Molecular Recognition by Induced Fit: How Fit is
the Concept?
Hans Rudolf Bosshard
Department of Biochemistry, University of Zurich, CH-8057 Zurich, Switzerland
Induced fit explains why biomolecules can bind together even if they are not optimized for binding.
However, induced fit can lead to a kinetic bottleneck and does not describe every interaction in
the absence of prior complementarity. Preselection of a fitting conformer is an alternative to
induced fit.
irtually all biological phenomena depend in one way or
another on specific molecular recognition. At the end of
the 19th century, Emil Fischer coined his famous lock-and-key
analogy to picture the specificity of enzyme reactions, which
are a molecular premise of life (4). The enzyme was considered
to be a rigid template in which the substrate had to fit as a key
into a lock. Over the years, however, it became apparent that
a rigid fit between preformed molecular structures cannot
explain all aspects of enzyme catalysis. For example, why
should a smaller substrate not fit into the active site of an
enzyme designed for a larger substrate? Or why are some
enzymes highly selective but others may accommodate several
structurally different substrate molecules?
It is in this context that, over 40 years ago, Daniel Koshland
formulated the concept of the induced fit (7). To facilitate the
enzymatic reaction in the absence of a precise fit, he postulated that “the substrate may cause an appreciable change in
the three-dimensional relationship of the amino acids at the
active site” (7). The idea of a precise fit was retained from the
lock-and-key image, but it was stated explicitly that the fit
“occurs only after the changes induced by the substrate itself”
(emphasis original). The concept soon became textbook
knowledge and has since been used to explain all kinds of
molecular recognition processes far beyond enzyme-substrate
reactions. Indeed, structural analysis of interacting biomolecules has established again and again that a complex and its
free component molecules may differ in fine details of structure, in apparent support of recognition by induced fit. A good
case in point is provided by several antigen-antibody complexes for which spatial adaptation has been demonstrated by
high-resolution crystal structure analysis (2). Structural plasticity is also evident in other protein-protein interactions (12).
Limits of induced fit
Why should one question a time-honored concept? My reason is that the induced fit paragon often is too ready at hand to
explain why molecules without apparent structural complementarity interact. The problem lies in the original assumption
that a favorable fit develops only after initial binding, which is
often taken too literally. Considering the kinetics and thermodynamics of a binding reaction, induced fit is possible only if
the match between the interacting sites is strong enough to pro0886-1714/99 5.00 © 2001 Int. Union Physiol. Sci./Am.Physiol. Soc.
vide the initial complex enough strength and longevity so that
induced fit takes place within a reasonable time. I wish to illustrate this crucial point by a simple model calculation based on
the thermodynamic cycle shown in Fig. 1. Consider molecules
A and B, which may be enzyme and substrate, antigen and antibody, hormone and receptor, or any other pair of interacting
molecules. For the sake of simplicity I also assume that the
induced fit occurs only in molecule B, which is changed into B*
to form the stable complex AB*. (This assumption does not
affect the result of the calculation but simplifies the mathematical formalism.) Binding by induced fit is described by reactions
1 and 2 of Fig. 1. In reaction 1, A and B interact to form the initial complex AB, which is a loosely bound pair of molecules.
Precise and energetically favorable interactions form afterward
in reaction 2, in which B is forced into conformation B* by
induced fit. The energy for “pulling” and “pushing” of B into the
fitting conformation originates from the optimized fit achieved
in the final complex AB*. The overall binding constant for complex AB* is K = K1 × K2 = k1 × k2/k1 × k2 (see Fig. 1 for definitions of binding constants and kinetic rate constants).
As a practical example, envision an antigen-antibody complex for which K typically is of the order of 108 M1. Since reaction 1 is unfavorable, I take K1 to be 1 M1, from which it follows that K2 = 108 M1. Because K2 >> K1, the equilibrium is
well on the side of the antigen-antibody complex AB*. For
example, if 1 × 106 M antibody reacts with 1 × 106 M antigen, the thermodynamic equilibrium is 91% on the side of the
antigen-antibody complex (calculated from K = 108 M1 and
[antigen]total = [antibody binding site]total = 1 × 106 M, where
square brackets indicate concentration). The important question, which is often overlooked, is how long it will take to
reach this equilibrium concentration. The time course is
described by
Downloaded from http://physiologyonline.physiology.org/ by 10.220.33.6 on June 17, 2017
V
d[AB]
= k1 [A][B] k1[AB]
dt
(1)
d[AB*]
= k2[AB] k2[AB*]
dt
(2)
To calculate the time to reach equilibrium, one needs reasonable estimates of k1 and k2 (k1 and k2 follow from the chosen values of K1 and K2; see Fig. 1). The rate of formation of a
very weak complex may be anywhere between 102 and 106
M1·s1. I choose k1 = 104 M1·s1; however, the value of k1 does
News Physiol. Sci. • Volume 16 • August 2001
171
not have much effect on the result of the calculation. Conformational changes in proteins have half-times of milliseconds, so
I choose k2 = 102 s1 (half-time = 7 ms, from ln2/k2). The corresponding values of the back reactions are then k1 = 104 s1 and
k2 = 106 s1. Numerical integration of equations 1 and 2 for
the above rate constants and 1 × 106 M starting concentrations
of antigen and antibody gives a half-time for the formation of
the antigen-antibody complex of 2.5 h. Thus reaching equilibrium takes 1 day (~10 half-times). The reason for this long
reaction time is that, once formed, the unstable antigen-antibody complex AB has only a very small chance to undergo the
induced fit transition to the stable antigen-antibody complex
AB*. The induced fit, although thermodynamically reasonable,
is too slow to be meaningful as a biological reaction.
Downloaded from http://physiologyonline.physiology.org/ by 10.220.33.6 on June 17, 2017
FIGURE 1. Thermodynamic cycle for the reaction of molecules A and B to
complex AB*, where B and B* are different conformational states of the same
molecule. The induced-fit pathway follows reactions 1 and 2. The initial complex AB formed in reaction 1 is not stable because the conformation of B is
not optimized. Induced fit reaction 2 brings B into the fitting conformation B*.
The conformational selection pathway follows reactions 3 and 4. Reaction 3
describes the conformational equilibrium between the nonfitting conformation B and the fitting conformation B*. Reaction 4 is the binding of the fitting
conformation B* to A. The induced-fit pathway is kinetically competent only
if complex AB has appreciable stability so that the induced fit has a reasonable chance to occur. If this is not the case and a small amount of the fitting
conformation B* is present in the absence of A, the conformational selection
pathway dominates.
conformational selection, I assume that one in a thousand molecules is in conformation B* (K3 = 103). It follows that K4 =
1011 M1 since the thermodynamic cycle of Fig. 1 has to be satisfied according to K1 × K2 = K3 × K4 = K = 108 M1. I further
assume that the conformational change B B* is as fast as the
induced fit (k2 = k3 = 100 s1). Furthermore, one needs a value
of the rate of association of B* with A in reaction 4. Measured
rate constants for antigen-antibody binding are in the range
104107 M1·s1 (8, 9, 11). I choose a middle value of k4 = 106
M1·s1. Under these conditions, the formation of the antigenantibody complex AB* through reactions 3 and 4 has a halftime of only 80 s; the equilibrium concentration of AB* is
achieved in <15 min.
These calculations demonstrate that binding by induced fit
makes sense only if there is a certain extent of preexisting complementarity between the interacting species; otherwise the
initial complex AB is too short lived (an assumption implicit in
Koshland’s initial paper). By using the above forward rate constants k1 and k2 and 1 × 106 M initial concentrations, one finds
that K1 has to be 104 M1 to attain the same half-time of 80 s
for the induced fit pathway as for the conformational selection
pathway.
Experimental demonstration of binding by conformational selection
Surprisingly few studies have attempted to show conformational selection despite the fact that it is more easily demonstrated than induced fit (1, 5, 8, 15). To demonstrate induced fit,
one would have to sample structural information over the
course of the binding reaction, a rather difficult task. To demonstrate conformational selection, one has to show that the rate of
formation of the complex AB* is linearly proportional to the
concentration of the fitting conformer B* and nonlinearly proportional to the total concentration of B + B*. If one can measure the conformational equilibrium reaction 3 in the absence of
Conformational selection is an alternative to induced fit
Other researchers, including myself (1, 3, 5, 8, 13, 14) have
pointed out that there is an alternative mechanism to induced
fit. The essence of conformational selection, described by reactions 3 and 4 of Fig. 1, is that the conformation change is not
assumed to occur after initial binding. This is a rather obvious
assumption. Take the antigen-antibody complex AB*. It dissociates into free antibody A and free antigen B* through reaction
4. Hence B* occurs in isolation even though it may be only a
short-lived minority conformation, rapidly equilibrating with
the major conformation B through reaction 3. To calculate how
long it takes to reach the equilibrium concentration of AB* by
172
News Physiol. Sci. • Volume 16 • August 2001
FIGURE 2. Kinetics of the cross-reaction of antibody c11L34Ser with the
leucine zipper GCN4. A: the rate of the cross-reaction with GCN4 is linearly
proportional to the concentration of the unfolded peptide calculated from K3
of reaction 3 in Fig. 1 („) and nonlinearly proportional to the total concentration of the antigen (z), as predicted for the conformational selection pathway.
B: the rate of reaction with the original peptide is faster than the rate of the
cross-reaction with GCN4 as predicted for the conformational selection pathway 3 J 4 in Fig. 1. The traces are for the reaction of 4 × 107 M antibody with
4 × 106 M GCN4-7P14P (peptide P) and 4 × 106 M GCN4, respectively. Data
adapted from Ref. 1 with permission.
The energy landscape model of protein conformation
Folded proteins do not have a single unique structure but are
better regarded as a large ensemble of similar structures having
similar energy contents. These so-called conformers are in
rapid fluctuation with each other (10). If the energy landscape
is smooth, the many conformers interchange rapidly. If it is
rugged, the ensemble may include conformers that may be
quite different and interchange more slowly. Thus selection
between the structures B and B* is a grossly oversimplified
view. In reality it is more like a selection among very many
more-or-less fitting structures (13). However, the end result of
conformational selection is the same: those conformers that
show the best fit bind best.
Conclusions
Conformational selection is a valuable alternative to
induced fit. This is not to say that binding by induced fit does
not occur. Actually, a combination of conformational selection
and induced fit would seem to be the best description of the
interaction between molecules that apparently do not optimally fit to begin with. We may envision selection between
partly fitting structures followed by minor readjustments to the
final stable complex (which according to the landscape theory
is itself an ensemble of similar conformers). The main point is
that induced fit cannot be a cure-all as is often purported in the
literature. Induced fit requires some prior molecular match to
provide sufficient affinity before conformational adaptation. If
this condition is not met, induced fit leads into a kinetic bottleneck, even if the overall reaction is thermodynamically feasible.
I am grateful to many of my past and present colleagues for helpful and
stimulating discussions.
Work from my laboratory has been supported in part by the Swiss National
Science Foundation.
References
Downloaded from http://physiologyonline.physiology.org/ by 10.220.33.6 on June 17, 2017
the binding partner A, one can calculate the concentration of
B* and predict the overall velocity of formation of AB*.
A well-studied example is the reaction of the single-chain
antibody fragment directed against the 33-residue-long prolinecontaining peptide GCN4-7P14P, called peptide P for short (6).
The peptide is very similar in sequence to the leucine zipper
domain of the yeast transcriptional activator GCN4, except that
it does not form a leucine zipper because it contains two proline
residues. (Prolines are not compatible with the helical conformation of a leucine zipper, which is a dimer of helices wound
around each other.) However, the antibody cross-reacts with the
GCN4 leucine zipper, and this cross-reaction clearly follows a
conformational selection pathway. The antibody reacts with the
unfolded peptide, which is supplied from the folded leucine zipper (reaction 3). Comparing the original reaction of the antibody
with peptide P to the cross-reaction with the leucine zipper
GCN4, one expects that the cross-reaction is slower since the
antibody has to select a small amount of unfolded peptide in
equilibrium with a large amount of folded leucine zipper. Moreover, the rate of the cross-reaction with the predominantly folded
leucine zipper will show a nonlinear dependence on the total
antigen concentration yet a linear dependence on the concentration of the small amount of unfolded peptide (8, 15). This
would not be the case for an induced-fit mechanism, which
should be biphasic, with the first phase corresponding to a
bimolecular association reaction whose rate depends linearly on
the total antigen concentration and the second phase to a concentration-independent conformational rearrangement. Figure 2
shows the actual data. The rate of formation of the antigen-antibody complex depends linearly on the concentration of the
small amount of unfolded peptide and nonlinearly on the total
antigen concentration (Fig. 2A). Figure 2B shows the kinetic
traces of the more rapid reaction with the original antigen and
the slower reaction with the GCN4 leucine zipper. Nonetheless,
conformational selection of the unfolded leucine zipper is still
orders of magnitude faster than an induced-fit pathway because
binding of the antibody to the folded leucine zipper (reaction 1
of Fig. 1) is extremely weak (1).
1. Berger C, Weber-Bornhauser S, Eggenberger J, Hanes J, Plückthun A, and
Bosshard HR. Antigen recognition by conformational selection. FEBS Lett
450: 149153, 1999.
2. Davies DR and Cohen GH. Interactions of protein antigens with antibodies. Proc Natl Acad Sci USA 93: 712, 1996.
3. Dürr E and Bosshard HR. A monoclonal antibody induces opening of a
coiled coilglobal protection of amide protons from H/D exchange
decreases by up to 1000-fold in antibody-bound triple-stranded coiled
coil. Eur J Biochem 249: 325329, 1997.
4. Fischer E. Einfluss der Configuration auf die Wirkung der Enzyme. Ber
Dtsch Chem Ges 27: 29842993, 1894.
5. Foote J and Milstein C. Conformational isomerism and the diversity of antibodies. Proc Natl Acad Sci USA 91: 1037010374, 1994.
6. Hanes J, Jermutus L, Weber-Bornhauser S, Bosshard HR, and Plückthun A.
Ribosome display efficiently selects and evolves high-affinity antibodies in
vitro from immune libraries. Proc Natl Acad Sci USA 95: 1413014135,
1998.
7. Koshland DE Jr. Application of a theory of enzyme specificity to protein
synthesis. Proc Natl Acad Sci USA 44: 98104, 1958.
8. Leder L, Berger C, Bornhauser S, Wendt H, Ackermann F, Jelesarov I, and
Bosshard HR. Spectroscopic, calorimetric and kinetic demonstration of
conformational adaptation in peptide-antibody recognition. Biochemistry
34: 1650916518, 1995.
9. Mason DW and Williams AF. Kinetics of antibody reactions and the analysis of cell surface antigens. In: Handbook of Experimental Immunology
(4th ed.), edited by Weir DM, Herzenberg LA, Blackwell C, and Herzenberg LA. Oxford: Blackwell Scientific, 1986, p. 38.138.17.
10. Onuchic JN, Nymeyer H, Garcia AE, Chahine J, and Socci ND. The energy
landscape theory of protein folding: insights into folding mechanisms and
scenarios. Adv Protein Chem 53: 87152, 2000.
11. Raman CS, Jemmerson R, Nall BT, and Allen MJ. Diffusion-limited rates for
monoclonal antibody binding to cytochrome c. Biochemistry 31: 10370
10379, 1992.
12. Sundberg EJ and Mariuzza RA. Luxury accommodations: the expanding
role of structural plasticity in protein-protein interactions. Structure 8:
R137R142, 2000.
13. Tsai CJ, Ma BY, and Nussinov R. Folding and binding cascades: shifts in
energy landscapes. Proc Natl Acad Sci USA 96: 99709972, 1999.
14. Van Regenmortel MHV. Transcending the structural paradigm in immunologyaffinity and biological activity rather than purely structural considerations should guide the design of synthetic peptide epitopes. Biomed Pept
Proteins Nucleic Acids 1: 109116, 1995.
15. Zeder-Lutz G, Van Regenmortel MHV, Wenger R, and Altschuh D. Interaction of cyclosporin A and two cyclosporin analogs with cyclophilin:
relationship between structure and binding. J Chromatogr B Biomed Appl
662: 301306, 1994.
News Physiol. Sci. • Volume 16 • August 2001
173