PROTEINS: Structure, Function, and Genetics 30:61–73 (1998)
`
`Exploring Hydrophobic Sites in Proteins
`With Xenon or Krypton
`Thierry Prange´ ,1* Marc Schiltz,1 Lucile Pernot,1 Nathalie Colloc’h,2 Sonia Longhi,3
`William Bourguet,4 and Roger Fourme1
`1LURE, Universite´ Paris-Sud, 91405-Orsay Cedex, France
`2LMCP, 75252-Paris Cedex 5, France
`3AFMB-CNRS, UPR 9039, 13402 Marseille Cedex, France
`4IGBMC, UPR de Biologie Structurale, 67404-Illkirch Cedex, France
`
`ABSTRACT
`X-ray diffraction is used to
`study the binding of xenon and krypton to a
`variety of crystallised proteins: porcine pancre-
`atic elastase; subtilisin Carlsberg from Bacil-
`lus licheniformis; cutinase from Fusarium so-
`lani; collagenase from Hypoderma lineatum;
`hen egg lysozyme, the lipoamide dehydroge-
`nase domain from the outer membrane protein
`P64k from Neisseria meningitidis; urate-oxi-
`dase from Aspergillus flavus, mosquitocidal
`d-endotoxin CytB from Bacillus thuringiensis
`and the ligand-binding domain of the human
`nuclear retinoid-X receptor RXR-a. Under gas
`pressures ranging from 8 to 20 bar, xenon is
`able to bind to discrete sites in hydrophobic
`cavities, ligand and substrate binding pockets,
`and into the pore of channel-like structures.
`These xenon complexes can be used to map
`hydrophobic sites in proteins, or as heavy-
`atom derivatives in the isomorphous replace-
`ment method of structure determination. Pro-
`teins 30:61–73, 1998.
`r 1998 Wiley-Liss, Inc.
`
`Key words: xenon; krypton; hydrophobic cav-
`ity; protein-ligand binding
`
`INTRODUCTION
`A large number of proteins are known to fix small
`exotic ligands, like organic solvent molecules, and
`several crystal structures of these complexes have
`been refined and deposited with the Protein Data
`Bank.2 The pioneering crystallographic studies of
`Schoenborn and co-workers demonstrated that sperm
`whale myoglobin and horse haemoglobin crystals
`can bind xenon under moderate pressure, through
`weak van der Waals forces.55,58,60,75 These xenon
`complexes were obtained by subjecting native pro-
`tein crystals to a relatively low gas pressure (2–2.5
`bar).§ Xenon was shown to bind to pre-existing
`atomic-sized cavities in the interior of these globular
`protein molecules. The interaction of xenon with
`
`§Non SI-units used in this article: 1 bar 5 105 Pa, 1 Å 5
`10210 m.
`
`r 1998 WILEY-LISS, INC.
`
`proteins is the result of non-covalent, weak-energy
`van der Waals forces,12,66,67 and therefore, the pro-
`cess of xenon binding is completely reversible.39,64
`The xenon complex with myoglobin has been studied
`in detail by Tilton and co-workers,64,66–68 who showed
`that the number and the occupancies of xenon bind-
`ing sites vary with the applied pressure. Thus, at a
`pressure of 7 bar, one major (almost fully occupied)
`and three secondary (half-occupied) sites were found.
`Xenon binds to myoglobin with very little perturba-
`tion of the surrounding molecular structure. Hence,
`the xenon complex is highly isomorphous with the
`native protein structure. Cyclopropane and dichloro-
`methane have also been shown to bind to the major
`xenon binding site in myoglobin,42,56,57 but the greater
`size of these molecules causes some distortions in the
`surrounding protein structure and rearrangement of
`some amino acid side chains. Even a single nitrogen
`molecule binds to this same site at a pressure of 145
`bar.65 Xenon was also described to bind to serum
`albumin,9 renin, and tobacco mosaic virus,57,59 but no
`structural studies have been undertaken on these
`complexes.
`Due to its anesthetic properties,11 xenon has been
`used extensively as a prototype for theoretical and
`experimental studies on the interactions of anesthet-
`ics with proteins.13,14 The structural investigations
`of xenon, cyclopropane, and dichloromethane bind-
`ing to myoglobin and hemoglobin, along with the
`crystallographic analysis of halothane binding into
`the enzymatic site of adenylate kinase,47 provide, so
`far, good examples of the interaction of general
`anesthetics to specific sites in proteins. The nature of
`the molecular site of general anesthesia is still a
`matter of important debate,18,38 but over the last
`decade, evidence has been accumulated in support of
`the theory that general anesthetics act by binding
`directly to proteins,17,18 rather than by perturbing
`lipid bilayers in synaptic membranes, as it was
`
`Contract grant sponsor: CNRS/MR/CEA (Action Concertee
`des Sciences du Vivant 5); Contract grant sponsor: NATO.
`*Correspondence to: Thierry Prange, LURE, Baˆt. 209d,
`Universite´ Paris-Sud, 91405-Orsay Cedex, France. E-mail:
`prange@lure.u-psud.fr
`Received 22 October 1996; Accepted 29 April 1997
`
`Exhibit 2068
`Page 01 of 13
`
`

`

`62
`
`T. PRANGE´ ET AL.
`
`thought before.76 More recently, the considerable
`development in medicine of Magnetic Resonance
`Imaging (MRI) techniques led to a re-investigation of
`xenon as a probe in imaging and diagnostic tech-
`niques.30,77
`Protein-xenon complexes can be used as highly
`isomorphous heavy-atom derivatives for solving the
`phase problem in X-ray crystallography.48,50,52,53,63,70
`Xenon has now been used successfully as a heavy-
`atom in the structure determination of the human
`nuclear retinoid-X receptor RXR-a ligand binding
`domain,3 the molybdoenzyme DMSO reductase from
`Rhodobacter spheroides,54 the oligomerization do-
`main of the cartilage oligomeric matrix protein
`(COMP),34 the Photosystem I from Synechoccus elon-
`gatus,25 the lipoamide dehydrogenase domain from
`the outer membrane protein P64k of the bacteria
`Neisseria meningitidis,33 and the enzyme urate-
`oxidase from Aspergillus flavus.8
`Until recently, structural studies on protein-xenon
`complexes had been limited exclusively to the cases
`of myoglobin and hemoglobin. Binding of xenon into
`the enzymatic site of serine proteinases has been
`reported by Schiltz et al.51 The present study de-
`scribes and compares the xenon and krypton binding
`sites in a variety of proteins: porcine pancreatic
`elastase, subtilisin Carlsberg from Bacillus licheni-
`formis, cutinase from Fusarium solani (wild type
`and S120A mutant), collagenase from Hypoderma
`lineatum, hen egg-white lysozyme, the lipoamide
`dehydrogenase domain of the outer membrane pro-
`tein P64k from Neisseria meningitidis, urate oxidase
`from Aspergillus flavus, mosquitocidal d-endotoxin
`CytB from Bacillus thuringiensis sp. kyushuensis
`and the ligand-binding domain of the human nuclear
`retinoid-X receptor RXR-a.† These are proteins whose
`native three-dimensional structures have been deter-
`mined by x-ray crystallography and refined to resolu-
`tions equal or better than 2.7Å. 3,4,8,26,31,33,35,37,78
`
`MATERIALS AND METHODS
`Sample Preparation and X-Ray Data
`Collection
`Standard procedures, described in the literature
`were used to grow crystals of elastase,37 Subtilisin,44
`lysozyme,62 and P64k.32 Crystals of cutinase,35 colla-
`genase,4 RXR,3 urate oxidase,8 and CytB31 were
`obtained from the crystallographic groups who solved
`the structures. For cutinase, the xenon binding was
`investigated on the wild type protein as well as on
`the mutant S120A, where the active site serine is
`replaced by alanine. Crystals of this mutant are
`isomorphous to the native ones.36
`
`The method used to prepare xenon derivatives has
`been described earlier.50 Native crystals, mounted in
`quartz capillaries, fitted to a specially designed cell,
`were submitted to gas pressure a few minutes before
`starting the data collection.‡ The gas pressure was
`maintained during the data collection. Native and
`derivative data were collected under similar condi-
`tions.
`Krypton complexes were prepared for elastase and
`lysozyme at pressures of 48–56 bars.52
`Diffraction data were collected at the DW32 wig-
`gler beam-line at the LURE synchrotron facility in
`Orsay, France16 using x-rays at a wavelength of 0.9 A
`and a MAR-Research image plate detector. Diffrac-
`tion data were processed with the MOSFLM pro-
`gram.29 Data reduction and merging as well as all
`subsequent crystallographic computations were car-
`ried out with programs of the CCP4 package.6 A
`summary of the X-ray data collections on all samples
`is given in Table I.
`
`Xenon and Krypton Site Determination
`The coordinates of the major xenon sites were
`located by difference-Fourier calculations using the
`0Fderiv 0 2 0Fnative 0 terms as amplitudes and fnative as
`phases (known native structures were from the
`Protein Data Bank2: codes for elastase, lysozyme,
`subtilisin, cutinase, collagenase, CytB and RXR were
`6EST, 1LSE, 1SCA, 1CUS, 1HYL, 1CBY and 1LBD
`respectively).
`
`Refinements of the Xenon/Krypton Sites
`Structure refinements of elastase/Xe, subtilisin/
`Xe, collagenase/Xe, urate oxidase/Xe, and lyso-
`zyme/Xe complexes were carried out by the method
`of simulated annealing with the XPLOR program.5
`The native structures inclusive of the previously
`located xenon/krypton atoms and exclusive of water
`molecules were used as starting models. At regular
`intervals, water molecules were identified by differ-
`ence Fourier-calculations and added to the model. As
`occupancies and thermal factors are highly corre-
`lated, the xenon atoms were assigned a fixed thermal
`factor equal to the average thermal factor ,BW. of
`the crystallographically refined water molecules.
`The xenon occupancy factor was adjusted periodi-
`cally, so that its B remains within a few Å2 equal to
`the recalculated ,BW.. So, there is a constant
`adjustment of the occupancy factor. The final stages
`of the refinements were carried out using the stereo-
`chemically restrained least-squares minimization
`method with the PROLSQ program.19 The data
`given in Table I represent only a brief summary of
`these refinements. The full details are reported in
`
`†For simplicity and clarity, the following short names will be
`used hereafter: elastase, subtilisin, cutinase, collagenase, lyso-
`zyme, P64k, urate-oxidase, CytB, and RXR. COMP is an
`abbreviation for cartilage oligomeric matrix protein.
`
`‡A detailed description of the pressure cell and accessories
`can be found at the Internet site http://www.lure.u-psud.fr/
`WWW_ROOT/DOCUMENTS/lure/sections/xenon/xenon.html
`
`Exhibit 2068
`Page 02 of 13
`
`

`

`PROTEIN HYDROPHOBIC SITE EXPLORATION
`
`TABLE I. Xenon-Protein Complexes: Summary of Structural Data*
`
`Refinement of Xe complexes
`
`Protein
`Elastase
`Subtilisin
`Collagenase
`Cutinase (wild-type)
`Cutinase (S120A)
`Urate-oxidase
`P64k
`Lysozyme
`RXR
`CytB
`
`No. of
`residues
`240
`275
`2 3 230
`196
`196
`4 3 301
`2 3 481
`129
`2 3 238
`2 3 259
`
`Xe gas
`pressure
`(bar)
`8
`12
`12
`12
`12
`8
`13
`12
`20
`10
`
`No. of
`sites
`1
`1
`2 3 1
`1
`1
`4 3 1
`2 3 2
`4
`2 3 2
`1
`
`Final R-factor
`(%)
`18.3
`19.4
`19.7
`n.r.
`n.r.
`21.0
`n.r.
`16.8
`n.r.
`n.r.
`
`High-resolution
`limit (Å)
`2.2
`2.08
`2.53
`—
`—
`2.3
`—
`2.1
`—
`—
`
`*The refined structures have been deposited with the Protein Data Bank (n.r. 5 not refined).
`
`63
`
`No. of
`reflections
`used
`10,937
`9,665
`17,432
`—
`—
`18,523
`—
`6,521
`—
`—
`
`the header of each deposited file within the Protein
`Data Bank, Brookhaven.
`For P64k, the xenon complex was used as a
`heavy-atom derivative for the resolution of the crys-
`tal structure,33 and during the phasing process, the
`xenon sites were refined with the program
`MLPHARE.43 Strictly speaking, however, the diffrac-
`tion data were not on an absolute scale, so that the
`reported occupancies of the xenon atoms may not
`serve for comparisons with sites in other proteins.
`For cutinase, the xenon atoms were also refined with
`MLPHARE, but in this case, high-resolution data
`had been collected (up to 1.6Å), so that the structure
`factor amplitudes could be put on an absolute scale
`by a Wilson plot.
`Xenon complexes of DMSO reductase54 and
`COMP34 were described in the literature.
`Dr. Schindelin (California Institute of Technology,
`Passadena, CA, USA) has kindly communicated the
`co-ordinates of the Xe atom in DMSO reductase to us
`to include the description of this site in Table II.
`
`RESULTS
`Table II summarizes the short contacts observed
`within a range of c.a. 4.5 Å around the xenon atom at
`each site. If we add the case of the DMSO-reduc-
`tase,54 we remark that hydrophobic contacts are
`dominant (57 over a total of 77). When the size of the
`cavities is rather small (i.e., in RXR), it can accommo-
`date only one single atom in a well-ordered position.
`The observed electron density in the difference-
`Fourier maps, corresponding to the xenon atom,
`usually has the shape of a sphere or ellipsoid. However,
`when the cavity is much larger than the volume of a
`single xenon or krypton atom, disordered positions
`are observed as in mutants of phage T4 lysozyme.45
`The various binding cavities corresponding to the
`data gathered in Table II are depicted in Figures 1 to
`7. The most frequent side chains in interaction are
`from Ala (6), Leu (21), Val (11), Ile (9), or Phe (5)
`
`residues, or Ser, in the special case of serine protein-
`ases.
`
`DISCUSSION
`One of the most puzzling features of xenon binding
`to proteins is the large diversity of the binding sites.
`Xenon binds to intra- as well as to inter-molecular
`sites, to closed inaccessible cavities, as well as to
`exposed pockets and even into channel-pores. The
`binding sites may be lined up exclusively by ali-
`phatic residues, or they also may include aromatic,
`or, polar groups. Xenon may bind into void sites, or
`replace existing water molecules. Below, we have
`attempted to classify the various binding sites into
`subsets according to their common features.
`
`Definition
`A cavity is a region in a protein that is not occupied
`by protein atoms and that is entirely closed off by the
`protein. By definition, a cavity is inaccessible from
`the outside, in the static description of a crystal
`structure. In contrast, a pocket (sometimes called
`surface invagination) is connected (accessible) to the
`outside. A pocket would only become a closed cavity if
`the atomic radii of the protein atoms that delimit the
`cavity were to be increased. In order to test whether
`binding sites are cavities or pockets, molecular sur-
`faces were calculated with the Connolly algorithm.10
`Cavities are delimited by closed surface patches that
`have no connection to the outer molecular surface.
`It is well known that the atoms in the interior of
`protein molecules are densely packed, and that these
`regions are more similar to solids than to organic
`liquids. Calculations presented by Klapper22 have
`estimated that there is twice as much free volume
`distributed throughout simple organic liquids than
`in proteins. However, in proteins, the free volume
`does not need to be distributed randomly (as in
`liquids), and empty intramolecular cavities exist in
`numerous proteins.21,23 These cavities exist at the
`
`Exhibit 2068
`Page 03 of 13
`
`

`

`TABLE II. Closest Contacts (Å) Around the Xenon (or Krypton) Atoms in the Various Binding Sites,
`Within a Sphere of 4.5–4.7 Å
`
`Atom
`Xe
`
`Kr
`
`Xe
`
`Xe
`
`Xe
`
`Xe
`
`Xe
`
`Kr
`
`Xe
`
`Xe
`
`Xe
`
`Protein
`Elastase
`
`Elastase
`
`Subtilisin
`
`Cutinase (native)
`
`Collagenase
`
`Urate-oxidase
`
`P64k
`Site #1
`
`Site #2
`
`Lysozyme
`
`Lysozyme
`Site #1
`(intermolecular)
`
`Site #2
`(intramolecular)
`
`Site #3
`(intramolecular)
`
`CytB
`(intermolecular)
`
`RXR
`Site #1
`
`Site #2
`
`DMSO-reductase
`
`Xe
`
`Atom (resid.)
`Og(S195)
`Cg2(V216)
`Og(S195)
`Cg2(V216)
`Og(S221)
`Ca(G154)
`Og2(T220)
`Ow(310)
`Og(S120)
`Cg2(T150)
`Cd2(L182)
`Subunit-A:
`Og(S195a)
`Cg2(V216a)
`Cg1(V213a)
`Oe1(Q37b)
`O(S214a)
`C(C191a)
`Ow(234)
`Cd2(L178)
`Cd1(F219)
`Cg2(T230)
`
`Cd2(L323)
`Cb(V319)
`Cd1(I179)
`Cb(N175)
`Cd(L327)
`C(W348)
`Og1(T318)
`Cg2(V92)
`Cg2(I55)
`Ce(M12)
`
`Og1(T43)
`N(R45)
`Cg2(T51)
`Ow(20)
`1 symmetry-related
`Cg2(V92)
`Cg2(I55)
`Ce(M12)
`Ca(I58)
`O(Q57)
`Cd1(I98)
`Cd1(L33)
`Cd1(I233)
`1 symmetry-related
`
`Cg1(V332)
`Ca(A337)
`Cg2(V342)
`Cd2(L441)
`Cd1(L370)
`Cd1(L425)
`Cz(F110)
`Cg1(V402)
`Cd2(L406)
`Ch2(W449)
`
`Dist.
`3.40
`3.89
`3.28
`3.75
`4.03
`3.92
`4.11
`4.43
`4.24
`4.19
`3.99
`
`3.67
`3.98
`4.22
`3.76
`4.53
`4.29
`3.87
`4.33
`4.12
`3.99
`
`3.42
`4.59
`4.60
`4.00
`3.60
`3.70
`3.25
`3.12
`3.46
`3.55
`
`3.65
`4.25
`4.55
`4.50
`
`3.18
`3.51
`3.54
`4.30
`4.03
`4.30
`3.4
`4.7
`
`4.09
`3.25
`3.41
`4.23
`2.92
`3.78
`3.23
`3.56
`4.21
`4.29
`
`Atom (resid.)
`O(C191)
`Cg2(T213)
`O(C191)
`Cg2(T213)
`Cb(A152)
`Nd2(N155)
`Ca(L126)
`
`Nd2(N84)
`Cg1(V177)
`
`Subunit B:
`Og(S195b)
`Cg2(V216b)
`Cg1(V213b)
`Oe1(Q37a)
`O(S214b)
`C(C191b)
`Ow(552)
`Cg2(T180)
`Cg1(V227)
`Cd2(L252)
`
`Cb(A176)
`O(L172)
`Cd2(L202)
`Cg(L172)
`Cz(F356)
`Ce(R355)
`N(L314)
`Cd1(L56)
`Cd1(I88)
`Cb(S91)
`
`Cb(T43)
`N(N44)
`Nh2(R68)
`
`Cd1(L56)
`Cd1(I88)
`Cb(S91)
`Ne1(W108)
`Cd1(W1080)
`O(A107)
`Cd1(I54)
`Cz(F237)
`
`O(S336)
`Cb(A340)
`O(K440)
`Cb(D444)
`Cb(R421)
`Cd1(L422)
`Cb(A428)
`Ce(M405)
`Cd2(L452)
`
`Dist.
`3.71
`4.0
`3.90
`3.85
`3.94
`3.73
`3.99
`
`3.33
`4.17
`
`3.51
`4.01
`4.33
`4.0
`4.58
`4.27
`4.40
`4.79
`4.14
`3.68
`
`3.88
`3.66
`3.81
`4.57
`2.65
`3.66
`3.45
`3.38
`4.16
`3.89
`
`4.00
`4.36
`4.49
`
`3.33
`4.19
`3.91
`3.75
`3.96
`3.78
`4.2
`3.9
`
`4.33
`3.36
`3.69
`3.94
`4.30
`4.62
`3.58
`4.05
`4.23
`
`B/Occ*
`22.6/0.81
`
`18.9/0.49†
`
`29.3/0.71
`
`19.1/0.81‡
`
`28.1/0.95 (in site A)
`
`31.8/0.80 (in site B)
`
`30.5/0.85
`
`22.5/0.55‡
`
`30.5/0.40‡
`
`19/0.49
`
`36.2/0.33
`
`32.9/0.28
`
`36.0/0.1
`
`n.r.¶
`
`n.r.
`
`n.r.
`
`n.r.
`
`*Occ 5 occupancy factor, B 5 temperature factor (refined, Å2) for Xe or Kr. Refined using PROLSQ.19 Standard deviations are within 2
`Å2 for B factor and 60.1 for Occ.
`†Refined with program SHARP.27
`‡Refined using program MLPHARE43 after absolute scaling of the data.
`¶n.r. 5 not refined.
`
`Exhibit 2068
`Page 04 of 13
`
`

`

`PROTEIN HYDROPHOBIC SITE EXPLORATION
`
`65
`
`Fig. 1. View of the binding sites of xenon atoms in hen egg lysozyme (lower) and RXR (upper).
`The accessible surfaces are calculated by the method of Connolly.10
`
`expense of considerable cost in free energy, so that it
`is unlikely that they are mere packing defects. The
`hypothesis that cavities are important for the confor-
`mational flexibility of protein molecules is supported
`by the characteristics of xenon binding to myoglobin.
`Tilton et al.66 have observed an overall reduction in
`temperature factors upon xenon binding. This effect
`was interpreted as a ligand-induced restriction of
`
`the number of conformational states. Such an inter-
`pretation would also explain why the rotational
`degrees of freedom of bound water molecules in the
`protein decrease upon xenon binding.79 Binding of
`xenon to myoglobin cavities also affects the func-
`tionality of the protein in a rather drastic way.61
`All these observations are evidences of the impor-
`tance of intramolecular cavities for the dynamics
`
`Exhibit 2068
`Page 05 of 13
`
`

`

`66
`
`T. PRANGE´ ET AL.
`
`Fig. 2. The distances around the xenon in site #2 of lysozyme (the inner cavity in Fig. 1). The
`xenon atom is represented at its van der Waals radius.
`
`and functionality of protein molecules. The prepara-
`tion of xenon complexes provides a unique experimen-
`tal method to detect and study such cavities.
`
`Xenon Binding Into Hydrophobic
`Intramolecular Cavities
`The main intramolecular binding site in lysozyme
`(site #2 in Table II), the binding sites in urate-
`oxidase and P64k, one of the binding sites in RXR
`(site #2 in Table II), and the binding site in DMSO-
`reductase belong to this category. These sites are
`truly closed cavities (Fig. 1), similar to the binding
`sites observed in myoglobin and hemoglobin. The
`observed cavities are usually small, accomodating
`space for no more than one xenon atom and they do
`not contain ordered water molecules. Thermody-
`namic considerations suggest that they are also void
`of disordered water.72,74 They are buried in the
`protein interior, sometimes at a distance of several
`angstro¨ms from the molecular surface, and they are
`thus built predominantly by hydrophobic side chains.
`Access to these sites is only possible via transient
`channels produced by concerted domain movements,
`as it was suggested earlier in the case of myoglo-
`bin,64,66–68 as well as for the binding of small, organic
`molecules,40,45 or molecules like benzene or indole to
`
`the interior of Phage T4 lysozyme mutants.15 The
`first of these studies demonstrates that there is a
`correlation between the geometry of the binding
`cavity, and the degree of occupancy and the disorder
`of the ligand. The second report illustrates the
`influence of the protein core fluctuations on the
`millisecond to microsecond time scale, which allow
`the molecules to easily reach the core region of the
`protein.
`The binding equilibrium of xenon to proteins has
`been analyzed both theoretically20,67 and experimen-
`tally.50,51,52,64,66
`
`Xenon vs. Krypton
`In elastase, krypton binds to the same site as
`xenon, but in order to achieve noticeable substitu-
`tion, substantially higher pressures have to be ap-
`plied.52 This is probably due to the fact that krypton
`atoms have a smaller electronic polarizability than
`xenon atoms. Thus, the van der Waals forces that
`exist between the rare gas atoms and the protein
`molecules are smaller with krypton gas. In lysozyme,
`krypton only binds into the major intramolecular
`site (site #2).52 Even at high pressures (48 bar), the
`three other xenon binding sites are not observed to
`
`Exhibit 2068
`Page 06 of 13
`
`

`

`PROTEIN HYDROPHOBIC SITE EXPLORATION
`
`67
`
`d-Endotoxin CytB: the intermolecular xenon binding site delimited by two symmetry-related
`Fig. 3.
`molecules. The two-fold axis runs horizontally in the plane; the symmetry related molecule is in light tracing.
`
`fix krypton. The smaller size of krypton atoms
`presumably also plays a role in this behaviour.
`
`crystal, this site lies on a crystallographic two-fold
`axis. (Fig. 3).
`
`Intra- vs. Inter-Hydrophobic Cavities:
`Lysozyme and CytB
`Xenon and krypton bind to hen egg lysozyme and
`CytB only to a moderate extent. In lysozyme, a
`difference-Fourier map reveals three sites and prob-
`ably a fourth with a very low occupancy. Two of them
`(sites #1 and 4) are inter-molecular sites, located
`onto (or near to) a crystallographic two-fold axis,
`(direction [110]) and are located in cavities that are
`formed by the contact of symmetry-related mol-
`ecules. The two other (sites #2 and 3) are intra-
`molecular sites. The site #2 is a small spherical
`cavity deep in the inner core of the enzyme (Figs. 1,
`2). In contrast, krypton binds solely to the intra-
`molecular site, visible on Figure 1, with an occu-
`pancy of about 50%. In the case of the mosquitocidal
`d-endotoxin CytB,31 xenon binds into a closed inter-
`molecular cavity at the dimer interface. In the
`
`Urate Oxidase Binding Site: A Probe for the
`Oxygen Binding?
`Urate oxidase from Aspergillus flavus is an en-
`zyme that converts uric acid to allantoin with the aid
`of molecular oxygen. The unique binding site of
`xenon (Fig. 4) is nearly fully occupied.8 The xenon
`atom is located near the active site of the enzyme.
`Hemoglobin was one of the first proteins shown to
`bind xenon. One might anticipate that many other
`enzymes, which are able to use molecular gas as one of
`their substrates like Ni/Fe hydrogenases, 71 nitroge-
`nases, or hemoglobins, and known to contain hydropho-
`bic channels or pockets, would be potentially good
`candidates for binding xenon, in (or near to) the regions
`involved in the enzymatic reaction or transport. As a
`consequence, xenon may be a useful probe in the
`elucidation of the mechanism of such enzymes.
`
`Exhibit 2068
`Page 07 of 13
`
`

`

`68
`
`T. PRANGE´ ET AL.
`
`Fig. 4. Urate-oxidase: the xenon binding site. This cavity is in
`the vicinity of the active site of the enzyme which converts uric acid
`to allantoin using molecular oxygen. In light tracing, the refined
`native coordinates, in bold tracing, the refined xenon complex.
`
`This illustrates the negligible steric modification induced by the
`binding in this case (the average distances for atoms of the 20
`residues in the vicinity of the xenon are c.a. 0.16 Å, only two times
`the r.m.s. deviations in the refined models: 0.08 Å).
`
`Inter-Helical Hydrophobic Pockets: The Case
`of P64k and RXR
`a-Helices in cross-interactions usually lead to the
`building of super secondary motifs that minimize
`interactions.7 Very often, hydrophobic side-chains,
`coming from different helices, are then brought into
`close contact to each other, delimiting pockets that
`may accommodate xenon atoms. Such pockets are
`observed in the P64k lipoamide dehydrogenase do-
`main and in the RXR retinoic binding domain (Figs.
`5, 6). In each crystal structure, two of these pockets
`are observed. In the case of P64k, a major (>50%)
`and a minor site (30%) were found. In the case of
`RXR, the binding was so favorable that both sites
`were nearly fully occupied and the structure was
`solved mainly with the aid of this doubly labeled
`derivative. It is important to note that one of these
`cavities (site #1) corresponds to the putative ligand-
`
`binding site (pocket B in the description given by
`Bourguet et al.3).
`
`Xenon Binding to COMP: A Molecular Model
`for the Action of General Anesthetics?
`COMP34 is a unique example of multiple xenon
`binding sites in a channel-like structural motif.
`COMP forms a pentameric coiled coil with subunits
`constituted by a-helical structures. The hydrophobic
`axial pore of the pentameric bundle is filled with
`water molecules in the native state, but under a
`pressure of 10 bar, xenon atoms bind into the pore
`and displace some of the water molecules. A total of
`eight xenon atoms were refined at discrete binding
`sites, lining up the pore axis. We suggest that this
`xenon complex may provide a molecular model for
`the action of general anesthetics in postsynaptic ion
`channels. The molecular targets of general anesthet-
`
`Exhibit 2068
`Page 08 of 13
`
`

`

`PROTEIN HYDROPHOBIC SITE EXPLORATION
`
`69
`
`Fig. 5. P64k: the major binding site of xenon, with the refined model superimposed to the
`electron density (calculated with \Fderiv 0 2 0 Fnative\ amplitudes and the refined phases; contouring is
`at 5 s above the mean density).
`
`ics in the organism are still unknown17,18,38 but there
`is evidence that anesthetic molecules inhibit the ion
`flux through the channel of ligand-gated postsynap-
`tic receptors by binding to discrete protein sites.80
`Site-directed mutagenesis studies on the nicotinic
`Acetylcholine Receptor (nAchR) suggest that anes-
`thetics bind directly to the channel pore.81 The
`structure of COMP has marked similarities with
`proposed models of the nAchR and the observed
`discrete xenon binding sites in the channel pore of
`COMP34 lend therefore strong support to the above-
`mentioned theory of the action of general anesthet-
`ics.
`
`The Special Case of the Serine Hydrolase
`Binding Sites
`Serine hydrolases possess a unique catalytic motif
`built by the so-called catalytic triad (most commonly
`Ser-His-Asp). Apart from a few examples, all triads
`have similar spatial configurations while the rest of
`the protein folding may be completely different. The
`arrangement of the chain around the catalytic site
`delimits a binding crevice for the amino-acid to be
`cleaved, and is at the origin of the specificity of the
`enzyme. The character of this zone is usually hydro-
`phobic, and indeed the xenon atom is observed
`
`bonded in this region. The four hydrolases we have
`investigated are able to fix xenon with a high occu-
`pancy in a single well-defined site. A superimposition
`of all catalytic triads complexed with the xenon
`positions is shown in Figure 7. For some of them,
`their binding sites have already been described51:
`the xenon (or krypton) atom is close to the Og of the
`active serine. It is located in the P1 site (following
`the nomenclature of Schechter and Berger49), with
`additional contacts towards a few hydrophobic resi-
`dues, or towards polar atoms. In contrast to the
`small, closed hydrophobic cavities observed in the
`majority of proteins, the xenon binding sites in
`serine hydrolases are large solvent-accessible pock-
`ets that, in the cases of elastase, collagenase and
`cutinase contain ordered water molecules which are
`displaced upon xenon binding. Thus, apart from
`mapping hydrophobic cavities that are important in
`protein dynamics, xenon can also be used to detect
`potential substrate or ligand binding sites in pro-
`teins.
`The role played by the catalytic serine is, in fact,
`not crucial for xenon binding. The S120A mutant of
`cutinase (where the active site serine is replaced by
`an alanine) still fixes the xenon at the same place
`and with equivalent occupancy.
`
`Exhibit 2068
`Page 09 of 13
`
`

`

`70
`
`T. PRANGE´ ET AL.
`
`Fig. 6. RXR: locations of the xenon binding sites. The site #2 is delimited by residues coming
`from two helices. Site #1: the side chains which delimit the cavity belong to three different helices,
`the cavity is very small, and the xenon atom has a full occupancy. The schematic ribbon diagram of
`the molecule above indicates the location of the two sites.
`
`Xenon and Krypton as Heavy Atoms
`The recent interest in protein-xenon complexes
`has been prompted by the possibility of using them
`as heavy-atom derivatives for phase determination
`in macromolecular crystallography.48,50,52,53,63,70 The
`crystal structures of RXR, DMSO-reductase, COMP,
`Photosystem I,25 P64k, and urate-oxidase have been
`solved by the MIR method with isomorphous deriva-
`tives including xenon complexes. The structure of
`RXR was initially solved with a 5Å MIR map, based
`only on the xenon derivative. The a-helices of the
`structure were clearly visible in this initial map and
`a discontinuous polyalanine model could be traced.
`At a latter stage, data from a classical mercury
`derivative were added to improve the map. Because
`of its high degree of isomorphism, the xenon deriva-
`tive has a significantly better phasing power than
`
`the mercury derivative.3 Xenon derivatives are likely
`to bind to sites that are different from those of
`standard heavy atoms, which, like Hg or Pt, bind
`predominantly to specific functional groups. The
`structure of urate-oxidase illustrates this advan-
`tage8: only Hg and Pb derivatives of good quality had
`been obtained by standard soaking techniques, but
`the major binding sites of both cations were very
`close to each other, thus limiting the phasing power.
`The heavy-atom site in the xenon derivative is far
`away from the Hg and Pb sites, and thus improved
`the quality of the MIR phases.
`The major advantage of xenon derivatives is their
`very high degree of isomorphism, but this notion
`should perhaps not be pushed too far. Local rear-
`rangements of side-chains and displacements of
`water molecules may occur upon rare gas binding, as
`
`Exhibit 2068
`Page 10 of 13
`
`

`

`PROTEIN HYDROPHOBIC SITE EXPLORATION
`
`71
`
`the catalytic triads of elastase,
`Fig. 7. Superimposition of
`subtilisin, cutinase (native and S120A mutant), and collagenase
`(two active sites). The least squares fit was performed using only
`the side chains of Ser195-His57-Asp102 (numbering refers to the
`standard chimotrypsin). The figure shows the refined position of
`the bonded xenon atoms in the five structures. Xenon positions
`are all within a sphere of 1Å, at a distance of >4Å away from the
`catalytic serine in proteinases (although not
`involved in the
`
`binding, the conserved Ser214 which lines the pocket border, is
`also represented). Note the special case of cutinase and its S120A
`mutant (light tracing); the hydrophobic pocket of this esterase is
`differently oriented with respect
`to the triad. The xenon still
`remains at close distance from Ser120 (or Ala120) but is no longer
`superimposable to the others. Xenon labeling: 1 5 elastase; 2 5
`subtilisin; 3 and 4 5 collagenase; 5 and 6 5 cutinase, native and
`mutant.
`
`was evidenced by a careful analysis of krypton
`binding to elastase.52 Though these structural
`changes do only marginally impede the quality of the
`calculated SIRAS-phases, they show that the con-
`cept of gas binding into a pre-existing site, without
`any perturbation of the surrounding protein struc-
`ture is only accurate in a first approximation. Bind-
`ing of dichloromethane and cyclopropane to myoglo-
`bin is also shown to be accompanied by local
`structural rearrangements.42,56,57 Stowell et al.63 re-
`port on two cases where xenon derivatives were
`non-isomorphous, due to changes in cell parameters.
`They noticed that cell parameters did not change
`upon pressurization with nitrogen gas, and con-
`cluded that this effect is due to a strong binding or
`interaction between xenon and the protein. In the
`absence of a more detailed analysis, one can only
`speculate about the cause of this effect, but one
`plausible explanation would be that xenon binds to
`intermolecular cavities (as in lysozyme or CytB),
`thus inducing changes (or disruption) in the crystal
`lattice.
`
`What size protein can be phased by xenon or
`krypton? The answer obviously depends on the num-
`ber of heavy-atom sites but a single xenon site gave a
`useful derivative for DMSO-reductase, a 85kDa pro-
`tein.54 On the other hand, careful data collection and
`processing, as well as statistically optimal heavy-
`atom refinement and phasing, allowed the determi-
`nation of a high-quality electron density map for
`elastase (26 kDa) from a single, half-occupied kryp-
`ton atom.52 From this, one can extrapolate that a
`single, fully occupied xenon atom can be used to
`produce a similar result with a 90 kDa protein. For
`larger structures, it is anticipated that more binding
`sites are requir