Oligomeric Proteins (Molecular Biology)Proteins that contain more than one polypeptide chain are said to be oligomeric or multimeric, as opposed to monomeric proteins with a single chain. The Greek "oligoi" means "few" and the Latin "multi" means "many," yet oligomeric and multimeric have become essentially equivalent. The first term, which was introduced by J. Monod and collaborators in the context of allostery (1), will be used here for proteins with two to a few tens of polypeptide chains, as opposed to large assemblies such as muscle fibers, which have hundreds or thousands. The individual polypeptide chains in an oligomer are also called subunits . The terms homo-oligomer and homo-multimer apply to oligomeric proteins built from only one type of polypeptide chain, the product of the same gene. Other oligomeric proteins are hetero-oligomers or hetero-multimers .Whereas most proteins have a primary structure (amino acid sequence), secondary structure (alpha-helices and beta-sheets), or a tertiary structure (three-dimensional), oligomers have an additional level called the quaternary structure . The quaternary structure of a homo-oligomer may be coded by the formula an, where n is the number of copies of the polypeptide chain. Many proteins are oligomers, and quaternary structures of all sorts are found in nature. Beyond formulae such as an that describe the number and nature of polypeptide chains, the quaternary structure is part of the three-dimensional structure and is determined along with it in X-ray crystallography or NMR studies.

1. Symmetry

The presence of multiple copies of the same chemical unit leads to the possibility that the three-dimensional structure has internal molecular symmetry. Molecular symmetry is easily established in crystal studies, where it can be either included in the symmetry of the crystal lattice or simply local. In electron microscopy, molecular symmetry is the basis for powerful image reconstruction methods (see Single Particle Reconstruction), and it can also be used to refine X-ray crystallography data (see Molecular Averaging). X-ray data indicate that symmetry is the rule in homo-oligomeric proteins, the lack of symmetry being the exception (2). An object with internal symmetry can always be divided into smaller identical units, the smallest of which is called the asymmetric unit by crystallographers. Monod and collaborators coined the word protomer to designate the same entity in an oligomeric protein, and it shall be used here to distinguish the asymmetric unit of the protein from that of the crystal, which may contain more than one protomer. The protomer itself may be more than one polypeptide chain. In hetero-oligomers, it must contain each type of chains; for instance, an immunoglobulin G molecule has one heavy and one light chain in its protomer. In the discussion below, we assume for simplicity that the protein is a homo-oligomer and the protomer a single polypeptide chain, but the same conclusions are readily extended to hetero-oligomers.Proteins are chiral objects that may not have inversion centers or mirror symmetry, both of which would invert the chirality. Protomers must be related to another by a rotation, translation, or screw rotation, which is a combination of a rotation and translation. When repeatedly applied to the protomer, these operations reconstitute the whole object. Their combination constitutes a group in the mathematical sense of the word. Groups of symmetry operations that generate objects of finite size are known as point groups. The number of asymmetric units is a characteristic of the group called its multiplicity. Chiral point groups belong to one of three families:

1. Cyclic Cn symmetry (also noted as n). Protomers are related by rotations of 360/ n degrees about a single axis c. The oligomer must therefore contain n protomers and the multiplicity is m = n.2. DihedralDn symmetry (also noted as n2). Protomers are related either by rotations of 360/n degrees about axis c, or by a 180° rotation about one of n two-fold axes in the plane orthogonal to c. The oligomer then contains m = 2n protomers.3. The cubic symmetries of a tetrahedron, an octahedron, or an icosahedron. All have multiplicities of 12 or a multiple of 12. Tetrahedral symmetry has nonorthogonal two-fold and three-fold axes; in addition, octahedral symmetry has four-fold axes, icosahedral symmetry five-fold axes.The symmetry of an oligomeric protein is closely related to the number of protomers and, therefore, polypeptide chains (Table 1). The only possible point group symmetry for a homo-dimer is C2, which has a single two-fold axis and m = 2. A homo-trimer must have a three-fold axis (120° rotation) and cyclic C3 symmetry (m = 3), if it is symmetric at all. On the other hand, a homo- tetramer can have two symmetries: either a four-fold axis in the cyclic point group C4, or three orthogonal two-fold axes in the dihedral point group D2 (also noted as 222). Both point groups have m = 4, yet they yield very different quaternary structures, and D2 is much more frequently observed than C4. Dihedral symmetry, which requires the number of subunits to be even (m = 2n), is very common in globular soluble proteins. Homo-hexamers generally have D3 symmetry. An example isEscherichia coli aspartate transcarbamoylase, in which each of the six protomers comprises one catalytic and one regulatory chain. Octamers have D4 symmetry, illustrated by hemerythrin in Figure 1. In contrast, membrane proteins often have cyclic symmetry and odd numbers of subunits. Porins of the bacterial outer membrane and the bacteriorhodopsin of Halobacterium halobium are homo-

trimers with C3 symmetry, whereas the a-hemolysin of Staphylococcus is a heptamer with C7 symmetry. Cubic symmetry is less common, but it is found in ferritin and large assemblies, such as the pyruvate decarboxylase complex or icosahedral viruses. The latter have capsids made of an assembly of one or several different polypeptide chains, all present in multiples of 60, the multiplicity of the icosahedral symmetry group.Figure 1. Octameric hemerythrin. Hemerythrin is a nonheme iron protein that transports oxygen in sipunculid worms. Sketch of the crystal structure (11) with subunits in different shades of gray and a-helices as cylinders. The assembly of eight identical chains has exact dihedral D4 symmetry (also called 42 symmetry). Here it is viewed along the fourfold axis (diamond); two-fold axes (arrows) run horizontal and vertical along the diagonals of the square-shaped tetramers.

Table 1. Symmetry of Oligomeric Proteins®Protein Quaternary Structure           Point Group

                                            

Cyclic

HIV proteinase

Hemoglobinb

Porin

Neuraminidase (flu virus)

Pentraxins

a-Hemolysin

Light harvesting complex II

Dihedral

Phosphofructokinase

Aspartate transcarbamoylase

Hemerythrin

GTP cyclohydroase

GroEL chaperonin

Cubic

Phaseolin Tetrahedral

Apoferritin Octahedral

Virus coats Icosahedral

a Adapted from ref. 2, where references to X-ray structures can be found. b Hemoglobin also has approximate D2 symmetry.

2. Approximate Symmetry and AsymmetrySymmetry requires the exact geometric repetition of chemically identical units, but approximate symmetry can nevertheless be observed in assemblies that do not satisfy this condition exactly. In mammalian hemoglobins, the a2b 2 oligomer displays the exact symmetry of point group C2 with the two-fold axis relating the two ab units; in addition, there is an approximate D2 symmetry equivalencing the very similar tertiary structures of the homologous a and b chains. Approximate symmetry between structural domains of a single-chain protein is also well documented, and it is usually interpreted as an indication that the protein derived from a symmetrical homodimer by gene duplication and gene fusion.

Because symmetry is so frequent, asymmetry is remarkable when it occurs, and it is usually reflected in the function. The F1 fragment of the ATP synthase of mitochondria and chloroplasts has the formula a3b3g. The three-fold symmetry of the assembly of a and b chains is broken by the presence of a single g chain in the middle. Contacts with g make the three active sites carried by the three b chains nonequivalent, an essential feature of the catalytic mechanism. Simpler examples are two dimeric proteins from the human immunodeficiency virus (HIV), the HIV proteinase and HIV reverse transcriptase. In both, X-ray crystallography studies have demonstrated departure from two-fold symmetry. Alone, the HIV proteinase dimer does display exact C2 symmetry, but this is incompatible with the binding of a peptide substrate and with the catalytic mechanism. A peptide cannot have internal symmetry, and the mechanism of hydrolysis requires one of two intrinsically equivalent active site aspartate residues to be protonated, the other deprotonated (see Carboxyl

Proteinase). In crystalline complexes of the enzyme with substrate analogues, the symmetry is broken. Minor structural changes occur in the protein to fit the asymmetric ligand, and the complex retains approximate symmetry. In contrast, HIV reverse transcriptase displays a major asymmetry. Its two chains are the product of the same gene, but one has undergone proteolytic processing. They adopt grossly different folds in the dimer and play very different roles in DNA synthesis. In this case, asymmetry exists in the chemical as well as the three-dimensional structure. It is not known whether or not the asymmetry preexists the proteolytic cleavage that yields the active dimer.Less often, oligomeric proteins display symmetries that do not belong to one of the point groups mentioned above and cannot be exact. The tetrameric E. coli lactose repressor is a case in point (Fig. 2). It is assembled from two dimers, each of which has C 2 symmetry. The dimers are related by an approximate two-fold axis that is not orthogonal to those of the dimers as D 2 symmetry would require. The result is that all four subunits have their DNA-binding headpieces on the same face of the tetramer, which they could not do in a D2 tetramer. On the opposite face, the C-terminal a-helices of each subunit form most of the dimer-dimer contacts. Remarkably, the four C-terminal a-helices assemble with D2 symmetry, whereas the rest of the protein does not, which a long connecting peptide makes possible by adopting different conformations in two of the subunits (3).Figure 2. The lactose repressor of E. coli. The homo-tetramer is made of two dimers, each with two-fold symmetry. However, their two-fold axes (arrows) are not orthogonal, and only the four-helix bundle made by C-terminal a-helices at the bottom of the molecule shows the usual D2 (or 222) symmetry of tetramers. Residues 1 to 61 of each 360-residue monomer form N-terminal DNA-binding domains located at the top of the molecule. In the absence of DNA, they are so

mobile that they cannot be observed in the crystal structure of the protein alone (3).

3. Subunit InteractionsThe stabilities of quaternary structures result from the contacts between subunits. Subunits form pairwise interfaces of two different characters, depending on whether or not they are related by a two-fold axis. Two-fold symmetry implies that the same surface and same amino acid residues of both subunits are in contact. Interfaces having other cyclic symmetries, or no symmetry at all, usually involve different sets of residues on the two partners. Following the nomenclature of Monod et al. (1) once again, isologous interfaces refer to those with two-fold symmetry, heterologous the others. All interfaces are isologous in C2 dimers and D2 tetramers; in other systems, at least one set of interfaces must be heterologous.

When the three-dimensional structure is known from X-ray or NMR studies, the extent and nature of the interactions between polypeptide chains can be analyzed in atomic detail. A convenient estimate of the extent of the contact is the interface area, the area B of the protein surface that is buried (made inaccessible to the solvent) as a result of the interaction. In a dimer, B is derived from the atomic coordinates of the structure by calculating the area of the solvent-accessible surface for the dimer and for the two isolated subunits, and subtracting the first value from the sum of the other two. Table 2 quotes interface areas observed in some oligomeric proteins. Btot is the total area of the surface buried in the quaternary structure. In all cases, it is at least 1400 A2, similar to the surface area buried in contacts between an antibody and a protein antigen, or proteinase and a proteinase inhibitor (see Protein-Protein Interactions), and this may be the minimum required for stable association. However, most oligomeric proteins have much larger subunit interfaces than this minimum. In bovine catalase, the quaternary structure buries as much as 70% of the protein accessible surface, in six large interfaces. All are isologous, and the D2 symmetry of the tetramer makes them two by two equivalent. Thus, only three pairwise interface areas are listed in Table 2.Table 2. Area of Interfaces between Subunits of Oligomeric Proteins®

ProteinMolecularWeight(Da)

Accessible Surface Area (A2) Btot

DimersAvian pancreatic 8500 5300 1400peptideUteroglobin 15,800 7500 3000Superoxide dismutase 31,400 13,800 1350

Triosephosphate 53,000 20,300 3180isomeraseAlcohol 79,800 29,000 3260dehydrogenase (ADH)Citrate synthase 96,000 28,500 9800TetramersMellitin 11,400 6300 4160 880 820 520Glutathione 83,600 28,600 6280 1520 1520 180peroxidasePhosphofructokinase 141,200 40,600 14,400 4500 2520 180Catalase 231,700 60,900 42,300 9260 9140 4120HexamerInsulin 34,600 13,100 8600 1280 1400 * 180Octomer

Hemerythrin 107,500 35,900 13,600 1780 260 *1420

a Adapted from ref. 4, where references to X-ray structures can be found. The molecular weight, accessible surface area, and total interface area are quoted for the whole molecule. Tetramers and larger oligomers contain several interfaces, with areas given per pair of subunits. Dimers in this table have C 2 symmetry, tetramers, D2 symmetry, with isologous interfaces only. Larger oligomers also contain heterologous interfaces marked with an asterisk.

In general, the buried surface area Btot is distributed among several pairwise interfaces, up

to  for an assembly of n subunits when each one is in contact with all others. In a

dimer,  , in a tetramer, Indeed, most D2 tetramers have six pairwise interfaces like catalase. In phosphofructokinase, let us label the four subunits A, B, C, and D. Symmetry-equivalent AB and CD pairs make extensive contacts burying 4500 A2 (A = 10-10m) each. As AC and BD pairs bury significantly less, and AD and BC pairs very little, phosphofructokinase appears, based on the size of the pairwise interfaces, to be a dimer of AB-like

dimers. It might be expected that AB-like dimers are actual intermediates in tetramer assembly, but this is not known. Dividing large oligomers into smaller assemblies on the basis of their three-dimensional structure is often an arbitrary decision. Hemerythrin, illustrated in Figure 1, is an octamer with D4 symmetry. Is it a dimer of cyclic tetramers, or a cyclic tetramer of dimers? Either description fits. Within a tetramer, each hemerythrin subunit makes two symmetry-equivalent heterologous interfaces with neighbors, eg, the green subunit of the top tetramer with the red and purple subunits, and two isologous interfaces with subunits of the other tetramer, eg, the green subunit with the yellow and brown bottom subunits. Both the green-yellow interface forming a dimer and the green-red or green-purple interfaces forming the tetramer are relatively large, making both types of substructures plausible.The chemical composition of the protein surface involved in subunit contacts is generally richer in aliphatic or aromatic groups than is the surface accessible to solvent, and it therefore is less polar. On average, nonpolar groups contribute 65% of the interface areas and only 57% of the accessible surface (4). From this point of view, subunit interfaces in oligomeric proteins tend to resemble protein interiors, whereas the interfaces of protein-protein complexes are more like the rest of the protein surface (see Protein-Protein Interactions). Hydrophobic (aliphatic or aromatic) amino acid residues are correspondingly overrepresented, and the most abundant residue at subunit interfaces is leucine. The well-known leucine zipper is an example of a coiled-coil dimerization motif formed by the surfaces of two leucine-rich a-helices, but there are many other situations where leucine residues contribute to interfaces. Remarkably, the second most abundant residue is arginine. Its very polar guanidinium group, which is almost always a surface group in monomeric proteins, is frequently involved in the quaternary structure of oligomers. This and other polar groups buried in subunit contacts contribute to the stability of the assembly by

forming intersubunit hydrogen bonds. Such bonds occur at the rate of about one per 250 A 2 of interface area (5), and the partner groups are favorite targets for site-directed mutagenesis aiming to perturb the quaternary structure.

4. Quaternary Structure ChangesThe quaternary structure of oligomeric proteins is often as stable as the tertiary structure itself, with the subunits dissociating only under conditions that denature the polypeptide chains. There are nevertheless cases where a dissociation equilibrium is observed under nondenaturing conditions. In the case of human hemoglobin, the a2b2 oligomer is normally stable when the protein is in the deoxy-form, but when oxygen binds, it dissociates into ab units with an equilibrium dissociation constant of the order of 1 |iM. This is much too low for dissociation to play a physiological role in red blood cells, where the hemoglobin concentration is greater than millimolar, yet the coupling between oxygen binding and tetramer dissociation provides a powerful biochemical tool for the study of the allosteric transition (6).While remaining tetramers, mammalian hemoglobins can still adopt more than one quaternary structure. This has a major functional significance, for the different structures have different affinities for oxygen and other ligands. The allosteric model of Monod et al. (1) makes quantitative predictions of how a quaternary structure change can lead to cooperative ligand binding in an oligomeric protein. The X-ray structures of deoxy- and oxy- or carbonmonoxy-hemoglobin give a structural basis for this model. Both forms of the protein are symmetrical a2b2 tetramers, yet the relative orientations of the ab units differ, and many contacts at the interface are disrupted or rearranged. Changes within ab units (tertiary structure changes) are of much lesser amplitude, but connect the heme sites to the interface. At least one other quaternary structure, called Y or R2, has been identified in mammalian hemoglobins, confirming the plasticity of the interface between the ab

units. In contrast, the interface between a and b within each unit is essentially invariant.Quaternary structure changes generally appear in X-ray structures as rotation or translation movements of the constituent subunits relative to each other, which perturb the contacts that hold together the assembly, without abolishing them. Such movements have been characterized in several allosteric proteins, for instance, in E. coli aspartate transcarbamoylase, but they also exist in oligomeric proteins that are not usually considered allosteric and display no cooperativity in ligand binding. In immunoglobulins, the interface between the variable domains of the heavy and light chains shows significant plasticity, and this may be important for antigen recognition, since the combining site spans that interface.

5. Folding and EvolutionThe oligomeric proteins listed in Table 2 are fairly globular, and their overall accessible surface area is correlated with their molecular weight (4). In many cases, the individual subunits are also reasonably globular and can remain folded in isolation, so these proteins assemble by the individual subunits folding, then associating. In other cases, however, the subunits themselves are often far from globular, and some parts of individual polypeptide chains may make more interactions with other subunits than with their own. In such cases, the subunit fold found in the oligomer is unlikely to preexist the assembly, and folding must therefore be tightly coupled with association. The protein folding of oligomeric proteins is less easily studied than that of small monomeric ones. Denaturation with urea or guanidinium salts is often irreversible, as the correct reassembly of the monomers competes with their aggregation, leading to insolubility. However, it has been achieved in a number of cases, showing that complete self-assembly can be reproduced in vitro (7). In some cases, a folded monomer or smaller oligomer can be characterized as an intermediate. For instance, E. coli aspartate

transcarbamoylase catalytic chains form active trimers in the absence of regulatory chains. But, intermediates are often too short-lived for detection during reassembly. This is illustrated by the possibly extreme case of the Arc repressor of bacteriophage P22, a small dimer that can refold very rapidly. No folded monomer is detected, and refolding and association appear simultaneously as a second-order reaction, even on a millisecond time scale (8).The evolution of oligomeric proteins offers a number of similarly remarkable features. Closely related proteins usually have the same quaternary structure, and residues forming the ab interface in mammalian hemoglobins are invariant like the internal residues. Nevertheless, the quaternary structure is much less well-conserved than the subunit fold. Globins can be monomeric (myoglobin, plant and insect hemoglobins), dimeric (lamprey hemoglobin), tetrameric (mammalian hemoglobins), or larger oligomers, while preserving the same fold and the same capacity of binding heme and oxygen. Other protein families are like the globins and, in general, the quaternary structure need not be the same in homologous proteins when the amino acid sequence identity is below 50%, a level at which tertiary structures hardly change. In globins, oligomerization is accompanied by a gain of function: Oxygen binding becomes cooperative. As mentioned above, the cooperativity results from an allosteric mechanism whereby ligation at the heme site is coupled to a change in quaternary structure. There is therefore a natural selection pressure on the quaternary structure distinct from that which maintains the tertiary structure of the subunits. It is focused on amino acid residues forming interfaces between ab units, and specific disorders in humans are associated with mutations that change these residues.The gain of function is even more evident in many enzymes where the active site is made of residues from more than one subunit. An example is the dimeric HIV proteinase mentioned above. Its catalytic mechanism is the same as in monomeric carboxyl proteinases, such as pepsin,

and involves two aspartate residues. In HIV proteinase, they are the same residue from each of the two subunits; in pepsin, they are different. The longer polypeptide chain of pepsin is a two-domain structure that displays approximate internal symmetry, each domain resembling the viral proteinase monomer. Thus, the single-chain enzyme reproduces the structure of the dimer, giving strong support to the hypothesis that single-chain carboxyl proteinases evolved by gene duplication and fusion from a shorter dimeric precursor. This hypothesis had been made on the basis of the sequence and structure of aspartate proteinases many years before the viral enzyme was known (9).Dimerization is often assumed to be an acquired feature in evolution. However, in the carboxyl proteinase family, the HIV-type of dimer is the ancestral form relative to the pepsin-type of monomer and, moreover, a monomeric ancestor could not have been a proteinase, for this function requires both aspartate residues. This shows that evolution can undo quaternary structure, as well as create it. In the globin family, the more complex dimeric or tetrameric hemoglobins are considered to have evolved from single-chain globins found in lower organisms, but in many other families there is no evidence for a monomeric precursor. Nevertheless, an evolutionary scenario for dimerization can be proposed. As pointed out by Monod et al. (1), isologous association should be easier to achieve, because two-fold symmetry duplicates any favorable (or unfavorable) interaction. Thus, a small number of mutations at the protein surface might yield a C2 dimer. An alternative model for dimerization is domain swapping (10), which does not even require mutations to form novel interactions. In domain swapping, domains A and B interact either within a monomer, or with domains A’-B’ of a second monomer, yielding a ( AB’) (A’B) dimer where noncovalent A-to-B’ and A’-to- B interactions are essentially identical to A-to-B interactions in the monomeric form. Domain swapping was named for the phenomenon in diphtheria toxin, where it can

be induced by temperature changes. There are some obvious cases in families of homologous proteins, such as the crystallin family of eye lens proteins in mammals (12). It includes monomeric g-crystallins and dimeric b-crystallins, all made of two domains that have very similar folds and related sequences, but swap their positions in the dimeric proteins, relative to the monomeric proteins.Domain swapping is a plausible evolutionary source of oligomers in which large parts of the polypeptide chain wander away from their subunit to interact with others, yielding some of the largest interfaces in Table 2. It could, in principle, yield more complex assemblies with cyclic symmetry, eg, a (AB") (A"B’) (A’B) trimer where each subunit donates a domain to its neighbor. However, assemblies such as hemoglobin or hemerythrin in Figure 1 show no indication of domain swapping, and many evolutionary mechanisms must coexist. Cells are very dense and complex mixtures of macromolecules. Any change on a protein surface must affect its capacity not only to form legitimate interactions with its siblings in an oligomer, but also illegitimate ones with other components of the cytoplasm, membrane, or organelle, and all these interactions compete in evolution.