Cell Signaling · Unit 7: Space, time and distance

Reverse Signaling: When the Ligand Is Also a Receptor

Many ligands are made as membrane proteins, and some start out as large as the receptors they bind. This lesson asks what the rest of the ligand is for and follows one answer, reverse signaling: when the receptor binds, a signal also passes into the cell that carries the ligand. You will be able to explain how B-type ephrins, GPI-anchored ephrins, membrane-bound TNF and a fly semaphorin do this, and to design an experiment that removes one direction of the signal and keeps the other.

Watch first A Grip That Signals Both Ways · 2:44

Epidermal growth factor is 53 amino acids long. The protein it is cut from is 1,207 amino acids long, and it spans the membrane. Why would a cell build something that large to release something that small? For EGF the answer is still incomplete. For several other ligands it is known, and it changes what the word “ligand” means: a ligand held in a membrane can receive a signal as well as give one.

A curiosity: the ligand starts out as large as its receptor

The EGF that Stanley Cohen purified (see the lesson on finding a signal) is a small peptide. When its messenger RNA was cloned from the mouse in 1983, the sequence predicted a protein more than twenty times longer. The human version has 1,207 amino acids. It begins with a signal peptide, the short stretch that sends a new protein into the secretory pathway. After that comes a long outer part. One segment crosses the membrane, and a tail lies in the cytosol. The 53 amino acids of EGF sit just outside the membrane, and proteases cut them free.

The precursor is about as large as the EGF receptor. The finished human EGF receptor has 1,186 amino acids, and the EGF precursor has 1,185 once its signal peptide is removed.

EGF is not a special case. Its relative TGF-alpha is 50 amino acids, cut from a membrane-spanning precursor of 160. Tumor necrosis factor (TNF) is made as a membrane protein of 233 amino acids, and a protease releases the outer 157. Fas ligand and the Notch ligand Delta are membrane proteins too. So a “ligand” can have a tail inside its own cell, and a tail in the cytosol is what you expect of a receptor.

What is Reverse Signaling? Both cells get a message

Picture two cells that touch. One carries a receptor. The other carries that receptor’s ligand, held in its membrane. When the two proteins bind, the receptor signals into its cell. Call this the forward signal. Reverse signaling is a second message, passing into the cell that carries the ligand. When both happen, the exchange is called bidirectional signaling.

A handshake is a fair picture. You cannot grip someone’s hand without your own being gripped. In biology the names were fixed by history: the protein with the kinase domain, or the one found first, was called the receptor.

A reverse signal needs two things. The ligand must stay attached to its cell, and it must have a route to the cytosol, through its own tail or through a neighbor in the same membrane. A useful hint when you meet a new ligand: if it stays in the membrane and its tail has changed little between species, suspect that the tail does something.

The processes affected make a long list: cell migration and proliferation, axon guidance, angiogenesis (the growth of new blood vessels), the immune response, apoptosis and development. Notch also needs two touching cells, but in the account you met in the Notch lesson the message runs one way, into the cell that carries Notch.

Ephrins: a tail that borrows a kinase

Ephrins are the clearest case. Their receptors, the Eph receptors, are the largest family of receptor tyrosine kinases, with 14 members in humans. Every ephrin is made attached to a membrane, in one of two ways. The five A-type ephrins are held by a GPI anchor, a lipid on the outer face of the membrane (see the lesson on lipid anchors). The three B-type ephrins cross the membrane once and end in a tail of 83 to 93 amino acids in the cytosol. The receptors fall into two matching classes, EphA and EphB, named for the ephrins they prefer, though some pairs cross the line: EphA4 also binds B-type ephrins.

The B-ephrin tail contains no enzyme. It has a few tyrosines whose positions are the same in all three B-type ephrins, and its last residues fit PDZ domains (see the lesson on protein interaction modules).

In 1996 and 1997 two laboratories, Tony Pawson’s in Toronto and Rüdiger Klein’s in Heidelberg, reported the key observation. When cells carrying a B-type ephrin met the outer part of an EphB receptor, the ephrin’s tail became phosphorylated on tyrosine. The receptor fragment had no kinase domain and the ephrin has none, so a kinase inside the ephrin’s own cell had to be responsible.

In 2002 Klein’s laboratory named it. Src-family kinases gather where B-type ephrins cluster, and they act as positive regulators of the tail’s phosphorylation. The same study found an off switch: a tyrosine phosphatase, PTP-BL, binds the PDZ-binding end of the tail and removes the phosphates after a delay. Mark Henkemeyer’s laboratory found that the adaptor Grb4 binds the phosphorylated tyrosines through its SH2 domain and connects the ephrin to proteins that reshape the actin skeleton.

The take-away fits in one line: “Ephrins use Src-family kinases to carry out reverse signaling."

"How could you signal via a GPI anchored protein?”

An A-type ephrin has no tail. It sits in the outer leaflet and never touches the cytosol. Try the question before you read on.

The answer the evidence supports is a partner. A-type ephrins collect in small patches of membrane rich in cholesterol, and Src-family kinases sit on the inner face of such patches, held there by their own lipid anchors. In 1999 Davy and colleagues reported that clustering ephrin-A5 with the outer part of an EphA receptor drew one of these kinases, Fyn, into the patches and changed how the cells adhered to their surroundings, a change that needed Fyn. Later work found partners that span the membrane. In the axons of the retina, ephrin-A associates with p75, a receptor for neurotrophins, and the reverse signal there needs p75.

Other GPI-anchored proteins solve the problem the same way. GFR-alpha-1 binds the growth factor GDNF and leaves the signaling to Ret, a kinase that spans the membrane. Which partner an A-type ephrin uses seems to depend on the cell, and how a protein in the outer leaflet switches on a kinase at the inner leaflet is not settled.

TNF and semaphorins: other ligands that listen

You met TNF as a ligand in the lessons on NF-kappa-B and on cell death. It is made as a membrane protein, three chains together, with a short tail of 35 amino acids in the cytosol. A protease called TACE can cut the outer part free. The part that stays in the membrane can act as a receptor: when a TNF receptor on another cell binds it, a signal passes into the TNF-bearing cell.

The tail has no enzyme activity. What the tails of several TNF-family ligands share is a short sequence of the kind that casein kinase 1 (CK1), a serine/threonine kinase, prefers. In membrane TNF a serine in the tail carries phosphate, and a 1999 study found that the phosphate was lost when a soluble TNF receptor bound. The take-away is worded with care: “CK1 Domain is likely an essential component in Reverse Signaling.” Note the word “likely”.

The clearest case in neurons comes from Alun Davies’s laboratory. Developing sympathetic neurons carry TNF on their axons, and the tissues those axons grow into make TNF receptor 1. In culture, a soluble form of that receptor made the axons grow and branch more, and the effect depended on TNF in the axon’s membrane. Mice lacking either TNF or the receptor had sparser sympathetic innervation in several tissues. Here the protein named “receptor” is the signal, and the protein named “ligand” receives it.

Semaphorins make the same point. The first was found in 1992 in the grasshopper embryo, as a membrane protein that guides growing axons. The family is now sorted into eight classes; some are secreted, some span the membrane, and one class is held by a GPI anchor. All share the sema domain, about 500 amino acids at the outer end: “Sema domain is highly conserved between all classes.” Their main receptors are the plexins.

In the fruit fly, the membrane-spanning semaphorin Sema-1a was first known as a ligand for plexin A. It also works the other way round. On the axons of the fly’s developing eye, Sema-1a receives a signal from plexin A that helps the axons hold to one another, and it needs its cytoplasmic part to do so. In motor axons, the tail of Sema-1a was found to bind two regulators of the small GTPase Rho, a GEF and a GAP (see the lesson on GTPase switches).

Reading a knockout: which direction did you remove?

Watch Which Direction Did You Remove? · 1:26

Suppose you delete the gene for a B-type ephrin and an organ fails to form. You have removed the forward signal into the receptor’s cell. You have also removed the reverse signal into the ephrin’s cell, and the plain grip that held the two cells together. The knockout cannot tell you which loss caused the defect. Deleting the receptor has the same problem, because its outer part was a ligand for the other cell.

Design problem

“How could we disrupt reverse signaling while leaving forward signaling intact?” Design the mouse, or the reagent, for an ephrin of the B class. Then say how you would do the opposite. Last, say what you would conclude if an animal with no ephrin at all had a defect that neither of your two animals had.

One way to do it

Remove the reverse signal. Change the ephrin’s gene so that the protein keeps its outer part and its membrane segment and loses its tail. Henkemeyer’s laboratory did this for ephrin-B2, putting the enzyme beta-galactosidase where the tail had been. The outer part can still bind and cluster EphB receptors on the next cell, so the forward signal survives. With no tail there is nothing for a Src-family kinase to phosphorylate and nothing for an adaptor to bind, so the reverse signal through the tail is gone. A finer version changes only the tail’s tyrosines to phenylalanine, or removes the last residue, which PDZ domains need. One control is required: show that the altered protein reaches the cell surface in normal amount. If it does not, you have weakened the forward signal as well.

The reagent. Add the outer part of an EphB receptor, fused to the Fc region of an antibody and clustered, to cells that carry the ephrin. This gives a reverse signal with no receptor cell present. The reagent asks whether a reverse signal is enough to cause an effect, and the mouse asks whether it is needed.

The opposite. Alter the receptor instead: keep its outer part and membrane segment, and remove or inactivate its kinase domain. Henkemeyer and colleagues made this mouse for EphB2 in 1996. Mice with no EphB2 had misrouted axons in a brain tract called the anterior commissure. Mice with the kinase-less receptor did not. The axons carried B-type ephrins and the cells beneath them carried the receptor, so the result pointed to a reverse signal into the axons. Since one ephrin binds several EphB receptors, you may need to alter more than one receptor gene.

The third animal. If the full knockout has a defect that neither altered animal shows, the defect needs neither the ephrin’s tail by itself nor the receptor’s kinase by itself. Either direction may be enough alone, or the cells may only need to be held together. A mouse carrying both alterations at once would tell these two apart.

How we measure it

Receptor-Fc fusion protein (a receptor body)

The outer part of a receptor is joined to the Fc region of an antibody and made as a soluble protein, which is often clustered further with an antibody against Fc. Added to cells that carry the membrane-bound ligand, it binds the ligand with no receptor cell present. Any response in those cells is a reverse signal.

Knock-in of a truncated protein

Gene targeting in the mouse replaces the normal gene with a version that lacks one part, such as the cytoplasmic tail, sometimes with beta-galactosidase in its place as a marker. The animal shows what the rest of the protein can do without the missing part. It needs a control showing that the altered protein reaches the cell surface.

Immunoprecipitation and phosphotyrosine blotting

The protein of interest is pulled out of a cell extract with an antibody, separated on a gel, and probed with an antibody that recognizes phosphotyrosine. A band shows that the protein carried phosphate on tyrosine when the cells were broken open.

Compartmented neuron culture

Neurons are grown in a device whose barriers let axons through and keep the fluid around the axons apart from the fluid around the cell bodies. A reagent added to one side reaches only that part of the neuron, which shows where on the cell a signal is received.

Finish line

Test yourself

8 questions

Practice and MCAT-style, with explained answers.

Next · Lesson 40

How Signaling Evolved

A human has 518 kinases and baker's yeast roughly 120 to 130. Where did the rest come from?

Continue →

← Lesson 38 · Exosomes and Extracellular Vesicles

EphB2 bound to ephrin-B2 — the outer parts that meet between two cells
EphB2 bound to ephrin-B2 — the outer parts that meet between two cells ● LIVE · RCSB PDB 1KGY
open on RCSB PDB ↗
Loading live structure 1KGY from RCSB PDB…

The receptor's binding domain grips a loop of the ephrin. The ephrin's tail and the receptor's kinase are not in the crystal: each lies on the far side of its own membrane.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
TNF — three chains together
TNF — three chains together ● LIVE · RCSB PDB 1TNF
open on RCSB PDB ↗
Loading live structure 1TNF from RCSB PDB…

Soluble TNF is a trimer. In the membrane-bound form each of the three chains continues through the membrane to a short tail in the cytosol.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
TNF with the outer part of TNF receptor 2 — the contact that can signal both ways
TNF with the outer part of TNF receptor 2 — the contact that can signal both ways ● LIVE · RCSB PDB 3ALQ
open on RCSB PDB ↗
Loading live structure 3ALQ from RCSB PDB…

Receptor chains sit in the grooves between TNF chains. The neuron experiments in this lesson used TNF receptor 1; this crystal contains receptor 2.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
A transmembrane semaphorin bound to a plexin — outer parts only
A transmembrane semaphorin bound to a plexin — outer parts only ● LIVE · RCSB PDB 3OKY
open on RCSB PDB ↗
Loading live structure 3OKY from RCSB PDB…

Each protein brings a sema domain, a propeller of seven blades. This is a class 6 semaphorin from the mouse; the fly Sema-1a in the lesson belongs to class 1. The tails that would reach into either cell are not in the crystal.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Mark Henkemeyer

Mark Henkemeyer

Made mice in which an Eph receptor, and later an ephrin, kept its outer part and lost its inner part. They showed that a receptor can do a job without its kinase, and took this to mean that it acts as a ligand for the ephrin on another cell.

As a postdoctoral fellow at the Samuel Lunenfeld Research Institute in Toronto, Henkemeyer made two lines of mice for an Eph receptor then called Nuk and now called EphB2. One line made no Nuk at all. The other made Nuk with its kinase domain and C-terminal tail replaced by the bacterial enzyme beta-galactosidase, so that the outer part still reached the cell surface. In the first line, most axons of one tract of the anterior commissure, a bundle that links the two sides of the brain, took the wrong path. In the second line they did not. The axons carried the membrane-spanning ligands for Nuk, and the cells beneath them carried Nuk. The simplest reading was that the receptor's outer part was acting as a ligand and the axons' ligands were acting as receptors. In his own laboratory at UT Southwestern in Dallas he turned the design around, replacing the tail of ephrin-B2 with beta-galactosidase, and the laboratory found that the adaptor Grb4 binds the phosphorylated tail. These mice gave the field a way to assign a defect to one direction of a two-way signal.

✓ verbatim Henkemeyer et al., Cell 86:35 (1996); Cowan & Henkemeyer, Nature 413:174 (2001).
Rüdiger Klein

Rüdiger Klein

His laboratory showed that the tail of a B-type ephrin is phosphorylated on tyrosine when an Eph receptor binds, and then that Src-family kinases add the phosphate and the phosphatase PTP-BL removes it.

Klein's laboratory, then at the European Molecular Biology Laboratory in Heidelberg, asked what the short tail of a membrane-spanning ephrin was for. In 1997, working with Elena Pasquale, the laboratory reported that the tail became phosphorylated on tyrosine when cells carrying the ephrin met the outer part of an Eph receptor. A paper from Tony Pawson's laboratory the year before had reported the same thing. The tail has no kinase of its own, so the next question was which kinase does the work. In 2002 Klein's laboratory answered it. Src-family kinases gather where B-type ephrins cluster and promote the phosphorylation, and a phosphatase called PTP-BL then binds the PDZ-binding end of the tail and removes the phosphates. Reverse signaling now had a named kinase and an off switch. The laboratory, by then at the Max Planck Institute of Neurobiology near Munich, went on to make mice with changes in single parts of the ephrin-B2 tail, so that each part could be tested in a living animal.

✓ verbatim Brückner, Pasquale & Klein, Science 275:1640 (1997); Palmer et al., Molecular Cell 9:725 (2002).
Alun Davies

Alun Davies

His laboratory showed that TNF on the axons of developing sympathetic neurons acts as a receptor, and that TNF receptor 1 in the target tissue acts as its ligand.

Davies's laboratory at Cardiff University studies how developing neurons grow into the tissues they supply. In 2013 Kisiswa and colleagues in the laboratory reported that the sympathetic neurons of young mice carry TNF on their axons while those axons are growing and branching in their target tissues, and that the target tissues make TNF receptor 1. In culture, a soluble form of TNF receptor 1 acted on the axons and made them grow and branch more. The effect depended on TNF held in the neuron's membrane. Mice lacking either TNF or TNF receptor 1 had fewer sympathetic fibers than normal in several tissues. The usual roles were swapped: the receptor in the target tissue was the signal, and the cytokine on the axon received it. Reverse signaling by the TNF family had been studied in immune cells. This work carried it into the developing nervous system and raised the question of which other members of the family act this way in neurons.

✓ verbatim Kisiswa et al., Nature Neuroscience 16:865 (2013).
Test yourself
Question 1 of 8 Score 0 / 0

Mature human EGF is 53 amino acids long. What is it cut from?
Sources & Integrity

Every structure, slide, simulation, and clip on this page is listed below with the source it came from. None of the 3D structures is drawn by hand or by AI: each one loads by accession ID from RCSB PDB or PubChem. This panel is generated from the lesson's own data file, so what it lists is what the page loads.

Molecular structures live data

  • EphB2 bound to ephrin-B2 — the outer parts that meet between two cells — RCSB PDB 1KGY · coordinates CC0 · Receptor ligand-binding domain with the ephrin's outer domain; Himanen et al., Nature 414:933 (2001)
  • TNF — three chains together — RCSB PDB 1TNF · coordinates CC0 · Soluble human TNF; Eck & Sprang, J. Biol. Chem. 264:17595 (1989)
  • TNF with the outer part of TNF receptor 2 — the contact that can signal both ways — RCSB PDB 3ALQ · coordinates CC0 · Mukai et al., Science Signaling 3:ra83 (2010)
  • A transmembrane semaphorin bound to a plexin — outer parts only — RCSB PDB 3OKY · coordinates CC0 · Mouse Sema6A with plexin A2; Janssen et al., Nature 467:1118 (2010)

Scientist profiles & quotes

  • Mark Henkemeyer — Henkemeyer et al., Cell 86:35 (1996); Cowan & Henkemeyer, Nature 413:174 (2001).
  • Rüdiger Klein — Brückner, Pasquale & Klein, Science 275:1640 (1997); Palmer et al., Molecular Cell 9:725 (2002).
  • Alun Davies — Kisiswa et al., Nature Neuroscience 16:865 (2013).

Quotes are verbatim from the cited source. Profile text is original to this site and was checked against the sources cited.

Animated explainer

A Grip That Signals Both Ways · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation: AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures, plots computed from their equations · captions from the narration script.

Section films

  • Which Direction Did You Remove? · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures · captions from the narration script.

Images & figures

This lesson carries no raster figures. Every structure is rendered live from RCSB PDB or PubChem, and every plot is drawn from its equation in the browser.

integrity Machines check the mechanics. The build confirms that every local file exists, that every question's answer key points at a real choice, and that any AI-generated scientific figure names the person who checked it; a weekly job re-resolves every accession ID against its database. Questions are original; MCAT-style items are written for this site and are not real AAMC questions. Banned sources (the Tymoczko textbook, publisher decks, copyrighted figures) are not used.

Like I'm five

When two friends shake hands, both of them feel the squeeze. Cells can touch like that too. One cell holds out a little flag, and the other cell grabs it with a little hand. The cell that grabs learns something, and the cell holding the flag learns something as well.
High school

Cells talk by making signal molecules called ligands, which fit into receptors on other cells. We usually picture a ligand as a message that floats from one cell to another. But many ligands stay stuck in the membrane of the cell that made them, with part of the protein reaching back inside that cell. When a receptor on a neighboring cell grabs such a ligand, the receptor's cell gets a message, called the forward signal. The ligand's cell gets one too, called the reverse signal. The two cells are having a two-way conversation, and the names "ligand" and "receptor" mostly record which protein was named first.
College

Many ligands are synthesized as membrane proteins. Some, such as EGF, are released by proteolysis from a much larger transmembrane precursor; others, such as the ephrins, act while still attached. A membrane-bound ligand can pass a signal into its own cell when its receptor binds, and this is reverse signaling. B-type ephrins are the best-understood case: binding and clustering by EphB receptors leads Src-family kinases to phosphorylate conserved tyrosines in the ephrin's short cytoplasmic tail, which then recruit SH2-domain adaptors such as Grb4, while the end of the tail binds PDZ-domain proteins. A-type ephrins are GPI-anchored and have no tail, so they rely on partners in the same membrane. Membrane-bound TNF and the transmembrane semaphorin Sema-1a also act as receptors. Because deleting a membrane ligand removes both directions at once, experiments that remove only the ligand's tail, or only the receptor's kinase, are needed to assign a phenotype to one direction.
Graduate

Eph receptors and ephrins assemble across the cell contact into 2:2 tetramers that pack into larger clusters, and clustering matters on both sides: soluble monomeric outer parts are poor agonists, and Fc fusions are usually clustered again before use. For B-type ephrins, Src-family kinases are recruited to the clusters and phosphorylate the tail quickly, and the PDZ-domain phosphatase PTP-BL is recruited with a delay and removes the phosphate. The authors of that work proposed a switch over time from phosphotyrosine-dependent output, which can run through the SH2 and SH3 adaptor Grb4, to PDZ-dependent output, which can run for example through PDZ-RGS3, a GTPase-activating protein for G-protein alpha subunits. Mouse alleles separate these parts: the tail replaced by beta-galactosidase, the tyrosines changed to phenylalanine, or the C-terminal valine removed. For the TNF family the proposed handle is a casein kinase 1 motif in the cytoplasmic tail; membrane TNF carries phosphate on serine at rest and was reported to lose it when soluble receptor binds. The tail of Sema-1a binds the Rho GEF Pebble and p190 RhoGAP.

One open question: how does a GPI-anchored ephrin, confined to the outer leaflet, activate Fyn at the inner leaflet? Transmembrane partners such as p75 answer it in some cells, but whether a patch of lipid can carry the signal across the bilayer without one is unresolved.
Named for a flag signal

The first semaphorin did not start with that name. In 1992 Alex Kolodkin and colleagues, in Corey Goodman's laboratory, described a membrane protein found on particular bundles of axons in the grasshopper embryo. They called it fasciclin IV, after the fascicles, or bundles, that it marked. The next year the same laboratory reported related genes in the fruit fly, in humans and in the genomes of pox viruses, and renamed the family the semaphorins. A semaphore is a way of sending messages with flags or movable arms. Some of these flags were later found to receive messages as well.

— Kolodkin et al., Neuron 9:831 (1992); Kolodkin, Matthes & Goodman, Cell 75:1389 (1993).
Two tail-less mice that disagreed

Ephrin-B2 is needed to build blood vessels in the mouse embryo. In 2001 Rüdiger Klein's laboratory reported a mouse that made ephrin-B2 without its tail. Its vessels failed much as they do with no ephrin-B2 at all, which pointed to the reverse signal. In 2004 Mark Henkemeyer's laboratory reported a mouse with beta-galactosidase in place of the tail, and its early vessels formed. That group had also made a plain tail-less protein, which stayed inside the cell and gave the same early failure, and it argued that tail-less proteins can be trapped this way, so a lost forward signal, not a lost reverse signal, could explain the first mouse. Klein's laboratory later found in cultured cells that its own protein did reach the surface, so the disagreement was not settled. Before you interpret an altered protein, find out where in the cell it is.

— Adams et al., Cell 104:57 (2001); Cowan et al., Developmental Biology 271:263 (2004); Mäkinen et al., Genes & Development 19:397 (2005).
A drug that is a receptor used as a ligand

Two kinds of drug block TNF. Infliximab is an antibody. Etanercept is the outer part of TNF receptor 2 joined to the Fc region of an antibody, which makes it a receptor body like the ones in this lesson. Both help in rheumatoid arthritis. In Crohn's disease infliximab works, and etanercept did not in a clinical trial. One proposed reason involves membrane TNF: the antibody binds TNF on immune cells in the gut wall, and a signal into those cells pushes them toward death. The proposal is still debated, and how well etanercept binds membrane TNF has been questioned.

— Eissner, Kolch & Scheurich, Cytokine & Growth Factor Reviews 15:353 (2004); Sandborn et al., Gastroenterology 121:1088 (2001).

 ↑ Top