12864
`
`Biochemistry 1993, 32, 12864-12874
`
`Mechanism for the Rotamase Activity of FK506 Binding Protein from Molecular
`Dynamics Simulationst
`
`Modesto Orozco,* Julian Tirado-Rives, and William L. Jorgensen•
`Department of Chemistry, Yale University, New Haven, Connecticut 06511-8118
`Received June 4, 1993; Revised Manuscript Received September 13, 19938
`
`ABSTRACT: Molecular dynamics (MD) and free energy perturbation (FEP) methods are used to study the
`binding and mechanism of isomerization of a tetrapeptide (AcAAPFNMe) by FK506 binding protein
`(FKBP). Detailed structures are predicted for the complexes of FKBP with the peptide in both ground-state
`and transition-state forms. The results support a mechanism of catalysis by distortion, where a large
`number of non bonded interactions act together to stabilize preferentially the twisted transition state. The
`two most important groups for the catalysis are suggested to be Trp59 and Asp37, but several other groups
`are identified as directly or indirectly involved in the binding and catalysis. However, the structural results
`do not support the notion that the keto oxygen of the immunosuppressive agents FK506 and rapamycin
`mimics the oxygen for the twisted peptide bond in the FKBP-transition-state complex.
`
`The isomerization of peptidic bonds is a difficult process
`with free energy barriers typically being 15-20 kcaljmol
`(Drankenberg & Forsen, 1971; Larive & Rabenstein, 1993).
`Most peptides correspond to secondary amides and are found
`exclusively in the trans conformation, which is about 2.5 kcal/
`mol more stable than the cis (Drankenberg & Forsen, 1971).
`However, the tertiary nitrogen in prolyl residues allows the
`coexistence of cis and trans isomers in proteins, which may
`require interconversion during protein folding. Considering
`the difficulty of this reaction, and the need for efficient folding
`mechanisms in vivo, the existence of enzymes which catalyze
`the isomerization of prolylpeptides (peptidylprolyl isomerases
`(PPis)) is not surprising. PPis were first characterized by
`Fischer and co-workers (Fischer et al., 1984), and since then
`different studies have demonstrated that these enzymes
`catalyze the cis++ trans isomerism of prolylpeptides during
`the folding of various proteins (Fischer & Bang 1985,
`Biichinger, 1987; Lang et al., 1987; Jackson & Fersht, 1991).
`The interest in two PPis, FKBP (FK506 binding protein)
`and cyclophilin, increased dramatically after the discovery
`that they are receptors for the immunosuppressant agents
`FK506 (Harding et al., 1989) and cyclosporin (Fischer et al.,
`1989a;Takahashi, eta!., 1989). Particularly, FKBPwasfound
`to be the receptor for a family of immunosuppressant drugs
`which includes FK506, rapamycin, and ascomycin. It was
`demonstrated that the binding of the drugs to the protein
`yields complexes which inhibit T-cell activation (see Rosen
`& Schreiber, 1992).
`FKBP is a small, and highly conserved cytosolic protein,
`which was isolated and purified by Schreiber and co-workers
`(Harding et al., 1989), who also cloned the protein (Standaert
`et a!., 1990), and determined the 3-D structure in solution
`(Michnick et al., 1991) and in the crystal (van Duyne et a!.,
`1991). Actually, 3-D structures of the protein as well as of
`the protein bound to FK506 (van Duyne et a!., 1991a),
`rapamycin (van Duyne eta!., 1991b), and ascomycin (Mead(cid:173)
`ows eta!., 1993) are available. The protein's structure features
`
`t This work was supported by NIH Grant GM32136 and a fellowship
`from the Fullbright Foundation.
`* To whom correspondence should be addressed.
`I On leave from the Department de Bioqulmica, Facultat de Qulmica,
`Universitat de Barcelona, Marti Franqu~ 1, Barcelona 08028, SPAIN.
`8 Abstract published in Advance ACS Abstracts, November I, 1993.
`
`a five-stranded antiparallel {3 sheet and a short amphipathic
`a helix, which partially covers one face of the {3 sheet. The
`active site, defined as the region where the protein binds the
`drugs, is located between the a helix and the {3 sheet in a
`hydrophobic pocket formed by the side chains of Trp59, Tyr26,
`Phe46, Phe36, Phe99, Ile56, Tyr82, and Val55,
`Both FK506 and rapamycin bind to FKBP as surrogates
`ofprolylpeptides inhibiting the PPI activity (Schreiber, 1991;
`Rosen & Schreiber, 1992). The low Ki and the presence of
`a C9-keto oxygen orthogonal to the pipecolinyl ring suggest
`that both drugs interact with FKBP as mimics of the transition
`state for cis++ trans isomerization. The fact that cyclosporin,
`a potent immunosuppressant drug, was a powerful inhibitor
`of the PPI activity of cyclophilin led to the speculation that
`the immunosuppressant effects of these drugs were associated
`with their inhibition of PPI activity (see Rosen & Schreiber,
`1992). This attractive hypothesis was ruled out by two
`observations: (i) rapamycin and FK506 have very different
`mechanisms for immunosuppressant activity, although their
`inhibition of PPI activity is similar (Bierer et al., 1990a), and
`(ii) the analog 506BD is a strong inhibitor of FKBP but is not
`an immunosuppressant (Bierer et al., 1990b). Accordingly,
`the relationship, if any, between the PPI and immunophilin
`activities of FKBP and cyclophilin remains to be determined.
`The mechanisms for the PPI action of the immunophilins
`are also obscure. Fischer and co-workers (Fischer et al.,
`1989b) proposed a mechanism based on the formation of
`covalent adducts between the peptide and protein. Kofron et
`al. (1991) suggested an alternative mechanism in which the
`protonation of the proline nitrogen would facilitate the
`isomerization process. Other authors (Albers et al., 1990;
`Harrison & Stein, 1990;Parket al., 1992; Rosen & Schreiber,
`1992) have suggested a mechanism based on the preferential
`stabilization of the transition state by non bonded interactions
`(catalysis by distortion). However, the lack of a structure for
`a complex between an immunophilin and a peptide has limited
`the experimental characterization of the rotamase mechanism.
`In this paper, we present a molecular dynamics (MD) study
`on the mechanism of peptide isomerization catalyzed by FKBP.
`Kinetic measurements have been made for cis -
`trans
`isomerization of tetrapeptides (Fischer et al., 1989; Albers et
`
`0006-2960/93/0432-12864$04.00/0
`
`© 1993 American Chemical Society
`
`

`
`Mechanism for Rotamase Activity
`
`Biochemistry, Vol. 32, No. 47, 1993
`
`I2865
`
`Ala2
`
`0
`
`~
`
`'1'4
`
`Ame6
`
`N/
`I
`
`Phe~0
`)l:Jr j~ Jt,~~"H H
`I'} tr lfvz ,
`
`Ace!
`
`H
`
`0
`
`Ala3
`Pro4
`FIGURE 1: Structure and nomenclature of the peptide used in the
`simulations.
`
`al., 1990; Harrison & Stein, 1990; Kofron et al., 1991; Harrison
`& Stein, 1992; Park et al., 1992). Consequently, structures
`for the complexes of a tetrapeptide with a cis peptide bond
`and of its transition state were sought and obtained here.
`Support for the model includes free energy results for mutation
`of the PI residue. Specific interactions responsible for selective
`stabilization of the transition state are identified. The model
`clarifies in detail the mechanism of rotamase action and is
`open to experimental testing. The binding insights should
`also be useful in the design of high-affinity ligands for FKBP
`with potential immunosuppressive activity.
`
`COMPUTATIONAL METHODS
`
`A peptide with the sequences Ace-Ala-Ala-Pro-Phe-Ame,
`where Ace (acetate) and Arne (methylamide) are theN- and
`C-terminal groups, was used to mimic the natural substrate.
`Figure I shows the structure and related nomenclature; in the
`following, plain numbering is used for the peptide residues
`and superscript numbering is used for protein residues. This
`sequence was selected because is resembles closely the synthetic
`peptides (Suc-Ala-X-Pro-Phe-pNA) used to measure the
`activity of PPis (Fischer et al., I989; Albers et al., I990;
`Harrison & Stein, I990; Kofron et al., 1991; Parket al., 1992).
`The peptide was built in both cis and twisted conformations
`around the Ala3-Pro4 peptidic bond. The peptide in the cis
`conformation is used as a model of the substrate, and the
`twisted conformation, as a model for the transition state.
`The molecular dynamics calculations were carried out with
`the AMBER program, version 4.0 (Pearlman et al., I991).
`The AMBER/OPLS force field (Weiner et al., 1984;
`Jorgensen & Tirado-Rives, I988; Jorgensen & Severance,
`I990) was used to describe both the peptide and the protein
`interactions; the only modifications (see Table I) were for the
`peptidic bond Ala3-Pro4, where the torsional barrier and
`improper torsion at nitrogen were removed (in both planar
`and twisted conformations), and the bond lengths and angles
`at N(Pro4) were assigned to average values for planar and
`twisted amides (Duffy et al., 1992). The stretching and
`bending parameters related to N(Pro4) were reduced in order
`to allow the peptide to accommodate easily to the protein
`environment. Four constraints were used to guarantee that
`the Ala3-Pro4 peptidic bond stayed in the cis or twisted
`conformation. The same force constants were used for both
`conformers in order to make comparable the Hamiltonians
`for the cis and twisted peptides (Table I). The contribution
`of these constraints to the total energy was not included in any
`of the simulations.
`Standard OPLS Leonard-Jones parameters were used for
`the cis and twisted structures (Jorgensen & Tirado-Rives,
`1988; Jorgensen & Severance, 1990). Standard OPLS charges
`for amides were also used to describe the planar structure
`(Jorgensen & Tirado-Rives, I988), while the charges for the
`
`Table 1: Force Field Parameters for the Ala3-Pro4 Peptide Bond•
`
`bond or angle
`C-NW
`C-NW-CH
`C-NW-CQ
`CH-NW-CQ
`
`equil. value
`1.39
`115.0
`115.0
`114.0
`
`force constant
`100
`5
`5
`5
`
`dihedral angle
`Ca(A3)-C(A3)-N(P4)-Ca(P4)
`O(A3)-C(A3)-N(P4)-Ca(P4)
`C(A3 )-N (P4 )-CD(P4 )-Ca(P4)
`Ca(A3)-C(A3)-0(A3)-N(P4)
`
`equil. value
`120/0
`-60/180
`-120/180
`180/180
`
`force constant
`200
`200
`100
`100
`
`q (cis)
`atom
`q (Twisted)
`Ca(A3)
`0.340
`0.200
`C(A3)
`0.500
`0.437
`O(A3)
`--0.457
`--0.500
`N(P4)
`--0.570
`--0.570
`0.285
`Ca(P4), CD(P4)
`0.225
`• The atom type NW corresponds toN (Pro4) of the twisted conformer;
`other atom types are from the standard AMBER/OPLS force-field
`(Weiner eta!., 1984; Jorgensen & Tirado-Rives, 1988). Distances are
`in A, angles in deg, charges in electrons, and force constants in kcaljmol
`A2, or kcalfmol rad2• The dihedral parameters are used as constrains
`to keep the cis or twisted conformation. Standard AMBER torsional
`parameters for the Ala3-Pro4 bond are removed.
`
`twisted amide bond were derived from recent calculations on
`twisted dimethylacetamide (see Table I and Duffy et al.
`(1992)). These charges correctly reproduce properties of
`amides in solution, including free energies of hydration
`(Jorgensen & Tirado-Rives, 1988; Duffy et al., 1992).
`The crystal structure of FKBP bound to rapamycin (van
`Duyne et al., 1991 b) was used as a starting point for the
`calculations. First, the peptide in the twisted form was overlaid
`on rapamycin. The conformation of the peptide was modified
`by rotation around single bonds to obtain the best fit with
`rapamycin and to reproduce well the environment of analogous
`hydrogen bond donor and acceptor groups. Once a suitable
`conformation of the peptide was obtained, the drug was
`removed, yielding a protein-peptide complex in which the
`peptide not only reproduces the general shape of rapamycin
`but also has a similar pattern of hydrogen bonds with the
`binding pocket.
`The protein-peptide model was solvated with a sphere of
`23-A radius containing TIP3P water (Jorgensen et al., 1983)
`centered on the center of mass of the peptide. This led, after
`removal of water molecules with unreasonably short contacts,
`to a solvation sphere with 1002 water molecules. A soft
`harmonic term with a force constant of 1.5 kcal/ (mol A2) was
`used to retreive water molecules drifting beyond the boundary
`of the sphere. The resulting system is shown in Figure 2,
`where it is clear that the sphere of water provides extensive
`hydration of the binding site for the MD calculations. The
`solvated system was then divided into two regions: (i) an
`inner part, which includes the peptide, the water, and all
`residues which have at least one atom closer than I2 A to any
`atom of the peptide, and (ii) an outer part, which contains the
`rest of the protein. The first part includes 78 residues of the
`protein and was free to move in all minimizations and MD
`simulations, while the outer part with 25 residues was kept
`rigid in all calculations. This initial system was then prepared
`through 2000 steps of energy minimization (500 where the
`peptide and the protein were kept rigid, 500 where only the
`peptide was fixed, and 1000 where the entire system was
`minimized), and 40 ps of MD for heating from T = 198-298
`K and equilibration. The final equilibrated structure was
`used to run 500 ps of constant temperature MD at 298 K. The
`
`

`
`12866 Biochemistry, Vol. 32, No. 47, 1993
`
`Orozco et at.
`
`Pc-,llja-Aia-Pn>-Phe-Arre
`ds
`
`~-1
`
`.Ao-.61a.Qy-Poo-l'he-Arre
`ds
`
`AG"...,
`
`Act..,
`
`Pc-,llja-Aia-i'n>-f'he-Arre
`Misled
`
`1~
`
`.Ao,tla~n>-Phe-Arre
`Misled
`
`f iGURE 3: Thermodynamic cycle used to determine the t:J.t:J.G• of
`isomerization between the Ala and Gly peptides. Dark lines are the
`peptide and light lines are the rapamycin molecule.
`
`~1•1 AX
`- kTln (exp(-(Vx,+~ - V~,)/kTJh,
`b.G = )
`~
`
`(I)
`
`wide sampling windows of 5 ps each (2 ps of equilibration and
`3 ps of averaging) for a total of 205 ps for each mutation. T he
`entire peptide was considered as the perturbed group in the
`mutation; not only the intergroup interactions, but also all the
`intragroup nonbonded interactions (including I~) were
`considered in the evaluation of the free energy changes. We
`had determined in pilot calculations that the explicit inclusion
`of bonded contributions in the free energy calculation led to
`increased noise in the final free energies, but did not alter
`significantly the final results. The averaging in eq I was
`performed by MD simulations at constant T = 298 K. where
`T is the temperature, k is Boltzmann's constant, >. is the
`coupling parameter, which controls the evolution of the Ala
`- Gly mutation, and v~ is the potential energy of the system.
`A similar simulation was performed to determine whether
`or not the structure of the cis peptide was able to explain
`experimental binding data (Parker et at., 1992). Thus, the
`difference in free energy of binding between Ala and Gly
`(cis}peptides (b.b.Gbind) was determined as the difference
`between the free energy changes in the mutation Ala- Gly
`in water (b.Gmutwat••), and in the protein complex (b.GmutP'01
`) .
`In order to compute b.Gmutwater the Ala (cis)peptide in the
`conformation adopted inside FKBP at the end of the MD was
`introduced into a cubic box (edge "" 30 A) of TIP3P water.
`The closest water molecules were removed to yield to a system
`with 807 water molecules. The system was minimized for
`3000 cycles (2000 with the peptide fixed and 1000 with the
`entire system free). The system was heated from 198 to 298
`K during 5 ps of NVT MD (during the first 3 ps only the
`water was allowed to move), and equilibrated during 20 ps of
`N PT M D (T = 298 K; P = I atm). The free energy change
`for the Ala- Gly mutation was computed using 41 double(cid:173)
`wide sampling windows of 2.5 ps ( 1.25 ps of equilibration and
`1.25 ps of averaging) for a total of I 02.5 ps. Periodic boundary
`conditions were used in all the simulations. All the remaining
`technical details of the simulations were identical to those
`noted above for the peptide-protein complexes.
`
`RESULTS
`
`Transition-State Structure. The structure of the twisted
`transition state model, which was obtained after the fitting,
`optimization, and equilibration process resembles the general
`
`FIGURE 2: FKBP- peptide complex inside the 23-A sphere of water
`used in the MD simulations.
`
`integration step was 0.002 ps, and the coordinates were stored
`every 0.2 ps. SHAKE (Ryckaert et al., 1977} was used to
`maintain all bonds lengths at their equilibrium values. A
`nonbonded spherical cutoff of 8 A and a residue-based
`non bonded list, updated every 25 integration steps, were used.
`The structure at t = 100 ps was used to generate the cis
`peptide- protein complex. The change between twisted and
`cis peptides was done in six steps with the angles of the
`constraints and the charges gradually modified from twisted
`to cis values (Table I). After each change, the peptide alone
`was optimized in vacuo for I 00 cycles; harmonic constraints
`were used to guarantee that mostoftheconformational changes
`would occur mainly in the Ala3- Pro4 area. This simple
`procedure mimics a fast change from twisted to cis, and we
`found it to be superior to other methods such as stepwise
`rotation for the full protein- peptide system and simulated
`annealing.
`The cis peptide was then placed inside the binding pocket
`of the solvated protein occupying the same location as the
`twisted conformer. The system was optimized, heated, and
`equilibrated during 40 ps using a procedure analogous to that
`described above. A 500-ps MD trajectory was then run for
`the cis system. It should be noted that the cis and twisted
`systems contain the same number of atoms, identical definition
`of the hydration shell, identical partitioning between the inner
`and outer regions, and, except for the charges, the same
`Hamiltonian.
`Free energy ca leu Ia t ions were used to predict the difference
`in isomerization rate between two peptides, Ace-Ala-Ala(cid:173)
`Pro-Phe-Arne and Ace-Ala -G ly-Pro-Phe-Ame, for comparison
`with experimental data on thecorrespondingSuc-Aia-X-Pro(cid:173)
`Phe-pNA peptides (Albers et al., 1990; Harrison & Stein,
`1992; Park et at., 1992). The thermodynamic cycle in Figure
`3 was utilized to obtain the change in free energies of activation.
`The free energy changes for the Ala3- Gly3 mutations was
`computed by statistical perturbation theory according to eq
`1 (Zwanzig, 1954}. These mutations were started with the
`structures and velocities of the cis and twisted systems at the
`end of the MD runs. The calculations were done in 41 double-
`
`

`
`Mechanism for Rotamase Activity
`
`Biochemistry, Vol. 32, No. 47, 1993 12867
`
`FIGURE 4: Superposition of the structures of bound rapamycin and the twisted peptide after the equilibration.
`
`ALA2
`180.0 .-----..----..-------,--~---....
`o +I
`•
`llf}
`
`120.0
`
`60.0
`
`e
`~
`~ 0
`
`-6o.o
`
`-120.0
`
`ALA3
`
`120.0
`
`60.0
`
`e
`~
`~ 0
`
`-6o.o
`
`-120.0
`
`140.0
`
`340.0
`240.0
`t(ps)
`
`440.0
`
`540.0
`
`140.0
`
`340.0
`240.0
`t(ps)
`
`440.0
`
`540.0
`
`PR04
`180.0 .-------,---,-~--.----r----,
`
`PHE5
`
`120.0
`
`60.0
`
`0.0
`
`-6o.o
`
`-120.0
`
`e
`~
`~ 0
`
`e
`! 0
`
`-6o.o
`
`-120.0
`
`140.0
`
`340.0
`240.0
`t(ps)
`
`440.0
`
`540.0
`
`-180.0 L-~--L..~~-'-----'-~.....L........L....-
`40.0
`140.0
`240.0
`340.0
`440.0
`540.0
`t(ps)
`
`FIGURE 5: Dihedral angle histories for the twisted peptide.
`
`structure of rapamycin, as shown in Figure 4. The main chain
`of both molecules overlaps well, and the H-bond donors and
`acceptors are also in proximal positions. In particular, O(Ala3)
`is placed very close to the keto oxygen of rapamycin, and the
`Pro ring shares the space of the pipecolinyl ring.
`Analysis of the 500 ps trajectory shows flexibility in the
`peptide, especially for the side chain of Phe5 and in the Ala3-
`Pro4 region. The histories for the main chain dihedral angles
`shown in Figure 5 reveal that most movements are short range
`oscillations, which do not lead to dramatic changes in the
`structure. However, there is also a systematic movement
`
`occurring for the Ala3-Pro4 region in the 80-120 ps period,
`which leads to a significant change in the relative orientation
`of the peptide inside the protein, as shown in Figure 6. The
`conformation of the 60 and 540 ps structures are similar, as
`noted in Figures 5 and 6 with a general "turn" topology
`stabilized by one or two intrachain hydrogen-bonds. The
`largest differences between the 60 ps and 540 ps structures
`are in the positioning of the proline ring and Ala3-Pro4 with
`respect to the binding site (see Figures 6 and 7). The reduction
`in 1/1 for Ala3 after 60 ps secures the hydrogen bonds for the
`turn structure and leads to better possitioning of the Ala3
`
`

`
`12868 Biochemistry, Vol. 32, No. 47, 1993
`
`Orozco et al.
`
`O(Aia3)
`
`Twisted 60 ps
`
`O(AJa3}
`
`Twisted 540 ps
`
`keto
`
`Rapamycin
`
`FIGURE 6: Representation of the structure adopted inside the binding site by rapamycin, and the twisted peptide at 60 and 540 ps. A common
`reference axis was used to orient the molecules.
`
`FIGURE 7: The twisted peptide bound to the active site of FKBP.
`carbonyl group with respect to theAspl7 side chain (see below).
`The distances for key interactions are shown in Figure 8.
`A hydrogen bond exists between the Asp37 side chain and the
`amino groups of Ala2 and often Ala3, which are proximal
`(Figures 6 and 7). Two intrachain hydrogen bonds are also
`
`common: Ame6- Acel, and Phe5- Ala2, which stabilize the
`turn structure. Another interesting interaction is the close
`contact (ca. 2.0 A) between the amide hydrogen of Phe5 and
`the N of Pro4. There are no hydrogen bonds between the
`oxygen of Ala3 and the protein. The analysis of Figure 8
`
`

`
`Mechanism for Rotamase Activity
`
`Biochemistry, Vol. 32, No. 47, 1993 12869
`
`6.0
`
`5.0
`
`g g 4.0
`i 3.0
`
`2.0
`
`d(O(P4)-NH(I56))
`
`d(0(3)-002(037))
`6.0 ,---,,.,....--.----n-----n
`
`d(C(I\3)-002(037))
`6.0 ,...--.,.--rr---.--------n
`
`5.0 < ..., 4.0
`g
`;! 3.0
`'6
`2.0
`
`5.0
`
`g g 4.0
`!!! 3.0
`.g
`
`2.0
`
`360.0
`200.0
`t(ps)
`
`520.0
`
`1.0 .'-:---:-:-"--:--:-:-':-::--:-:-':"
`520.0
`360.0
`200.0
`40.0
`t(ps)
`
`1·~.0
`
`360.0
`200.0
`t(ps)
`
`520.0
`
`d(HN(Ame)-O(Ace))
`6.0 I"'T"r----r--....----,-,
`
`d(HN(A2)-002(037))
`6.0 ,-----,...---.------.,
`
`5.0 g g 4.0
`
`~ 3.0
`'6
`
`2.0
`
`5.0
`
`d(HN(I\3)-002(037))
`6.0 r-----r--.--------n
`< ..., 4.0
`g
`~ 3.0
`'6
`
`5.0
`
`2.0
`
`g
`«> 4.0
`g
`~ 3.0
`'6 2.0 ~~..,_ . . . . . . . ~
`1.0 .'-:---:-:-"--:--:-:-':-::--:-:-':"
`360.0
`200.0
`520.0
`360.0
`200.0
`40.0
`t(ps)
`t(ps)
`FIGURE 8: Histories for selected interatomic distances during the MD simulation for the twisted peptide.
`
`520.0
`
`360.0
`200.0
`t(ps)
`
`520.0
`
`reveals that the movements during the 80-120 ps period have
`two major consequences: first the hydrogen bond between
`IleS6 and Pro4, which mimics one of the hydrogens bonds in
`the crystal structure of the FKBP- rapamycin complex is lost,
`and second, the intrapeptide hydrogen bond between Ame6
`and Ace! is formed. Two less dramatic, but particularly
`significant changes, are in the distances between the Asp37
`side chain and the C= O moiety of Ala3. In the starting
`conformation, one of the negatively charged ox ygens of Asp37
`is about 3.6 A from O{Ala3) and 4.0 A from C{Ala3), while
`in the final conformation these distances are around 4.7 and
`3.9 A, respectively. Thus, the conformational change in the
`80-120-ps period transforms a clearly bad charge~ipole
`contact into a much more favorable one. It is also notable
`that after the conformational change, the position of the
`carbonyl oxygen of Ala3 is more similar to that of the amide
`oxygen rather than the keto oxygen of rapamycin.
`T he movement of the O(Aia3) from the starting to the final
`conformation leads to a modest reduction in the number of
`aryl interactions (a contact O{Ala3)-H (aromatic) is con(cid:173)
`sidered as an aryl interaction when the Q ... H distance is Jess
`than 3.2 A) between O(Ala3) and the aromatic side chains
`of Phe36, Phe99, Tyr26, and T yr82 from 2.6 to 2.0. Analysis
`of the hydration of the peptide used a geometric definition of
`a hydrogen bond requiring an N···H or O···H distance less
`than 2.5 A, and a donor- hydrogen- acceptor angle between
`120 and 180°. With this definition, there a re initially 5.4
`waters hydrogen bonded to the peptide on average. This
`number increases to 5.8 in the last 240 ps of the trajectory.
`One water is bound to O(Aia2), 1.3 to O(Acel ), and 1.5 to
`O(Phe5); however, the latter's intra peptide hydrogen bonding
`partner, the amino group of Ame6, is hydrogen bonded to a
`water only 10% of the time. 0(Pro4) is not solvent exposed
`in the starting conformation where it is H-bonded to IJe56, but
`in the final part oft he trajectory (100- 540 ps) it is hydrogen
`bonded to an average of 1.1 water molecules. Finally, O(Ala3)
`is in a purely hydrophobic place during the first 320 ps of the
`trajectory, but at this point, it captures a water molecule,
`which remains hydrogen bonded during 80% ofthe remaining
`time as in Figure 7 (an averageof0.4 water is bound to0(Aia3)
`when the complete trajectory is considered).
`
`O(Ala3)
`
`Cis 540 ps
`FIG URE 9: Representation of the conformation adopted inside the
`active site by the cis peptide. Only the structure at the end of the
`MD run is shown, since no major changes occur for the cis peptide
`during the simulation.
`
`The last step in the analysis of the protein-peptide
`interactions was identifying water bridges connecting polar
`groups of the peptide and the protein. There is a water bridge
`between 0(Phe5) and the carbonyl of Glu54 during 55% of
`the time, mimicking another one of the hydrogen bonds in the
`FK BP-rapamycin structure. A water bridge is also detected
`43% of the time between O(Ala2) and His87, and one exists
`40% of the time in the 120-320-ps period between O(Pro4)
`and T yr82, T he latter bridge is lost in the final part of the
`MD, when the water on O(Ala3) makes a bridge to the
`hydroxyl group of Tyr82 during 80% of the time (Figure 7).
`Cis Conformation. The cis form of the bound peptide also
`has a turn structure, with obvious similarities to the structures
`of ra pamycin and the twisted peptide (Figure 9). The position
`of the peptide in the binding site is very similar to that of
`rapamycin, with the proline ring replacing the pipecolinyl
`ring. The MD provides significant sampling of the config(cid:173)
`urational space of the peptide, but no systematic changes occur,
`as evident in the distributions of main chain dihedral angles
`shown in Figure I 0. Key interactions are indicated in Figures
`II and 12. Thus, the Asp37 side chain is again hydrogen
`bonded to the amino group of Ala2, and often to the amino
`group of Ala3. Hydrogen bonds also form ocasionally between
`
`

`
`12870 Biochemistry, Vol. 32, No. 47, 1993
`
`Orozco et a!.
`
`0'
`CD e Cl
`
`CD
`~
`CD
`"6l c:
`cu
`...
`iii
`'0
`CD
`.J:;
`0
`
`ALA2
`180.0 ..-~--,-~---,-----,---.,.------,
`
`120.0
`
`60.0
`
`0.0
`
`,_.
`
`XX
`
`X
`.
`
`-60.0
`
`-120.0
`
`-180.0
`40.0
`
`140.0
`
`~)~~--~~~;;.~
`; -~ •' "~. ~-. :'.-~.··. . -~·':.
`.... ~. :
`.
`
`240.0
`340.0
`t{ps)
`
`440.0
`
`540.0
`
`180.0
`
`PR04
`
`120.0
`
`0'
`CD e
`~ 60.0
`~
`CD
`"6l
`c: cu
`...
`iii
`'0
`CD
`.J:;
`i5
`
`0.0
`
`-60.0
`
`-120.0
`
`ALA3
`
`0 +2
`'lf2
`X
`
`180.0
`
`120.0
`
`60.0
`
`0.0
`
`-60.0
`
`-120.0
`
`-180.0
`40.0
`
`140.0
`
`340.0
`240.0
`t{ps)
`
`440.0
`
`540.0
`
`PHE5
`
`180.0
`
`120.0
`
`60.0
`
`0.0
`
`-60.0
`
`-120.0
`
`0 +4
`'lf4
`X
`
`0'
`~
`
`Cl
`CD
`~
`CD
`"6l
`c: cu
`e
`
`'0
`CD
`.J:;
`0
`
`0'
`~
`
`Cl
`CD
`~
`CD
`"6l
`c: cu
`... -g
`iii
`
`.J:;
`
`i5
`
`-180.0
`40.0
`
`140.0
`
`240.0
`340.0
`t{ps)
`FIGURE 10: Dihedral angle histories for the cis peptide.
`
`440.0
`
`540.0
`
`-180.0
`40.0
`
`140.0
`
`340.0
`240.0
`t(ps)
`
`440.0
`
`540.0
`
`""i
`
`FIGURE 11: The cis peptide bound to the active site of FKBP.
`NH(Ame) and O(Ala2) and between NH(Phe5) and O(Ala2).
`Other H-bonds (not shown in Figure 12) that are present
`some of the time are between NH(Phe5) and the imidazole
`ring of His87 and between NH(Ame) and O(Pro4). A
`hydrogen bond exists between O(Ala3) and the hydroxyl group
`of Tyr26 only during short periods of time, for instance in the
`160-180-ps period, but most of the time this H-bond is not
`present. It is very interesting to look at the relative positions
`of Asp37 with respect to the carbonyl group of Ala3; the OD2-
`0(Ala3) distance is around 3.3 A, and the OD2-C(Ala3)
`distance is about 4.0 A. These reveal a repulsive charge(cid:173)
`dipolecontact, which contrasts the favorable interaction found
`
`for the transition state. Intra peptide hydrogen bonds that are
`present in the twisted conformation, but are not present in the
`cis structure are between NH (Phe5) and N (Pro4) and between
`NH(Ame) and O(Ace).
`The carbonyl oxygen of Ala3 is near the aromatic side chain
`of Phe99 and Tyr26 (Figure 11); the average number of aryl
`interactions is 1.2 for this oxygen, which is ca. 1 less than in
`the twisted conformation. The cis peptide is well hydrated,
`with an average of 5.0 waters hydrogen bonded (0.4-0.Sless
`than in the twisted structure). A detailed analysis of the
`hydration shell shows that the oxygens of Ace and Phe5 are
`the best hydrated, with an average of 1.9 and 1.8 water
`
`

`
`Mechanism for Rotamase Activity
`
`Biochemistry, Vol. 32, No. 47, 1993 12871
`
`d(HN(F5)-0(A2))
`6.0 .--~-.---,..---r.
`
`d(HN(A2)-002(037))
`6.0 .-----.---....---.,..,
`
`d(HN(A3)-002(037))
`6.0 .------.---....---.,.,
`
`5.0
`
`g g 4.0
`.i 3.0
`
`., 2.0
`
`6.0
`
`5.0
`
`g
`g 4.0
`
`~ 3.0
`'6
`
`2.0
`
`200.0
`360.0
`t(ps)
`
`520.0
`
`d(0(3)-002(037))
`
`5.0 g g 4.0
`~ 3.0
`'6 2.0 ......... u~''lllli"'•l.IJoL.ll(llt.'lllllll•olii.'IJ
`
`1.0 L..-_
`40.0
`
`__._ __ ...._ _ _u
`200.0
`360.0
`520.0
`t(ps)
`
`d(C(A3)-002(037))
`
`6.0
`
`5.0
`
`g
`g 4.0
`
`~ 3.0
`'6
`
`2.0
`
`5.0
`
`g
`~ 4.0
`~ 3.0
`'6 2.0
`
`6.0
`
`5.0
`
`g
`g 4.0
`
`; 3.0
`'6 2.0
`
`200.0
`360.0
`t(ps)
`
`520.0
`
`d(HN(Ame)-O(A2))
`
`1.0
`40.0
`
`200.0
`360.0
`t(ps)
`
`520.0
`
`1.0
`40.0
`
`200.0
`360.0
`t(ps)
`
`520.0
`
`360.0
`200.0
`t(ps)
`
`520.0
`
`FIGURE 12: Histories for selected interatomic distances during the dynamics for the cis peptide.
`
`molecules bound. The oxygens of Pro4 and Ala2 are less well
`solvated with 0.5 and 0.7 water molecule bound on average.
`The amino group of Arne is hydrogen bonded to a water only
`10% of the time. O(A1a3) which is hydrogen bonded to a
`water molecule in the last part of the MD for the twisted
`conformer, is not solvated in the cis conformation. Three
`water bridges exist sporadically; the most common involve
`GluS4 which has water bridges with O(Pro4) and O(Phe5)
`during 31% and 22% of the simulation, respectively. A water
`bridge between O(Ala2) and the imidazole ring of His87 also
`exists 38% of the time during the first 200 ps, but it is rare
`in the remaining 340 ps. Finally, there are no water bridges
`involving Tyr82 and O(Pro4) or O(Ala3), in contrast to the
`twisted conformer.
`Free Energy Calculations. The structures determined for
`both the cis and twisted complexes seem reasonable from the
`viewpoints of stability, interactions, and function. Never(cid:173)
`theless, doubt remains about whether or not they are the real
`conformations. Comparison of total potential energies dem(cid:173)
`onstrates that the twisted form is slightly lower in energy than
`the cis, in good agreement with the proposed stabilization of
`the transition state by the enzyme, but the difference (around
`7 kcaljmol) is not significant considering the statistical noise
`in the simulations (::1::70 kcaljmol). Additional MD calcu(cid:173)
`lations, graphical manipulation, and simulated annealing
`procedures support the correctness of the structures, but it is
`obvious that more quantitative results are desirable. Though
`an experimental 3-D structure of the FKBP-peptide complex
`is not available, attempts could be made to reproduce kinetic
`data for the isomerization catalyzed by the enzyme. One
`possibility was to compute the potential of mean force (PMF)
`for rotation about the Ala3-Pro4 peptidic bond, and to compare
`the tlG of this process with the observed tlG of activation for
`the enzymatic reaction (tlG*). Unfortunately, several at(cid:173)
`tempted PMF calculations using holonomic constrains en(cid:173)
`countered major technical difficulties. Development of
`alternative procedures for such perturbations is under con(cid:173)
`sideration.
`However, some support can be sought by comparison of the
`relative tlG* of peptides with different sequence. Park et al.
`( 1992) found a difference in kcat favoring Ala over Gly at the
`
`P1 position in the Suc-Ala-X-Pro-Phe-pNA peptide. Their
`kca1 ratio of 1200 (T = 278 K) translates to a lower tlG* for
`the Ala(P 1) peptide by 3. 9 kcal/ mol. Harrison & Stein (1992)
`reported a kcatf Km ratio for Ala and Gly peptides of 44 (T
`= 283 K), which considering the Km for these peptides (Park
`et al., 1992) translates to a lowering of tlG* for the Ala(P1)
`peptide by 2.7