Cell Biology · Unit 2: Membranes and compartments

Transport Across Membranes

A lipid bilayer stops ions and most polar molecules, so every cell relies on proteins to move them: channels, carriers and pumps. This lesson sorts those proteins by how they work, shows how to predict which way an ion will flow from its concentrations and the membrane voltage, and teaches you to read a single-channel recording and a tracer-uptake experiment. Afterwards you can compute a Nernst potential, decide the direction of passive ion flux, and tell a channel from a carrier from the data.

Watch first Voltage Can Reverse an Ion's Flow · 3:01

Hold a cell at −70 millivolts with 140 mM potassium inside and 5 mM outside, open potassium channels, and potassium leaves. Hold it at −100 millivolts and potassium comes in, against a 28-fold concentration difference.

The concentrations did not change between those two cases. Only the voltage did. That is the first thing this lesson will teach you to calculate. The second is how to tell what kind of protein is carrying an ion or a sugar across a membrane, using the two kinds of data that answer that question: a recording from one channel, and a count of labelled molecules.

The idea: a bilayer is a wall with designed doors

A lipid bilayer has an oily core about as thick as two lipid tails. Gases such as oxygen and carbon dioxide and small nonpolar molecules dissolve in that core and cross it by simple diffusion, needing no protein. Water crosses too, though slowly. Ions and larger polar molecules such as glucose cross almost not at all, because leaving water for oil costs them too much energy. You met the chemistry behind this in membrane structure and function.

So a cell is a walled town, and every door is a protein. The doors are specific: a potassium channel passes potassium and not sodium, and a glucose carrier passes glucose and not fructose. A cell decides what crosses by choosing which doors to build, where to put them and when to open them. The sidedness of those proteins in the membrane, covered in membrane organization and topology, is what makes a door point one way.

Doors, ferries and pumps

Membrane transport proteins fall into three groups by mechanism.

Channels form an aqueous pore that, when open, connects both sides at once. Ions flow through, often in single file, down their electrochemical gradient. A channel can switch between open and closed (gating) in response to voltage, a ligand or stretch, but while open it has no cycle to complete, so it is fast: millions of ions per second through one pore. Water channels, the aquaporins, are channels for water.

Carriers (also called transporters) are ferries. A carrier binds its solute on one side, changes shape, and releases it on the other; at no moment is a pathway open to both sides. This is alternating access. Because each crossing needs a shape change, a carrier works far more slowly than a channel, and because it has a limited number of binding sites, its rate saturates as solute rises, like an enzyme with Michaelis-Menten kinetics. A uniporter moves one solute, as GLUT1 moves glucose into red blood cells. A symporter moves two solutes the same way. An antiporter moves them opposite ways, as the Na+/Ca2+ exchanger brings sodium in and sends calcium out.

Pumps couple transport to a direct energy source, usually ATP hydrolysis, and so move ions uphill. This is primary active transport. The Na+/K+-ATPase in the plasma membrane of animal cells exports three Na+ and imports two K+ for each ATP it splits. It moves one net charge out per cycle, so it contributes a little to the voltage directly. Mostly it builds the sodium and potassium gradients that other proteins spend.

Spending a gradient is secondary active transport. SGLT1 in intestinal cells is a symporter that brings glucose in with sodium. Sodium runs down its steep inward gradient, and that downhill run pays for glucose to go uphill. The pump never touches glucose, but block the pump and glucose uptake fails as the sodium gradient runs down. You will see SGLT1 again in digestion, and the same logic with protons instead of sodium in the proton-motive force.

Two cautions. A protein’s family does not tell you its mechanism: CFTR, the chloride channel mutated in cystic fibrosis, belongs to the ABC family, most of whose members are ATP-driven transporters, yet it is a channel. And ligand-gated channels that act as receptors belong to ion channels as receptors in the signaling course.

Electrochemical driving force and the Nernst equation

An ion feels two pushes: its concentration difference, and the voltage across the membrane. Membrane potential, Vm, is defined as the inside minus the outside. A negative Vm pulls cations in and pushes anions out.

For each ion there is one voltage at which the two pushes balance. That is its equilibrium potential, given by the Nernst equation:

E_X = (RT / zF) × ln([X]out / [X]in)

R is the gas constant, T the absolute temperature, F the Faraday constant, and z the ion’s charge with its sign (+1 for K+, −1 for Cl−). At E_X the ion’s net passive flux is zero. Ions still cross; equal numbers cross each way.

To predict direction, compare Vm with E_X. The free energy change for moving one mole of the ion from outside to inside is:

ΔG_in = zF × (Vm − E_X)

If ΔG_in is negative, net passive flow is inward; if positive, outward. Keep z’s sign: for an anion the rule flips.

Worked problem. Take a membrane permeable only to K+, at 37 °C (310.15 K), with [K+]in = 140 mM and [K+]out = 5 mM. These are teaching values close to those of a typical mammalian cell, and you may use concentrations in place of activities.

Step 1: find E_K. RT/F = (8.314 × 310.15) / 96 485 = 0.02673 V = 26.73 mV. The ratio 5/140 = 0.0357, and ln(0.0357) = −3.332. So

E_K = 26.73 mV × (−3.332) = −89.1 mV

Using base-10 logs gives the same answer: 61.5 mV × log10(5/140) = 61.5 × (−1.447) = −89.0 mV, the small difference being rounding.

Step 2: clamp at −70 mV. Vm − E_K = −70 − (−89.1) = +19.1 mV. ΔG_in = (+1)(96 485)(+0.0191) = +1.84 kJ/mol. Moving K+ in is uphill, so net K+ flux is outward. The inside is not negative enough to hold potassium in against its concentration gradient.

Step 3: clamp at −100 mV. Vm − E_K = −10.9 mV, so ΔG_in = −1.06 kJ/mol, and net K+ flux is inward. This is the reversal promised at the top: the voltage now outweighs the concentration gradient.

Step 4: one channel. Suppose a single K+ channel has a conductance γ of 20 pS and behaves ohmically. At Vm = −40 mV its current is

i = γ × (Vm − E_K) = 20 pS × 49.1 mV = 0.98 pA, outward

Dividing by the charge on one ion (1.602 × 10^−19 C) gives about 6.1 × 10^6 ions per second. A carrier cycling hundreds or a few thousand times per second could not carry that. Rate is one way to tell a channel from a carrier.

Step 5: a shortcut to watch. Many books quote 58 mV per tenfold change in concentration. That figure holds near room temperature. At 37 °C it is 61.5 mV.

Do not conclude that a cell rests at E_K. A real membrane is permeable to several ions, and the resting potential lies between their equilibrium potentials, nearest the one with the highest permeability. Many cells rest less negative than E_K for exactly this reason.

Watching one channel and counting tracer

In 1976 Erwin Neher and Bert Sakmann recorded current from a small patch of membrane on denervated frog muscle and saw it switch between levels in square steps of a few picoamperes. Each step is a current through one open channel: a flow of millions of ions per second, not one ion. The 1981 gigaseal methods of Hamill, Neher, Sakmann and colleagues made such records clean enough to use routinely.

From a one-channel record you read two numbers. The unitary current, i, is the height of the step, set by the channel’s conductance and the driving force. The open probability, P_o, is the fraction of time the channel is open. A drug that changes P_o without changing i alters gating; one that changes i alters conduction or driving force. Across many identical channels, the mean current is I = N × i × P_o, so a whole-cell current alone cannot separate these three.

Single-channel record: step height and open time⚙ original · interactive
Unitary current i = γ × (Vm − EK)
Ions per second through the open channel
Direction of K+ flow
Fraction of time open: measured here, set
Mean current, I = N × i × Po: expected, from this record

The step height i comes from the driving force and sets which way ions move; the open probability Po is read from how long the channel stays open. In this sketch, Po is set separately from Vm; in a real voltage-gated channel Vm can change both. Assumes the lesson’s ohmic 20 pS K+ channel; EK = −89.1 mV (140 mM in, 5 mM out, 37 °C). Dwell times are drawn at random. Time is in relative units, so this record is a sketch of the logic, not a prediction for a real channel.

Orientation matters. In an inside-out patch, pulled off the cell, the cytoplasmic face meets the bath; in an outside-out patch, the outer face does. Applying a substance to the bath of each tells you which side it acts from. Bandwidth matters too: brief closures shorter than the recording can resolve vanish from the trace.

Carriers are slower and are usually studied by tracer flux: add radioactive substrate, stop at timed points, wash and count. Three traps follow.

First, tracer entry is not net uptake. A carrier can swap a labelled molecule outside for an unlabelled one inside, so label enters while total substrate stays put. In 2023 a study of the E. coli carrier DgoT found a mutant, E133Q, that kept robust substrate exchange although Glu133 is needed for net flux.

Second, saturation identifies a carrier. Initial rates fit v = Vmax[S] / (Km + [S]), levelling off at high substrate, while flux through a simple pore rises roughly in proportion to concentration over a wide range.

Third, a carrier’s Km is not a binding constant. It depends on the conditions on the far side, on the voltage and on any coupled ion, and different protocols give different values for the same protein.

Design problem

You inject RNA for an unknown membrane protein, Q, into Xenopus oocytes, and the eggs become more permeable to chloride. Oocytes have chloride channels of their own. Is Q itself a chloride pore, or a protein that switches on an endogenous channel? And how would you tell whether Q is a channel or a carrier?

One way to do it

Start with controls: compare Q-injected eggs with eggs injected with the same volume of water, from the same frog and on the same day, so that any change is due to Q.

Then take Q out of the oocyte. Purify it, put it into lipid vesicles and fuse them with a planar bilayer. If chloride current appears with purified Q and no other protein, Q is enough to form the pathway. This is how Christine Bear, John Riordan and colleagues showed in 1992 that purified CFTR is itself a regulated chloride channel.

Next, look for single-channel events: square current steps of fixed height whose reversal potential tracks E_Cl as you change chloride on one side. Then mutate residues you suspect line the pore. If a mutation changes the unitary conductance or which anions pass, Q forms the pore. A mutation that changes only P_o is consistent with Q being a regulator, or a channel whose gating you hit, so it decides less.

To separate a channel from a carrier, weigh rate (millions of ions per second points to a pore), saturation (a carrier levels off at low concentrations), discrete unitary events, and exchange (does chloride entry need another ion leaving?). Remember ClC-ec1: a family name will not answer this for you.

How we measure it

Patch clamp

A glass pipette sealed onto a cell isolates a small patch of membrane, and an amplifier records the current through it at a voltage you set. With one channel in the patch you see its openings as square steps; excised patches let you change the solution on the cytoplasmic face (inside-out) or the outer face (outside-out).

Radiotracer flux

You add a radioactively labelled substrate, stop uptake at timed points by filtering and washing, and count the label retained by cells or vesicles. Subtracting a matched control and dividing by specific activity gives an amount; a time series gives a rate. Tracer entry alone does not show net accumulation, since a carrier can exchange labelled for unlabelled substrate.

Reconstitution into vesicles and planar bilayers

A purified membrane protein is placed in artificial lipid vesicles, which can be assayed directly or fused with a flat bilayer for electrical recording. If the activity appears with the purified protein and nothing else, that protein is enough to carry it.

Expression in Xenopus oocytes

RNA injected into large frog eggs is translated into protein that reaches their surface. Comparing injected eggs with matched control eggs, by swelling in dilute solution or by electrical recording, reveals what the new protein adds.

Finish line

Test yourself

8 questions

Practice and MCAT-style, with explained answers.

Next · Lesson 6

Organelles as Distinct Environments

In HeLa cells the mitochondrial matrix measured near pH 8 and the Golgi near pH 6.6. How does one cell keep both?

Continue →

← Lesson 4 · Membrane Organization and Topology

Na+/K+-ATPase — the sodium pump Skou found
Na+/K+-ATPase — the sodium pump Skou found ● LIVE · RCSB PDB 3B8E
open on RCSB PDB ↗
Loading live structure 3B8E from RCSB PDB…

The large alpha subunit carries the ion sites within the membrane and the ATP-binding domains in the cytosol. Find the beta subunit beside it, and notice how far the phosphorylation site lies from the ion sites.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
KcsA potassium channel — a pore that picks potassium
KcsA potassium channel — a pore that picks potassium ● LIVE · RCSB PDB 1K4C
open on RCSB PDB ↗
Loading live structure 1K4C from RCSB PDB…

Four identical subunits surround a central pore. Look along the axis at the narrow selectivity filter near the outer end, where carbonyl oxygens line up to hold potassium ions in single file.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Aquaporin-1 — the water channel once called CHIP28
Aquaporin-1 — the water channel once called CHIP28 ● LIVE · RCSB PDB 1J4N
open on RCSB PDB ↗
Loading live structure 1J4N from RCSB PDB…

Each of the four subunits has its own narrow water pore. Find the constriction in the middle of one pore, which lets water through in single file while excluding protons.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
CFTR with ATP bound — a channel built like a pump
CFTR with ATP bound — a channel built like a pump ● LIVE · RCSB PDB 5W81
open on RCSB PDB ↗
Loading live structure 5W81 from RCSB PDB…

The two nucleotide-binding domains in the cytosol have come together around ATP. Follow the membrane helices to the outer side: in this structure the extracellular mouth of the pore is still closed.

Drag to rotate · scroll or pinch to zoom · right-drag to pan
Jens Skou

Jens Skou

Found an ATP-splitting enzyme in crab nerve membranes that needed sodium and potassium together for full activity, and suggested that it might be part of the active transport of those ions: the enzyme now called the Na+/K+-ATPase, or sodium pump.

By the 1950s physiologists knew that nerve and muscle cells push sodium out and take potassium in, against both gradients, and that the work needs metabolic energy. No one had found the machine. Skou, working in Aarhus, Denmark, looked for it among membrane enzymes that split ATP. In membrane preparations from crab nerve he found an ATPase whose activity needed magnesium and was stimulated when sodium and potassium were both present. An enzyme that needs both of the ions a pump moves, on the membrane where the pumping happens, was a strong candidate for the pump itself. He published the result in 1957 in Biochimica et Biophysica Acta and suggested that the enzyme might be part of the active transport system. Later work by many laboratories showed that the same enzyme, studied in many animal cells, couples the hydrolysis of one ATP to the export of three sodium ions and the import of two potassium ions. Skou shared the 1997 Nobel Prize in Chemistry. His enzyme made the sodium pump a protein that could be purified, inhibited and, much later, crystallized.

✓ verbatim Skou, Biochim. Biophys. Acta 23:394–401 (1957); Skou, Nobel Lecture in Chemistry, 8 December 1997.
Erwin Neher

Erwin Neher

With Bert Sakmann, recorded the current through individual ion channels in the membrane of denervated frog muscle (1976), and with colleagues developed the gigaseal methods of 1981 that made such recordings routine in many cell types.

Neher trained as a physicist. According to his Nobel autobiography, a Fulbright scholarship took him to the University of Wisconsin in 1966, and he later moved to neurobiology. He and Bert Sakmann began working together in Göttingen in the 1970s. Membrane currents measured across a whole cell were thought to be the sum of many small pores opening and closing, but the steps of a single pore were too small to see against the noise of a large membrane. Their answer was to record from a tiny area. They pressed the tip of a glass pipette against a denervated frog muscle fibre, where receptors for acetylcholine are spread over the surface, and recorded current from the patch under the tip. With an acetylcholine-like agonist in the pipette they saw square pulses of current a few picoamperes high, the openings of single channels. Their 1976 paper in Nature reported it. In 1981, with Owen Hamill, Alain Marty and Fred Sigworth, they described seals of 10^9 to 10^11 ohms that cut the noise and allowed excised patches and whole-cell recording. Neher and Sakmann shared the 1991 Nobel Prize in Physiology or Medicine.

✓ verbatim Neher & Sakmann, Nature 260:799–802 (1976); Hamill, Marty, Neher, Sakmann & Sigworth, Pflügers Arch. 391:85–100 (1981); Neher, Nobel Lecture, 9 December 1991.
Peter Agre

Peter Agre

With Gregory Preston, Thomas Carroll and William Guggino, showed that a red-cell membrane protein called CHIP28, expressed in frog oocytes, made them highly permeable to water: a water channel protein, later named aquaporin-1.

Water crosses a pure lipid bilayer slowly. Red blood cells and kidney tubule cells pass water far faster than that, and mercury compounds slow them down, which suggested a protein pore. Agre's laboratory at Johns Hopkins had found a membrane protein of about 28 kilodaltons in human red cells and named it CHIP28. Its function was unknown. The test was simple. Gregory Preston, Thomas Carroll, William Guggino and Agre injected RNA encoding CHIP28 into eggs of the frog Xenopus laevis, which are naturally poor at passing water, and compared them with control eggs. When moved into a dilute solution, the CHIP28 eggs took up water much faster, and mercuric chloride reversibly blocked the effect. They reported it in Science in 1992. CHIP28 was renamed aquaporin-1, and a family of related water channels has since been found in many organisms. Agre shared the 2003 Nobel Prize in Chemistry with Roderick MacKinnon. The work gave water transport a protein that could be studied by mutation and structure.

✓ verbatim Preston, Carroll, Guggino & Agre, Science 256:385–387 (1992); Agre, Nobel Lecture in Chemistry, 8 December 2003.
Christine Bear

Christine Bear

With John Riordan and colleagues, purified the cystic fibrosis protein CFTR and showed in artificial membranes that it forms a regulated chloride channel by itself, settling whether CFTR is a channel or a regulator of one.

The gene mutated in cystic fibrosis was identified in 1989, and it encoded a membrane protein named CFTR. Its sequence placed it in the ABC family, many of whose members use ATP to pump substances across membranes, while cells from people with cystic fibrosis were known to have defective chloride transport. Was CFTR the chloride channel, or a protein that controlled some other channel? Expressing it in a cell could not settle this, since any cell has its own channels. Bear, working with John Riordan's group, produced CFTR in insect cells, purified it and put it into lipid vesicles. Fusing the vesicles with a planar lipid bilayer gave chloride channels that opened when the protein had been phosphorylated and ATP was present. The 1992 paper in Cell made a purified protein the channel, with no other cell protein available to do the job. It established CFTR as an ABC protein that works as a channel rather than a pump.

✓ verbatim Bear, Li, Kartner, Bridges, Jensen, Ramjeesingh & Riordan, Cell 68:809–818 (1992).
Test yourself
Question 1 of 8 Score 0 / 0

In one cycle of its usual forward reaction, what does the Na+/K+-ATPase of an animal cell move?
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

  • Na+/K+-ATPase — the sodium pump Skou found — RCSB PDB 3B8E · coordinates CC0 · Pig kidney Na+/K+-ATPase, X-ray crystal structure (Morth et al., Nature, 2007)
  • KcsA potassium channel — a pore that picks potassium — RCSB PDB 1K4C · coordinates CC0 · Streptomyces lividans KcsA, high-resolution X-ray crystal structure (Zhou et al., Nature, 2001)
  • Aquaporin-1 — the water channel once called CHIP28 — RCSB PDB 1J4N · coordinates CC0 · Bovine aquaporin-1, X-ray crystal structure (Sui et al., Nature, 2001)
  • CFTR with ATP bound — a channel built like a pump — RCSB PDB 5W81 · coordinates CC0 · Phosphorylated, ATP-bound zebrafish CFTR, cryo-electron microscopy (Zhang, Liu & Chen, Cell, 2017)

Simulations

  • One channel, two numbers · Original interactive — Biochemistrypedia (Deppmann Lab).

Scientist profiles & quotes

  • Jens Skou — Skou, Biochim. Biophys. Acta 23:394–401 (1957); Skou, Nobel Lecture in Chemistry, 8 December 1997.
  • Erwin Neher — Neher & Sakmann, Nature 260:799–802 (1976); Hamill, Marty, Neher, Sakmann & Sigworth, Pflügers Arch. 391:85–100 (1981); Neher, Nobel Lecture, 9 December 1991.
  • Peter Agre — Preston, Carroll, Guggino & Agre, Science 256:385–387 (1992); Agre, Nobel Lecture in Chemistry, 8 December 2003.
  • Christine Bear — Bear, Li, Kartner, Bridges, Jensen, Ramjeesingh & Riordan, Cell 68:809–818 (1992).

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

Animated explainer

Voltage Can Reverse an Ion's Flow · 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.

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

A cell is wrapped in an oily skin that keeps salty and sugary things from leaking in or out. To let the right things through, the skin has special doors. Some doors are tunnels that stay open for a moment, some are revolving doors that carry one thing at a time, and some are little machines that push things the hard way, using energy.
High school

Small oily molecules and gases can slip straight through the lipid bilayer, but ions and sugars cannot. Proteins carry them. Channels are pores that let ions flow quickly when open. Carriers bind a molecule on one side and release it on the other. Pumps use ATP to push ions against their gradients, like the sodium pump that sends three sodium ions out and brings two potassium ions in. Ions move in response to both their concentration difference and the voltage across the membrane, so you have to consider both to predict which way they go.
College

Passive transport runs down an electrochemical gradient through channels (a continuous aqueous pore, millions of ions per second) or carriers (alternating access, far fewer cycles per second, saturable). Uniporters move one solute; symporters and antiporters couple two. Primary active transport uses ATP directly, as the Na+/K+-ATPase does (3 Na+ out, 2 K+ in). Secondary active transport uses an ion gradient, as SGLT1 uses Na+ to bring in glucose. For an ion X, E_X = (RT/zF) ln([X]out/[X]in), and the free energy of moving it inward is zF(Vm − E_X); negative favours influx. Single-channel recordings report unitary current and open probability separately, and tracer fluxes measure one direction at a time, so exchange can look like uptake.
Graduate

The Nernst relation gives the thermodynamic direction of net passive flux for one permeant species; it says nothing about rate, and real cells rest at a potential set by several permeabilities and by electrogenic pumps, so Vm need not equal any single E_X. Macroscopic current I = N i P_o confounds channel number, unitary current and occupancy; single-channel records separate the last two, within the limits of bandwidth, which can hide brief closures. For carriers, an apparent Km from initial-rate fits depends on trans conditions, membrane potential and coupling ions, and differs between zero-trans and equilibrium-exchange protocols, so it is not a binding constant. Family membership does not fix mechanism: E. coli ClC-ec1 is a H+/Cl− exchanger in a family that includes channels, and CFTR is an ABC protein that conducts chloride. An open question is how alternating-access carriers and gated pores grade into each other: what structural features make a protein that conducts in bursts rather than in strictly coupled cycles?
Discrete steps before the frog muscle

The 1976 frog-muscle recording is often called the first single-channel record, and that overstates it. In 1970 Stephen Hladky and Denis Haydon reported discrete, step-like conductance changes in artificial lipid membranes containing certain antibiotics that form small pores. What Neher and Sakmann added was a recording of single channels in a living cell's own membrane, made by a protein the cell itself had built. Both results matter. The fair way to state priority is to say which kind of membrane each group recorded from.

— Hladky & Haydon, Nature 225:451 (1970); Neher & Sakmann, Nature 260:799 (1976).
A chloride channel family with an exchanger in it

The ClC proteins were known as chloride channels, and the bacterial member ClC-ec1 from Escherichia coli became the model whose crystal structure guided work on the whole family. In 2004 Alessio Accardi and Christopher Miller tested it directly in reconstituted vesicles and found it was not a channel at all: it exchanged protons for chloride ions across the membrane. Similar sequence in one family does not decide whether a protein is a pore or a carrier. Only a functional test does.

— Accardi & Miller, Nature 427:803 (2004).
ATP closes a potassium channel from the inside

In cultured mouse pancreatic beta cells, Patrik Rorsman and Gerhard Trube recorded potassium channels in cell-attached patches and found that raising glucose to 20 mM suppressed their activity. Excising the patch inside-out let them put solutions directly on the cytoplasmic face, and 3 mM ATP there reversibly blocked the channels. The excised patch was the key: it let the authors test a metabolite on the side of the membrane it normally reaches from inside the cell.

— Rorsman & Trube, Pflügers Arch. 405:305 (1985).

 ↑ Top