Protein Quaternary
Structure: Subunit–Subunit
Interactions
Susan Jones, University College, London, England
Janet M Thornton, University College, London, England
The quaternary structure of proteins is the highest level of structural organization observed
in these macromolecules. The multimeric proteins that result from quaternary structure
formation involve the association of protein subunits through hydrophobic and
electrostatic interactions. Protein quaternary structure has important implications for
protein folding and function.
Introduction
Proteins are organized into a structural hierarchy. The
polypeptide chain at the primary structural level comprises
a linear, noncovalently linked amino acid residue sequence.
Secondary structure is the level at which the linear
sequences aggregate to form structural motifs such as
helices and sheets. The tertiary structure is formed by
packing of the secondary structural elements into one or
more compact globular domains. In many cases proteins
are composed of only a single polypeptide chain that has
tertiary structure as its highest level of organization, e.g.
lysozyme. These are termed monomeric proteins. However,
many proteins are composed of more than one
polypeptide chain, associated into assemblies possessing a
specific quaternary structure, e.g. dimeric interleukin 8 and
tetrameric a2b2 human haemoglobin (Figure 1). The most
complex assemblies are those observed in the higher order
structures, such as theicosahedral viruses, that comprise 60
monomers in identical symmetrical positions. The quaternary
structure of a protein describes the stoichiometry
and stereochemistry of assemblies of noncovalently linked
subunits, characterized by the lower levels of structural
organization (Jaenicke, 1987).
In a discussion of protein quaternary structure it is
important to adhere to a single set of definitions. Those
widely used in the literature, and adopted here, were
derived by Monod et al. (1965) in their theoretical model of
allosteric effects in protein structures. An oligomer is
defined as a protein assembly containing a finite, relatively
small number of identical subunits. Protomers are defined
as the identical subunits associated within an oligomeric
protein. A monomer is defined as the fully dissociated
protomer, or any protein that is not made up of subunits. A
subunit is purposely undefined, and may be used to refer to
any chemically or physically identifiable submolecular
entity within a protein, whether identical to or different
from, other components. Using these definitions, the
haemoglobin tetramer (comprised of two a and two b
polypeptide chains) is defined as an oligomer consisting of
twoprotomers, each consistingoftwomonomers,i.e. onea
and one b polypeptide chain. The definition of a subunit
allows the term to be used for either the a- or b-monomer,
or for the ab-protomer. The term multimer is also widely
used in the literature and is defined here as a protein with a
finite number of subunits that need not be identical.
The quaternary nature of some proteins was first
identified from centrifugation experiments devised in the
1920s to calculate the molecular weights of proteins. Since
then, combinations of association and hybridization
techniques have led to the discovery of a large number of
proteins possessing quaternary structure. To date, the
three-dimensional structures of many hundreds of multimeric
proteins with identical subunits (oligomers) and
nonidentical subunits have been solved by X-ray crystallography.
An example of an oligomeric protein solved by
this method is aspartyl protease from human immunodeficiency
virus 1 (HIV-1) retrovirus (Navia et al., 1989). This
protein functions to release structural proteins and
enzymes (such as reverse transcriptase and integrase) from
viral polyprotein products. HIV-1 protease is a homodimer,
in which each subunit comprises almost exclusively
b sheet, turn and extended polypeptide structural elements
(Figure 1). Each subunit contributes one highly conserved
Asp-Thr-Gly catalytic triad sequence to form a symmetric
active site in the dimer. An example of a multimeric protein
with nonidentical subunits solved by X-ray crystallography
is the glycoprotein hormone, human chorionic
gonadotrophin (Lapthorn et al., 1994). This protein is
secreted by the placenta in the early weeks of pregnancy
and stimulates the secretion of the steroid progesterone.
The protein is an ab-heterodimer in which each subunit has
a similar extended topology that includes a cysteine knot
motif, common to a number of growth factors. The a and b
Article Contents
Secondary article
. Introduction
. Quaternary Structure Assembly
. Folding and Function
. Protein–Protein Recognition Sites
. Concluding Remarks
1ENCYCLOPEDIA OF LIFE SCIENCES © 2001, John Wiley & Sons, Ltd. www.els.net
subunits are integrally associated with a segment of the b
subunit wrapped around the a subunit and covalently
linked by a disulfide bond (Figure 1).
Quaternary Structure Assembly
Protein stoichiometry
The stoichiometry of proteins possessing a quaternary
structure considers the number of subunits involved in the
assemblies. The association of subunits in quaternary
structure can lead to the formation of closed structures, of
which dimers and tetramers are by far the most frequently
observed. In addition, quaternary proteins can exhibit
open elongated polymer structures. The way in which
different numbers of protein subunits associate to form
aggregates has been defined as macroassociation.
The process of macroassociation is divided into three
modes, namely heterologous, isologous and pseudoisologous.
Each can be defined using two terms: binding set (the
residues of one protomer involved in binding to one other
protomer), and domain of bonding (the two, linked
binding sets) (Monod et al., 1965). In heterologous
associations the domain of bonding is made up of two
different binding sets, and in isologous associations the two
binding sets involved are identical (Figure 2). In pseudoisologous
associations, the domain of bonding comprises
two almost identical binding sets. In isologous associations
the binding set of each protomer is ‘covered’ by the
equivalent binding set on the other protomer, hence these
associations tend to lead to finite closed structures.
Figure 1 Molscript diagrams depicting the secondary structure elements and the quaternary structure of (a) homodimeric interleukin 8; (b)
heterodimeric human chorionic gonadotrophin; (c) homodimeric HIV-1 protease; (d) heterotetrameric human haemoglobin. In each diagram the protein
subunitsaredifferentiatedbycolour,andin (d)oneab-protomerof haemoglobin iscolouredred andonegreen.Thehaem groupsin eachsubunitshownin
(d) are depicted by ball-and-stick representations.
a b
b a
b a b a b a
d c d c d c
(a) (b)
Figure 2 Modes of association in multimeric proteins: (a) isologous, in
which the binding sets (indicated by the letters ‘a’ and ‘b’) are identical; (b)
heterologous, in which the binding sets (indicated by the letters ‘a’, ‘b’, ‘c’
and ‘d’) are not identical. Adapted from Monod et al., 1965.
Protein Quaternary Structure: Subunit–Subunit Interactions
2
Heterologous associations can involve multiple binding
sets on a single protomer that can lead to infinite open
structures. If an oligomer has an odd number of equivalent
protomers, then the associations between them must be
heterologous; however, if an oligomer has an even number
of equivalent protomers, the associations can be either
isologous, heterologous or a mixture of the two. Some
elongated proteins, such as actin, are formed by indefinite
heterologous association of globular protomers; however,
heterologous associations can also lead to closed structures,
such as those observed in the trimeric bacteriochlorophyll
protein and the tetrameric manganese superoxide
dismutase.
The definition of different modes of association raised
the question of which one is most prevalent amongst
proteins. Monod et al. (1965) proposed that the exclusive
use ofisologous associations would lead only to dimers and
tetramers. In support of this is the prevalence of dimers and
tetramers amongst the proteins solved, and by implication
isologous associations. However, thermodynamic calculations
on the possibilities of all-isologous, all-heterologous
and mixed structures give no indication that isologous
associations are more energetically favourable than
heterologous.
Protein stereochemistry
Stereochemistry (the spatial arrangements of subunits
within a structure) at the quaternary structure level
involves theconceptof symmetry. Aninitial understanding
of the importance of symmetry in oligomeric proteins was
derived principally from the comparative studies of
myoglobin and haemoglobin. The importance of symmetry
in terms of protein structures was also introduced in the
theoretical model of the allosteric effects of enzymes
(Monod et al., 1965). From these, and the increasing
number of protein structures solved by X-ray crystallography,
it was found that many subunits of oligomeric
proteins (thosewith identical subunits)were organizedinto
stable arrays with high symmetry. For example, in general,
proteins with two identical subunits (e.g. malate dehydrogenase
and triosephosphate isomerase) have their subunits
arranged with twofold rotational symmetry. In a similar
way, proteins with three identical subunits (e.g. bacteriochlorophyll
protein and 2-keto-3-deoxy-6-phosphogluconic
aldolase) generally have their subunits arranged with
threefold rotational symmetry. Tetrameric proteins (e.g.
glyceraldehyde-3-phosphate dehydrogenase and lactate
dehydrogenase) commonly exhibit dihedral 222 symmetry.
There are, however, oligomeric proteins that do not exhibit
symmetry at thequaternarylevel of structure. One example
of this is observed in yeast hexokinase. This structure
contains two identical subunits related by a rotational and
a transitional symmetry element (not the expected 1808
rotation). In addition it is possible that in some structures
the association itself may be symmetrical but minor
structural changes between the subunits may exist. Such
changes generally only exist in the crystal form of the
protein, where identical, symmetrically related subunits
might be in an anisotropic crystal environment.
The symmetry of an oligomeric protein affects its
properties and functions, and can be crystallographic or
noncrystallographic. The unit cell of a crystal is the basic
building block, repeated infinitely by translation in three
dimensions. The asymmetric unit is the basic repeating
unit, which is related to all the other identical units in the
unit cell by the operation of the symmetry elements. If the
volume of the asymmetric unit of a crystal accommodates
just one subunit of an oligomeric protein, then the other
subunit(s) will be related to it by the same symmetry
operation(s) that relate the asymmetric units to each other.
In this case the symmetry of the oligomer is expressed in the
crystallographic symmetry. Alternatively, the asymmetric
unit of the crystal may accommodate the whole oligomer
or more than one oligomer. In this case the symmetry of the
protein will not be determined by the crystallographic
symmetry but by a noncrystallographic symmetry opera-
tion.
All proteins exhibiting quaternary structure can be
termed biological complexes, in that they are associations
known to exist in solution and hence, by inference, in the
cell. A completely different set of protein–protein complexes
are those represented by associations observed in
crystal packing, termed crystallographic complexes. The
problem of distinguishing between crystallographic complexes
and true biological complexes is a difficult one.
Protein–protein interactions in crystal packing differ
significantly from biological complexes. Crystal packing
contacts have no biological role and hence are not subject
to evolutionary pressures.
Folding and Function
The subunits of multimeric proteins can be considered as
independent folding units. In these structures, protein
folding probably begins with the folding of the independent
subunits (the same as in monomeric proteins) and
continues until the formation of a specific recognition site
that can be identified by another monomer. At this stage
the folding pathway shifts from being intramolecular to
intermolecular, to yield a dimer structure. The dimeric
structure may then undergo further folding steps to form a
native protein or a folding intermediate, with a further
recognition site that permits a second association step to
occur. In this way the folding of multimeric structures is a
succession of monomolecular folding steps and bimolecular
association steps (Jaenicke, 1987). The high specificity
of the association step is fundamental to the correct
folding of multimeric proteins.
Protein Quaternary Structure: Subunit–Subunit Interactions
3
Multimeric proteins are, in general, functionally more
versatile than proteins comprising a single polypeptide
chain. It is important to consider the difference between the
sum of the isolated subunits and the complete monomer. In
multienzyme complexes individual subunits catalyse distinct
consecutive reactions. In such structures the proximity
of the reaction sites leads to enhanced activity by
providing higher local concentrations of substrate about
the active site of the subunit catalysing the second reaction.
An example of this is observed in tryptophan synthetase,
an enzyme that catalyses the final reactions of tryptophan
biosynthesis. The a2b2-multimer is in equilibrium with its
constituent a and b subunits. The subunits show two
distinct functions: the a subunit catalyses the conversion of
indole-glycerol phosphate to indole, while the b subunit
(usually present as a dimer b2) catalyses the formation of
tryptophan from indole. The complete multimer exhibits a
higher rate of the reactions than the isolated subunits.
Also significant are protein functions that result from
intersubunit contacts, such as allosteric interactions.
Aspartate transcarbamoylase in Escherichia coli is a wellstudied
allosteric enzyme that shows cooperative effects in
substrate binding and is subject to feedback control. The
quaternary structure of this enzyme includes two types of
subunit, a catalytic subunit and a regulatory subunit. In the
native enzyme there are two catalytic subunits and three
regulatory subunits present, and the cooperative mechanism
is achieved through the interaction between one pair of
catalytic subunits connected by a bridging regulatory
subunit.
As well as catalytic functions, quaternary structure may
also serve to confer additional stability to protein
structures. It is also possible that proteins form quaternary
structures in certain conditions to avoid an excessive
osmotic pressure. Another possible function is that a
greater size of macromolecules may be important in
compartmentalization within the cell or in protein turnover.
The association of protein subunits is also the basis of
some common diseases. In sickle cell anaemia a single
mutation in the b subunit of haemoglobin causes the
deoxygenated form of haemoglobin to polymerize into
long fibres. Alzheimer disease is characterized by the
association of b-amyloid proteins to form brain lesions
termed senile plaques. These last two examples illustrate
how protein associations can result in adverse as well as
advantageous effects in the cell.
Protein–Protein Recognition Sites
Protein–protein associations that occur in the quaternary
structure formation involve the specific complementary
recognition of two macromolecules to form a stable
assembly. The recognition process involves factors favouring
and opposing the stable association. Hydrophobic and
electrostatic interactions favour the association. The loss
of translational and rotational freedom of amino acids on
binding opposes the association. The affinity for two
molecules is determined by the change in energy and
entropy of a system that contains the two proteins and
solvent and the complex and solvent; however, the lack of
experimental binding-association data has meant that the
relative contributions of factors contributing to the
binding energy of association remains unclear.
Hydrophobic interactions
The hydrophobic interaction is considered to be the
primary driving force in the stabilization of protein
associations. The term ‘hydrophobic interaction’ is used
to describe the gain in free energy upon the association of
nonpolar residues of proteins in an aqueous environment.
The process of folding and protein–protein aggregation
reduces the surface of a protein in contact with water. This
is the structural basis of the hydrophobic effect in proteins.
The folding of polypeptide chains and aggregation of
subunits buries the hydrophobic residues of the proteins,
and hence minimizes the number of thermodynamically
unfavourable solute–solvent interactions. The quantitative
evaluation of exactly how much hydrophobic interactions
contribute to the stabilization of protein–protein
associations is controversial. The controversy is based on
different definitions and interpretations of the hydrophobic
effect in proteins. Empirical calculations have led to
energy values of between, 25 and 72 calories per A˚ 2
(1 A˚ 2
5 0.01 nm2
) of accessible surface area gained on
association. These energy values are important when
considering the minimum size of recognition sites in
multimeric proteins and other protein–protein complexes.
Electrostatic interactions
Electrostatic interactions, in addition to hydrogen bonds
and van der Waals interactions, are considered of
secondary importance in protein associations (Chothia
and Janin, 1975); however, the hydrogen bond (a polar
interaction between donor and acceptor electronegative
atoms) is an intrinsic component of protein–protein
interactions. Hydrogen bonds between protein molecules
are more favourable than those made with water, and
hence intermolecular hydrogen bonds contribute to the
binding energy of association. It has been proposed that
whereas hydrophobic forces drive protein–protein interactions,
hydrogen bonds and salt bridges confer specificity
(Fersht, 1987). It has been observed that the geometry of
hydrogen bonds across protein–protein interfaces (such as
those in multimeric proteins) are generally less optimal and
have a wider distribution than those observed in the
interior of proteins. This leads to the proposal that
intermolecular hydrogen bonds are weaker than those in
Protein Quaternary Structure: Subunit–Subunit Interactions
4
protein interiors. Salt bridges across the binding interface
of multimeric structures can also significantly enhance
stability in some complexes.
Van der Waals interactions occur between all neighbouring
atoms, but those interactions at the interface are
not more energetically favourable than those made with
the solvent; however, they are more numerous, as the
tightly packed interfaces are more dense than the solvent.
Hence these interactions also contribute to the binding
energy of association.
Shape complementarity
The complementarity of protein interfaces is derived from
both electrostatic interactions and shape. Shape complementarity
has been characterized by the size of the buried
surface, and the packing density of interface atoms. Many
methods (Chothia and Janin, 1975; Lawrence and Colman,
1993; Jones and Thornton, 1996) have been employed to
measure packing of protein–protein subunits, with the
general conclusion that subunits in multimeric proteins are
tightly packed. Such methods have also revealed that there
are differences in interface packing between different types
of protein–protein complex. Proteins exhibiting quaternary
structure have protein–protein interfaces that are more
closely packed than other types of protein–protein
associations, such as enzyme–inhibitor and antibody–
antigen complexes. Such differences possibly reflect the
evolutionary time scale of these structures (Jones and
Thornton, 1996).
Recognition site properties
The number and type of interactions in multimeric proteins
are generally considered in relation to the area of the
subunit interface. This can be measured in terms of
accessible surface area (ASA). The native structure of
proteins exists only in the presence of water, and the ASA
describes the extent to which protein atoms can form
contacts with water. Lee and Richards (1971) were the first
to propose the concept of ASA, defining it as the area of a
sphere of radius R, on each point of which the centre of a
solvent molecule can be placed in contact with an atom
without penetrating any other atoms of the molecule. The
radius R, is given by the sum of the van der Waals radius of
the atom and the chosen radius of the solvent molecule
(Figure 3). The problem with this definition is that it implies
that the system is static: it does not account for any
movement or flexibility that an atom or group may possess
within the molecule.
The deposition of the three-dimensional coordinates of
protein structures (solved by X-ray crystallography and
nuclear magnetic resonance) in the Protein Data Bank
(PDB) Bernstein et al., 1977), has permitted the analysis of
relatively large numbers of multimeric proteins. Many
computational studies (Miller et al., 1987; Argos, 1988;
Jones and Thornton, 1996) have analysed the properties of
the recognition sites of multimeric proteins in comparison
with the protein exterior and protein interior. Such studies
have looked for common trends in terms of the size and
shape, amino acid composition, hydrogen bonding and
secondary structure.
Size and shape
Subunits in protein dimers contribute 6–40% of their ASA
to the contact interface; the mean is 12%. For trimers and
tetramers the means are 17% and 21%, respectively (Miller
et al., 1987). It has been predicted that 5–6% of ASA of the
subunit must be contributed to the contact interface as a
minimum requirement for its stabilization (Argos, 1988).
In terms of absolute ASA buried by each subunit, the
ranges for dimers is very large, with small areas recorded
forstructures suchas434repressor(368 A˚ 2
)andlarge areas
for structures such as in citrate synthase (4746 A˚ 2
) (Jones
and Thornton, 1995). In homodimers the ASA is
approximately linearly related to the molecular weight of
the protomer. Thus the larger the protomer, the larger the
interface site required to stabilize its interaction with a
second protein subunit.
When viewed as an overall or global cross-section,
interfaces are generally flat; however, exceptions have been
noted. These include proteins in which the two subunits
twist together across the interface (e.g. isocitrate dehydrogenase)
or proteins that have subunits with ‘arms’ that
clasp the two halves of the structure together (e.g. aspartate
aminotransferase).
Amino acid composition
The protein–protein interfaces in multimeric proteins are
largely hydrophobic. These interfaces have been shown to
be more hydrophobic than the exterior but less hydrophobic
than the interior. Calculating residue interface
R
Atom Atom
Solvent
probe
Accessible
surface
Figure 3 The accessible surface area of a protein. The diagram shows
just two atoms, with a probe sphere (with a radius of R) defining the
accessible surface.
Protein Quaternary Structure: Subunit–Subunit Interactions
5
propensities, which give an indication of the relative
importance of different amino acids in the interface
compared with the protein surface as a whole, reveals that
specific amino acids have high probabilities of being
present in protein–protein interfaces compared with their
frequency on the exposed surface of the protein (Jones and
Thornton, 1996). The hydrophobic residues occur frequently
in the interfaces, along with the single aromatic
residues, histidine, tyrosine and phenylalanine, which
make particularly good ‘glue’ for sticking together protein
subunits (Argos, 1988).
Electrostatic interactions
The number of intermolecular hydrogen bonds is approximately
proportional to the ASA buried in the interface. In
homodimers there are, on average, 0.88 hydrogen bonds
per 100 A˚ 2
of ASA buried (for interfaces covering
4 1500 A˚ 2
per subunit); but the number of hydrogen
bonds varies from zero in some complexes (e.g. uteroglobin)
to as many as 46 in variant surface glycoprotein (Jones
and Thornton, 1995). Side-chain hydrogen bonds represent
approximately 76–78% of the interactions. Salt
bridges have also been observed between subunits of
multimeric proteins, but only 56% of homodimeric
proteins were found to possess such interactions, many
having none or, at the most, five. Intermolecular disulfides
are rarely seen in dimeric proteins, as they only occur in
oxidizing environments; however, when intermolecular
disulfides do occur they often play an important role in
structural stabilization. Protein engineering experiments
on two structures, platelet-derived growth factor B and
thymidylate synthase, have shown in both that the
introduction of intermolecular disulfides increases the
stability of the protein associations.
Secondary structure
Interfaces in multimeric proteins occur between helix,
sheet and coil motifs, with both like and nonlike
interactions observed across the interface. Interfaces
commonly have a central area of extended sheet, helix–
helix packing or sheet–sheet packing decorated at the
edges by loop interactions. The loop interactions contributed
on average 40% of the interface contacts (Miller,
1989). The loops commonly interact with other loops and
with the ends of secondary structures, and are stabilized by
large numbers of hydrogen bonds. Motifs are often shared
across interfaces; stability within interfaces is enhanced by
converting loops within motifs into linkers across inter-
faces.
Concluding Remarks
Quaternary structure is the highest level of protein
organization. The quaternary structure of a protein is
fundamentally important to its functional role in the cell.
Chemical and physical properties, including hydrophobic
interactions, electrostatic interactions and shape complementarity,
play complex roles in the interaction of one
protein subunit with another. The importance of these
properties varies depending upon the type of complex and
its function.
References
Argos P (1988) An investigation of protein subunit and domain
interfaces. Protein Engineering 2: 101–113.
Bernstein FC, Koetzle TF, Williams GJB et al. (1977) The protein data
bank: a computer-based archival file for macromolecular structures.
Journal of Molecular Biology 112: 535–542.
Chothia C and Janin J (1975) Principles of protein–protein recognition.
Nature 256: 705–708.
Fersht AR (1987) The hydrogen bond in molecular recognition. Trends
in Biochemical Sciences 12: 301–304.
Jaenicke R (1987) Folding and association of proteins. Progress in
Biophysics and Molecular Biology 29: 117–237.
Jones S and Thornton JM (1995) Protein–protein interactions: a review
of protein dimer structures. Progress in Biophysics and Molecular
Biology 63: 31–65.
Jones S and Thornton JM (1996) Principles of protein–protein
interactions. Proceedings of the National Academy of Sciences of the
USA 93: 13–20.
Lapthorn AJ, Harris DC, Littlejohn A et al. (1994) Crystal structure of
human chorionic gonadotropin. Nature 369: 455–461.
Lawrence MC and Colman PM (1993) Shape complementarity at
protein/protein interfaces. Journal of Molecular Biology 234: 946–950.
Lee B and Richards FM (1971) The interpretation of protein structures:
estimation of static accessibility. Journal of Molecular Biology 55: 379–
400.
Miller S (1989) The structure of interfaces between subunits of dimeric
and tetrameric proteins. Protein Engineering 3: 77–83.
Miller S, Lesk AM, Janin J and Chothia C (1987) The accessible surface
area and stability of oligomeric proteins. Nature 328: 834–836.
Monod J, Wyman J and Changeux J (1965) On the nature of allosteric
transitions: a plausible model. Journal of Molecular Biology 12: 88–
118.
Navia MA, Fitzgerald PMD, McKeever BM et al. (1989) Threedimensional
structure of aspartyl protease from human immunodeficiency
virus HIV-1. Nature 337: 615–620.
Further Reading
Banaszak LJ, Birktoft JJ and Barry CD (1981) Protein–protein
interactions and protein structures. In: Protein–Protein Interactions,
pp. 31–128. New York: Wiley.
Chan WW (1976) The relationship between quaternary structure and
enzyme activity. Trends in Biological Sciences 11: 258–260.
Duquerroy S, Cherfils J and Janin J (1991) Protein–protein interaction:
an analysis by computer simulation. In: Chadwick DJ and Widdows K
(eds) Protein Conformation, pp. 237–252. Chichester, UK: Wiley.
Garel JR (1992) Folding of large proteins: multidomain and multisubunit
proteins. In: Creighton TE (ed.) Protein Folding, pp. 405–454.
New York: Freeman.
Garel JR, Martel A, Muller K et al. (1984) Role of subunit interactions in
the self-assembly of oligomeric proteins. Advances in Biophysics 18:
91–113.
Protein Quaternary Structure: Subunit–Subunit Interactions
6
Klotz IM, Darnell DW and Langerman NR (1975) Quaternary structure
of proteins. In: Neurath H and Hill RL (eds) The Proteins, 3rd edn, pp.
25–62. New York: Academic Press.
Matthews BW and Bernhard SA (1973) Structure and symmetry of
oligomeric enzymes. Annual Review of Biophysics and Bioengineering
2: 257–317.
Riddihough G (1994) The evolution of oligomerization. Nature
Structural Biology 1: 411–412.
Weber G (1992) Protein Interactions. London: Chapman and Hall.
Protein Quaternary Structure: Subunit–Subunit Interactions
7