, as Roberts and Lalage Wakefield set out in 2003. Early on, TGF-β acts as a tumour suppressor, because the cells still obey the stop order. Later the tumour cells no longer stop, and the TGF-β around them works in the tumour’s favour: it quiets nearby immune cells and can push tumour cells to loosen from their neighbours and move.

Design problem

TGF-β stops a healthy epithelial cell from dividing, yet many late tumours make large amounts of it. Propose two different changes in a tumour cell that would let it ignore the stop signal. For each, say whether the cell would also lose TGF-β’s other effects, and which change you would expect a tumour to gain from more.

Three brakes, and what breaks

A signal this strong has to be kept off most of the time.

The ligand is stored locked. TGF-β is secreted still wrapped in the front portion of its own precursor chain, which has been cut off but stays bound and covers the surfaces that receptors need. This latent complex is usually tethered to the extracellular matrix, the protein mesh between cells. Active TGF-β is released locally, for example when a cell’s integrins (adhesion receptors) pull on the wrapper or a protease cuts it.

The signal makes its own inhibitor. One of the genes that Smads switch on encodes Smad7, which binds the active type I receptor and keeps R-Smads away from it. This is negative feedback, as in the signaling circuits lesson. Smad6 does a similar job, with a preference for the BMP branch.

The parts are destroyed. Smad7 also brings ubiquitin ligases called Smurfs (Smad ubiquitination regulatory factors) to the receptor. Smurfs tag the receptor, and some Smads, with ubiquitin for destruction, as covered in the lesson on ubiquitin and proteolysis.

Pancreatic cancer. Smad4 was found in 1996, in Scott Kern’s laboratory, as a gene missing from pancreatic tumours, and was first named DPC4, for deleted in pancreatic cancer. It is inactivated in about half of pancreatic cancers.

Muscle size. Myostatin, a TGF-β relative made by muscle, tells muscle to stop growing. Mice made without the gene in Se-Jin Lee’s laboratory in 1997 have muscles two to three times the normal weight.

Bone in the wrong place. In fibrodysplasia ossificans progressiva, muscle and connective tissue progressively turn to bone. Nearly all patients with the classic form carry the same change of one amino acid, histidine for arginine at position 206, in a BMP type I receptor called ACVR1, as the laboratory of Eileen Shore and Frederick Kaplan reported in 2006. The change sits in the GS domain, the part of the receptor that keeps the kinase off until type II phosphorylates it.