Proteins and sugars embedded in the bilayer

A lipid bilayer is mostly a barrier. Nearly everything a membrane does beyond acting as a barrier is carried out by the proteins set into it, and by the sugar chains attached to some of those proteins and to some lipids. This topic is about how those components are attached, how their arrangement can be read from a sequence, and why their orientation is a permanent feature of the cell. The jobs they perform (moving solutes, sticking cells together, relaying signals) are the subject of later topics, and get only a naming here.

What membrane proteins do, in outline

Membrane proteins fall into a few broad functional classes. Some are transporters, which carry solutes across. Some are adhesion molecules, which link a cell to its neighbors or to the surrounding matrix. Some are receptors, which detect a signal on one side and pass on a message to the other. Some are enzymes working at the surface, and some are anchors to the cytoskeleton. Every one of these jobs depends on knowing which end of the protein faces which compartment, so the structural questions below come first.

Three ways to be attached

Proteins are classified by how tightly they are bound and by what they are bound to.

Integral proteins

An integral protein has at least one stretch that sits within the hydrophobic core. The core is roughly 3 nm thick, and a polypeptide crossing it faces a problem. Backbone amide N-H and carbonyl C=O groups are polar, and a polar group stripped of water and surrounded by hydrocarbon is costly unless it finds a partner. The solution is to satisfy nearly all backbone hydrogen bonds inside the core, by folding the segment into a regular secondary structure.

Single-pass helices. In the simplest case, one alpha helix of about 20 mostly hydrophobic residues crosses the bilayer once. In an alpha helix each backbone C=O bonds to an N-H four residues further along, so the backbone groups in the buried stretch are paired with each other. The exceptions are local: the first and last turns of a helix lack a full set of partners, and a proline in the middle of a segment removes an N-H and interrupts the pattern. The side chains point outward and contact the lipid tails, so the surface that meets the core is hydrophobic. A single-pass protein may have a large domain on each side, or a very short one.

Multi-pass helical bundles. Many proteins cross the bilayer several times, packing their helices side by side. They are named here by class only: seven-helix receptors and twelve-helix carriers are both common patterns. Some helices of a bundle may carry polar or charged side chains on the face that points toward the other helices, so a bundle can build a hydrophilic channel or pocket in its interior while the outside surface remains hydrophobic.

Beta-barrels. A second solution is the beta-barrel, a sheet rolled into a cylinder. Neighboring strands hydrogen-bond to each other along their length, so again nearly every backbone group is paired, with the two edge strands closing the cylinder by bonding to each other. The side chains alternate in direction along a strand, with one set facing the lipid and the other lining a central pore. Barrels are characteristic of the outer membranes of Gram-negative bacteria, mitochondria and chloroplasts, where they form water-filled pores called porins. They are not typical of the plasma membrane of animal cells.

The structural facts needed here are only these: a helix is a coil with hydrogen bonds between nearby backbone groups, a strand is a nearly extended stretch, and a sheet is made of strands bonded side to side. Full treatment of folding belongs to the protein-structure topics.

Residues at the edges

Not all of a transmembrane segment is equally hydrophobic. The tryptophan and tyrosine side chains, which are aromatic and slightly polar, tend to cluster at the two interfaces, where the core gives way to the head-group region. Positively charged residues (lysine and arginine) are more common on the cytosolic flanks of segments than on the exterior ones, a bias summarized as positive inside. This bias is a correlation used to guess orientation, not a law.

Worked example: helix length

The hydrophobic core is about 3.0 nm thick. An alpha helix rises 0.15 nm for each residue along its axis. The number of residues needed to cross is

$$\frac{3.0\ \text{nm}}{0.15\ \text{nm per residue}} = 20\ \text{residues}$$

which matches the usual figure of about 20 residues per spanning helix.

A beta-strand is much more extended, rising about 0.33 nm per residue. Running straight through the core, it needs

$$\frac{3.0}{0.33} \approx 9.1$$

so about 9 to 10 residues. The strands of real barrels are tilted relative to the membrane normal, which lengthens the path a little, but the point holds: a barrel needs fewer residues per crossing than a helix.

Peripheral proteins

A peripheral protein does not enter the core. It is held to the surface by ionic and hydrogen-bonding contacts, either with lipid head groups (negatively charged head groups are common on the cytosolic leaflet) or with an exposed part of an integral protein. Because these contacts are electrostatic, they can be weakened by raising the ionic strength, since salt ions compete for the charges, or by changing the pH enough to alter the charges involved. Such treatment lifts the protein off and leaves the bilayer intact. An integral protein, in contrast, cannot be freed without disrupting the lipid environment around its buried segments, and so detergent is needed. Detergent molecules replace the lipids around the hydrophobic surface and keep the protein soluble in water.

Lipid-anchored proteins

A third group is bound to the membrane only by a covalently attached lipid. The protein itself may be entirely water-soluble. Which side it sits on depends on the type of lipid:

• GPI anchor. A glycolipid built on phosphatidylinositol is attached to the C-terminus, and its lipid tails sit in the outer leaflet. The protein hangs out into the exterior space. GPI-anchored proteins tend to collect in ordered, tightly packed lipid domains.
• Myristoylation. A 14-carbon saturated fatty acid joined to the N-terminus, facing the cytosol. Attachment is stable.
• Palmitoylation. A 16-carbon saturated fatty acid joined to a cysteine side chain, facing the cytosol. The bond can be cleaved and re-formed, so the attachment is reversible and can be regulated.
• Prenylation. An isoprenoid chain (such as farnesyl or geranylgeranyl) joined near the C-terminus, also facing the cytosol.

Anchoring changes behavior. Saturated chains favor ordered domains, so an anchored protein may be steered into a raft. A reversible lipid such as palmitoyl allows a protein to move between membrane and cytosol, or between domains, as conditions change. A GPI-anchored protein can also be released from the cell surface by an enzyme that cuts the anchor, a fact the extraction test below depends on.

Worked example: extraction test

Three proteins are prepared from a membrane fraction.

• Protein A is released by 1 M NaCl and the membranes stay intact. Salt breaks ionic contacts, so A is held by electrostatic attachment: peripheral.
• Protein B stays with the membrane after salt and is released only when detergent dissolves the bilayer. Detergent release alone is not decisive, because lipid-anchored proteins are also solubilized by detergent. B is called integral only with independent evidence of a membrane-spanning segment, such as a hydropathy peak (a prediction) confirmed by a tested membrane-spanning segment. Protease protection alone would not settle it, because it reports which side a segment is on, not whether it spans the bilayer; soluble proteins inside sealed vesicles are protected too. With that evidence, its hydrophobic surface is buried in the lipid core: integral.
• Protein C is released by a phospholipase that cuts the GPI anchor. The enzyme can only reach the anchor if the anchor lies in the outer leaflet, and the protein needs no detergent. C is lipid-anchored, and because GPI anchors lie in the outer leaflet, its location is the exterior face. That location comes from the identity of the anchor. Accessibility tests are reserved for intact cells or sealed vesicles of known orientation, since an isolated membrane fraction does not by itself show which leaflet is exposed.

The inference rests on reasoning about what each treatment can reach and which bonds it breaks.

Reading topology from a sequence

How can one guess which parts of a new protein cross the membrane? Each amino acid has a hydrophobicity value on a standard scale. Plotting single values would be noisy, so the values are averaged over a sliding window of about 19 residues. A window about that long is close to the length of a membrane-spanning helix, so averaging smooths away isolated hydrophobic residues but lets a continuous hydrophobic stretch stand out. A horizontal threshold is set, and a sustained peak of about 20 residues above it is a candidate membrane-spanning helix.

The method has limits. It predicts helices, so it misses beta-barrels, whose strands alternate between facing the lipid and facing the pore and so are never uniformly hydrophobic. It also can mistake amphipathic helices (one face hydrophobic and one polar) for non-spanning, or hydrophobic signal sequences and buried hydrophobic cores of ordinary soluble proteins for spanning segments. The result is a prediction to be tested.

Worked example: counting segments and placing the ends

A 450-residue protein shows 7 qualifying peaks on its hydropathy plot. Separately, an N-linked glycosylation site on the N-terminus is found to be used. Because N-linked sugar is only added on the luminal side of the ER, which becomes the exterior of the cell, the N-terminus must be outside.

Each crossing swaps the side. Starting outside, after 1 crossing the chain is inside, after 2 outside, and so on. After an odd number of crossings the chain ends up on the opposite side from where it began, so with 7 crossings the C-terminus is inside.

Now suppose one segment is deleted, leaving 6. An even number of crossings brings the chain back to its starting side, so the C-terminus would now end up outside, provided the N-terminus is still outside. The rule is general: the terminus on the same side means an even count, and opposite sides mean an odd count.

Orientation is fixed

For proteins entering the eukaryotic secretory pathway, orientation is established when the protein is inserted into the ER membrane as it is made, and it is normally preserved through trafficking. (Bacterial and mitochondrial proteins are inserted at their own membranes by different machinery.) Moving a hydrophilic domain through the core would cost far too much energy, and no mechanism exists to reverse it. The result is permanent sidedness.

Vesicles carry this orientation along. The lumen of the ER, the Golgi and transport vesicles corresponds topologically to the exterior of the cell: when a vesicle fuses with the plasma membrane, its luminal face becomes the outer surface. A domain that faced the lumen therefore ends up facing the outside, and the cytosolic domain remains in the cytosol. The sidedness of the lipids, too, is maintained across these steps.

Three experiments test the topology of a protein:

• Protease protection. Sealed vesicles with the protein inserted are treated with a protease. Segments on the outside of the vesicle are digested, whereas segments inside are shielded. Adding detergent to break the vesicles then exposes the rest. This test establishes which side a segment faces (accessibility and sidedness), not whether the protein is integral, since soluble luminal proteins are protected as well.
• N-linked glycosylation sites. N-linked sugar is added on the luminal side only, so a site that carries sugar was on the luminal (exterior) side.
• Antibody access. An antibody against a tag reaches the tag on intact cells only if the tag is outside. After the membrane is permeabilized, a tag inside becomes reachable as well. A tag that is stained only after permeabilization is therefore cytosolic.

Movement and fences

Lipids diffuse quickly in the plane of the membrane. Many proteins diffuse more slowly, because they are larger and because they can be tethered. In the red blood cell, a network of the protein spectrin lies against the cytosolic face and is bound to some integral proteins, restricting their motion. Tight junctions form a fence between the apical and basolateral domains of an epithelial cell, so proteins and outer-leaflet lipids stay on their own side. The next topic covers junctions in detail.

The carbohydrate coat

Sugar chains are attached to some membrane proteins (glycoproteins) and some lipids (glycolipids). In the mature cell-surface coat the sugar is on the exterior face, never on the cytosolic side. For N-linked glycoproteins this follows from sugar being added in the ER and Golgi lumen. Some glycolipid precursors differ: glucosylceramide, for example, receives its first sugar on the cytosolic surface of the Golgi and is then translocated, after which its chain is extended in the lumen. Either way, the finished chains face the lumen, and that sidedness is preserved to the cell surface.

Together, these chains form a coat at the cell surface with two broad functions:

• A protective, hydrated layer. The many hydroxyl groups hold water and help cushion the surface.
• Recognition between cells, since particular sugar arrangements are read by binding proteins on other cells.

The ABO blood-group antigens give a structural example of how small differences matter. They are sugar chains on surface lipids and proteins. Two versions of a chain can differ by one terminal sugar, which changes the shape and the hydrogen-bonding pattern at the tip of the chain. A binding protein can distinguish the two shapes. The point here is only that a single sugar unit alters the chemical identity of a surface.

Putting the pieces together

For a new membrane protein, ask in order: how is it attached (extraction behavior), how many segments does the sequence suggest, which side is the terminus on, and is sugar present. Because sugar marks the exterior, glycosylation closes the loop: a detected sugar tells you which face you are looking at, and the crossing count then places every other segment.