Membranes: The Impermeable Sheet That Builds Itself

One protein, one disease

Cystic fibrosis comes down to a single protein in the wrong place. The gene is CFTR, a chloride channel that should sit in the membranes of cells lining the lungs and gut. In the most common mutation, ΔF508, one phenylalanine is deleted, the protein misfolds, and it is destroyed in the endoplasmic reticulum before it ever reaches the surface. The channel is not broken at the membrane. It never arrives. Everything downstream — the thick mucus, the chronic infections, the shortened life — follows from that one missing protein.

That is the throughline of this chapter, because it forces a useful question. To understand why one absent channel can reorganize an entire physiology, you first have to understand what a membrane is and what it does.

Physically, a membrane is thin and self-organizing. It runs about 7 nanometers thick (roughly 70 Angstroms): two leaflets, one lipid thick in each, with the tails meeting in the middle. For scale, a ribosome is about 25 nm across, so a membrane is roughly a quarter of that. It always forms closed compartments with no free edges, and if you puncture it, it reseals, because noncovalent forces hold it together rather than covalent bonds.

Functionally, four properties matter. The membrane is asymmetric: the outer leaflet differs from the inner one. Phosphatidylserine, for instance, stays on the inside until a cell dies, when flipping it outward becomes an “eat me” signal for the immune system. The membrane is fluid, its lipids drifting laterally. It is selectively permeable, letting oxygen through while blocking sodium. And it is electrically polarized, the inside sitting around 60 to 70 millivolts negative relative to the outside. A sheet of plastic would have none of these.

A membrane that builds itself

Phospholipids form bilayers spontaneously, and the energy comes from water, not ATP. Each lipid has a polar head and two greasy tails. Drop them in water and the tails cluster to escape it, the heads face out, and a bilayer (or a sealed sphere, a liposome) forms on its own. The driving force is the hydrophobic effect, and it is entropic: when the tails huddle together, the water that had been caging them is released into disorder. The cell spends nothing to assemble its boundary.

This self-assembly is also useful. Liposomes carry drugs. Load a water-soluble drug in the aqueous core, a fat-soluble one in the bilayer, and you have a delivery vehicle. Doxil packages the chemotherapy drug doxorubicin in liposomes and spares the heart much of its toxicity. The COVID-19 mRNA vaccines used lipid nanoparticles to protect fragile mRNA long enough to deliver it.

The same physics that builds the membrane makes it nearly impermeable to ions, and the reason is desolvation. An ion in water wears a tight shell of water molecules. To cross the oily membrane core it would have to shed that shell, and stripping water off an ion costs roughly 40 to 80 kJ/mol. What blocks the ion is its charge, not its size: it travels wrapped in water it must leave behind. Consider indole and tryptophan: nearly the same aromatic ring, but tryptophan carries charged groups and crosses orders of magnitude more slowly. The ring crosses fine; the charge is the toll.

Stack everything up and permeability spans about twelve orders of magnitude. Water crosses a membrane roughly 10 billion times faster than sodium does. The order runs gases, then water, then small polar molecules, then ions far at the bottom.

Water has one official exception. Aquaporins are channels that pass about 3 billion (10⁹) water molecules per second through a pore just 2.8 Angstroms wide, single file. The remarkable part is what they keep out. Protons normally race through water by hopping along hydrogen bonds, but an asparagine in the pore breaks that chain, and a positively charged arginine repels any proton that gets close. The channel passes water in a torrent and still excludes the proton. In the kidney, aquaporin-2 is switched on by the hormone vasopressin; when that regulation fails, the kidneys cannot concentrate urine and a person can pass up to about 20 liters of dilute urine a day, the condition called diabetes insipidus.

Tuning the fluidity

A membrane has to stay fluid, and the cell tunes that fluidity with chemistry. The melting temperature, Tm, is the point where half the lipids have melted from a tight gel into a fluid state. Two features set it. Longer fatty-acid tails pack more snugly and raise Tm. A cis double bond puts a roughly 30-degree kink in the tail that prevents tight packing and lowers Tm. Saturated tails are straight sticks that stack neatly; unsaturated tails have a permanent bend that keeps them loose. Trans fats straighten the tail back out, so they pack like a saturated fat despite being technically unsaturated.

Cholesterol is the membrane’s thermostat, and it works in both directions. Its rigid four-ring body jams between lipids and stops them from packing too tightly when the cell gets cold, keeping the membrane from freezing solid. Its flexible tail fills gaps and stops the membrane from getting too runny when the cell warms up. One molecule buffers against both rigidity and excess fluidity. This may also bear on why statins, which block the mevalonate pathway, can cause muscle pain in some people: that pathway makes not only cholesterol but isoprenoids and coenzyme Q10 that muscle cells depend on. The mechanism is still debated.

Cholesterol and sphingolipids also cluster into lipid rafts, small dynamic patches about 50 to 200 nm across. Rafts act as gathering points where signaling proteins find their partners. HIV and influenza both exploit rafts to assemble and bud.

The proteins in the sheet

Most of what a membrane does, it does through protein. Most membrane proteins are integral, embedded in the bilayer; the rest are peripheral, perched on the surface. The test for telling them apart is operational: peripheral proteins come off with high salt or a pH change, while integral proteins release only when detergent dissolves the membrane around them.

Proteins cross the membrane in a few stereotyped ways. An alpha-helix needs about 20 hydrophobic residues to span the bilayer once, and multi-pass proteins string several together. A beta-barrel rolls a beta sheet into a tube, found in outer membranes. Some proteins skip spanning entirely and attach by a lipid anchor: a GPI anchor on the outside, a prenyl or fatty-acyl group on the inside. The signaling protein Ras, for example, is prenylated so it can dock at the inner membrane.

One clinically loaded example: the enzyme COX-1 sits embedded in the membrane and pulls arachidonic acid in through a hydrophobic channel to make prostaglandins. Aspirin works by reaching into that channel and irreversibly acetylating the active site. Because the block is permanent, it lasts until the cell makes fresh enzyme, which is the basis of low-dose aspirin’s lasting antiplatelet effect. Ibuprofen competes for the same site reversibly, so taking ibuprofen before aspirin can actually shield the site and blunt aspirin’s heart protection.

What stays put, what moves

Membrane proteins and lipids move, and FRAP lets us watch them do it. Fluorescence recovery after photobleaching means exactly what it says: tag the membrane with fluorescent dye, bleach a small spot dark with a laser, and watch unbleached molecules diffuse back in. How fast the spot recovers gives the diffusion coefficient. How completely it recovers tells you the mobile fraction; if recovery stalls at 50%, half the labeled molecules are anchored to the cytoskeleton and cannot move.

The two kinds of motion are not equal. Lateral diffusion is fast: a lipid can travel across a whole cell in seconds. Transverse motion, a lipid flipping from one leaflet to the other, is glacially slow without a dedicated enzyme called a flippase. That asymmetry of rates is what preserves the membrane’s chemical asymmetry. Phosphatidylserine stays inside precisely because flip-flop almost never happens on its own.

Crossing on purpose: pumps, channels, and the selectivity trick

When something charged or large has to cross, a protein does the work, and three questions sort out how. Does it need a protein at all? If so, is it moving down its gradient or against it? And if against, does ATP power it directly or indirectly? The answers define three machines. Channels are highways: fast, around a million ions per second, but passive, moving things only downhill. Carriers are slower and regulated, changing shape to ferry their cargo. Pumps are escalators: slow, around a hundred cycles per second, but they spend ATP to push against the gradient.

The flagship pump is the Na⁺-K⁺ ATPase, a P-type ATPase that ejects 3 sodium ions and imports 2 potassium ions per ATP. It runs constantly and burns a large fraction of your resting ATP — on the order of a quarter, and up to half in neurons — to maintain the gradients that almost everything else depends on. The “P” marks its trick: it phosphorylates one of its own aspartate residues, and that phosphoaspartate intermediate drives the conformational flip from inward-facing to outward-facing.

That gradient is a battery, and secondary active transport spends it. Couple a molecule running downhill to another being dragged uphill, and you get free uphill transport, the cellular version of a water wheel stealing energy from one current to drive another. Symporters move both in the same direction (SGLT1 hauls glucose into intestinal cells on the back of sodium); antiporters move them oppositely (the Na⁺/Ca²⁺ exchanger uses sodium to push calcium out). Every one of these depends on the gradient the Na⁺-K⁺ ATPase builds. The diabetes drugs called gliflozins block SGLT2 in the kidney so glucose spills into the urine and blood sugar falls.

Pull on this thread and a folk remedy appears: digitalis, from the foxglove plant, inhibits the Na⁺-K⁺ ATPase. Sodium then backs up inside the heart cell, the Na⁺/Ca²⁺ exchanger slows, calcium accumulates, and each heartbeat hits harder — a treatment for failing hearts, with a narrow therapeutic window, since too much triggers calcium-overload arrhythmias. (William Withering worked the rest out from an old woman’s foxglove brew in 1785; that story is in the margins.)

CFTR belongs to a family that explains a lot of this machinery. ABC transporters share an architecture: two transmembrane domains that form the path across, and two nucleotide-binding domains that act as the ATP-powered engine. When ATP binds, the two engine domains clamp together and flip the protein from inward-facing to outward-facing. P-glycoprotein is the family’s promiscuous pump, throwing structurally diverse drugs out of cancer cells. Tumors that overexpress it become multidrug-resistant, evolving a pump that ejects chemotherapy before it can work. CFTR is the family’s odd member: same domains, but it functions as a channel rather than a pump. And that is what makes ΔF508 a missing channel rather than merely a broken one. The drug Trikafta, approved in 2019, combines two correctors (elexacaftor and tezacaftor) that help the protein fold and reach the surface with a potentiator (ivacaftor) that keeps it open once it gets there. It is projected to add decades of expected survival in cystic fibrosis — precision medicine aimed at the protein-level consequence of a mutation.

Then there are channels that gate. Voltage-gated channels open in response to membrane voltage; ligand-gated ones open when a molecule binds. Two famous toxins, tetrodotoxin from pufferfish and saxitoxin from the dinoflagellates behind paralytic shellfish poisoning, both plug the outer mouth of voltage-gated sodium channels, so nerves cannot depolarize and the victim is paralyzed. TRP channels sense the environment: TRPV1 opens to heat and to capsaicin, TRPM8 opens to cold and to menthol. Capsaicin activates the heat-sensing channel directly, which is why chili registers as heat and mint as cold even though nothing has changed temperature.

The most elegant problem in this whole chapter is how a potassium channel rejects sodium, because sodium is the smaller ion. A simple sieve would let the smaller ion through, so size cannot be the answer. The selectivity filter solves it thermodynamically. It lines the path with backbone carbonyl oxygens that present, at each site, 8 oxygens to the passing ion, closely mimicking the cage of water K⁺ carries in solution. K⁺ sheds its water and the oxygens take water’s place almost perfectly, so it pays little net energy to enter. Na⁺ is too small to touch all 8 oxygens at once, so it cannot pay back the energy it spent shedding its own water, and it stalls. The result is about 10,000-fold selectivity for the larger ion. The channel offers an energy landscape only K⁺ can afford.

That raises one more puzzle. If the filter binds K⁺ tightly enough to be that selective, why does K⁺ ever leave? The answer is crowding. Multiple K⁺ ions sit in the filter at once and repel each other electrostatically, so the entry of one ion shoves the next one out, like a Newton’s cradle converting entry energy directly into exit energy. K⁺ moves at near the diffusion limit, around 10⁸ ions per second, selective and fast at the same time. Roderick MacKinnon solved this structure and won the 2003 Nobel Prize in Chemistry for it.

The action potential is where all of this comes together: voltage-gated sodium and potassium channels firing in sequence, the Na⁺-K⁺ pump resetting the gradients, a ball-and-chain plug swinging in to inactivate the sodium channel, and a refractory period that keeps the signal traveling in one direction. The region the impulse just passed cannot fire again yet, so the signal does not run backward. The membrane — the impermeable sheet that builds itself for free — turns out to be the substrate on which a nerve fires.