Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
Reconciling the lock-and-key and dynamic views of canonical
serine protease inhibitor action
Zoltán Gáspária,*, Péter Várnaib, Balázs Szappanosa,1 & András Perczela,c
a
Structural Chemistry and Biology Laboratory, Institute of Chemistry, Eötvös Loránd University,
Budapest, Hungary
b
Department of Chemistry and Biochemistry, University of Sussex, Brighton, UK
c
HAS-ELTE Protein Modelling Group, Institute of Chemistry, Eötvös Loránd University, Budapest,
Hungary
*
Corresponding author.
E-mail: [email protected],
Structural Chemistry and Biology Laboratory, Institute of Chemistry,
Eötvös Loránd University, Pázmány Péter sétány 1/A, 1117 Budapest,
Hungary
Tel.: +36-1-3722500 ext. 1408,
Fax: +36-1-3722620
1
Present address: Evolutionary Systems Biology Group, Institute of Biochemistry, Biological Research
Center, Szeged, Hungary
1
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
ABSTRACT
The efficiency of canonical serine protease inhibitors is conventionally attributed to the rigidity of their
protease binding loop with no conformational change upon enzyme binding, yielding an example of the
lock-and-key model for biomolecular interactions. However, solution-state structural studies revealed
considerable flexibility in their protease binding loop. We resolve this apparent contradiction by
showing that enzyme binding of small, 35-residue inhibitors is actually a dynamic conformer selection
process on the nanosecond-timescale. Thus, fast time-scale dynamics enables the association rate to be
solely diffusion-controlled just like in the rigid-body model.
Keywords: Canonical serine protease inhibitor, nanosecond-timescale conformer selection, protease
binding, internal dynamics, dynamic conformational ensemble, population shift
2
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
1. INTRODUCTION
Canonical, standard mechanism serine protease inhibitors belong to several evolutionarily unrelated
families, yet possess a structurally similar segment, the protease binding loop, mainly responsible for
their interaction with the target enzyme. This loop binds into the substrate binding cleft of the cognate
protease(s) and is even cleaved in a substrate-like manner, albeit with a low rate and an equilibrium
constant near unity [1]. Canonical inhibitors are traditionally viewed as classical examples of the 'lockand-key' inhibitory mechanism, in which the free inhibitor is fairly rigid and structural changes are
restricted upon complexation [2-4]. This view is based on the observation that X-ray structures of
several free and complexed inhibitors show little structural differences in their protease biding loop, as
measured by the deviation of backbone dihedral angles [3]. However, heteronuclear NMR studies have
revealed that the interaction site of canonical protease inhibitors in complex with the enzyme is flexible
on the ps-ns time scale [5-9]. These observations are in line with recent experimental and theoretical
results emphasizing the role of internal dynamics in protein function in general [10-12] and are in
apparent contradiction with the rigid lock-and-key model widely accepted for canonical serine protease
inhibitors.
To assess the role of ps-ns timescale dynamics in canonical serine protease inhibition, we performed
dynamically restrained molecular simulations on small, 35-residue inhibitors that exhibit picomolar
inhibition of their target enzymes. These highly potent inhibitors are therefore ideal models to study the
role of internal flexibility in protein binding. Interestingly, generalized order parameters (S2 values)
derived from NMR spectroscopy data indicate low motional restriction of backbone N-H moieties in
the range of 0.6-0.7 throughout the backbone, not just in their primary protease binding region (0
corresponds to totally unrestricted, while 1 to restricted motion; [13]. Similar values are characteristic
of interaction sites in proteins, including the enzyme binding regions of larger, unrelated canonical
inhibitors. One of the inhibitors studied here, SGCI (Schistocerca gregaria chymotrypsin inhibitor),
3
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
has shown overall changes upon binding to its target enzyme as followed by NMR spectroscopy,
indicating that residues far from the primary contact site are influenced by complex formation [14].
Another inhibitor, SGTI (Sch. gregaria trypsin inhibitor) possesses an extended enzyme binding site
adjacent to the strictly defined protease binding loop [15]. A recent phage-display study revealed that
residues involved in structural stabilization affect the enzyme selectivity of SGTI variants [16]. Thus,
these inhibitors can be regarded as 'whole-molecule binding loops', globally influenced by the
interaction with the enzyme, rendering the recognition and analysis of the role of internal dynamics
easier compared to systems where only limited local changes occur.
2. METHODS
Dynamically restrained molecular dynamics simulations were performed in GROMACS [17] as
described previously [18, 19]. Briefly, 8 replicas of the inhibitors were simulated in parallel in explicit
water. The simulations were restrained to reproduce the generalized order parameters (S2), calculated
across all replicas at each simulation step, and the NOE distance restraints that were treated in a
pairwise manner. Four 230 ps long simulated annealing cycles yielded 32 final structures in the
dynamic ensembles. Principal component analysis was done with the PCA_NEST [20] server,
neighbor-joining analysis on different metrics with the PHYLIP package and all other analyses with inhouse programs. Detailed description of the methods can be found in the Supplementary material.
3. RESULTS AND DISCUSSION
We have generated dynamically restrained ensembles of SGCI and SGTI by the MUMO (minimal
under-restraining minimal over-restraining) protocol [18, 19] using NOE [21] and backbone N-H S2
data [13]. The structural diversity in the resulting ensemble primarily stems from the internal dynamics
of the inhibitor on the ps-ns time scale. These dynamic structural ensembles exhibit excellent
4
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
agreement with experimental data including chemical shifts that were used as independent measures for
cross-correlation analysis (Fig. 1). In fact, correlation of back-calculated Hα shifts and experimental
data improved considerably, compared with the structures determined using 'conventional' singleconformer NOE refinement (referred to as SCR) protocol (Fig. S1). The structural ensembles of SGCI
and SGTI determined by the MUMO protocol show considerable heterogeneity in accord with the low
experimental S2 values, mapping a remarkably larger conformational space than their SCR counterparts
(Fig 2.). Nevertheless, variations in the structures do not adversely affect the recalculation of S2 values
by procedures considering the anisotropy of N-H vector distributions [22], suggesting the robustness of
the computational procedure even for such flexible molecules (Fig. S2). Moreover, our detailed
analysis of selected sidechain-sidechain interactions in the two molecules show that the dynamic
ensembles are not simply overly relaxed conformer sets with a uniformly high RMSD. Residues
considered pivotal in structure stabilization, the Phe10-Ala26 pair in SGCI and the Lys10-Trp25 pair in
SGTI [23] are, on average, more restricted in the dynamic ensembles compared with structures
obtained by SCR (Fig 2).
To assess the role of ps-ns timescale dynamics in protease binding, we compared the MUMO
ensembles of SGCI and SGTI with available X-ray structures of their enzyme-bound counterparts [15]
(for SGCI, the chymotrypsin-bound structure of one of its close homologues is used; [24]). Neighborjoining analysis using pairwise conformer distances based on both RMSD data and backbone dihedral
angle (φ and ψ) differences show that complexed inhibitors cluster within the conformational space of
the dynamic structures (Figs S3 and S4). Principal component analysis [20] strongly supports the
observation that backbone conformations present in the bound state are also accessible in solution. In
contrast, structures of the same inhibitors determined by using the SCR protocol represent a conformer
set distinctly different from the complexed states (Fig 3). We note that the present analysis of backbone
motions is robust since solely the backbone is restrained by both NOE and S2 data in the simulations.
5
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
Analysis of the distributions of backbone dihedral angles in the protease binding loop region of SGCI
and SGTI shows that dynamic structures sample regions of the conformational space also adopted in
the bound state. On the other hand, many of the φ and ψ angles are not constrained to the 'canonical
values', derived from the analysis of static X-ray structures of free and enzyme-bound inhibitors [4]
which forms the basis of the rigid-body-interaction hypothesis. Therefore, we propose that
conformational changes inevitably occur upon enzyme binding, in contrast to the lock-and-key
mechanism. Solution-state NMR and computational studies on a variety of unrelated canonical
inhibitors corroborate this finding [5-9,25].
Picosecond-to-nanosecond timescale dynamics, deciphered from backbone S2 values, is faster than
molecular rotation and intermolecular association rates and thus has not generally been considered to
play significant roles in molecular recognition of canonical inhibitors by serine proteases. Motions on
this time scale are generally small-amplitude local fluctuations, in contrast to larger-amplitude
rearrangements on slower time scales (from μs to ms) typically associated with biological function
[26]. Our study reveals the presence of enzyme-bound conformers in the structural ensemble of the free
inhibitors in solution which is also consistent with earlier predictions of the rigid body model. In either
case, conformational motions do not adversely affect the association rate with proteases, measured to
be in the order of 106 M-1s-1 for the (unrelated) canonical inhibitor eglin c [27]. Interconversion from a
non-productive conformation to an active one takes place faster than interaction with the partner
protease since no significant energetic cost is required for conformational rearrangements in an
equilibrium ensemble. However, in contrast to the rigid binding loop model, different conformational
states may be selected by different enzymes from the dynamic pool of inhibitor structures in solution
[28]. Our results support an equilibrium shift model of the inhibitors upon enzyme binding, which also
explains the large chemical shift changes observed for SGCI during its interaction with chymotrypsin.
Motional restriction is expected to result in entropy loss for the inhibitor in action (for a pictorial
6
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
representation see Fig S5). A recent computational study [25] including an unrelated canonical
inhibitor is in line with this expectation but with the notable result of a net entropy gain for the whole
complex formed with protease. It should be noted that supra-ns dynamics of the free and complexed
states [14,29], together with site-specific interactions [3] and factors governing the dissociation rate of
the enzyme-inhibitor complex all contribute to the observed high efficiency of canonical serine
protease inhibitors.
Nanosecond-timescale conformational selection may well be a general phenomenon, as relatively low
S2 order parameter values indicate the presence of considerable motions of interaction sites in a number
of proteins. This also suggests that fast internal motions should also be considered to account for the
full details of fine-tuned biomolecular interactions and exploited in inhibitor design. Alas, neither the
locks nor the keys are actually rigid as envisioned by the early pioneers of structural biology [30].
4. ACKNOWLEDGEMENTS
The authors thank László Gráf for the initiation of research on locust protease inhibitors in the mid 90s
and for his suggestion of the role of internal dynamics in their mode of action. The useful discussions
with Gábor Pál, Michele Vendruscolo and Barbara Richter are gratefully acknowledged. This work was
supported by the Hungarian Scientific Research Fund (OTKA F68079, K72973 and NI68466) and
ICGEB (CRP/HUN09-03). Z.G. also acknowledges a FEBS Short-term Fellowship and a János Bolyai
Research Fellowship.
APPENDIX A. Supplementary data
Detailed description of the applied methods, as well as 5 figures and 3 tables are supplied as
supplementary material.
7
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
5. REFERENCES
[1] Otlewski, J., Jelen, F., Zakrewska, M., Oleksy, A. (2005). The many faces of protease-protein
inhibitor interaction, EMBO J 24, 1303-10.
[2] Berg, J.M., Tymocko, J.-L., and Stryer, L. (2006) Biochemistry (6th ed.) WH Freeman, New York
[3] Laskowski, M., and Qasim, M.A. (2000). What can the structures of enzyme-inhibitor complexes
tell us about the structures of enzyme-substrate complexes? Biochim Biophys Acta 1477, 324-37.
[4] Bode, W., and Huber, R. (1992). Natural protein proteinase inhibitors and their interaction with
proteinases. Eur. J. Biochem. 204, 433-51.
[5] Song, J., and Markley, J.L. (2003). Protein inhibitors of serine proteinases: role of bacbone
structure and dynamics in controlling the hydrolysis constant. Biochemistry 42, 5186-97.
[6] Liu, J., Prakash, O., Cai, M., Gong, Y., Huang, Y., Wen, L., Wen, J.J., Huang, J.-K., and
Krishnamoorthi, R. (1996). Solution structure and backbone dynamics of recombinant Cucurbita
maxima trypsin inhibitor-V determined by NMR spectroscopy. Biochemistry 35, 1516-24.
[7] Liu, J., Gong, Y., Prakash, O., Wen, L., Lee, I., Huang, J.-K., and Krishnamoorthi, R. (1998).
NMR studies of internal dynamics of serine proteinase protein inhibitors: binding region mobilities
of intact and reactive-site hydrolyzed Cucurbita maxima trypsin inhibitor (CMTI)-III of the squash
family and comparison with those of counterparts of CMTI-V of the potato I family. Protein Sci.
7, 132-41.
[8] Shaw, G.L., Davis, B., Keeler, J., and Fersht, A.R. (1995). Backbone dynamics of chymotrypsin
inhibitor 2: effect of breaking the active site bond and its implications for the mechanism of
inhibition of serine proteases. Biochemistry 34, 2225-33.
[9] Clarkson, M.W., and Lee, A.L. (2004). Long-range dynamic effects of point mutations propagate
through side chains in the serine protease inhibitor eglin c. Biochemistry 43, 12448-58.
[10] Jörgensen, W.L. (1991). Rusting of the lock and key model for protein-ligand binding. Science
254, 954-55.
[11] Eisenmesser, E.Z., Millet, O., Labeikovsky, W., Korzhnev, W., Wolf-Waltz, M., Bosco, D.A.,
Skalicky, J.J., Kay, L.E., and Kern, D. (2005). Intrinsic dynamics of an enzyme underlies catalysis.
Nature 438, 117–21.
[12] Henzler-Wildman, K.A., Lei, M., Thai, V., Kerns, S.J., Karplus, M., and Kern, D. (2007). A
hierarchy of timescales in protein dynamics is linked to enzyme catalysis. Nature 450, 913-6.
[13] Szenthe, B., Gáspári, Z., Nagy, A., Perczel, A., and Gráf, L. (2004). Same fold with different
mobility: backbone dynamics of small protease inhibitors from the desert locust, Schistocerca
8
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
gregaria. Biochemistry 43, 3376-3384
[14] Gáspári, Z., Szenthe, B., Patthy, A., Westler, W.M., Gráf, L., and Perczel, A. (2006). Local
binding with globally distributed changes in a small protease inhibitor upon enzyme binding.
FEBS J. 273, 1831-42.
[15] Fodor, K., Harmat, V., Hetényi, C., Kardos, J., Antal, J., Perczel, A., Patthy, A., Katona, G., and
Gráf, L. (2005). Extended intermolecular interactions in a serine protease–canonical inhibitor
complex account for strong and highly specific inhibition. J. Mol. Biol. 350, 156–69.
[16] Szenthe, B., Patthy, A., Gáspári, Z., Kékesi, A.K., Gráf, L., and Pál, G. (2007). When the surface
tells what lies beneath: combinatorial phage display mutagenesis reveals complex networks of
surface-core interactions in the pacifastin protease inhibitor family. J. Mol. Biol. 370, 63-79.
[17] Van der Spoel, D., Lindahl, E., Hess, B., Groenhof, G., Mark, A.E., Berendsen, H.J. (2005). GROMACS: fast, flexible and free. J. Comput. Chem. 26, 1701-18.
[18] Batta, G., Barna, T., Gáspári, Z., Sándor, S., Kövér, E.K., Binder, U., Sarg, B., Kaiserer, L.,
Chhillar, A.K., Eigentler, A., Leiter, É., Hegedüs, N., Pócsi, I., Lindner, H., and Marx, F. (2009).
Functional aspects of the solution structure and dynamics of PAF – a highly-stable antifungal
protein from Penicillium chrysogenum. FEBS J. 276, 2875-90.
[19] Richter, B., Gsponer, J., Várnai, P., Salvatella, X., and Vendruscolo, M. (2007). The MUMO
(minimal under-restraining minimal over-restraining) method for the determination of native state
ensembles of proteins. J. Biomol. NMR 37, 117-35.
[20] Yang, L.-W., Eyal, E., Bahar, I., and Kitao, A. (2009). Principal component analysis of native
ensembles of biomolecular structures (PCA_NEST): insights into functional dynamics.
Bioinformatics 25, 606-14.
[21] Gáspári, Z., Patthy, A., Gráf, L., and Perczel, A. (2002). Comparative structure analysis of
proteinase inhibitors from the desert locust, Schistocerca gregaria. Eur. J. Biochem. 269, 527-37.
[22] Dosset, P., Hus, J.-C., Marion. D., and Blackledge, M. (2000). Efficient analysis of
macromolecular rotational diffusion from heteronuclear relaxation data. J. Biomol. NMR 16, 23-8.
[23] Gáspári, Z., Ortutay, C., and Perczel, A. (2004). A simple fold with variations: the pacifastin
inhibitor family. Bioinformatics 20, 448-51.
[24] Roussel, A., Mathieu, M., Dobbs, A., Luu, B., Cambillau, C., and Kellenberger, C. (2001).
Complexation of two proteic insect inhibitors to chymotrypsin’s active site suggests decoupled
roles for binding and selectivity. J. Biol. Chem. 276, 38893–8.
9
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
[25] Grünberg, R., Nilges, M., and Leckner, J. (2006). Flexibility and conformational entropy in
protein-protein binding. Structure 14, 683-93.
[26] Lange, O.F,, Lakomek, N.-A., Farès, C., Schröder, G.F., Walter, K.F.A., Becker, S., Meiler, J.,
Grubmüller, H., Griesinger, C., and de Groot. B.
(2008). Recognition dynamics up to
microseconds revealed from an RDC-derived ubiquitin ensemble in solution. Science 320, 1471-5.
[27] Mellet, P., Boudier, C., Mely, Y., and Bieth, J.G. (1998). Stopped flow fluorescence energy
transfer measurement of the rate constants describing the reversible formation and the irreversible
rearrangement elastase-α1-proteinase inhibitor complex. J Biol Chem 273, 9119-23.
[28] Goh, C.-S., Milburn, D, and Gerstein, M. (2004). Conformational changes associated with proteinprotein interactions. Curr. Opin. Struct. Biol. 14, 104-9.
[29] Patthy, A., Amir, S., Malik, Z., Bódi, A., Kardos, J., Asbóth, B., and Gráf, L. (2002). Remarkable
phylum selectivity of a Schistocerca gregaria trypsin inhibitor: The possible role of enzymeinhibitor flexibility. Arch. Biochem. Biophys. 398, 179–87.
[30] Fischer, E. (1894). Einfluss der configuration auf die wirkung der enzyme. Ber. Dt. Chem. Ges.
27, 2985–93.
10
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
FIGURE LEGENDS
Fig.1. Dynamic ensembles of inhibitors SGCI and SGTI. Left: superimposed structures with the
protease binding loop darkened and the N- and C-termini labeled. Right: Correlation of calculated N-H
S2 order parameters and back-calculated Hα chemical shifts with experimental data.
Fig.2. Heterogeneity of the dynamic and SCR inhibitor ensembles. From left to right: structural
ensembles mapped to the first two largest PCA modes (calculated on the combined SCR and dynamic
ensembles); backbone RMSD values (with respect to the mean structure) of the ensembles calculated
for residues in the protease binding loop (residues 28-33 for SGCI and 27-32 for SGTI) and the
scaffold of the molecule (residues 4-27 for SGCI and 4-26 for SGTI); heavy atom RMSD of the coreforming residues calculated separately for the different ensembles; schematic representation of the
inhibitor structure with the β-strands, disulfide bridges, protease binding loop and core interaction site
indicated.
Fig. 3. Comparison of the dynamic and SCR ensembles with the bound-state conformations of the
inhibitors. Top: Structural superpositions of the dynamic ensembles with the complexed structures are
shown along with the first two largest PCA modes for the combined dynamic+complexed and
SCR+complexed ensembles. Bottom: Backbone dihedral angles of the binding loop in the dynamic
ensembles and the complexed structures are shown along with their 'canonical ranges' [4].
11
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
Figure 1
12
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
Figure 2
13
Preprint of: FEBS Letters 584 (2010) 203–206, doi:10.1016/j.febslet.2009.11.058
Figure 3
14