[Frontiers in Bioscience 9, 3453-3478, September 1, 2004]

INSIGHTS INTO THE CATALYTIC MECHANISM OF PEPTIDYL

PROLYL CIS/TRANSISOMERASES
Jrg Fanghnel and Gunter Fischer
Max-Planck-Forschungsstelle fr Enzymologie der Proteinfaltung Weinbergweg 22, D06120 Halle, Saale, Germany
TABLE OF CONTENTS
1. Abstract
2. Introduction
3. Mechanism of catalyzed prolyl cis/trans isomerization
3.1. Non enzymatic acceleration of prolyl cis/trans isomerization
3.1.1. Environmental effects
3.1.2. Intramolecular catalysis
3.1.3. Intermolecular catalysis
3.2. PPIase catalyzed prolyl cis/trans isomerization
3.2.1. Cyclophilins
3.3.2. FKBP
3.3.3. Parvulins
4. Perspective
5. Acknowledgment
6. References
1. ABSTRACT
A large body of physiological, cell biological, kinetic and structural data about peptidyl
prolylcis/trans isomerases (PPIases) has been accumulated during the past 20 years,
but despite the simplicity of the catalyzed reaction the question of how the enzyme
action is performed is still not fully answered. In this review the center of attention is the
molecular background of the catalytic mechanism of PPIases and the spontaneously
occurring peptidyl prolyl cis/trans isomerization. We summarize and compare the
available kinetic, structural and amino acid sequence data of all three PPIase families,
the cyclophilins, FKBP and parvulins. Different catalytic mechanisms that have been
suggested in the literature are discussed. A comprehensive comparison of enzyme
active site structures reveals a hitherto unnoticed similarity between the three PPIase
families and might suggest that PPIases utilize mechanisms that are more similar than
previously suspected.
2. INTRODUCTION
The pioneering work of L. Pauling over half a century ago prepared the ground for

understanding the atomic structure of proteins and the molecular basis of enzyme
catalysis which should be especially valuable in studying catalysis of chemically simple
reactions (1). Twenty years later, J. F. Brandts proposed that the slow phases observed
in the unfolding kinetics of proteins might be due to the cis/trans isomerization of
peptidyl prolyl bonds (the terms prolyl bond and prolyl isomerization are used
throughout this review for the peptide bond preceding a proline residue and the peptidyl
prolyl cis/trans isomerization, respectively) (2). An enzyme class discovered 9 years
later with its first member isolated from pig kidney represented a powerful catalyst of
prolyl isomerization operating on a level of one of the simplest chemical reactions: the
rotation about a single bond (3). According to their substrate specificity
these cis/trans isomerases were named peptidyl prolyl cis/trans isomerases (PPIases).
These enzymes have opened up a broad new area into the investigations of assisted
protein folding. Recently the secondary amid peptide bondcis/trans isomerases
(APIases) have been found which are actively engaged in lowering the rotational barrier
of a secondary amide peptide bond C(O)-NH- in oligopeptide- and protein substrates(4).
Another 20 years have passed since the first PPIase was characterized and what have
we learned about these enzymes in that time? A great deal of data has been published
about the physiological significance and the enzyme mechanism of PPIases. The best
characterized enzymes among the PPIases are human cyclophilin 18 (human Cyp18)
and human FKBP12 (human FKBP12). Both are prototypic members of their
corresponding PPIase families known as cyclophilins and FK506 binding proteins
(FKBP). The family names derive from the ability of their respective members to bind to
cyclosporine A or FK506, both highly active immunosuppressive compounds. A third
PPIase family, the parvulins, was discovered in 1994 (5). The most prominent and best
investigated member of this family is human Pin1 (human Pin1). The finding that
cyclophilins and FKBP participate in immunosuppressive processes initiated a rush of
scientific interest. Since then PPIases have been found to take part in a great number of
physiological processes such as in vivo protein folding, heat shock response,
transcription and translation, channel gating, virus assembly, signal transduction, tumor
metastasis, pathogen virulence, cell cycle control and others (6-10). These topics have
been reviewed exhaustively and shall not concern us here. In this review we want to
focus on publications related to the catalytic mechanism and try to combine all available
data to give a complete overview of this subject.
Despite the amount of data, the molecular basis of the PPIase mechanism is still only
poorly understood. For biochemical investigation, the simplicity of the reaction is a
blessing and a curse at the same time. Since the reaction is so undemanding, evolution
did not "bother" to equip prototypic PPIases with cofactors; which can often be exploited
to unveil the enzyme mechanisms. Furthermore the difference between the substrate
and the product state of a polypeptide chain is very subtle and therefore sometimes
difficult to investigate. On the other hand the acceleration of the peptide bond rotation is
already achieved by globular low-molecular-mass proteins, making these enzymes a
perfect subject of structural investigations. Since the reaction is, at least for oligopeptide
substrates, completely reversible, PPIases can be studied under equilibrium conditions,
an advantage if one wants to employ NMR spectroscopic tools which often require long
measuring times under biological conditions.
There exists no sequence homology between the three PPIase families, but they all
catalyze the same chemical reaction. A fundamental question arises from this fact:

Do PPIases facilitate their enzymatic action by utilizing a common catalytic pathway?

3. MECHANISM OF CATALYZED PROLYL CIS/TRANS ISOMERIZATION
3.1. Non enzymatic acceleration of prolyl cis/trans isomerization
To fully understand the enzyme mechanism of PPIases, a profound knowledge of the
chemical properties of prolyl bond isomers is unavoidable. Therefore a short excursion
into the nonenzymatic acceleration of prolyl bond isomerization is necessary here.
Although a full description requires a more complex explanation, the classical concept
of resonance stabilization", where the lone electron pair of the peptide bond nitrogen
is delocalized over the entire peptide bond, is now widely accepted. This model explains
many of the peptide bond features such as the limited number of stable ground state
conformations, the high barrier to rotation about the C-N bond, a shortened C-N bond
length, the kinetic stability of peptide bonds towards a nucleophilic attack and the
isomer-sensitive carbonyl stretching frequencies in the IR-spectrum, and it defines the
chemical means to lower the rotational barrier catalytically (Figure 1). Reviews on this
topic are available (11, 12). Here we will focus on the data directly linked to the
acceleration of the prolyl bond isomerization, supplemented in some cases with results
obtained from secondary amides.
3.1.1. Environmental effects
The amide group shows a greater charge separation when planar than in its proposed
orthogonal transition state, therefore the isomerization rate should be faster in less polar
solvents. Accordingly the isomerization rate of N,N-dimethylacetamide (DMA) increases
by a factor of 60 as the solvent is changed from water to cyclohexane (13). A similar
increase in isomerization rates were observed for peptides containing 2-C substituted
proline-like oxazolidine and thiazolidine derivatives, when the solvent was shifted from
DMSO to CDCl3 (14). Contrarily, it was found that the isomerization rate of Ac-GP-OMe
did not correlate with the solvent dielectric constant or the solvent polarity but depended
on the ability of the solvent to donate hydrogen bonds. Further analysis showed that the
barrier to rotation is proportional to the strength of hydrogen bonds formed to the
peptide bond oxygen. From the experimental results it was calculated that desolvation
of the prolyl bond can reduce the free energy of activation by 1.3 kcal/mol, which is only
little when compared with the decrease achieved by PPIases (15). A pH shift from 7.0 to
1.8 increases the rate of isomerization of DMA 130-fold (16). Although the preferred
protonation site of a peptide bond is the carbonyl oxygen, the rotational barrier of
amides can be lowered by a small but kinetically significant amount of an amide
nitrogen protonated species. The direct participation of hydroxide ions to the
acceleration of the cis/trans isomerization of DMA at pH 11.8 has been observed. The
rate enhancement was attributed to the formation of a tetrahedral hemiorthoamide, in
which the amide resonance is disrupted (16). Such extreme pH conditions are usually
not found in biological systems under native conditions but reactive hydroxide species
are frequently observed in enzymes utilizing Zn 2+ ions like carbonic anhydrase (17). The
findings that the isomerization rate of short proline containing peptides in buffered
solutions is independent of the pH within the range of 5 to 9 apparently indicates the
lack of general intermolecular acid/base catalysis (18). In this case the observed
deuterium kinetic solvent isotope effect (KSIE) is nearly in unity and therefore it was

postulated that no proton movement during transition state formation occurs (19).
Not mentioned in either of the categories above are effects on the isomerization rate of
prolyl bonds due to modifications of the prolyl bond itself or of the two neighboring
amino acid residues. In general, one finds that electron donating substituents near the
carbonyl moiety accelerate the isomerization rate whereas a deceleration is observed
when the electron density increases at the peptide nitrogen. Only a few modifications
are of physiological relevance, such as hydroxylation of proline in position 4 or Oglycosylation and phosphorylation of serine or threonine side chains preceding proline.
Proline homologues with differing ring size and heteroatom substitution in the proline
ring influence the imidic rotational barrier of oligopeptides containing these analogues.
For Aze, Pip, 4-Oxa and 2-Thz (Figure 2) a substantial acceleration of
the cis/trans isomerization was detected (20). A similar effect was observed for
fluoroproline derivatives (Ac-(4R)-FPro-OMe, Ac-(4S)-FPro-OMe and Ac-4,4-F2ProOMe), where the inductive effect of the fluorine substitutions leads to a weakening of
the Ac-Pro amide bond (21). Higher isomerization rates of the imide bond have also
been reported for short peptides containing different 2-C substituted proline-like
oxazolidine and thiazolidine derivatives (Figure 2) (14). On the other hand, the presence
of 3-C alkylated proline derivatives in the chain leads to a marked decreasein the imidic
isomerization rates in water. These effects were mainly attributed to sterical restrictions
arising from these bulky side chains (22). O-glycosylation of serine residues preceding
proline and hydroxyl substitution of proline were shown to have no effect on the
isomerization rate, whereas serine or threonine phosphorylation decreased the
isomerization rate up to 7 fold (22-25). The substitution of the prolyl bond carbonyl
oxygen by sulfur in short peptides reduces the isomerization rate by a factor of about
100. This is in accordance with the resonance theory since the thioxo group increases
the electron density along the C-N bond (26).
3.1.2. Intramolecular catalysis
Intramolecular catalysis of prolyl bond isomerization was first observed during the
refolding reaction of denatured dihydrofolate reductase (27). Proline 66 undergoes
a trans to cisisomerization catalyzed by the guanidinium group of arginine 44, which
interacts with the imide nitrogen of the Gln65-Pro66 peptide bond. Analyzing protein
structures available from the RCSB data bank revealed that in almost 6% of all
presently available structures at least one arginine guanidinium group is within 4 of a
proline imide nitrogen (Wille, G. et al., in preparation) suggesting a general concept for
accelerating proline-limited protein folding. It was also found that a hydrogen bond
between the imide nitrogen and the adjacent amidic NH within a five-membered ring, a
so called 5-NH- -Na hydrogen bond, can accelerate cis/trans isomerization up to 260fold, depending on the used solvent (28). A tenfold increase of the isomerization rate
was reported for Cys-Pro bonds in disulfide bonded cyclic peptides (Ac-CPPC-NH 2 and
Ac-TCPPCR-NH2) as compared to the corresponding acyclic compounds. With the help
of Monte Carlo molecular mechanics simulations it was concluded that the NH proton of
the residue following Pro can establish a hydrogen bond to the sp 3 lone electron pair of
the imide nitrogen in the transition state (29). An about tenfold increased rate of prolyl
isomerization was found in peptides containing a His-Pro moiety after a pH jump from
basic to acidic conditions. It was proposed that the protonation of the imidazol ring
promotes the isomerization by either providing an intramolecular hydrogen bond or by
localizing a positive charge close to the prolyl bond. Since the KSIE under acidic
conditions was determined to be 2.0, it was concluded that a direct proton transfer from

the protonated histidine moiety to the imide nitrogen during transition state formation
occurs (30). The sequence specific catalytic effect to the neighbouring prolyl bond
rotation suggests that His-Pro may play a general role in preventing proline-limited slow
folding phases in proteins..
3.1.3. Intermolecular catalysis
Monoclonal antibodies raised against -keto dicarbonyl containing haptens (Figure 3)
were described to accelerate the cis/trans isomerization of short proline containing
peptides and the refolding rate of RNase T1, which is limited by prolyl
bond cis/trans isomerization. The authors suggested that the used -keto dicarbonyl
function might mimic a twisted amide bond of a PPIase bound substrate. It was further
suggested that this hapten might result in antibodies capable of forming a tetrahedral
adduct to the electrophilic carbonyl group of the -keto functionality. However, such a
specific catalytic mechanism was ruled out because no pH dependence or KSIE was
observed and the small rate enhancement was comparable to that achieved by
hydrophobic solvents alone (31, 32). Similar acceleration factors were observed for
vesicle- and micelle-entrapped oligopeptides. Thermodynamic results as well as the
effects of side chain variations of the investigated peptide substrates will be discussed
in the FKBP section of this review (33).
3.2. PPIase catalyzed prolyl cis/trans isomerization
Among PPIases are single and multidomain enzymes, a review summarizing the
domain structure of human PPIase is available (34), here we focus on the enzymatic
activity and therefore only the function of PPIase domains will be discussed. Figure 1
shows five possible catalytic mechanisms to accelerate the cis/trans isomerization of
prolyl bonds. Many of them have been proposed to be the driving force behind the
catalytic power of PPIases but only few have been shown to be at least involved in the
catalytic cycle of these enzymes.
3.2.1. Cyclophilins
After the development of an in vitro PPIase assay based on the isomer-specific
proteolytic cleavage of peptide bonds it became feasible to investigate the catalytic
mechanism of PPIases (3). Since then a number of different activity assays have been
developed, including a protease free UV/Vis assay, dynamic NMR based methods,
fluorescence based assays and protein folding/unfolding assays. A complete summary
and descriptions of all available PPIase assays has been given in a recent review
published by Fischer and Aumuller (34). In the last two decades a large number of
cyclophilins from different organisms have been purified and kinetically characterized.
Almost all cyclophilins investigated so far show a second order rate constant
kcat/KM between 105 and 107M-1s-1 when measured using the oligopeptide Suc-AAPFpNA as substrate. Under optimal conditions the prototypic human Cyp18 is a perfectly
evolved catalyst that approach the limit of diffusion control for enzyme catalyzed
reactions; for the most suitable oligopeptide substrates it has a high turnover number
(kcat > 600 s-1) combined with a low ground state affinity (KM > 80 M). Table 1 gives
an overview of the catalytic constants (kcat/KM) for thecis to trans isomerization of
cyclophilins, mostly estimated with the oligopeptide substrate Suc-AAPF-pNA.

To date the three dimensional structures of 11 cyclophilins from 9 different organisms

have been determined (Table 2), most of these in complex with a tight-binding inhibitor
such as CsA. All these cyclophilin domains consist of an eight-stranded antiparallel barrel capped at either end by two -helices (Figure 4A). Despite the large quantity of
structural data only three cyclophilins from different organisms, namely human
Cyp18, E.coli CypA and C.elegance Cyp3 have been co-crystallized with a substrate
(35-42). The first view of the active site of human Cyp18 in complex with a substrate
(Ac-AAPA-amc) was provided by Kallen et al. 1992. Amino acid residues of human
Cyp18 in close contact (< 3.8 ) with the substrate are R55, I57, F60, Q63, A101, N102,
Q111, F113, L122, H126, and R148 (38). These residues form a channel on top of two
antiparallel -strands (Figure 4B). The binding pocket for the proline pyrrolidine ring is
formed by residues F60, M61, F113 and L122 (Figure 4D).
The same hydrophobic pocket is occupied by the MeVal-11 moiety of the tight binding
inhibitor cyclosporin A (Figure 5A). The shape of the cavity is optimized to interact with a
five-membered ring. The catalytic efficiency of human Cyp18 is greatly reduced if the
prolyl moiety of peptides is replaced by four- or six-membered ring derivatives of
proline, and completely abolished in cases of a secondary amide peptide bond (18, 43).
This finding makes it clear that the correct spatial orientation of the proline residue in the
transition state plays a crucial role in catalysis.
A pairwise amino acid sequence alignment of cyclophilin domains originating from
cyclophilins that have been reported as active, shows that all residues comprising the
binding pocket are conserved, with the exception of R148 (Table 3). The highest degree
of variability is found at the three positions 57, 102 and 126. The exchange in position
57 of isoleucine with valine in six sequences is conservative, but nevertheless four out
of these six cyclophilins posses a strongly reduced catalytic activity. The side chain of
N102 is not involved in substrate interactions but the backbone atoms of this residue
are in close contact with the substrate. The observed substitutions are tolerated
accordingly; only OvCyp16, in which this residue is exchanged with a serine residue,
possesses a low PPIase activity. The exchange of H126 with tyrosine found in four
cyclophilin sequences appear not necessarily to impair the catalytic activity since both
PPIases E.coli CypA and E.coli CypB are described as highly active. The same
substitution seems to be responsible for the reduced activity of Cyp-10 from C. elegans.
Mutational analysis of seven active site residues of human Cyp18 showed that only the
variants R55A, F60A, and H126Q posses less than 1% residual activity (Table 4) (44).
The phenylalanine residue in position 60 and arginine in position 55 are completely
conserved among the cyclophilins. The side chain of F60 forms part of the prolyl binding
pocket and is therefore essential for proper positioning of substrates. The importance of
R55 for the catalytic power of human Cyp18 will be discussed later. On the basis of the
first structure of a human Cyp18/substrate complex, it was suggested that H126 might
attack the prolyl bond, but further refinement showed that H126 is not involved in any
direct or indirect hydrogen bonds with the substrate (38, 45). It was pointed out that the
hydrophobic interaction of this residue is essential for substrate binding (39). It was
further concluded that the hydrophobic properties of H126 are more important than its
ability to participate in hydrogen bonds, which was supported by the strongly reduced
but still significant activity of the H126Q variant (Table 4).
The structural data in combination with kinetic experiments and site directed
mutagenesis resulted in a variety of theories describing the molecular mechanism of
cyclophilin catalysis. From the observed inactivation of human Cyp18 by sulfhydryl

group modifying agents, an involvement of an active site thiol was proposed. The
suggested mechanism assumed a nucleophilic attack of an activated sulfhydryl group to
the prolyl bond carbonyl carbon resulting in hemithioorthoamides as a covalently bound
tetrahedral intermediate (Figure 1, path C) (46). For this intermediate the resonance
stabilization of the former peptide bond is destroyed, resulting in a greatly reduced
rotational barrier about the C-N bond. This mechanism was supported by the
observation that the kinetic secondary deuterium isotope effect of the cis to trans
isomerization of Suc-AG(d2)PF-pNA changed from kH/kD >1 to 0.91 0.01 when
cyclophilin was added to the reaction (47). A number of experiments disagree with this
isotope effect derived mechanism. Site directed mutagenesis showed that all four
cysteine residues (C52, C62, C115, and C161) of human Cyp18 are dispensable for
activity (48). In accordance with these results is the observation that no cysteine is in
the proximity of the substrate, the closest cysteine (C115) is, with 8 , too far away to
build the proposed hemithioorthoamide intermediate (38).
Harrison and Stein suggested a mechanism which they called "catalysis by distortion",
where the enzyme induces strain or distortion in the substrate by utilizing the binding
energy provided by the enzyme/substrate complex formation (Figure 1, path A) (18, 19).
Cyclophilin was proposed to tightly bind a transition state in which the prolyl bond is
characterized by a partially rotated C-N imide bond. In
agreement with this mechanism is the observed small enthalpy of activation ( H* =4.3
kcal/mol) and the relatively large negative entropy of activation ( S* = -47 eu). It should
be pointed out that the Eyring plots used to calculate the activation parameters were
nonlinear and that two interchangeable enzyme forms were proposed to account for it
(49). This mechanism was also supported by a KSIE on k cat/KM and kcat close to unity for
the cyclophilin catalyzed isomerization of Suc-AAPF-pNA (19, 50) and a normal
temperature independent secondary deuterium isotope effect of 1.12 0.02 using
Suc-AG(d2)PF-pNA as substrate. The authors suggested that the transfer of the
substrate into the hydrophobic active site results in a reduced hyperconjugation of the CH electrons generating the normal -deuterium isotope effect
They also attribute a part of the catalytic power to the reduced stabilization of the polar
resonance structure of the prolyl bond in the apolar active site. Structural analyses of
human Cyp18 substrate complexes did not provide further proof for this theory. Neither
in crystal structures nor in solution do short peptides such as Ala-Pro, Ac-AAPA-amc or
Suc-AAPF-pNA adopt a twisted conformation (36, 37, 39, 40, 51). Further investigation
using dynamic NMR spectroscopic techniques revealed that human Cyp18 binds to
both ground state conformers with similar affinities (52). Studies using modified bicyclic
lactam inhibitors and phosphinic Ala-Pro surrogates, which were thought to resemble
the above mentioned twisted transition state, failed to display significant binding affinity
towards the active site of human Cyp18 (Figure 6) (53, 54). Peptides harboring a thioxo
prolyl bond or peptides with a reversed chirality at the C carbon of peptides in P1 and
P1 position (nomenclature according to Schechter and Berger (55)) can not activate
the catalytic pathway of human Cyp18 (26, 56). This suggests that the pathway from
ground state to transition state is sterically and electronically demanding, thereby ruling
out the above mentioned enzyme mechanisms like ground state distortion and
desolvation by binding to a hydrophobic cavity.
New grounds for speculation were given by a water molecule observed in a human

Cyp18/Ala-Pro complex. This water molecule is hydrogen bonded to the Q63 active site
residue of human Cyp18. Since the water is placed in the right orientation to stabilize a
twisted transition state via a hydrogen bond to the carbonyl oxygen of the prolyl bond,
the authors concluded that the human Cyp18 catalysis is "solvent-assisted" (Figure 1,
path B). The necessary energy to distort the peptide bond should arise from
thermodynamic fluctuations (39). Later it was found that the Ala-Pro dipeptide is not a
substrate but rather a competitive inhibitor of human Cyp18 (37). The authors discussed
that the additional charge of the free proline carboxyl group disturbs the active site of
human Cyp18. It forces the guanidinium group of R55 away from the proline nitrogen
and into an additional hydrogen bond with the unprotected C-terminus of the peptide. It
could therefore be concluded that an additional amino acid residue C-terminally located
to the proline might be a prerequisite for catalysis. The authors also suggested that the
guanidinium group of R55 might be the hydrogen donor proposed earlier by Kofron et
al., which lowers the rotational barrier of the prolyl bond by forming a hydrogen bond to
the proline imide nitrogen (Figure 1, path D) (50). Consistent with this mechanism is the
greatly reduced activity caused by the R55A mutation and the finding that intramolecular
general acid catalysis by NH groups will lead to an increased isomerization rate of prolyl
bonds (30, 44). The distance between the guanidinium NH2 group and the nitrogen of
the prolyl bond ranges from 3.3 to 3.8 depending on the structure used and is
therefore too far away to form a hydrogen bond. A shortening of this distance while the
prolyl bond proceeds from ground to transition state was suggested (36, 41) and
confirmed by a computational investigation (57, 58). Further support for this mechanism
was provided by the observation that the R44 side chain of dihydrofolate reductase
intramolecularly catalyzes the rate limiting step of folding by accelerating
the cis to trans isomerization of the Q65-P66 peptide bond (Figure 7) (27). The spatial
arrangement of the R44 guanidinium group and the imide nitrogen resembles the
situation found in cyclophilin/substrate complexes. On the other hand, the observation
that the peptide benzyl-Phe-Ala-Pro, which has the same free carboxy terminus as the
dipeptide, acts as a substrate for human Cyp18 as shown by NMR spectroscopic line
shape experiments casts doubt on this mechanism (59). Human Cyp18 activity (k cat/KM)
has been shown to be pH dependent; a pKa value of 5.9 was calculated (26). The
enzyme is almost two orders of magnitude more active at neutral or basic conditions
than at pH 4.5. Since the observed decrease of activity originates mainly from a lowered
kcat value it is feasible that ionizable active site side chains are involved in catalysis. The
guanidinium group of R55 is expected to be protonated within the pH range investigated
thereby making this residue unlikely to be the source of the pH dependency.
All experiments mentioned so far utilized short peptides to gain information on the
catalytic mechanism, whereas cyclophilins will act on proteins in biological systems. As
it is obvious from the contradicting enzyme mechanisms derived from crystal structures
harboring dipeptides compared to structures in complex with oligopeptides, it is
necessary to ask whether peptides utilize the complete enzymatic machinery of
cyclophilins. A new tool to help answer this question was provided by the interesting
finding that human Cyp18 is incorporated into the human immunodeficiency type 1 virus
(HIV) by interacting with the HIV Gag protein (60). Several structures showed that
human Cyp18 binds specifically to a G89-P90 motive, which is situated within a solvent
exposed loop of the capsid protein (35, 41, 61). The G89-P90 bond in this complex was
shown to be in the trans conformation, unlike all other human Cyp18 peptide substrate
complexes, where the prolyl bond was found exclusively in the cis conformation. An
exchange of G89 to alanine resulted in a reduced binding affinity (62). The binding
affinity depended not only on the amino acid moiety preceding the proline residue (P1

position) but also on the chain length of the substrate used, suggesting an extended
substrate binding site of cyclophilin (63). NMR exchange spectroscopy experiments of
the native N-terminal domain of the capsid protein in the presence of catalytic amounts
of human Cyp18 showed clearly that the isomerization of the G89-P90 bond is
catalyzed by human Cyp18 in the native state of the capsid protein (64). Structures of
variants of the HIV capsid protein co-crystallized with human Cyp18 resulted in
complexes in which both cis and trans conformations of the X-P90 bond occurred in
crystallographically equivalent positions (41). The authors proposed a mechanism
where the transition state deviates only minimally from the cis and trans ground states.
During reaction progress the C-terminal part of the substrates including the proline
remains fixed relative to the enzyme, while the carbonyl oxygen of the amino acid
preceding proline performs a flip-over. This model is consistent with a computational
analysis (57). However, it is inconsistent with NMR relaxation rate changes and an NMR
line shape analysis of human Cyp18 main chain amides during catalysis, where a
substantial movement of substrate and enzyme residues, especially in the C-terminal
region of the tetrapeptide-4-nitroanilide substrate, was observed (52, 65).
3.3.2. FKBP
Even though FKBP are a conserved family of PPIases, their PPIase domains show a
much higher degree of sequence variability than the cyclophilins. Some FKBP, like
the E.coli trigger factor andE.coli SlyD can not be inhibited by FK506 or rapamycin. As
described for higher molecular mass cyclophilins, the catalytic domain of FKBP was
found to be complemented with other domains in several cases. That these additional
domains can influence the PPIase activity by binding to their designated interacting
proteins has been shown for human FKBP38 (Edlich F et.al, submitted). Some of the
multidomain FKBP harbor up to four nonidentical FKBP PPIase domains within the
same polypeptide chain. This domain copying is a feature not found within the
cyclophilin family of PPIases. The functional significance of these multiple FKBP PPIase
domains is still unclear, especially since it has been reported that like in the case of
human FKBP51, human FKBP52, and rabbit FKBP59 the PPIase activity is mainly due
to one single N-terminally located FKBP PPIase domain when tested with oligopeptide
substrates (66-68). It remains to be discussed whether the inactivity of the other FKBP
PPIase domains is due to the experimental conditions used, due to autoinhibitory
segments within these proteins or if the inactivity is indeed an intrinsic property of these
domains.
The majority of FKBP accelerate the cis to trans isomerization of the prolyl bonds with a
second order rate constant between 104 and 107 M-1 s-1 (Table 5). Several FKBP exhibit
kcat/Km values in the range of enzyme variants point mutated in the active site of human
FKBP12 (Table 6). A few exceptions are found where only little or no activity could be
detected (69-72). FKBP achieve the observed rate enhancement by lowering the
rotational barrier by app. 6.6 kcal/mol (50, 73). In contrast to the uncatalyzed reaction,
where the Eyring activation parameter is characterized by a large enthalpic contribution
of H*= 18.9 kcal/mol and a small entropic part of S*= -1.16 eu , the FKBP catalyzed
reaction is characterized by a large entropic ( S*= -43.95 eu) and small enthalpic
( H*=5.85 kcal/mol) contribution to G* (74). The marked entropic contribution to the
free energy of activation led the authors to the conclusion that the rate limiting step is a
physical rather than a chemical one. They proposed that the enzyme selectively
stabilizes a twisted amide by using the binding energy to compensate for the loss of

amide resonance. In contrast to cyclophilins, the catalytic efficiency of FKBP depends

strongly on the side chain in position P1 of the substrate. Peptides containing large
hydrophobic residues like leucine or phenylalanine in this position show an up to 100- to
1000-fold increased second order rate constant compared to substrates with charged or
small side chains in this position (73, 74). The difference is not due to an impaired
substrate binding since the KM value is not affected within the experimental error. It is
rather an effect on kcat which decreases more than 1000-fold when the leucine residue
in the P1 position of Suc-ALPF-pNA is changed to glycine (73). These data in
combination with the observed pH independence of the FKBP from pH 5 to 10, the
solvent isotope effect close to unity and the reported, mainly entropic contribution to the
activation parameter led to an enzyme mechanism called "catalysis by distortion" which
had also been proposed for the cyclophilin catalyzed reaction. The energy necessary to
destabilize the substrate was thought to arise from the binding energy. Within the active
site the substrate is then distorted by electrostatic and/or geometric means towards the
transition state. In addition to the induced strain, the transition state is further stabilized
by the hydrophobic environment of the active site (18). The normal secondary kinetic
deuterium isotope effect observed for the specificity constant of the substrate SucAG(d2)PF-pNA with FKBP12 has been used to argue in favor of a transition state
characterized by a partial rotation about the C-N bond. However, since this -G-Psubstrate might not be able to activate the complete machinery of the biocatalytic
pathway, the mechanistic significance of the isotope effect remains questionable (73). A
so called "unassisted conformational twist mechanism" was proposed on the basis of
this isotope effect in combination with a slight inverse solvent isotope effect ((k cat/KM(H2O))/
(kcat/KM(D2O)) = 0.92) (73). This mechanism is a modified version of the "assisted
conformational twist mechanism" also suggested for cyclophilins, but it assumes that
catalysis by FKBP does not require favorable hydrogen bonding to the lone electron pair
of nitrogen from an active site residue.
The central question that needed to be addressed was what kinds of enzyme/substrate
interactions were used by FKBP to lower the rotational barrier in a FKBP-like manner.
Some answers could be found by investigating the FKBP structures reported. As many
as 51 structures of FKBP from 12 different organisms have been determined (Table 7),
none of them with a standard peptide substrate bound. They all show a very similar
general fold which consist of a four to six stranded -sheet wrapped around a short helix (Figure 8A). Inhibitor/enzyme complexes have been used to localize the active site
of FKBP. The high resolution structure of the FK506/human FKBP12 complex revealed
that the pipecolinyl moiety of FK506 (Figure 5B) is deeply buried in a hydrophobic cleft
of human FKBP12 (Figure 8B). It was concluded that the pipecolinyl moiety might
substitute for prolyl moieties of putative peptide or protein substrates (75). In
accordance with this proposal is the finding that in the crystal structure of the GPI1046/human FKBP12 complex, the five-membered pyrrolidyl moiety of GPI-1046 is
located similar to the pipecolinyl function of FK506 (76) (Figure 8C & D). One of the
most striking properties of the proposed active site of FKBP is its hydrophobic character.
In case of human FKBP12 thirteen residues (Y26, F36, D37, R42, F46, F48, Q53, E54,
I56, W59, Y82, H87, and F99) make direct contacts with FK506 (74, 75, 77). Four
residues (E54, I56, Y82 and D37) stabilize the drug/enzyme interaction via hydrogen
bonds. The residues in direct contact with the pipecolinyl ring have an exclusively
aromatic character (Y26, F46, W59, and F99). In contrast to earlier proposals (78) the
pipecolinyl-C8 carbonyl bond exhibits a trans conformation in its enzyme bound state
and does not mimic a twisted amide bond (77). Later it was proposed that the

orthogonal conformation of the -dicarbonyl moiety of FK506 might act as a twisted

amide surrogate (74), but the very low activity of catalytic antibodies derived from dicarbonyl containing haptens could not provide further evidence (31, 32). To determine
how conserved the thirteen drug binding residues are among the FKBP tested for
PPIase activity, a pairwise sequence alignment of these enzymes with human FKBP12
was performed and analyzed (Table 5). The degree of conservation is quite low for all
positions. It appears that most of the side chains within the active site are dispensable
for FKBP activity. Even P.horikoshii PhFKBP29, which has only one active site residue
in common with human FKBP12, was reported to be enzymatically active. For some
FKBP the low PPIase activity coincides with a low sequence identity in comparison to
human FKBP12, namely
for M.jannaschiiMjFKBP26, M.thermoautotrophicum FKBP28.3, M.genitalium trigger
factor, P.horikoshiiPhFKBP29, A.thaliana AtFKBP42, and Thermococcus sp. TcFK. On
the other hand - the E.colitrigger factor - that has only very few active site residues in
common with human FKBP12 posses a high catalytic activity. It appears as though the
enzymatic activity does not correlate directly with sequence conservation. In contrast,
FK506 binding constants provide a better measure of sequence similarity. This might
indicate that the residues involved in FK506 binding are not necessarily part of the
active site. The active site residues conserved best among 51 FKBP domains aligned
with human FKBP12 are Y26, F36, D37, I56, Y82, and F99 (Table 5). Two of these (Y26
and F99) form the hydrophobic cavity thought to accommodate the substrate proline
residue. The importance of D37 for the enzymatic activity will be discussed later. From
the 15 mutants of human FKBP12 depicted in Table 6 only the D37L, the W59A, the
F99Y and the Y26F/Y82F double variant posses less than 10% catalytic activity as
compared to the wild type enzyme. The tyrosine mutations do not influence the
formation of the Michaelis-Menten complex since the K M value remained unchanged
while kcat is greatly reduced (73). Since a single point mutation Y82L diminished the
slight pH dependence observed between pH 8 and 10.5 of human FKBP12, it was
concluded that the tyrosine hydroxyl group of Y82 might be acting as a hydrogen bond
donor to the prolyl nitrogen and thereby assist the prolyl cis/trans isomerization (18, 79).
One has to bear in mind that the single Y82L mutation of human FKBP12 still leaves a
residual activity of app. 10%, indicating that human FKBP12 stabilizes the transition
state by hydrogen bonding only to a limited amount. Further mutations of potential
nucleophiles and hydrogen bond donors (S8A, S38A, S67A, S77A, C22A, T75A and
T96A) led only to insignificant reduction of the catalytic function (73). Quantitative
modification of the cysteine thiols did not impair the catalytic activity (80). Since all
nucleophilic residues close to the active site of human FKBP12 are dispensable for
activity a side chain assisted covalent mechanism can be been ruled out (73).
Similar to FKBP, a preference for large hydrophobic side chains in the P1 position of
oligopeptides was found for the prolyl bond isomerization catalyzed by micelles. Various
surfactants have been shown to lower the free activation energy of isomerization by D
G*= 1.8 kcal/mol. In contrast to PPIases, micelles also accelerate the isomerization rate
of thioxoproline containing peptides (33). These results proved that only a small part of
the catalytic power of FKBP arises from the desolvation of the peptide bond and that the
rate enhancement of FKBP depends on a finely tuned spatial and electronic
arrangement within the active site. The same conclusions were drawn from the
observation that human FKBP12 cannot catalyze the cis/trans isomerization of
oligopeptides which harbor D-amino acids in position P1, P1 or P2 while ground

state binding was found to occur (56).

Molecular dynamics and free energy perturbation calculations where the peptides AcAAPF-Ame and Ac-LPF-Ame were fitted into the active site of human FKBP12 provided
evidence that these peptides adopt a VIa -turn type conformation. This allows
intrachain hydrogen bonding between the backbone nitrogen proton of the
phenylalanine residue and the proline imide nitrogen. This interaction is believed to
lower the rotational barrier via an "enzyme-induced autocatalysis" mechanism (81, 82).
Against such a mechanism stands the observation that an acidification of the respective
NH proton in thioxopeptide derivatives does not lead to a substantial lowering of the
rotational barrier neither for the free thioxopeptide nor for the FKBP12/thioxopeptide
complex. For example for the human FKBP12 catalyzed cis to trans isomerization of
AGP-psi[CS-N]-F-pNA a kcat/KM value of 67 M-1 s-1 was estimated which compares to the
value of 920 M-1 s-1 of the oxo congener. In addition to the above mentioned transition
state stabilization, examples for ground state destabilization were calculated as well. In
agreement with the observed reduced activity of the D37L variant an unfavorable
ground state interaction between the carbonyl moiety of the prolyl bond and the D37
side chain was calculated. This interaction changes to a more favorable charge-dipole
interaction when the substrate proceeds from the planar ground state to the twisted
transition state. The simulation also predicted a catalytically relevant water molecule,
which is bound between the carbonyl oxygen of the prolyl bond and the hydroxyl oxygen
of Y82 during transition state formation. The computational analysis could not pinpoint
the catalytic efficiency to certain side chains or favor one of the proposed enzyme
mechanisms (73). It was concluded that a combination of factors like desolvation of the
imide bond by the hydrophobic active site, ground state destabilization, preferential
transition state binding and enzyme induced autocatalysis contribute to the enzyme
activity. The theoretical methods mentioned here all suffer from the same limitations, in
particular the lack of FKBP/substrate complex structures to provide valid starting points
for the calculations. The methods used rely on the assumption that the conformation of
complexed FK506 or rapamycin resembles the substrate conformation in a FKBP bound
state. Some doubt is cast on the validity of this assumption by the observation that
proline containing peptides quench the fluorescence of W59 located at the bottom of the
drug binding site of human FKBP12 differently from FK506 (73).
3.3.3. Parvulins
The first member of the parvulin family of PPIases was identified in E.coli in 1994 (5).
The substrate specificity of the 92 amino acid long E.coli Par10 is similar to the FKBP
family of PPIases but the second order rate constant is at least one order of magnitude
higher and the activity could not be inhibited by either FK506 or cyclosporin. The
differences in the substrate specificity for human Pin1 and E.coli Par10 point out that
within the parvulin family two subfamilies exist. Whereas the enzymatic activity of all
phosphate specific parvulins reported so far is exceptionally high, the observed
kcat/KM values among the parvulins that show no such preference varies within a range
from 103 to 107 M-1 s-1 (Table 8). The sequence alignment shows that all phosphate
specific parvulins have an identical set of active site residues. On the other
hand, E.coli Par10 type parvulins show greater sequence variability and both residues
which have been proposed to attack the prolyl carbon (C113 and S154 in human Pin1)
are not conserved, questioning the described enzymatic mechanism.

A first glimpse of the catalytic mechanism of parvulins was provided by the crystal
structure of the human parvulin-like human Pin1 in complex with an Ala-Pro dipeptide
(83). Since then, 10 structures of parvulins from human, A.thaliana and E.coli have
been solved (Table 9). The human Pin1 consists of a C-terminally located PPIase
domain and an N-terminally located WW-domain, which was found to interact with
phosphoserine (pS) or phosphothreonine (pT) containing peptides and proteins. The
presence of the WW-domain has no influence on the PPIase activity of the whole
enzyme when the standard in vitro assay is applied (84). Whereas no sequence
homology between FKBP and parvulin exists, a similar general fold can be observed
(Figure 9). The PPIase domain of human Pin1 consists of a four stranded antiparallel -sheet surrounded by four -helices (Figure 10A). The active site of human
Pin1 is formed by 10 residues (H59, K63, R68, R69, C113, L122, M130, F134, S154
and H157). A basic cluster of side chains (K63, R68 and R69) in the 1/1 loop region
binds a sulfate ion in the crystal structure. The sequestered sulfate is in close proximity
to the bound dipeptide and has led to the assumption that human Pin1 possesses a
strong preference for negatively charged side chains in position P1 of proline containing
substrates. Accordingly it was found that the second order rate constant of Suc-AEPFpNA is at least two orders of magnitude higher than that of an uncharged peptide like
Suc-AQPF-pNA. Later it was observed that substrates with phosphorylated serine or
threonine residues in this position were catalyzed most effectively and that the
glutamate moiety preceding proline can not activate the full catalytic power of Pin1.
Especially the remarkable pH dependence of Pin1 catalyzing
thecis to trans isomerization of Ac-AAS(PO3H2)PR-pNA was not found to the same
extend using the substrate Suc-AEPF-pNA, pointing out that glutamate can not mimic
the phosphorylated serine residue completely (25, 85). The rate enhancement for
substrates with phosphorylated side chains vs. their unphosphorylated counterparts is
about 1300-fold. Most effective catalysis was observed at pH values where the
phosphorylated side chain is in its dianionic form (25). Simultaneous substitution of R68
and R69 with alanine reduced the catalytic efficiency to the level observed with
unphosphorylated peptides. The main contribution arises from the positive charge of the
R69 side chain, since the R69L variant lowers the enzymatic activity about 50-fold more
efficiently than the R68L variant (84). The finding that the 1/1 sulfate/phosphate
binding loop is extended into the solvent when no ligand is bound to the PPIase domain
led to the hypothesis that this loop might act as a lid which opens and closes during the
catalytic cycle (86). Structural analysis of human Pin1 in solution could not strengthen
this induced fit mechanism and the authors concluded that the observed open
conformation might be due to the crystallization conditions or crystal contacts (87).
The prolyl moiety of the dipeptide in the crystal structure of human Pin1 is located within
a hydrophobic pocket similar to that which is occupied by the pipecolinyl moiety of
FK506 in FKBP12/FK506 complexes (Figure 10A). It is formed by residues L122, M130
and F134 (Figure 10C). A minimum of three peptide bonds is essential to enable
activation of the catalytic machinery of Pin1. Dipeptides as used in the crystal structure
act as weak competitive inhibitors. These results are in agreement with those found for
the other two PPIase families. It seems that the extended substrate/enzyme interaction
is necessary to stabilize the transition state. Similar to human FKBP12 D-amino acid
residues in position P1 are not tolerated in substrates for human Pin1 (88). Close to the
dipeptide prolyl bond are the residues H59, C113, S154 and H157. A covalent
mechanism was suggested, where the deprotonated side chain of His59 abstracts a
proton from the C113 thiol, the resulting thiolate than attacks the carbonyl carbon of the

prolyl bond. According to path C (Figure 1) the isomerization barrier of the formed
intermediate is greatly reduced and therefore the rate of cis/trans isomerization is
increased (83). Active site variants C113A and H59A showed a 123-fold and a 17-fold
reduced activity respectively towards phosphorylated and unphosphorylated substrates
(85). In accordance with this mechanism is the bell-shaped pH dependence of
kcat/KM with apparent pKa values of 5.6 and 7.5, both pKa values are consistent with the
titration of the active site residues H59 and His157. Further investigations confirmed
that the side chain of His59 is responsible for the single dissociation step found below
pH 6. No pH dependence was found with substrates containing an Ala-Pro bond and it
was concluded that peptides that interact only unspecifically with the active site of
human Pin1 are catalyzed via a desolvation mechanism (25).
Although the model of covalent catalysis via a cysteine residue is consistent with a great
deal of work, some results argue against it. Juglone (Figure 11A) has been found to
covalently modify the active site cysteine of E.coli Par10, the reported rate of
modification was about 5 times faster than the observed enzyme inactivation. In
accordance with the reported CD-spectroscopic data, it was suggested that the juglonmodified E.coli Par10 is destabilized and that the misplacement of catalytic residues
leads to partial unfolding and thereby inactivation of the enzyme (89). Interestingly, it
was also observed that juglone derived inhibitors (Figure 11B & C) could inhibit human
Pin1 as well as human Par14 in a competitive and noncovalent manner (90).
The solution structure of a human Pin1 homolog from A.thaliana (Pin1At) also
questioned the participation of an activated cysteine in the catalytic mechanism (91).
The protein is highly homologous to human Pin1, but lacks the WW domain, like all
plant human Pin1 homologues reported so far. All above mentioned active site residues
are present in Pin1At and the general fold reveals a high level of similarity. The
substrate binding pocket was mapped using NMR chemical shift perturbation
experiments initiated by addition of a phospho-threonine peptide to Pin1At. Whereas it
was found that the residues R21, R22, M87 and F91 (corresponding to R68, R69, M130
and F134 in human Pin1) were clearly involved in substrate binding, no or only slight
changes in chemical shifts were observed for L79, C70, H12 and H114 (L122, C113,
H59 and H157 in human Pin1) upon addition of saturating concentrations of substrate.
Structural comparison of the active sites showed that the C70 side chain of Pin1At
pointed away from the hydrophobic prolyl binding pocket, making it unlikely to interact
with the prolyl bond. The authors suggested that the serine residue S71 (corresponds to
S154 in human Pin1), which showed large chemical shift perturbation upon ligand
binding, might instead be involved in catalysis.
Human parvulin 14 displays only a weak PPIase activity (k cat/KM = 103 M-1 s-1) with a
preference towards positively charged residues in P1 position (92). This preference is
due to a shortened loop region where the positively charged side chains K63, R68 and
R69 of the human Pin1 are replaced by a negatively charged area made of two
negatively charged side chains D74 and E73 of human Par14 (93). The active site C113
of human Pin1 is exchanged to D74 in human Par14 and S154 is exchanged to F120.
Only the two histidine residues in the active center of human Pin1 are conserved in
human Par14. It is thought that the negative charges provided by the E73-D74 patch
might be necessary to support catalytic assistance of other active site residues or
enzyme bound water molecules.

4. PERSPECTIVE
The question asked in the introduction - "Do PPIases facilitate their enzymatic action by
utilizing a common catalytic pathway?" is still unanswered, not even the nature of the
catalytic pathway of a single PPIase family is yet fully understood. As it seems none of
the mechanisms depicted in Figure 1 can explain all of the published data, but some
can be ruled out. A direct involvement of a tetrahedral intermediate formed by a
nucleophilic amino acid sidechain originating from the PPIase active site seems not to
be supported by the data. Furthermore, it has been shown for different members of the
three PPIase families that a number of atomic features in the substrate chain are
necessary to activate the full catalytic power of these enzymes, therefore mechanisms
which do not require specific enzyme/substrate interactions can also be ruled out. On
grounds of the existing data and new results obtained by investigating the influence of
cosolvents and heavy water on PPIase catalysis (Fanghnel et al., in preparation) we
favor a concerted mechanism in which the electron pair is stabilized on the amide
nitrogen bond by an H-bond donor of the protein. A simultaneous electrostatic transition
state stabilization by an enzyme bound general base-polarized water molecule of the
perpendicularly oriented carbonyl group is thought to occur (Figure 1, path D and E).
One major handicap to elucidate a common catalytic pathway of PPIases is obviously
the lack of structures of FKBP and parvulins in complex with substrates. In addition, the
danger posed by cyclophilin/substrate complexes is the potential problem of analyzing
dead complexes which do not map to the catalytic pathway. On a molecular level we
have only limited knowledge of how these two enzyme families bind to their natural
protein ligands. That the three enzyme families might catalyze the prolyl isomerization in
a similar way became evident in structural comparisons.
The active site residues of human Cyp18 have been superimposed with some of the
known drug binding residues of human FKBP12 (Figure 12). Four of the depicted active
site residues of human Cyp18 have identical counterparts in human FKBP12, the other
three residues are conservatively exchanged (L122 to I56, H126 to Y26 and Q63 to
D37). Interestingly, all residues crucial for the activity of human FKBP12 (Y26, D37,
W59 and F99) coincide with active site residues found to be essential (H126, Q63, F60)
or at least important (W121) for human Cyp18 catalysis. The pipecolinyl residue of
FK506 does not superimpose with the proline residue of the human Cyp18 bound
substrate, underlining the doubt that it occupies the active site of human FKBP12. It
should also be noted that in the crystal structure of the human FKBP12 in complex with
the cytoplasmic domain of the type I TGF-beta receptor the proposed proline binding
pocket of human FKBP12 is occupied by a leucine residue of the receptor (94).
When the same human Cyp18 residues are superimposed with the active site of human
Pin1 a similar picture is observed (Figure 13). Four residues of human Cyp18 have
identical counterparts in the active site of human Pin1 (R55 to R68, F113 to F134, Q63
to Q131 and L122 to L122). Again, among them are three residues necessary for
human Cyp18 activity (R55, Q63, F113), and F60 and H126 show a mirror symmetry to
the human Pin1 residues H157 and F125. The superimposed active site residues of
human Pin1 also include R68 and His59, which have been shown to be important for
efficient human Pin1 catalysis. In contrast to the superimposed human Cyp18 and
human FKBP12 structures, the proline residues of the bound ligands of human Cyp18
(Suc-AAPF-pNA) and human Pin1 (Ala-Pro) share a common space.

Despite the lack of a common general fold cyclophilins, FKBP and parvulins have a very
similar active site structure. This observation leads us to the assumption that the
catalytic pathway utilized by the different PPIase families is closely related.

INSIGHTS INTO THE CATALYTIC MECHANISM OF PEPTIDYL

PROLYL CIS/TRANSISOMERASES
Jrg Fanghnel and Gunter Fischer
Max-Planck-Forschungsstelle fr Enzymologie der Proteinfaltung
Weinbergweg 22, D-06120 Halle, Saale, Germany
FIGURES

Figure 1. Canonical structures of the prolyl bond and possible catalytic

mechanism for thecis/trans isomerization. Catalysis by the hydrophobic
enzyme environment which stabilizes a more apolar twisted prolyl bond (A).
Catalysis by a solvent assisted mechanism in which a enzyme bond water
stabilizes the intermediate by a hydrogen bond to the carbonyl oxygen (B).
Nucleophilic catalysis (C). Catalysis by an intra- or intermolecular protonation
of the imide nitrogen (D). An active site structural water molecule assists bond
rotation by electrostatic transition state stabilization (E).

Figure 2. Haptens (R = CO(CH2)3CO2H or R = CHO) used to raise catalytic

antibodies.

Figure 3. Proline analogues used to investigate the uncatalyzed

prolyl cis/trans isomerization. A) (S)-azetidine-2-carboxylic acid, Aze; B) (S)oxazolidine-4-carboxylic acid, 4-Oxa; C) (R)-thiazolidine-2-carboxylic acid, 2Thz; D) R)-thiazolidine-4-carboxylic acid, 4-Thz; E) 2-C substituted
thiazolidine derivatives, F) 2-C substituted oxazolidine derivatives.

Figure 4. Ribbon representationof the global fold of cyclophilin domains, here

shown for human Cyp18. The protein consists of a -barrel formed by eight
antiparallel -sheets comprising residues 5 to 12 (1), 15 to 24 (2), 53 to 57
(3), 61 to 64 (4), 91 to 100 (5), 112 to 115 (6), 127 to 134 (7) and 156 to
163 (8) and two amphipathic -helices (1 residues 30 to 41 and 3
residues 136 to 145) and a short 3-helix (residues 120 to 122) (A). Two of
the eight -strands (4, 6) form the cavity that accommodates the proline
residue (B). This binding pocket is formed by four hydrophobic residues
(Phe60, Met61, Phe113, and Leu122). The cavity accommodates the MeVal11 moiety of the inhibitor CsA (C) as well as the prolyl moiety of a substrate
(D).

Figure 5. CsA (A) and FK506 (B); residues that penetrate the prolyl binding
pocket (MeVal-11 of CsA and the pipecolinyl moiety of FK506) are indicated.

Figure 6. Compounds used as surrogates of the prolyl bond in their ground

state (A) or a twisted transition state (B, C and D).

Figure 7. Distances from the NH2 nitrogen of arginine to proline in DHFR and
human Cyp18/substrate complexes. Panel A depicts the interactions in DHFR
where R44 is involved in intramolecular catalysis of the Gln65-Pro66 bond
cis/trans isomerization. Panel B shows the active site R55 residue of human
Cyp18 which interacts with a ground state bound peptide substrate.

Figure 8. General fold of a FKBP domain, which is exemplarily shown for

human FKBP12 (A). The six antiparallel -sheets exhibit +3, +1, -3, +1 side
chain topology; the -sheets are capped with an amphiphilic -helix. The
pipecolinyl ring protrudes into the drug binding site of human FKBP12 (B). The
four hydrophobic residues F99, Y26, F46 and W59 are located in 2, 4, 6
strands and the1 helix respectively (C). The shape of the cavity can
accommodate five- and six-membered rings as found in the structures of
FK506 (C) and GPI-1046 (D). The four residues are shown in spacefilling
mode and labeled; the inhibitors are depicted as stick models.

Figure 9. Arrangement of secondary structure elements of human FKBP12

(left), human Pin1 (middle) and human Par14 (right). Conserved secondary
structures are in black; grey structures are found to be variable within the
individual structures. Bars represent -strands; wires correspond to -helices
and lines to loops and turns.

INSIGHTS INTO THE CATALYTIC MECHANISM OF PEPTIDYL

PROLYL CIS/TRANS ISOMERASES
Jrg Fanghnel and Gunter Fischer
Max-Planck-Forschungsstelle fr Enzymologie der Proteinfaltung Weinbergweg 22, D06120 Halle, Saale, Germany
TABLES
Table 1. Cyclophilins described as active using Suc-AAPF-pNA as substrate in the
standard in vitro PPIase assays

Organism

Name as in publication or data

bank/generic name

Swissprot

(kcat/KM)
M-1s-1

References

A. nidulans

CypB / Cyp23

O94190

active

95

A. niger

CypA / Cyp19

O94184

active

96

A. niger

CypB / Cyp23

Q8X166

active

97

A. thaliana

AtCyp22

P34790

5.7 106

98

A. thaliana

AtCyp28

P34791

active

99

B. malayi

Cyp-1 / Cyp98

Q27450

7.5 106

100

B. malayi

Cyp-2 / Cyp19

Q17246

1.23 107

101

B. subtilis

PPiB / Cyp15

P35137

1.1 106

102

B. taurus

CypA / Cyp18

P04374

1.3 107

50

B. taurus

CypB / Cyp23

P80311

3.0 106

103

C. albicans

Cyp1 / Cyp18

P22011

active

104

C. elegans

cyp-1 / Cyp21

P52009

7.0 104

105

C. elegans

cyp-2 / Cyp18.5

P52010

6.1 105

105

C. elegans

cyp-3 / Cyp18.6

P52011

3.6 105

105

C. elegans

cyp-4 / Cyp59

P52012

1.8 104

105

C. elegans

cyp-5 / Cyp22

P52013

7.4 104

105

C. elegans

cyp-6 / Cyp21.9

P52014

8.4 106

105

C. elegans

cyp-8 / Cyp54

P52016

1.95 104

105

C. elegans

cyp-9 / Cyp36

Q09637

1.5 104

105

C. elegans

Cyp-10 / Cyp18

P52017

1.9 104

105

C. elegans

Cyp-11 / Cyp20

P52018

1.5 104

105

C. elegans

CeCyp-16 / Cyp25

Q9XXI7

2 103 a)

106

D. discoideum

CypE / Cyp17

Q9NI62

Active

107

D. melanogaster

Moca-CypA / Cyp113

Q8ISE5

5.6 104

108

D. immitis

Dicyp-3 / Cyp60

O61300

3.95 105

109

E. coli

CypA / Cyp20

P20752

5.71 107

110

E. coli

CypB / Cyp18

P23869

6.74 107

110

E. histolytica

EhCyp / Cyp18

O15729

active

111

H. cutirubrum

Cyp19

O50586

active

112

H. sapiens

NKCR_HUMAN / Cyp166

P30414

7.5 105

113

H. sapiens

Cyp 40

Q08752

1.9 106

114

H. sapiens

SnuCyp-20

O43447

active

115

H. sapiens

Cyp18

P05092

1.3 107 b)

116

H. sapiens

CypF / Cyp22

P30405

2.3 107

117

H. sapiens

CypB / Cyp23

P23284

1.1 107

117

L. esculentum

CypA / Cyp18

P21568

active

L. major

LmCyp19

O02614

1.6 106

119

L. pneumophila

lpCyp18

Q48822

4.6 106

120

N. crassa

CPH / Cyp24

P10255

2.8 106

121

N. crassa

NcCyP41

Q9P3X9

6.5 105

122

O. volvulus

OvCyp16

Q8IA80

5.2 102 c)

100

Orpinomyces sp.

CypB / Cyp22

Q01490

9.3 106

123

P. falciparum

PFCyP / Cyp25

Q8I6S4

active

124

P. falciparum

PFCyP / Cyp22

Q8IIK8

2.3 106

125

R. norvegicus

Matrin CYP / Cyp88

O55035

1.0 106

126

R. norvegicus

CypF / Cyp22

P29117

0.9 106

127

S. cerevisiae

Cpr1 / Cyp17

P14832

active

128

S. cerevisiae

Cpr3 / Cyp20

P25719

5.8 106

129

S. cerevisiae

Cpr6 / Cyp42

P53691

5 105

130

S. cerevisiae

Cpr7 / Cyp45

P47103

7 104

131

S. mansoni

SmCypB / Cyp23

Q26551

8.2 105

132

S. mansoni

SmCypA / Cyp31

Q26548

3.65 105

132

S. mansoni

Smp17.7

Q26565

active d)

133

S. pombe

SpCyp3 / Cyp19

O74729

1.5 106

134

S. chrysomallus

ScCYPA / Cyp18

Q06118

3.73 106

135

S. chrysomallus

ScCYPB / Cyp19

P77949

7.5 106

136

T. cruzi

TcCyP19

AI021872e)

active

137

T. gondii

Cyp18.5

Q26994

active

138

T. gondii

Cyp20

Q26995

active

138

T.mentagrophytes

Cyp13

B019518

active

139

T. tridentatus

CypG / Cyp24

O44073

1.8 105 f)

140

T. inflatum

no name/ Cyp25

Q99009

active

141

V. faba

CypB / Cyp27

Q41651

active

142

X. laevis

XlCyp / Cyp17

AJ496795g)

1.1 107

143

Z. mays

zmCyp18

P21569

1.1 107

144

Z. mays

CypB

Q10724 h)

2.5 107

144

The reported enzymatic constants were obtained as described in the respective

publication. The generic name consists of the abbreviation of the respective PPIase family
followed by the rounded molecular weight of the full length protein. a) Suc-AVPF-pNA was
used as substrate, b) active site titration for this enzyme has been performed (145), c)
Suc-ALPF-pNA was used as substrate, d) no data about the substrate used, e) no
swissprot entry, SRS EMBL nucleotide accession number, f) Suc-AAPF-mca was used as
substrate, g) no swissprot entry, SRS EMBL protein accession number, h) protein
fragment, no sequence data available.
Table 2. Cyclophilin structures available in RCSB Data Bank

Organism

Name as
used in
publication

Number
of
structures

H. sapiens

human
Cyp18

43

PDB identifier

1ak4, 1awq, 1awr, 1aws, 1awt, 1awu, 1bck, 1cwa,

1cwb, 1cwc, 1cwf, 1cwh, 1cwi, 1cwj, 1cwk, 1cwm,
1cwl, 1cwo, 1cya, 1cyb, 1fgl, 1m63, 1m9c, 1m9d,
1m9e, 1m9f, 1m9x, 1m9y, 1mf8, 1mik, 1nmk, 1oca,
1rmh, 1vbs, 1vbt, 2cpl, 2cyh, 2rma, 2rmb, 3cyh,
3cys, 4cyh, 5cyh

H. sapiens

hCypB

1cyn

H. sapiens

SnuCyp-20

1mzw, 1qoi

B. malayi

CYP-1

1a58, 1a33, 1c5f

M.
musculus

CypC

2rmc

C. elegans

Cyp-3

1dyw 1e3b

C. elegans

Cyp-5

1hop

P.
falciparu
m

Cyp18

1qng, 1qnh

B. taurus

Cyp40

1ihg, 1iip

E. coli

CypB

1clh, 1csa

E. coli

CypA

1lop, 2nul

S.
cerevisiae

Cpr-1

1ist

Table 3. Conserved active site residues of cyclophilins; activity has been measured using
Suc-AAPF-pNA if not stated otherwise

Residue (human
Cyp18
nomenclature)

Number
of
exchanges

I57

Organism/Name as
used in publication

Found
exchange

(kcat/KM)
M-1s-1

Swissprot
entry

C.elegans / Cyp-9

1.5 104

Q09637

C.elegans / Cyp-16

2 103 a

Q9XXI7

R.norvegicus / Matrin
CYP

1.0 106

O55035

7.5 105

P30414

H.sapiens /
NKCR_HUMAN

7 104

P47103

S.cerevisiae / Cpr7

5.6 104

Q8ISE5

E.coli / CypA

5.71 107

P20752

E.coli / CypB

6.74 107

P23869

B.subtilis / PPiB

1.1 106

P35137

L.pneumophila /
Cyp18

4.6 106

Q48822

5.2 102 a

Q8IA80

D.melanogaster /
Moca-CypA

N102

O.volvulus / Cyp16

L122

C.elegans / Cyp-9

1.5 104

Q09637

H126

E.coli / CypA

5.71 107

P20752

E.coli / CypB

6.74 107

P23869

C.elegans / Cyp-10

1.9 104

P52017

P.falciparum / CyP

no number

Q8I6S4

a) Enzyme was inactive with Suc-AAPF-pNA, the value was obtained by using Suc-ALPFpNA. Residues R55, F60, Q63, A101, Q111, and F113 were conserved throughout all
sequences
Table 4. Relative activities of recombinantly expressed variants of human Cyp18

Variant

% rel. activity

WT

100

H54Q

15.0

R55A

0.1

F60A

0.32

Q111A

15.0

F113A

3.0

W121A

8.7

H126Q

0.53

Table 5. FKBP examined for their in vitro PPIase activity

Organism

Name as in Swissp (kcat/K

publication/g
rot
M)
eneric name
M-1s-1

Active site residues as found in a

pair wise alignment with human
FKBP12, +) identical residue, -)
residue not present

X FK50 Re
a)
6
f.
bindin
g

H.
sapiens

FKBP12 b)

P2007 1.6
1
106

Y F D R F F Q E I W Y H F L Ki=0.5 14
2
nM
6
6 3 3 4 4 4 5 5 5 5 8 8 9
6 7 2 6 8 3 4 6 9 2 7 9

H.
sapiens

FKBP12.6

Q1664 6.2
5
105

+ + + + + + + + + F + + + L Kd=0. 14
55 nM 7
14
8

H.
sapiens

FKBP13

P2688 3.1
5
106

+ + + Q + + G Q + + + A + L Ki=74 14
nM
9

H.
sapiens

FKBP25

Q0068 active + + + K L + G K + + + Q + A IC50=4 15

8
00 nM 0

A.
thaliana

AtFKBP42

Q9LD inactiv + + E E I L E K L L + N Y L no
69
C0
e
bindin
g

B. malayi FKBP13

O9633 6.3
5
104

+ + + + + + G Q + + + S + L IC50 = 15
1.7
1
M

B. taurus

FKBP12

P1820 6.6
3
105

+ + + + + + + + + + + + + L Ki =
0.25
nM

50

B. taurus

FKBP25

P2688 8 105
4

+ + + K L + G K + + + Q + A Ki =
160
nM

15
2

C.
burnetii

CbMip /
FKBP26

P5175 active + + + + A + K S + + + A + F no
15
2
numbe 3
r

C. tracho- chl-mip /
matis
FKBP27

P2662 active + + + N I L - - + F + Q + F no
15
3
numbe 4
r

E. coli

trigger
P2225 6.5
factor/FKBP4 7
105
8

V + T K + L G R + F F G S L no
15
inhibit 5
ion

E. coli

FKBP22

P39311 1.3
106

+ + + + A + - - + + + A + L Ki =
15
25 nM 6

E. coli

SlyD/FKBP21 P3085 1.4

6
104

G + + A Y Q + R P V A + + L no
15
inhibit 7
ion

H. cutiru- HcFKBP33
brum

Q9P9H 9.7
4
105

+ L + E R I E H F V + + + A no
15
inhibit 8
ion

M. genita- trigger factor / P4748 7.2

lium
FKBP51
0
103

F L A Q Y L N S + F + L + A no
15
inhibit 9
ion

M. janna- MjFKBP18
schii

Q5772 9.2
6
105

I Y Y Y I + G + + F + I + L IC50 = 71
170
nM

M. janna- MjFKBP26
schii

Q5823 6.4
5
102

L I N F V + E + V I T L I L 70% at 71
20 M

M.
FKBP28.3
thermoautotrophi
cum

O2719 3.6
7
102

E E E A I V G H + L F + M L IC50 > 72
20 M

M.
thermo-

O5298 3.5
0
105

D A G + L + G Q + F + I + L IC50=
250

MtFK /
FKBP17

16

lithotrophicus

nM

+ + + + + + G Q + + + - + F Ki =
4.5
nM

N. crassa

NcFKBP22

O6004 6.9
6
105

16
1

N. crassa

NcFKBP /
FKBP13

P2008 active + + + + L + G Q + + + V + A no
16
0
numbe 2
rs

N. menin- NmFKBP /
gitidis
FKBP12

P2513 active + + + + L I G Q + + + A + A no
16
8
numbe 3
rs

P. horikoshii

PhFKBP29

no
entry

S. cerevisiae

yFKBP12

P2008 8.2
1
107

+ + + + + C G Q + + + F + L Kd =
0.9
nM

S. cerevisiae

yFKBP13

P3247 5.4
2
107

+ + + + I + G R + + + V + L Kd = 16
18 nM 5

S. cerevisiae

yFKBP70

P38911 active + + + - + + G + + + + L + L no
16
numbe 6
rs

S. frugiperda

FKBP46 /

Q2648 7.8
6
106

+ M F K + + K + + + + S + L IC50 = 16
5 M 7

O9377 3.5
8
102

+ L V Y M V G + + L + K + L IC50 = 70
7 M

Thermo- TcFK /
coccus sp. FKBP18
KS-1

1.6
103

D I Y I V I G H + L E Q Y L 75% at 16
20 M 4

16
5

V. faba

FKBP15

Q4164 active + + + + I + G Q + + + S + L Ki =
16
9
30 nM 8

H.
sapiens

hFKBP51

Q1345 1.24 1 + + + + + + G Q + + + S + L no
16
1
06
numbe 9
rs

domain I

H.
sapiens

FKBP51

Q1345
1

L + + + + V H D P I F K Y

Q0279 3.8
0
105

+ + + + + + G + + + + S + L Ki =
17
10 nM 0

domain II

H.
sapiens

FKBP52
domain I

H.
sapiens

FKBP52

Q0279 inactiv L + + + + I N D P L F K Y
0
e

domain II

M.
musculus

FKBP51

Q6437 4.8
8
105

+ + + + + + G Q + + + + + L Ki=
14
10-15 8
nM

Q6437
8

L + + + + V H D P I F K Y

P2712 1.2
4
106

+ + + + + + G + + + + S + L no
66
numbe
rs

P2712 0.02
4
106

L + + + + V L D P L F K Y L no
66
numbe
rs

Q9M3 1.5
104

C V E + I D S K + L + A +

domain I

M.
musculus

FKBP51

no
17
numbe 1
rs

domain II

O. cuniculus

rFKBP59
domain I

O. cuniculus

rFKBP59
domain II

A.
thaliana

FKBP72

no
17
numbe 2

A.
thaliana

domain I

26

FKBP72

Q9M3
26

I I F - Y + S + P L L L +

Q9M3
26

+ Y + N L + G L P F + R W

domain II

A.
thaliana

FKBP72
domain III

M.
musculus

FKBP60
domain I

M.
musculus

FKBP60

Q9Z24 active + + + + + V G Q + M + V + L no
17
7
numbe 3
rs

Q9Z24
7

+ + + + Y T G W+ M+ D +

Q9Z24
7

+ + + + + T G Y + M+ R +

Q9Z24
7

+ L + L Y I G Q V M+ V +

Q6157 6.5
6
105

+ + + + V I G R + M + V + A IC50 = 17
45 nM 4

Q6157
6

+ + + + Y T G W+ M+ Y +

Q6157
6

+ + + + Y T G Y + M+ T +

domain II

M.
musculus

FKBP60
domain III

M.
musculus

FKBP60
domain IV

M.
musculus

FKBP65
domain I

M.
musculus

FKBP65
domain II

M.
musculus

FKBP65

domain III

M.
musculus

FKBP65

Q6157
6

+ L F Y Q I N K + L H A +

domain IV

Activity was measured with the substrate Suc-AXPF-pNA. The reported enzymatic
constants were obtained as described in the respective publication. All FKBP domains of
multi domain FKBP have been aligned individually; the activity shown was determined for
the full length protein, if not stated otherwise. The alignment of MjFKBP18, MjFKBP26 and
FKBP 28.3 showed only very weak homologies, the given result should therefore be
interpreted with caution. The statement "active" means that in the publication PPIase
activity assays were performed but no (kcat/KM) values were provided. The generic name
consists of the abbreviation of the respective PPIase family followed by the rounded
molecular weight of the full length protein, a) X Stands for the amino acid proceeding
proline in the used substrate (Suc-AXPF-pNA), b) active site titration for this protein has
been performed (175).
Table 6. Catalytic activity of recombinantly expressed human FKBP12 active site variants
determined using Suc-ALPF-pNA as substrate

Variant

Wt

kcat/KM M-1s-1

KM

kcat

References

1.2 106 b)

5 10-4 M b)

600 s-1 b)

176

3.5 106 a)

0.9-1.3 10-3 M c)

1000-1300 s- 1 c)

176

1.2 0.4 106 c)

155

2.2 0.2 106

177

3.6 106 c

79

Y26F

~ 1 105 a)

176

F36L

~ 4 106 a)

176

D37V

3 105 a)

176

D37L

6.6 0.4 104 c)

155

R42A

1.1 0.2 106

177

F46L

~ 5 106 a)

176

F48L

~ 0.9 106 a)

176

Q53A

1.8 0.4 106

177

W59A

~ 1 105 a)

176

Y82L

3.6 105 c)

0.7-1.0 10-3 M c)

20-24 s-1 c)

79

Y82F

2.6 105 b)

2.9 10-4 M b)

75 s-1 b)

176

Y26F/Y82F

6.4 104 b)

7.5 10-4 M b)

48 s-1 b)

176

H87A

1.9 0.2 106

177

F99Y

6.4 104 c)

155

a) numbers estimated from diagram, assay performed at 15 C, b) assay performed at 5

C, c) assay performed at 10 C
Table 7. FKBP structures available in the RCSB Data Bank

Organism

Name as used in
publication

Number of
structures

PDB identifier

B. taurus

FKBP12

1fkk, 1fkl

C. elegans

FKB-6

1r9h

E. coli

FKPA

1q6h, 1q6i, 1q6u

H. sapiens

FKBP12

33

1a7x, 1aui, 1b6c, 1bkf, 1bl4,

1d6o, 1d7h, 1d7i, 1d7j,
1eym, 1f40, 1fab, 1fkb,
1fkd, 1fkf, 1fkg, 1fkh, 1fki,
1fkj, 1fkr, 1fks, 1fkt, 1j4h,
1j4i, 1j4r, 1nsg, 1qpf, 1qpl,
1tco, 1fap, 2fke, 3fap, 4fap

H. sapiens

FKBP12.6

1c9h

H. sapiens

FKBP25

1pbk

H. sapiens

FKBP52

1n1a

H. sapiens

FKBP51

1kto

L. pneumophila

LpMip

1fd9

M. genitalium

Trigger factor

1hxv

M.

MtFK

1ix5

thermolithotrophicus

O. cuniculus

FKBP59

1rot, 1rou

S. boliviensis

FKBP51

1kt1

S. cerevisiae

FKBP12

1yat

T. cruzi

TcMip

1jvw

Table 8. Parvulins examined for their in vitro PPIase activity

Organis
m

Name as swisspro (kcat/KM

in
t
)
publicatio
n or data
M-1s-1
bank/
generic
name

H.
sapiens

Pin1 /
Par18

Q13526

H.
sapiens

Par14

Q9Y237 3.9 103

B.
subtilis

PrsA /
Par33

P24327

6 103

E. coli

Par10

P39159

1.9 107

Active site residues as found in a

pair wise alignment with human
Pin1, +) identical residue, -)
residue not present

substrate Ref
.

H K R R C M F S H AApSPR- 85
5 6 6 6 11 13 13 15 15 pNA a)
9 3 8 9 3 0 4 4 7

+ + -

SucARPFpNA

92

+ A -

SucAKPFpNA

178

1.35 107 + + -

SucALPF-

92

pNA

E. coli

SurA /
Par47

P21202

1.9 104

+ P P T D

SucALPFpNA

92

+ + -

SucAEPFpNA

179

Domain I

E. coli

SurA /
Par47

P21202

Domain II

3.4 106

E. coli

PpiD /
Par68

P77241

A.
thaliana

AtPar13

Q9SL42 2.7 106

+ + + + +

Ac180
AApSPFpNA a)

D. lanata DlPar13

Q9LEK8 1.5 107

+ + + + +

AcApSPYpNA a)

M.
domestica

Q94G00 3.1 106

+ + + + +

Ac180
AApSPFpNA a)

MdPar13

Q -

181

N. crassa Ssp1 /
Par21

O60045

6.5 106

+ + + + +

Ac182
AApSPFpNA a)

S.
cervicea

Ptf1 /
Par22

P22696

1.7 107

+ + + + +

AcApSPYpNA a)

181

X. laevis

Pin1 /

Q9I9K6 active

+ + + + +

no

183

Par18

informatio
n

The activity was measured with the indicated substrate. The reported enzymatic constants
were obtained as described in the respective publications. The two parvulin domains
of E.coli SurA have been aligned individually; the activity shown was determined for the
full length protein. The statement "active" means that in the publication PPIase activity
assays were performed but no (kcat/KM) values were provided. The generic name consists
of the abbreviation of the respective PPIase family followed by the rounded molecular
weight of the full length protein, a) the substrate contains a phosphorylated serine (pS)
preceding proline
Table 9. Parvulin structures available in the RCSB Data Bank

Organism

Name as used in
publication

Number of structures

PDB identifier

H. sapiens

Pin1

1nmv, 1nmw, 1pin,

1f8a

H. sapiens

Par14

1eq3, 1fjd

A. thaliana

Pin1At

1j6y

E. coli

Par10

1jns, 1jnt

E. coli

SurA

1m5y