Involvement of DPP-IV catalytic residues
`in enzyme-saxagliptin complex formation
`
`"\\1LLIAM J. METZLER} JOSEPH YANCHL"NAS/ CAROLYN WEIGELT,1 KEVIN KISH, 1
`HERBERT E. KLEI, 1 DIANLIN XIE, 1 YAQIT'r ZHANG, 1 MARTIN CORBETT, 1
`JAMES K. TAMURA, 1 BIN HE,2 LAWRENCE G. HAMANN, 3 MARKS. KIRBY?
`AND JOVITA MARCINKEVICIENE1
`1Deparlment of Molecular Biosciences, Bristol-Myers Squibb Re~carch and Development. Princeton,
`New Jersey 08543-4000, USA
`2Departmenl of Metal:tolic Diseases, Btistol-Myers Squibl:t Research and Development. Princeton,
`New Jersey 08543-4000, USA
`3Department of Discovet:y Chemistry, Bristol-Myers Squibb Research and Development, Princetoa,
`New Jersey 0~:543-4000, USA
`(RECEIVED September 18, 2007; FlNAL REVISION Noveml:ter 8, 2007: ACCEPTED November 8, 2007)
`
`Abstract
`The inhibition of DPP-IV by saxagliptin has been proposed to occur through fonnation of aeMalent but;;;
`l\W¢t$il)le con'l:p]..ey;.. To evaluate fmther the mechanism of inhibition, we detennined the X-my crystar
`structure of the DPP-N: saxaglipti n complex. Th:is·structtire reveals covalent {ltta,c;!n,ne~t bet,vet}n 1)630 ~
`th¢inhlbitorrtillile.carb6rf(C~O:dis~aricek:liiJ,Af: To iove::-;tigate whetheT this serine addition is assisted by
`the catalytic His-Asp dyad, we generated twD mut<1nts of DPP-IV, S630A and H740Q, and assayed lhem for
`ability tu bind inhibitor. DPP-IVH7
`.mQ bound saxagliptin with an ~I 000-fold reduction in affinity relative to
`DPP-Nwr, while DPP-IVs630
`A showed no evidence for binding inhibitor. Ap.al),'l\l,'l}g.Qf!i;l)(agltptin laO,Jd:n&
`tl1!;ll)itrile.gtou'p·i!h0wed unchMg~qbiodiqgpr{)pertieiftotneboth mut:iiht protei~, highlighting th;; essential
`role 5630 and H740 play in covalent bond formation between S630 and saxagliptin. Further suppm1mg
`mechanism-based inhibition by saxagliptin, NMR spectra of enzyme-saxagliptin complexes revealed the
`presence of tluee downfield resonances with low fractionation factors characteristic of short and strm1g
`hydrogen bonds (SSHB). Comparison of the NMR spectra of various wild-type and mutant DPP-IV:ligand
`complexes enabled as~ignment of a resonance at -14 ppm Lo H740, Two additional DPP-IV mutants, Y547F
`and Y547Q, generated to probe potential stabilization of the enzyme-inhibitor complex by this residue, did
`not show any differences in inhibitor binding either by ITC or ::--n\fR. Together with the previously published
`enzymatic data, the structural and binding data presented here strongly support a histidine-assisted covalent
`bond formation between 5630 hydroxyl oxygen and the nitrile group of snxagliptin.
`Keywords: DPP-IV; X-ray crystal structure; mutant; ITC; proton NI\1R; short, strong hydrogen bond;
`serine protease; saxagliptin
`
`Dipeptidyl peptidase IV is a serine protease that modu(cid:173)
`lates the biological activity of specific circulating peptide
`hormones, chemokines. cytokine,~. anrl neuropeptides
`
`(Lambeir et al. 2001) by specifically cleaving two N(cid:173)
`terminal amino acids. DPP-IV attenuates postprandial
`blood glucose comro1 through inactivation of GLP-1
`
`Reprint r~quests la: Wilham J. Metzle1·, Mail Stnp H23-02, P.O. Box
`4000, P1'incoton, NJ 08543-4000, USA; e·mail: william.me!zler@bms.
`com; fax: (609) 252-60 (2,
`Abbreviations: ANS, 1-ar:ilino-8-naphthalene sulfonate; DPP-IV,
`dipcptidyl peptida~e IV; GLP·L glucagon-like peptide-!; ITC. iso(cid:173)
`thermal titration calorimetq; pNA, p-nitroaniline; P.~P,•, amino acid
`re&idues of the sub~trate which numericnlly indicnte the position
`relati"ely to the ;;cissile bond, n being the rc«idues toward the N
`
`terminus (this
`terminu" and n' being the residues toward the C
`nomendature was first defined by Schechter and Berger [1967]);
`SEC-MALS, &ize-exclu&ion chromatography~multiple angle light scat(cid:173)
`tering; SKIE, sohent kinetic isotope effc<:t; SSHB, sLort strong
`hydrogen bond; SIHB, "hort ionic hydrogen bond; TSE, theriTllll
`stability enhancement.
`Article and publication a'" at http://www,protcinscicncc.org/cgi/doif
`10.111 Ofps.0/3253208.
`
`240
`
`Prorein St:ieJrrP (20(181, 17:24(~250. Ptblisbed tty C.old Spring Harbor Lat:toratory Press. Copyright© 100E The Pmtein Society
`
`AstraZeneca Exhibit 2072
`Mylan v. AstraZeneca
`IPR2015-01340
`
`Page 1 of 11
`
`

`
`(Ahren ct a!. 2002), suggesting that modulation of DPP(cid:173)
`lV activity can serve as a potCJltial treatment for diabetes.
`Toward that end, the phannacological proof of concept
`for DPP-IV inhibitors in the treatment of type II diabetes
`is now well-established, and the first member of this
`class, sitagliptin (Kim et at. 2005), is now approved by
`the FDA; the second member, vildagliptin (Ahren et al.
`2004, 2005), awaits approval.
`DPP-IV is a prototypic l>erine protease whose substrate
`cleavage is driven by activation of a Ser-His-Asp triad.
`Numerous reports describing the discovery of structurally
`diverse DPP-IV inhibitors have appe11red
`in the la~t
`decade (Hughes et al. 1999; Villhauer et al. 2002; Augeri.
`et a!. 2005; Kim et al. 2005). All known inhibitors reported
`to date occupy the SJ-S2 pocket at the DPP-IV active
`site and make extensive hydrophobic, van der Waals, and
`hydrogen-bonding interactiom with residnes lining this
`pocket. l'(lte!l'l8tingly; ·- cyail.Opytrolidine derivatives·. were
`si:1ggested to Yi.eld ·a covalent .adducr with ·the active.Osite•t'
`serine· (S630) · of DPP~Iv; fomring: an iniidate ·. :otdduct;.
`(Hughes et al. 1999; Oefner et a1. 2003). Mechanism-based
`enzyme-inhibitor complex fonnation between DPP-IV and
`saxagliptin (Fig. l) has been extensively described in our
`previous publication (K;im.ettl! .. :'.?-0()6)). We suggested initial
`inhibitor binding .is followed by serine addition to the inhib(cid:173)
`llor nitrile carbon, catalyzed by histidine. This suggestion
`was supported by the observation of an inverse solvent kinetic
`isotope effect (SIGE) for the onset of inhibition and by the
`NMR detection of a proton fractionating b1 a short and
`strong hydrogen bond in the DPP-IV:saxagliptin complex.
`In this repmt, we fmther investigate _the mechanism
`of inhibition of DPP-IV by saxagliptin. The structure of
`the DPP-IV:saxagliptin complex was determined by X(cid:173)
`ray crystallography. To understand more fully the origin
`of the downfield NMR resonances and to elaborate the
`involvement of the active-site residues in the inhibitor
`binding, we pelformed mutagenesis studies targeting
`residues in the active site of the enzyme. Human DPP(cid:173)
`JV mutants DPP-1V 864(1A_ DPP-lVm4oo, DPP-IVY547Q,
`and DPP-IVY547
`F were expressed in baculovirus. purified
`to homogeneity, and tested for activity in a dipeptide
`(Gly-Pro--pNA) cleavage a~say. The effect of the muta(cid:173)
`tions on saxagliptin binding was characterized by NMR,
`
`A
`
`B
`
`Ho-&J,A_
`H2N~Ny
`° CN
`
`Figul"e 1. Chemical stnlctures o:F (A) saxagliptin nnd (B) BMS-538305.
`
`Binding of saxagliptin to DPP-I.V variants
`
`ITC, and TSB. The role of the nitrile moiety wa~ a~sessed
`by monitoring the binding of BMS-538305, an analog of
`saxagliptin lacking the nitriJc group (Fig . .1 ).
`
`Results
`
`Characterh:.atio11 of mutant proteins
`All mutant proteins were expressed to the same level and
`purified to the same purity (Fig. 2}. As expected, all were
`found to be devoid of catalytic activity when measured
`at 1 O-f old excess of enzymes and 1 0-fold higher sub(cid:173)
`strate ccmcentration compared with the wild-type protein.
`Hence, specific activity (;Ould not be used as a gauge of
`protein quality. Instead, various biophysical methods
`were used to assess whether the proteins were properly
`folded. TSE and CD thermal denaturation data indicated
`that the melting temperatures CTm) for DPP-IVwT, DPP(cid:173)
`IV8630A, DPP-IVY547F, and DPP-IVY5470 were identical
`(70°C), whereas the Tm for DPP-IVH7400 was ,-10°C lower
`(600C). Comparison of tile CD spectra for wild-type and
`mutant DPP-IV proteins revealed that the &econdary ~truc­
`ture was unperturbed by the mutations (data not shown).
`Similarly, the upfield region of the 1H NMR spectra were
`indistinguishable, indicating the hydrophobic cores of the
`prorems were unperturbed (data not shown). To evaluate the
`quaternary states of the proteins, SEC-1.-IALS studies were
`performed. All of the DPP" IV mutants were found to be
`dimeric with the same solution MW as wild type (Tah1e 1),
`providing furthe1· evidence that the mutants were well
`folded and suitable for binding studies. Finally, ITC studies
`demonstrated that all proteins were competent for binding
`inhibitor (see below).
`
`Crystal stmcture ofwild-type DPP-IV complex
`with saxagliptin
`
`The three-dimensional structure of DPP-IV:saxagliptin
`complex was detennined by X-ray crystallography (Fig. 3).
`Statistics for the data refinement are provided in Table 2.
`'J:Qe ~tructure revealed, that saxagliptin is> coval~tltly bouri(l
`··.to the PPP"IVactive.,sit~ withtb~ nitri!eJorroinga-c'ovaleirt
`~1~*~4~~£}, ~!~~~~j~~~~~l~fJ~io~~~~Jt;ie~~:~~f~<-·.
`~~~~;~~~:er~t:!~~~:~:-~:z:e~;:~~h:~j!~i~~~~~;:~:;
`
`4;5"fl1'1'}tb!tP()py:n:o)ir;li!}e )iog is. buried in th~ hydf:OJ:ihobicr.
`
`fqqp,cp~,; pocl<e~; V7 L l, V656, Y662, Y666, W659, and
`Y547. In the S2 pocket, the side-chain NH?: of N71 0
`hydrogen bonds with the carbonyl oxygen of saxagliptin.
`The prink'lY)' amine of saxagliptjn, which is a critical
`functionality for potent inhibitory activity in this series,
`participates in a symmetrical hydrogen-bonding network
`with Y662 and glutamic acids, E205 and E206. The
`
`www.proteinscicnce.org
`
`241
`
`Page 2 of 11
`
`

`
`Metzler et al.
`
`li'igut·e 2. SDS-PAGI1 of DPP-[Vw' (lane I), DPP-IVH7400 (Jane 2). DPP(cid:173)
`IV'6:li)A (lHn~ 3), DPP-IVY5~7F (lane 4), and DPP-IVY547Q (lnne 5). The
`4%~1.2% NuPAGE gel was .run under nonrcducing conditions. Each l<me
`contains 5 11!! of protein.
`
`\'~ant aile ~g'~.rxtepds• into the SZ pock~t enabling two
`hydrogen bonds with the side-chain hydroxyl of Y547: one
`with the adamantane hydroxyl and the second witb the
`imidate nitrogen. Although DPP-IV exist<> as a dimer in the
`unit cell, only monomer A is shown in Figure 3 for
`simplicity because the protein~inhibitor interactions are
`much the same for both molecules in the asymmetric unit.
`One notable difference is the hydrogen bond between the
`imidate nitrogen and the side-chain oxygen of Y547
`(distance of 3.1 A). This hydrogen bond is seen in
`monorner A but not in JnonouHor B. Iu monomer B, lhe
`imidate nitrogen points toward Rl25 and may interact with
`this
`residue
`though water-mediated hydrogen bonds
`(unmodeled; data not shown).
`
`Binding studies by fTC and TSE
`
`To evaluate the ability of the mutant proteins to bind
`saxagliptin and BMS-538305, we performed isothermal
`calorimetry and thermal stability enhancement studies.
`tightly by lTC
`Saxagliptin bound DPP-IVwT very
`(Kd « 5 nM, Table 3) but did not bind to DPP-IV 5630
`A
`(Fig. 4A,B; Tables 3, 4). This result was verified by
`thermal stability enhancement experiments (Fig. 5}, in
`which no change in Tm was ohserved at saturating
`sax:Jgliptin concentrations relative to protein alone. In
`contrast, the saxagliptin analog BMS-538305 bound
`tightly to both DPP-IVwr and the DPP-IVS630A. Saxa(cid:173)
`10
`gliptin also bound poorly (Kd > l [LM) to DPP-IVH7
`'
`Q,
`whereas BMS-538305 showed almost unaltered affinity
`{Kct = 6.4 :!:: 2.2 nYI). Binding of both compounds was
`essentially unaffected by mutation of Y547 (Q or F). In
`all lTC experiments in which binding was observed, the
`
`242 Protein Science, vol. 17
`
`apparent stoichiometry was approximately one compound
`molecule per DPP-N monomer (observed molur ratio
`:+: 0.1 0, N = 23
`0.93
`titrations ), providing further
`evidence that the inactive variants were properly folded,
`
`Detection of SSHB by NMR
`
`\Vhen serine protease~ bind lUeGhanism-baseJ inhibitors
`capable of fanning Ser adducts analogom; to the tetrahe(cid:173)
`dral intermediate in catalysis, the Asp .. ·His H-bond of the
`catalytic triad frequently becomes a SSHB (Cassidy et al.
`1997; Lin et al. 1998). The proton in this hydrogen bond,
`the His nitrogen (W 1
`) proton, located between the
`imidazole and aspartate groups, becomes unusually stable
`and can often be detected in the low-field 1·egion of the 1H
`NMR spectrum, typically in the range of 14-20 ppm
`(Robillard and Schulman 1972). We had previously
`noticed that the NMR spectrum for DPP-IVWT showe-d
`a re-sonance at 16.1 ppm, even in the absence of bound
`inhibitor. By comparison to literature reports, we had
`assigned this resonance to the H740 N°1 (Kim et al.
`2005). To confirm the as:signment of this peak, we
`collected the 1H Nl\1R spectrum of DPP-IVH7400
`. Unex(cid:173)
`pectedly, the resonance at 16.1 ppm remained present
`(Fig. 6A), indicating that this resonance does not arise
`from the active-site histidine. This result led us to conduct
`a more detailed investigation of the origin of the down(cid:173)
`field resonances.
`We collected the 1Il NMR spectrum of all four mutant
`proteins, DPP-IVH740Q, DPP-IV8630A, DPP-TVYS<l?F, and
`DPP-IVY5470
`Q, bolll in the absence (Fig. 6A) and pres(cid:173)
`ence of BMS-538305 (Fig. 6B) or saxagliptin {Fig. 6C).
`Examination of the NMR spectra for the mutant proteins
`indicates the resonance near 16.1 ppm is present in each
`mutant protein (Fig. 6A). Moreover, observation of this
`resonance is independent of the presence or absence of
`the ligands.
`The addition of BMS-538305 to the DPP-IV pruteim;
`resulted in the appearance of new resonances near 17.2
`and 14.7 ppm in all spectra, except that of H740Q (Fig.
`
`Table 1. Quaternary srm.cture of DPP-IV variants
`by SEC/MALS
`
`DI'P-IV
`val'iant
`
`Solution
`Peptide
`MW (kOa]" MW {kDa)~
`
`Quaternary slmcturc
`
`WT
`S630A
`H740Q
`Y547Q
`Y547P
`
`84.8
`84.8
`84.8
`84.&
`84.&
`
`186 :t 3
`191 ± 1
`190 :t 1
`189 :t 2
`190 ± 1
`
`Din:cr with 9% carbohyd.l11to
`Dimer ·;v:ilh 11 % carbohydrate
`Dircer with 11% carbohydr~te
`Din:er with 10% carbohydrate
`Din:cr with 1 I% carbohydrate
`
`•cal.cttlated based on sequc1tcc.
`b Fmm MALS anulysis; value.< a:ro ~hown :!: &tandatd error> from fits to
`individual da.t.a sets.
`
`Page 3 of 11
`
`

`
`Binding of saxagliptin to DPI'-IV variants
`
`and appears to be split into two resonances. (Note that
`and DPP,IV5530A-saxagliptin
`spectm for DPP-IV8530A
`complexes collected on two different protein prepru·ations
`months apart yielded the same reproducible spectrum.)
`Comparison of the 1H NMR spectra of DPP-IVS 530A in
`the presence and absence of saxagliptin indicates very
`Httle difference between the spectra; resooances are ob(cid:173)
`served near 16.1 and 14.4 ppm in each spectrum. Because
`the resonance at 14.4 ppm is present in the- spectra of both
`free and complexed DPP-lVSSJDA. it is unlikely to be
`related to the observed resonance at 14.1 ppm (or 14.7
`ppm for the BMS-538"05 complex). lnstead, tbe 1H NMR
`spectrum suggests a structural reOL·gani;;-;ation of re,~idues
`in the active site of DPP-IV5530
`A. It is clear from the
`spectrum of DPP-IV8530A-BMS-538305 and the corre(cid:173)
`sponding ITC data, however, tl1at this mutant protein
`remains competent for binding inhibitors. Finally, the
`NMR spectra for DPP-1Vy547 F and DPP-IVv547Q indicate
`both mutants bind saxagliptin and BMS-538305 similarly
`to the wild-type protein.
`
`Discussion
`
`The details of the catalytic and inhibition mechanisms of
`serine proteases remain of great interest due both to the
`historical importance of serine protcascs in studies on
`enzyme catalysis and to the continuing medical interest
`
`Table 2. Wild-type DPP-IV complex with saxagliptin data
`collEction and refinement statistics
`
`Data collection'
`
`Space gwup
`Unit ccoll pamme>ters
`a, b, c (A}
`Cl, p, 'I (0>
`ReS(>]ution rat1ge (A)
`Rr:uc::rgO!' (%)
`Average Ifcr(J)
`Complctcnos8 (%)
`Number of obse1vntions
`Number of unique reflection>
`Average redtmdancy
`
`Refjnernent
`Resolution runge (A)
`Number of reflectiom
`RwmJR&= (%)
`N1..unber of atoms
`Protein
`Ligand/carbohydrate
`Water
`RMSD.<
`Bond lengU1s (},)
`Bond angle» (0
`
`)
`
`65.3, 68.0, 423.8
`90.0, 90.0, 90.0
`50.0-2.4 (2.43-2.35)
`15.7 (58.7)
`10.8 (3.0)
`98.4 (39.3)
`604.55-:3 (2:41,466)
`7iU\23 {7 ,046)
`7.7 (6.4)
`
`50.0--2.4
`78,763
`21.4/25.3
`
`11,926
`16fi
`154
`
`0.006
`l.l
`
`a Data were collected from one cryo-motmt~d crystal. Nn v·sihk rlecay of
`the diffraction pattem w;,s observed over thB course of the experiment.
`
`www.proteinscience.org
`
`243
`
`trJgurc 3. fllustmtions of X-my co-cry.,tal structure of saxagliptin
`(magenta) complexed with DPP-fV. (A) Protein colored from blue (N
`terminus) to red (C terminus). Catalytic rc,idues (Ser630, His740, and
`A>p708) ure shown in gray. All alor.B are ~!10wn for Trp629, Set630, and
`Tyr631 (green but :for Ser630) to outline the oxy>mioJl hole. Orher active"
`site rc,idues are colored per the ribbons b·ace. Selected hydrogen-bonded
`water molecules are sbown as red spheres. (B) Same as above in stereo and
`zoomed in on BMS-477118. The electron density (2F" - Fe, lrr, cya<J) is
`consistent with covalent attachment to Ser630. Figure created with
`PyMOL (DeLano Scientific).
`
`6B). These data suggest BMS-53lD05 binds similarly to
`each of the mutant and wild-type proteins. The absence
`of a 14.7-ppm peak in all DPP-IVH74
`0Q spectra allows us
`to assign this resonance to the H740 imidazole proton.
`Similarly, the addition of saxagliptin to tl1e DPP-IV
`proteins resulted in the appearance of new resonances
`near 17.2 and 14.1 ppm in the 1HNMRspectrum. As seen
`for the complex with BMS-538305, the resonance near
`14.1 ppm is absent in the DPP-IVH7400-saxagllptin com(cid:173)
`plex, again consistent with this resonance arising from
`the H740 imidawle proton.
`A strong resonance i~ observed at 17.2 ppm in the 1H
`NMR spectra of all complexes except for DPP-IVssJuA_
`saxagliptin. In the DPP-IV 553{)A_saxagliptin complex, this
`resonance is significantly reduced in intensity, is shifted,
`
`Page 4 of 11
`
`

`
`Metzler et al.
`
`Table 3. Thermody1wmic parameters measured by iTC for
`inl1ibitor binding to DPP-.JV mriants
`
`DPP-lV
`varlam
`
`WT
`
`S630A
`
`H'740Q
`
`Y547Q
`
`Y547F
`
`Inhibitor
`
`Kd (nM)'
`
`11H.,)N..
`(kcal/mol)b
`
`!{'
`
`Saxl'!gliptin
`BMS-538305
`Suxagliptin
`BMS-538:105
`Saxagliptin
`BMS-538305
`Saxagliptin
`llMS-538305
`SHxaghptin
`BMS-538305
`
`:S5
`14 0:: 3
`NB'1
`31 1:: ll
`1,:00 ± 100
`6.4 :!:: 2.2
`:S5
`21 ±2:
`:s;j
`35 ± 3
`
`-17.6
`~8.3
`.Nl:!<l
`~8.5
`~11.3
`
`~12.0
`
`~20.4
`---12.5
`~14.0
`- 10.4
`
`3
`3
`]
`4
`2
`2
`2
`2
`2
`2
`
`• Kct valne~ given :!:: smn<iard deviation for e:<[Jeriments with three or more
`replicateH (N] or ± range for duplicate litratiuns.
`b En·ms for obHcn•ed liH values were ±0. 9 l:cal/mol.
`"Number of replicate litmlions.
`J_r..;<J binding observed for Htxugl.iptin to DPP-IV8"30".
`
`in their inhibition. Considerable excitement surrounding
`new pharmu;:;ologicaJ agents based on DPP-lV inhibition
`has emerged recently with several molecules undergoing
`clinical evaluation (Ahren et al. 2005). All currently
`known DPP-IV inhibitors occupy the Sl-S2 pockeL in
`the enzyme active site. Although interaction with the
`catalytic residue(s) is not a necessary prerequisite for
`inhibitor potency and eftkacy (Kim et al. 2005), an
`i.ntriguing feature of cyanopyrrolid.ine-containing inhib(cid:173)
`itors is their potential for transient covalent bonding with
`S630 (Oefner et al. 2003). Inhibition of DPP-IV by these
`compounds should logically involve the presence of the
`catalytic triad. To test this hypothe>is, we mutated three
`residues of human DPP-IV (S630, H740, and Y547;
`Fig. 3). These atnino acids were proposed to facilitate
`substrate proteolysis, and they project into the S 1/S2
`pockets.
`As expected, all mutants were catalytically ina;:;tive at
`cleavage of the stundard pseudosubstrate. To ensure that
`the observed lack of pmteolytic activity was the resuh of
`impaired catalytic potential rather than from misfolding
`or dimer disruption (Chien et al. 2004), we characterized
`the mutants using several biophysical methodg. Taken
`together with our CD and 1H NMR spectral data, the
`SEC/MALS and binding data strongly suggested that all
`mutant protejns bear similar folding and dimerization
`patterns
`to the DPP-IVwT (Table I). However, the
`inhibitor bimling abilities uf the prute.in~ fur two inhib(cid:173)
`itors (measuced by lTC) we1·e significantly altered in the
`mutilllts (Table 3). Whereas we did not observe any
`binding of saxagliptin to S630A either by ITC or TSE
`(Figs. 4A,B, 5), binding of BMS-538305 was only
`minimallv affected (K1 of 31 :!: 11 nM for DPP-IV 3630
`A
`vs. 14 ± .3 nM for DPP-lVwT; Table 3).
`
`244 Protein Science, vol. 17
`
`The only structural difference between thb saxagliplin
`and BMS-538305 is the pr~sence (saxagliptin) or:.:al;\"S,ej:lcy
`(B;JM;!),,~f3~.;1..0~) of a nitrile group (Fig. 1 ); Ji4'1!.¥~J.l!~
`q;;yst.ai . stru~turec. show!> ;this.. group covalently: ·bound . to
`~2~!i"l~:!~~~~1f~:~:~i~~!n~llh~ahs~:~1~:s;it'
`
`(Ki .•. 'T 50Q.pMJU1d'··.·,l(}nM fotsaxag;liptin·atid DM'Il-
`538~fJ5,,re:Jl~<:tivtJy;data nots'hOW11)~ By eliminating the
`
`nuclevphilic hydroxyl in DPP-l'ls630·", we envisioned
`that the covalent complex formation would be obviated
`while maintaining the rest of the affinity determinants. Vle
`thus expected that DPP-NsfiJCA would bind saxagliptin and
`BMS·538305 with similar affinities. UneApectedJy; our
`results show that binding of saxagliptin to this mutant
`was totally obliterated, whereas binding of BMS-538305
`was only mildly pertnrherl (Ti!hle 3). To better understand
`these findings, we modeled the DPP-IV8630
`A ~saxagliptin
`complex using the X-ray stru;:;ture of rhe DPP-IVwT_
`saxagliptin complex. Our modeling studies support that
`saxagliptin ;:;annot bind to the DPP-IV5630A. In the complex
`with DPP-IVWT, the nitrile carbon of saxagliptin adopts sp 2
`geometry as it fom1s a covalent bond with the serine
`hydroxyl, However, unable to form a covalent interaction
`in DPP-IV5630A, the nitrile carbon retains its sp character_ In
`Lhis hybriJizaLiun geomell:y, the nitrile of saxagliptin causes
`a large steric clash with the Alu630 side chain of DPP(cid:173)
`IV8630A, which is reflected in the Ki value.
`Study of the DPP-IVH740
`Q mutant confirmed the hy(cid:173)
`pothesis that enhanced nuclcophilicity of Scr630 by the
`basicity of H740 is an essential component of the mech(cid:173)
`anism by which saxagllptin covalently binds to DPP-IV.
`Despite tile presence of the Ser630 hydroxyl in DPP(cid:173)
`IVH140Q for potential covalent interaction, its ability to func(cid:173)
`tion as a nudeophile was compromised by a dysfunctional
`catalytic triad as a result of the H740Q mutation. The Kd for
`
`Time ~mini
`50
`10()
`150
`
`200
`
`Tirile{min)
`
`&G
`
`10~ 150
`
`200
`
`........ n.lilllt•
`
`. -- ........ ~ ..
`. . .
`
`...
`
`Figure 4. Binding of Raxagli[ltin to (A) DPP-1VW'l- and (B) DPP-TV 86~DA
`by 1TC"
`
`[saxagl iptin]I[S630A]
`
`Page 5 of 11
`
`

`
`Binding of saxagliptin to DPP-IV variants
`
`Table 4. Pairwise comparison of binding entlwlpies measured by JTC for inhibitor binding ro DPP IV wtriams
`
`DPP-IV Variant
`
`WT
`S630A
`ll740Q
`Y54iQ
`Y547P
`
`61iH0 b, (1renllmol)
`(SaK<tgliptiu minu&
`BMS-538305"1
`
`D.t:.Hoo, (Jccalimol)
`(Mutaut minus wild l)'pe
`fN ~;axagliptinb)
`
`6.tiH00, O:kcLI/mol)
`!Mutant minu• wild type
`for BMS-53~3051>)
`
`-9.6
`NIA
`-t0.7
`··-7.9
`-3.G
`
`X
`NIA
`+6.3
`-2.8
`+J-6
`
`X
`····0.2
`3.7
`-4.2
`·2.1
`
`Negative AMI in th.i s tnble reflects:
`• more favorable ob,erved binding enthalpy of sax•gliptin compared to BMS-538~05;
`I> or more favorable obsen-ed binding enthalpy of e<1mpound to mutant cornpar~ to \'IT.
`
`saxagHptin decreased by three orders of magnitude (Table
`3), whereas DPP-IVHHOQ bound the nitrile-deficient inhib(cid:173)
`itor RMS-5:1R305 with similar affinity as DPP-IVwr. As
`3
`with DPP-IV 5
`0A, lhis chaLLge in :mx11gliptin Kd is lilrgerthan
`ti
`expected; similartoDPP-IVs630
`A, theinabi1ityofth.eenzyrne
`to prepare the inhibitor for covalent interaction translates into
`a suboptimal geometric configumtion and stelic hindrance.
`Examination of the X-ray co-crystal structure of
`saxagliptin bound to DPP-IVwT reveals that a saxagliptin(cid:173)
`derived
`imidate nitrogen
`is within hydrogen-bonding
`distance to the side-chain hydroxyl of Y547 (Pig. 3).
`This ub~ervaliun i:s consistent wirh reports that Y547
`plays a major role in catalysis, presum<tbly by stabili:.dng
`the oxyanion hole througb hydrogen bonding (Aertgeerts
`et a!. 20011). Loss of proteolytic activity has been ob(cid:173)
`served upon mut.1.tion of Y547 (BjeJke et al. 2004) and
`in our present studies. To dete=ine whether Y547 plays a
`role in stabilizing the enzyme-inhibitor complex, the
`binding of saxagliptin was measmed for both DPP(cid:173)
`IVY547Q and DPP-IVY547F. Ideally an increased Kd of
`saxagliptin in DPP-IVY:;47
`.,: but not in DPP-IVY547
`Q,
`would indicate the importance of this residue in stabiliza(cid:173)
`tion of the enzyme-inhibitor complex. We found both
`mutant proteins bound saxagliptin very tightJy ( <5 nM),
`agreeing well with its potent Ki of 0.5 nM for wild-type
`protein (Kim et al. 2006). Thus, replacing this residue
`either by phenylalanine or glutamine did not alter the
`binding potency significantly. While this result initially
`seems to disagree with the activity data, we cannot rule out
`potential changes in Kd beyond the detection limit of our
`rrc assay. For example. even if a lO-fold reduction in the
`Kct could occur. it would be undistinguishable in our assay,
`and thus the role of Y547 cEUmot be assessed unambig(cid:173)
`uously from our data. Alternatively, mutation of tyrosine
`to phenylalanine could allow for an additional water mol(cid:173)
`ecule to fill the space. previously u~.:cupitd by the hydroxyl,
`thet·eby potentially compensating for ~ome uf the lu~L
`binding energy remlting fmm the dismpted hydrogen(cid:173)
`binding patterns (Bjelke et a1. 2004) with the inhibitor,
`bnt not the oxyanion of intermediate during the catalysis.
`
`The thermodynamic parameters measured by ITC for
`inhibitor binding to DPP-1V variants a(e shown in Tables
`3 and 4. Comparison of the differences in obr>erved heats
`of hinding (l!.Hobs: Table 3) is somewhat compromised by
`the moderate (4.88 kcal!mol) heat ofioni7:ation of HEPES
`(Goldberg et al. 2002). However, we expect that, when
`the differences in the observed heats of binding between
`saxagliptin and BMS-538305 are relatively large, such a
`comparison may still be informative. We also note that
`the heats are uncorrected for potential release or uptake of
`protons during binding. The observed enthalpy for bind(cid:173)
`ing DPP-Ivw-r is nearly ~ l 0 kcal/mol gr-eater for sax(cid:173)
`agliplin than for BMS-533305, reflecting the formation of
`the weak covalent bond between the cyano group of
`saxagliptin and Ser630 (see Fig. 3); this cyano group is
`absent in BMS-538305. When S630 wag mutated to
`alanine, binding of saxagliptin was abolished, and no
`binding enthalpy was detected. In contrast, BMS -538305
`bound DPP-IVWT and DPP-IVs630
`A with similar affinities
`and observed errthalpies, indicating tbat binding of BMS-
`538305 was unperturbed by the mutation. Binding of
`saxagliptin to DPP-IVH74
`DQ was reduced 220-fold relative
`to DPP-IVWT; this reduction in binding is retlected in the
`reduction of observed binding heat for saxagliptin for
`
`10
`s
`~--=="
`e -~
`G
`/
`""
`~ 4
`E
`~ 2.
`
`a
`
`2oo
`so
`150-
`100
`[Compound} (~tM)
`
`Figure 5. Effect of srumgliptin (squares) and BMS-538305 (circles) rm
`the melting temper.:tture of DPP-IVWl" (filled symbols) and DPP-IV5630
`A
`(open symbols).
`
`www.prl>teinsdence.org 245
`
`Page 6 of 11
`
`

`
`Metzler et a!.
`
`n
`A
`y~r_q __ ------- ~~"--~ --~-~-~~--~ --- __ )' j\ \..-~-------"----
`
`Y547.F
`
`/
`-------------- --~---~-- -------,)
`
`\
`
`\,.----··-----~---,..__..,,,,
`
`/\
`J"'
`r
`'\.
`i
`\
`'-'
`S630A
`""'"''•....,'~-Y;n,~-
`./1.-+~JV.,~"i-w"J..""~""tr'Y"'--c.;A.tV~.-......~~w,~k·JB;· .... ..._~w,~"uv,r'lr
`'\r..;j~o
`''-;.~¥/
`f,
`J
`I \
`H74DQ
`_!
`'.-f-.r''..,._r~,..,.., ..... ~.,
`~,....,...•,-"+~..j"~..._-"",~11·,.-~ri..A-,_.,-_,....._,,...r'tv--.. ~-.;wr='"'·rr<"',•-..,of\.,._i...__,J'•'\l
`
`24
`
`2,3
`
`22
`
`21
`
`20
`
`19
`ppm
`
`18
`
`17
`
`16
`
`15
`
`14
`
`B
`
`S630:A
`
`~
`(11
`A 'i\ I\
`_r \P,
`Y547Q
`----"--~-----~-~-~-- -~------ , ____ --------\ \, \_) 1\\~j .~'\ ---~---
`! I
`! ! \ ! 1
`Y547F
`-~ ~~ __ . .___.-•• w:.1f-l
`'.....-·'
`'-·--·--'
`Jr·<
`_,
`---~ ....... ,_.,.,_,..,.,_..-~--..-,._, .... ,~ ...... 1 \ \ .. _y /\, -~-,~) \~,~,........,..v-
`j\,
`j -\
`1 '• \
`
`1
`
`}\
`
`--·--
`
`·----·-···-----...... -
`
`! ~\ \ I
`H740Q
`, . ,_ ........ ~ . . . c _ -->••~···-·-· ........ -~.;....-':_;_,,_~•••••-• •"'-u.r-•.0~ I~ \~.t'"
`/\
`i\
`w~ ........ ---................. ~~-...~ .... ~-'L...-~"h~~r'-........---."1'""'11""'-r>c•'l.o· ......... --..,.-;.,/
`\""""'-..r-~"-.. ...... ..,_,_
`\.~ ......... ,_.-,
`19
`ppm
`
`\,
`
`'•,,•__.,_.,,~..--~~~.,.....__,
`
`24
`
`23
`
`22
`
`21
`
`20
`
`18
`
`17
`
`16
`
`15
`
`14
`
`24
`
`23
`
`22
`
`21
`
`20
`
`19
`ppm
`
`fS
`
`17
`
`16
`
`15
`
`14
`
`Figure 6, Expansion of t11e- downfield 1H NMR spectra of wild type and
`the tom DPP-IV mutant proteins in the absence o:f any inhibitor (A} :u:Jd in
`the presence of L2-molar excess BMS-538305 (B) or 'axagliptin (C).
`
`DPP-IVHNDQ ( ~ 11.3 kcal/mol) compared to DPP-TV""T
`( ~ 17_6 kcal/mol). Moreover, the ohservecl tlH of bincling
`of saxagliptin to DPP-IVH74
`rlQ was e-ssentially the same
`as that for BMS-538305, consistent with tile absence of
`17
`covalent bond formation. For DPP-1'/Y5
`Q, the high
`'
`favorable observed tJ.tliJ of binding ( ~7.9 kcalfrnol) for
`saxagliptin re-lative to BMS-538305 is also consis€ent
`with covalent bond formation. Overall, the comparisons
`show that large changes in observed binding enthalpy are
`
`246 Prot-ein Science, vol. 17
`
`consistent with fonnaliun of U1e covalent cornplex with
`~axagliptin observed in the DPP-IVw-r crystal structure.
`Mechanism-based serine protease
`inhibitors when
`cornplexed with the enzyme often feature short strong
`hydrogen bonds readily detectable by 1H l\'MR (Cassidy
`ct al. 1997). In our previous publication (Kim eta!. 2005),
`we reported our observation of several resonances which
`could be related to the protons of the catalytic triad
`featuring SSHB in both free, as well as saxagliptin(cid:173)
`cornplexed, DPP-IV In addition
`to large downfield
`chemical shifts, these resonances had reduced fraction(cid:173)
`ation factors (-0.5) with re~pect to the fractionation
`factors of protons engaged in "normal" hydrogen bonds.
`By compari8on to literature spectra, we postulated that
`the resonance at 16 ppm in the 1H NMR spectrum was
`the N 1H proton of H740_ Tn an attempt to confirm our
`Rss1gnments, we examined the- spectra of the mutant
`proteins. Surprisingly, the resonance at 16 ppm was
`present in H740Q mutant as well. We found the same
`resonance ]n all of our mutant proteins, free and com(cid:173)
`plexed with either inhibitor, suggesting that this signal
`does not ctrise from an active-site residue. A similar obser(cid:173)
`vation has been reported for serine prot.eases helonging
`to the class of proiyl oligopeptidases: prolyl oligopep(cid:173)
`tidase (PO) and oligopeptidase B of E.~cherichia coli (OpB)
`(Kahayaoglu et aJ. l997). Downfield chemical shifts (-17
`and -16 ppm) were observed for the wild type and an
`active-site H652A mutant of OpB. Nuclear Overhauser
`enhancement (:-.TOE) experiments provided strong evi(cid:173)
`dence that these two resonances belonged to two different
`noncatalytic histidines_ Because DPP-IV contains 19
`
`Table 5. Summary of the dmtmjiefd NMR resonance shifts
`for WT and nmtant DPP-IV proteins
`
`Condition
`
`6 {rpm)
`
`S (ppml
`
`8 (ppm)
`
`WT
`No inhibitor
`S='-'gliptin
`BMS--'i'lR:IOS
`S630A
`No inltibitor
`Saxz.gliplin
`BMS-538305
`H740Q
`No inhibitor
`s~x"gliptin
`BMS-538305
`Y547Q
`No inhibitor
`SJD:agliptin
`BMS-5JKHJ5
`Y547F
`No jnhibitor
`Suagliplin
`IJMS-5JSJ05
`
`17.29
`17.1\J
`
`17.15
`
`17.23
`17.21
`
`17.14
`t7.ml
`
`17.17
`17.12
`
`16.10
`16.08
`1607
`
`1607
`16.05
`16.07
`
`16.04
`16.0;J
`16.07
`
`16.10
`16.10
`16.10
`
`16 11
`16.11
`16.08
`
`14.02
`14.66
`
`14.42
`14.49
`14.24
`
`13.96
`14.86
`
`14.02
`14.47
`
`Page 7 of 11
`
`

`
`histidine residues, further mutagenesis studies to assign this
`resonance would be impractical to execute. Examination of
`the structure uf DPP-IV reveals no His-Asp (or Glu) pairs
`that are sufficiently close to form a SSHB. A
`likely
`candidate for the resonance at 16.1 ppm is a hydrogen
`bond that is observed in
`the active site between the
`carboxyl groups of E206 and D663. ,Not only is
`the
`hydrogen-bonding distance short (2.5 A), the