Natural Products as Probes of Cellular Function : Studies of Immunophilins
`
`By Michael K. Rosen and Stuart L. Schreiber"
`One of the great mysteries of cell biology remains the mechanism of information transfer, or
`signaling, through the cytoplasm of the cell. Natural products that inhibit this process offer a
`unique window into fundamental aspects of cytoplasmic signal transduction, the means by
`which extracellular molecules influence intracellular events. Thus, natural products chemistry,
`including organic synthesis, conformational analysis, and methods of structure elucidation, is
`a powerful tool in the study of cell function. This article traces our understanding of a group
`of natural products from the finding that they inhibit cytoplasmic signaling to their current
`recognition as mediators of the interaction between widely distributed protein targets. The
`emphasis of the discussion is primarily structural. The interactions between the natural-
`product ligands and their protein receptors are analyzed at a molecular level in order to shed
`light on the molecular mechanisms of the biological functions of these compounds. In the
`process we hope to illustrate the power of chemical analysis as applied to biological systems.
`Through chemistry we can understand the molecular basis of biological phenomena.
`
`1. In trod uction
`
`Signal transduction refers to the process by which extra-
`cellular molecules influence intracellular events. For exam-
`ple, the actions of the hormone insulin are initiated at the cell
`surface, when extracellular insulin binds to the membrane-
`bound insulin receptor. This binding triggers a series of
`membrane-associated events, followed by a cascade of poor-
`ly defined intracellular events that result in the transcription
`of genes that encode metabolic enzymes. Thus, a signal that
`originates outside the cell is transmitted through the cell
`membrane, along intermediate carriers in the cytoplasm, and
`into the nucleus, causing a change in the status of the cell. In
`recent years, much has been learned about the mechanisms
`of signal transduction at the membrane and in the nucleus of
`the cell. In contrast, very little is known about the mecha-
`nisms of signal transduction through the cytoplasm of the
`cell. Despite the importance of such processes, the detailed
`mechanisms of cytoplasmic signaling remain among the
`great mysteries of cell biology, a fact that has led the cyto-
`plasm to be referred to as the "black box" of signal trans-
`duction.
`Natural products chemistry has long played an important
`role in the elucidation of biological mechanisms. Pioneering
`synthetic and mechanistic studies of molecules such as
`steroids, prostaglandins, and porphyrins, to name only a
`few, have led to fundamental insights regarding the biologi-
`cal functions of these important classes of compounds. Cer-
`tain natural products are also uniquely suited for the study
`of the mysterious processes of the cell, including cytoplasmic
`signaling, due to their interference with these processes. By
`studying inhibitory natural products bound to their biologi-
`cal receptors, we may gain a detailed understanding of the
`function of these receptors. Research in this area relies on a
`combination of the powerful, complementary tools of syn-
`thetic chemistry, molecular biology, cell biology, and meth-
`ods of structure elucidation, including both NMR spec-
`troscopy and X-ray crystallography. This review highlights
`recent advances in the structural and mechanistic under-
`
`[*] Prof. Dr. S. L. Schreiber, M. K. Rosen
`Department of Chemistry, Harvard University
`Cambridge, MA 02138 (USA)
`
`standing of cytoplasmic signaling that have arisen through
`application of these techniques to the study of the immuno-
`philins, a family of cytosolic proteins that bind natural prod-
`ucts. These advances have culminated in the identification
`and characterization of pentameric complexes composed of
`two normally noninteracting protein constituents and a nat-
`ural product "glue" that binds the proteins together in a
`biologically significant manner. In the process, it has been
`discovered that a protein phosphatase is a key cytoplasmic
`component of a family of signal transduction pathways.
`
`2. Background
`
`FK506".
`and cyclosporin A[31 (CsA) are fungal natural
`products (Fig. 1) that inhibit CaZ +-dependent[**] signaling
`pathways in a variety of cell types.[4* 51 In T cells, both agents
`inhibit the transcription of a number of genes, including that
`encoding interleukin-2 (TL-2), which are normally activated
`by stimulation of (by binding of certain extracellular mole-
`cules to) the T cell receptor (TCR).r61 In mast cells they
`inhibit the exocytosis (i.e. movement to the cell surface and
`fusion with the cell membrane) of secretory vesicles that
`normally results from stimulation of the IgE re~ept0r.L~~ A
`variety of biochemical and biological data indicate that in
`both cases inhibition occurs within the cytoplasm and not at
`the cell surface or in the nucleus.
`On a biochemical level FK506 and CsA are also quite
`similar. Through synthesis of radiolabeled and immobilized
`derivatives of these agents for use in protein purification, it
`was discovered that both bind with high affinity to soluble,
`cytoplasmic receptor proteinsta1 (immunophilins;[8* 91 this
`term is used to denote immunosuppressant binding proteins,
`as both FK506 and CsA are
`immunosuppressive
`drugs['* ''I). The FK506 receptor has been named FKBP,["]
`and the CsA receptor has been named cyclophilinrlz] (CyP).
`Both proteins catalyze the isomerization of cis and trans
`amide-bond rotamers of peptide and protein substrates."
`
`' 3
`
`['*I The terms Ca'+-dependent and -independent in this context refer to
`signaling pathways that are characterized by the presence or absence,
`respectively, of an immediate rise in the concentration of intracellular
`cytoplasmic Caz
`following binding to the cell surface receptor.
`
`+
`
`384 8 VCH Verlugsgesellschqfl mhH, W-6940 Weinhein?, 1992 Oj70-0~3~/92/0404-0384 $3.50+ ,2510
`
`Angew. Chrm. Int. Ed. Engl. 3f (1992) 384-400
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 1 of 17
`
`

`
`The enzymatic activity of FKBP is potently inhibited by
`binding of FK506 (Ki = 0.4 nM), but not CsA,[”] and the
`activity of CyP is inhibited by CsA (Ki = 6 nM),[”] but not
`FK506.
`
`FK506
`
`Rapam ycin
`
`Abu2 MeBmt MeVal MeLw
`10
`11
`1
`
`Me,
`
`Val5 MeLeu6
`
`Alal
`
`D-Ala 8
`
`Cyclosporin A (CsA)
`Fig. 1. Chemical formulas of the immunophilin ligands.
`
`H
`
`506BD
`
`Although FK506 is approximately 100 times more effec-
`tive than CsA in cellular assays, in virtually all other respects
`the two molecules behave identically in both T cells and mast
`cells. This similarity in biological function, coupled with the
`discovery that the molecules inhibit two distinct rotamase
`
`enzymes,[*] led to speculation that the biological effects of
`these agents were due to their rotamase inhibition. It was
`hypothesized that a necessary step in activation of unknown
`proteins required for IL-2 transcription in T cells was rotam-
`ase-catalyzed isomerization of a peptidyl-prolyl amide
`bond. Inhibition of this catalysis by FK506 or CsA would
`then result in inhibition of IL-2 transcription.“ 31
`Strong evidence against this “rotamase hypothesis” came
`from biochemical and biological studies of two structurally
`related molecules: rapamy~in,“~. 5 1 an immunosuppressive
`fungal natural product and FKBP ligand, and 506BD,[’” a
`synthetic FKBP ligand (Fig. 1). Given the structural similar-
`ities between the three ligands, it is not surprising that, like
`FK506, rapamycin and 506BD also bind tightly to FKBP
`and inhibit the rotamase activity of the enzyme (Ki(rapa-
`mycin) = 0.2 nM; Ki(506BD) = 5 nM).[lsx
`In fact, NMR
`and crystallographic studies have demonstrated that FK506
`and rapamycin interact with a common domain of FKBP
`through their common structural elements.[”] Fascinatingly,
`although FK506 and rapamycin both bind to FKBP and
`inhibit its rotamase activity, rapamycin does not inhibit the
`same TCR-mediated signaling pathway that is affected by
`FK506 and CsA. Rather, it blocks a later Caz +-independent
`pathway associated with T cell activation, which is mediated
`by the IL-2 receptor (IL-2R)[153 18,
`(Fig. 2). Even more
`striking is the fact that 506BD has no inhibitory effect on
`either of the signaling pathways inhibited by FK506 or ra-
`pamycin. Furthermore, it was shown that FK506, ra-
`pamycin, and 506BD can all block the actions of the others,
`
`[*I The expression rotamase has the same meaning as PPIase (for peptidyl
`prolyl cis-trans isomerase) an acronym used by others for the description of
`proteins that catalyze the isomerization of cis and fruns amide-bond
`rotamers of peptides and proteins.
`
`Stuart L. Schreiber was born in 1956 and was raised in Virginia ( U S A ) . He studied as an
`undergraduate in Charlottsville, Virginia, where he obtained a B.A. in chemistry in 1977 at the
`University of Virginia. Following Ph.D studies under the direction of R. B. Woodward and
`Yoshito Kishi at Harvard University, he began his independent career in 1981 in the Department
`of Chemistry at Yale University. In 1988, he accepted a Professorship at the Department of
`Chemistry at Harvard. His group is engaged in the study of biological processes through the
`combined application of synthetic organic chemistry and molecular and structural biology.
`
`Michael Rosen was born in 1965 in Philadelphia, Pennsylvania ( U S A ) . He received B.S. degrees
`in chemistry and chemical engineeringfrom the University of Michigan in 1987. As a Winston
`Churchill Scholar during 1987-1988 he earned a C.P.G.S. in the Natural Sciences from the
`University of Cambridge (England) under the direction of Professor Alan R. Battersby. Since
`that time he has been a graduate student in the research group of Projessor Stuart L. Schreiber
`at Harvard University. His primary interests lie in the structural analysis ofproteins andprotein-
`ligand complexes by N M R .
`
`Angew. Chem. Inf. Ed. Engl. 31 (1992) 384-400
`
`385
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 2 of 17
`
`

`
`presumably through competitive binding to a common re-
`ceptor, FKBP. In contrast, the actions of FK506 and ra-
`pamycin are unaffected by CsA, which does not bind FKBP;
`the actions of CsA are also unaffected by FK506, rapamycin,
`and 506BD, which do not bind CYP.['~,'~,
`19,201 If the bio-
`logical properties of FK506 and rapamycin were simply due
`to inhibition of the rotamase activity of FKBP, then they
`should both interfere with the same pathways. 506BD should
`
`upon binding to CyP a yet-to-be discovered CyP ligand may
`provide a complex that has properties indistinguishable from
`those of the FKBP-rapamycin complex.
`One obvious, if seemingly unlikely prediction from the
`above biological data is that the structurally unrelated
`FKBP-FK506 and CyP-CsA complexes should act, either
`directly or indirectly (through other proteins), upon a com-
`mon target molecule that is distinct from the target acted
`upon by the FKBP-rapamycin complex. The FKBP-FK506
`and CyP-CsA target should be a component of Ca2+-depen-
`dent signaling pathways such as TCR-mediated transcrip-
`tion in T cells and IgE receptor-mediated exocytosis in mast
`cells; the FKBP-rapamycin target molecule should be a
`component of CaZ +-independent signaling pathways such as
`IL-2R-mediated proliferation in T cells. In both cases, the
`target should be unaffected by either protein or ligand alone,
`since only the immunophilin-ligand complexes are active
`inhibitors (Fig. 3 top).
`In fact, we have recently demonstrated that calcineurin
`(CN), also referred to as protein phosphatase 2B, a calmod-
`din-dependent serine/threonine protein ph~sphatase,['~]
`possesses all the predicted biochemical properties of a target
`common to both FKBP-FK506 and C ~ P - C S A . [ ~ ~ ] CN is a
`heterodimeric protein composed of two subunits, calcineurin
`A (CNA), which contains the calmodulin-binding and phos-
`
`Fig. 2. FK506 and CsA inhibit Ca2+-dependent signaling pathways, while ra-
`pamycin inhibits Ca'+-independent pathways. These are illustrated in the T cell
`by the pathways emanating from the T cell receptor and the IL-2 receptor,
`respectively, but include signaling pathways in a variety of cell types, including
`mast cells and neurons.
`
`also inhibit these same pathways and thus be biologically
`active. Because this is not the case, it implies that FK506 and
`rapamycin do not act by eliminating a function of FKBP;
`rather they act by adding a function to the protein. In a sense,
`FK506 and rapamycin are prodrugs that are activated by
`binding to FKBP. CsA is similarly inactive until it binds
`CyP. This explanation of the immunophilin-ligand complexes
`as the species responsible for signal inhibition has been termed
`the "active complex" hypothesis.[*, 9, 15, 16, ''I Genetic stud-
`ies have also provided strong support for this idea by demon-
`strating that FKBP is both necessary and sufficient to medi-
`ate the actions of rapamycin in yeast,[z2~231 and that CyP
`mediates the actions of CsA in this organism.[21, 241 Paradox-
`ically, although FK506 and rapamycin are structurally simi-
`lar and bind to a common protein, FKBP, their FKBP com-
`plexes have different biological actions. Furthermore, while
`CsA is structurally dissimilar to FK506 and binds to CyP,
`which is unrelated to FKBP, the biological actions of the
`CyP-CsA complex are indistinguishable from those of the
`FKBP-FK506 complex. It is interesting to speculate that
`
`386
`
`Fig. 3. Top: The CyP-CsA and FKBP-FK506 complexes act on a common
`target molecule that is not affected by FKBP or CyP alone, or by the FKBP-
`rapamycin or FKBP-506BD complexes. Bottom: Formation of the pentameric
`immunophilin-drug-CNA-CNB-calmodulin complex (see text).
`
`Angew. Chem. Int. Ed. Engl. 31 (f992) 384-400
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 3 of 17
`
`

`
`phatase active sites, and calcineurin B (CNB), which is a
`Ca2+-binding protein with, as yet, unknown function. The
`binding of calmodulin to CNA results in a tenfold increase
`in the phosphatase activity of the enzyme. In direct binding
`assays, CN binds to both the FKBP-FK506 and CyP-CsA
`complexes, but not to FKBP or CyP alone, or to the FKBP-
`rapamycin complex. In addition, the binding of CN to
`FKBP-FK506 is inhibited by CyP-CsA (but not by CyP or
`CsA alone), suggesting that the two complexes compete for
`the same, or two interacting binding sites. The phosphatase
`activity of calcineurin, when measured with a phosphopep-
`tide substrate, is potently inhibited by the FKBP-FK506
`and CyP-CsA complexes, but is unaffected by FKBP, CyP,
`FK506, rapamycin, 506BD, or the FKBP-rapamycin com-
`plex. Interestingly, it was recently reported that the translo-
`cation of a cytosolic component of the transcription factor
`NF-AT, which regulates IL-2 transcription and is sensitive to
`FK506, into the cell nucleus due to a rise in intracellular
`calcium levels is blocked by both FK506 and CSA.['~] The
`common dependence on intracellular calcium, location in
`the cytosol, and sensitivity to FK506 and CsA have led us
`and others to speculate that this component of NF-AT may
`be a substrate for CN, and that its location in the cell may be
`dependent on its phosphorylation state.[', 28, 291 Thus, the
`inhibition of IL-2 transcription by FK506 and CsA may be
`the result of their indirect inhibition of the dephosphoryla-
`tion of a component of NF-AT.
`An important question arises from this body of data: How
`does the binding of FK506 and rapamycin to FKBP and of
`CsA to CyP change both the ligands and the receptors, en-
`abling the complexes to perform functions the individual
`components are incapable of performing alone? The remain-
`der of this review will focus on the aspects of ligand-receptor
`interactions most relevant to this question. In particular, we
`will discuss the structures of both the natural products and
`the proteins, and the changes that occur to each upon
`binding. The pentameric immunophilin-drug-CNA-CNB-
`calmodulin complexes (Fig. 3 bottom) will
`then be
`analyzed in light of the known biochemical data on the inter-
`actions between the immunophilin-ligand complexes and
`calcineurin.
`
`3. Structural Studies of FKBP
`and its Complexes with FK506 and Rapamycin
`
`3.1. Initial Work
`
`Early structural studies of the FKBP-FK506 and CyP-
`CsA complexes centered on analysis of the interactions be-
`tween ligand and receptor. These studies were motivated, in
`part, by two conflicting proposed mechanisms of rotamase
`catalysis, and hence of rotamase inhibition by the ligandsL3'I
`(Fig. 4). One mechanism involved initial formation of a te-
`trahedral enzyme-substrate adduct similar to that formed
`during the hydrolysis of the amide bond by serine or cysteine
`proteases. Rotation about the C-N bond in the adduct, fol-
`lowed by expulsion of the enzyme nucleophile would result
`in amide-bond isomerization. This mechanism was support-
`ed by studies of Fischer et al.,[lZbl who showed that in CyP
`modification of a cysteine in the active site with p-hy-
`
`Angew. Chem. Int. Ed. Engl. 31 (1992) 384-400
`
`I
`
`Fig. 4. Two proposed mechanisms of rotamase catalysis. Left: Initial formation
`of a tetrahedral intermediate. Right: Catalysis by stabilization of a twisted
`amide bond. enz = enzyme.
`
`droxymercuribenzoic acid eliminated the rotamase activity
`of the enzyme. Fischer et al.13'] also found an inverse sec-
`ondary deuterium-isotope effect with substrates containing
`[a,~-~H]Gly-Pro, indicating a change in hybridization
`(sp' 3 s p 3 ) and thus formation of a covalent bond in the
`transition state of the reaction. These kinetic data were, how-
`ever, disputed by Harrison and Stein,["] who found a nor-
`mal secondary deuterium-isotope effect using the same sub-
`strates. Site-directed mutagenesis studies further demon-
`strated that none of the cysteine residues in CyP are required
`for catalysis.1331 These results, coupled with measurement of
`thermodynamic activation parameters for C Y P ' ~ ' ~ and
`FKBP,'34. "I
`led to an alternative mechanism of rotamase
`activity involving binding of a transition-state structure that
`contains a twisted, or distorted amide bond. In this view, the
`energy needed to overcome amide bond resonance and iso-
`merize the C-N bond would come from favorable noncova-
`lent interactions between the enzyme and a peptide substrate.
`The mechanism involving a tetrahedral intermediate led to
`the consideration that FK506 and rapamycin might bind to
`FKBP through nucleophilic attack of a side chain of the
`enzyme on one of the two electrophilic carbonyl carbon
`atoms of the ligand, C8 or C9.
`NMR studies of the
`complex of fully synthetic [8,9-13C]FK506[361 and recombi-
`nant human FKBP,1371 however, showed no evidence for
`formation of a tetrahedral add~ct.'~'] These studies also
`demonstrated that FK506 binds to FKBP in a single confor-
`mation, although the unbound ligand exists in organic solu-
`tion as a 2: 1 mixture of cis and trans amide-bond rotamers.
`The finding that FK506 and, by inference, rapamycin do
`not bind covalently to FKBP suggested a possible explana-
`tion for the strong interaction between the ligands and their
`receptor. In the solid state, both FK506 and rapamycin pos-
`sess a dihedral angle of approximately 90" about the C8 -C9
`bond. This conformation is also maintained in the complexes
`of both ligands with FKBP. A dihedral angle of 90" between
`C8 and C9 and a planar N7-C8 amide group place the keto
`carbonyl roughly perpendicular
`to the plane of
`the
`pipecolinyl ring (pipecoline = methylpiperidine). Because
`the pipecolinyl ring most probably mimics the proline ring in
`natural peptide substrates, the keto carbonyl of FK506 or
`rapamycin is in the same position as would be a twisted
`amide carbonyl group of a peptide undergoing rotamase
`catalysis (Fig. 5). Thus, the perpendicular keto carbonyl
`groups of FK506 and rapamycin allow the ligands to mimic
`
`387
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 4 of 17
`
`

`
`3.2. Three-Dimensional Structures
`
`In order to further explain the enzymatic, ligand-binding
`and biological properties of the immunophilins, we under-
`took the structure determination of free human FKBP by
`NMR.[39-411 Our success in this effort was due significantly
`to the cooperative nature of FKBP, which proved to be
`soluble and stable, and gave beautiful NMR spectra. Struc-
`tural studies of the human FKBP-ligand complexes were
`also undertaken in a fruitful collaboration with Professor
`Jon Clardy’s group at Cornell University. This work resulted
`in high-resolution crystal structures of both the FKBP-
`FK506 and FKBP-rapamycin complexes.[42 - 441 Analyses
`of the three FKBP structures, showing both the bound and
`free forms of the protein, have yielded new insights into
`many aspects of immunophilin function.
`
`3.2.1. Overview of the Structures
`
`All three FKBP structures show the same fold of the
`protein (Fig. 7 left). The structure is characterized by a five-
`stranded antiparallel j sheet with a novel + 3, + 1, - 3, + 1
`loop topology. The strands of the sheet, which run roughly
`
`Fig. 7. Richardson diagrams of FKBP (left) and CyP (right). A ball-and-stick
`model of FK506 is positioned in the ligand-binding site of FKBP[42,44]. A
`similar model of CsA[69,70] is used to indicate schematicully the location of the
`ligand-binding site in CyP[67,68]. We prepared an approximate model of the
`CyP-CsA complex by docking the structure of bound CsA(691 into the struc-
`ture of free CyP[68] using reported intennolecular NOES observed between
`MeLeu 9 of CsA and Trp 121 of CyP[73]. We thank Professor Ke for providing
`the CyP coordinates.
`
`perpendicular to the long axis of the molecule, are composed
`of residues 2-8, 21-30, 35-38 with 46-49 (this strand is
`interrupted by a loop at residues 39-45), 71 -76, and 97-
`106. A short amphipathic a helix containing residues 57-63 is
`aligned with the long axis of the protein and lies against the
`sheet, forming a tightly packed hydrophobic core. The core
`is composed entirely of aliphatic and aromatic residues, with
`all but one of the aromatic residues clustered at one end of
`the molecule. The conserved aromatic and aliphatic side
`chains of Tyr 26, Phe 36, Phe46, Val 55, Ile 56, Trp 59, Tyr 82,
`and Phe99 line a shallow cleft at the N-terminus of the a
`helix, forming the FK506 and rapamycin binding site. The
`side chains of these residues are well defined in both the
`
`Fig. 5. FK506 and rapamycin may mimic a twisted peptido-prolyl amide bond
`in a peptide substrate. Left: Model of a twisted amide bond in a peptide
`substrate. Right: Portion of the crystal structure of free FK506. The analogous
`carbonyl groups in the two structures are colored red.
`
`a transition-state structure involving a twisted amide
`
`The view of FK506 and rapamycin as twisted amide pep-
`tidomimetics was further extended by studies of the substrate
`specificity of FKBP.[341 In the peptide series succinyl-Ala-
`Xaa-Pro-Phe-(p-nitro)anilide, it was found that peptides
`with branched hydrophobic amino acids Leu, Ile, and Val as
`Xaa were greatly favored (up to 1000-fold over peptides with
`Xaa amino acids with charged side chains as determined
`by measurement of Kca,Kil values) with Leu > Ile
`> Val. Analysis of the structures of FK506 and rapamycin
`suggested a possible rationale for these observations. Begin-
`ning with the pipecolinic acid moiety and proceeding in the
`“N-terminal” direction, both ligands possess a dicarbonyl
`group, a tertiary hydroxyl group, and a branched aliphatic
`chain. As illustrated in Figure 6, these can be mapped onto
`the amide carbonyl group, amide nitrogen atom, and the side
`chain of a branched aliphatic amino acid residue N-terminal
`to a proline. Thus, the binding of FK506 and rapamycin to
`FKBP was proposed to result from the ability of the ligands
`to mimic a Leu-Pro dipeptide with a twisted amide bond.r381
`
`Fig. 6. FK506 (left) and rapamycin may mimic a leucine-proline substrate with
`a twisted amide bond (right). Analogous atoms are colored alike. Leu is the
`preferred P1 residue.
`
`388
`
`Angew. Chem. Int. Ed. Engl. 31 (1992) 384-400
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 5 of 17
`
`

`
`carbon atoms C18-C23 of its macrocycle, including the ally1
`group, do not contact FKBP and are exposed to solvent.
`Similarly, the binding domain of rapamycin consists of the
`region from the C28 hydroxyl group through the pyranose
`ring; carbon atoms C15-C27 are exposed to solvent. For
`rapamycin the orientation of the cyclohexyl ring (C34-C42)
`is different from that in FK506, and makes several contacts
`to FKBP. The pipecolinyl ring is the most deeply imbedded
`portion of both ligands (Fig. 9). This moiety is in van der
`Waals contact with the indole ring of Trp59 at the back of
`the binding pocket, and with the side chains of Tyr26,
`Phe 46, Val 55, rle 56, and Phe 99 at the sides of the pocket. In
`the NMR spectra of the FKBP-FK506 complex the ex-
`tremely high-field chemical shifts (6 = - 2 to 0) of the
`pipecolinyl methylene hydrogens, and several NOE cross-
`peaks in the aromatic region also reflect the proximity of this
`group to aromatic residues of the protein."'] A number of
`additional hydrophobic contacts are made between the lig-
`ands and the protein. These include interactions between the
`pyranose ring and residues Tyr 26, Asp 37, Tyr 82, His 87, and
`Ile90, and between the region C24-C26 of FK506, and
`C28-C32 of rapamycin, and residues Phe46 and Glu 54 of
`FKBP. Thus, the nanomolar binding constants of FK506
`and rapamycin may be due, in part, to the complementary
`hydrophobic surface of FKBP.
`The favorable van der Waals interactions between FKBP
`and both FK506 and rapamycin are complemented by simi-
`lar arrays of intermolecular hydrogen bonds. Both ligands
`have identical hydrogen bonds from the C1 ester carbonyl
`group to Ile56 NH, the C8 amide carbonyl group to the
`Tyr 82 phenolic OH group, and the C10 hemiketal OH group
`to the Asp 37 carboxylate. Despite the differences between
`the C24-C26 region of FKS06 and the C28-C33 region of
`rapamycin, both also make analogous hydrogen bonds to
`Glu54 CO, FK506 from the C24 hydroxyl group, and ra-
`pamycin from the C28 hydroxyl group. Rapamycin makes
`one additional hydrogen bond not seen in the FK506 com-
`plex, from the C40 hydroxyl group to Glu 53 side chain CO
`group. The pattern of hydrogen bonds involving Ile 56 NH,
`Glu 54 CO, and Gln 53 CO is reminiscent of the antiparallel
`P-sheet-like contacts often seen between proteins and pep-
`tide ligands. These hydrogen bonds then extend the portions
`of FK506 and rapamycin that may be described as pepti-
`domimetic to include the region between the C1 ester car-
`bony1 group and the C24 (C28 for rapamycin) hydroxyl
`group. The structural analogy between the ligands and a
`tetrapeptide is illustrated in Figure 10. The pipecolinyl ester
`carbonyl group acts as the proline amide carbonyl group in
`a Leu-Pro peptide fragment. The region from 01 to C25
`(C29 for rapamycin), including the cyclohexyl ring, is roughly
`equivalent to an aromatic amino acid residue. The C24 hy-
`droxyl group (C28-OH for rapamycin) then represents the
`amide NH group of the following residue in the chain. Thus,
`FK506 and rapamycin may bind to FKBP by acting as tran-
`sition-state analogues of a Leu(or Val)-Pro-Xaa-Yaa peptide
`substrate. We are currently investigating this hypothesis
`through synthesis of peptides and peptide analogues de-
`signed based on the FKBP-FK506 and FKBP-rapamycin
`complex structures.[451 Biochemical analyses of these mole-
`cules, coupled with structural analysis of their complexes
`with FKBP (through X-ray crystallography and/or NMR
`
`389
`
`Fig. 8. Stereoview of selected residues in the ligand-binding pocket of FKBP.
`An overlay of 15 structures of the free protein generated by NOE-restrained
`molecular dynamics simulation is shown. The orientation is approximately
`perpendicular to that in Figure 7. (Note the a helix at the upper right and the
`/I strands lining the lower left.) Residues are identified by the one-letter amino
`acid code. Side chains are colored as follows: Phe = red; Tyr = green; Arg,
`Asp, Trp = white; Ile, Val = purple. The loops Ser38-Pro45 and Gly83-
`His94 have been removed for clarity.
`
`ligated and unligated forms of FKBP (Figs. 8 and 9). Both
`ligands fit tightly into the binding pocket and have a number
`of similar hydrophobic and hydrogen-bonding interactions
`with the protein. The FKBP-binding domain of FK506
`extends from the C24 hydroxyl group, through the diketo
`pipecolinyl moiety and the pyranose ring, to the C15
`methoxy group. This finding was anticipated on the basis of
`structural analyses of FK506 and rapamycin and on the
`biological properties of the two molecules, and led to the
`design of the FKBP ligand 506BD (see Fig. 1).Ii6] Most of
`the trisubstituted cyclohexyl ring (C26-C34) of FKS06 and
`
`Fig. 9. Stereoview of selected residues in the binding pocket of the FKBP-
`FK506 complex. Orientation and colors of the side chains are the same as in
`Figure 8. Bound FK506 is yellow. Hydrogen bonds (see text) are white.
`
`Anyew. Chem. I n t . Ed. Enpl. 31 (1992) 384-400
`
`NOVARTIS EXHIBIT 2010
`Par v Novartis, IPR 2016-00084
`Page 6 of 17
`
`

`
`Fig. 10. Schematic illustration of the bonding between FKBP and the ligands rapamycin Oeft) and FK506 (middle) and a peptide (right). Analogous atoms have the
`same color.
`
`spectroscopy) will probe the energetic contributions of the
`various receptor-ligand interactions.
`The hypothesis that FK506 and rapamycin are transition-
`state analogues of an FKBP substrate predicts that in addi-
`tion to the interactions discussed above there should be sig-
`nificant contacts between the protein and the keto carbonyl
`groups of the ligands. These same contacts would then stabi-
`lize a twisted amide bond in the transition state for rotamase
`catalysis. Initially we had expected, based on the many known
`structures of protein-ligand complexes, that the keto car-
`bony1 group (at C9) would be bound by a network of hydro-
`gen bonds or typical electrostatic interactions with charged
`residues. However, the structures of the two receptor-ligand
`complexes clearly show that the keto carbonyl group does
`not make any hydrogen bonds to the protein, nor does it
`contact any charged residues of the protein. Instead, it is in
`close proximity to the C'H atoms at the edges of three aro-
`matic rings, Tyr26, Phe36, and Phe99 (Fig. 9). Such
`C-H . . . O interactions have been observed in small organic
`
`m o l e c ~ l e s ~ ~ ~ ~ and proteins,[471 and it has been proposed that
`these contribute to the stability of such systems. Theoretical
`calculations, too, support the notion that C-H . . .O interac-
`tions can stabilize small molecules as well as proteins.[481
`However, to our knowledge the FKBP complexes would be
`the first examples in which this type of interaction con-
`tributes significantly to the binding of a ligand to its recep-
`tor. It would also represent a novel mechanism of catalysis,
`wherein transition-state stabilization is achieved through
`arene-carbonyl interaction.
`
`3.2.2. Structural Consequences of Binding
`
`Since neither FKBP nor FK506 or rapamycin are able to
`bind to or inhibit the phosphatase activity of CN, it is impor-
`tant to analyze the differences between the free and bound
`forms of both protein and ligands in order to understand the
`biological actions of the two complexes. How does binding
`activate FKBP and its ligands, allowing FKBP-FK506 to
`
`390
`
`recognize CN, and FKBP-rapamycin to recognize a differ-
`ent target protein? The existance of both free and bound
`structures of all components of these complexes now allows
`this question to be addressed.
`
`3.2.2.1. Comparison of free and bound FKBP
`
`Comparison of the structure of unbound FKBP with
`those of either of the FKBP-ligand complexes reveals that
`the majority of residues in the protein binding pocket are not
`greatly perturbed by ligand binding. In particular, the aro-
`matic and aliphatic residues making direct contact to ligand
`(Tyr 26, Phe 46, Val 55, Ile 56, Trp 59, and Phe99) show only
`small conformational differences between the bound and
`free structures (compare Figs. 8 and 9). In addition, regions
`of the protein that are not directly involved in binding (for
`e