Crystal structure of human dipeptidyl peptidase
`IV in complex with a decapeptide reveals
`details on substrate specificity and tetrahedral
`intermediate formation
`
`KATHLEEN AERTGEERTS, SHENG YE, MIKE G. TENNANT, MICHELLE L. KRAUS,
`JOE ROGERS, BI-CHING SANG, ROBERT J. SKENE, DAVID R. WEBB,1 AND
`G. SRIDHAR PRASAD
`Syrrx Inc., San Diego, California 92121, USA
`(RECEIVED September 26, 2003; FINAL REVISION October 25, 2003; ACCEPTED October 27, 2003)
`
`Abstract
`Dipeptidyl peptidase IV (DPPIV) is a member of the prolyl oligopeptidase family of serine proteases.
`DPPIV removes dipeptides from the N terminus of substrates, including many chemokines, neuropeptides,
`and peptide hormones. Specific inhibition of DPPIV is being investigated in human trials for the treatment
`of type II diabetes. To understand better the molecular determinants that underlie enzyme catalysis and
`substrate specificity, we report the crystal structures of DPPIV in the free form and in complex with the first
`10 residues of the physiological substrate, Neuropeptide Y (residues 1–10; tNPY). The crystal structure of
`the free form of the enzyme reveals two potential channels through which substrates could access the active
`site—a so-called propeller opening, and side opening. The crystal structure of the DPPIV/tNPY complex
`suggests that bioactive peptides utilize the side opening unique to DPPIV to access the active site. Other
`structural features in the active site such as the presence of a Glu motif, a well-defined hydrophobic S1
`subsite, and minimal long-range interactions explain the substrate recognition and binding properties of
`DPPIV. Moreover, in the DPPIV/tNPY complex structure, the peptide is not cleaved but trapped in a
`tetrahedral intermediate that occurs during catalysis. Conformational changes of S630 and H740 between
`DPPIV in its free form and in complex with tNPY were observed and contribute to the stabilization of the
`tetrahedral intermediate. Our results facilitate the design of potent, selective small molecule inhibitors of
`DPPIV that may yield compounds for the development of novel drugs to treat type II diabetes.
`Keywords: Dipeptidyl peptidase IV; DPPIV; CD26; crystal structure; adenosine deaminase binding pro-
`tein; serine protease; tetrahedral intermediate
`
`The type II transmembrane serine protease, DPPIV, also
`known as CD26, or adenosine deaminase binding protein
`(ADAbp), is highly expressed on endothelial cells, differ-
`entiated epithelial cells and lymphocytes (Hegen et al. 1997;
`
`Reprint requests to: G. Sridhar Prasad, Syrrx Inc., 10410 Science Center
`Drive, San Diego, CA 92121, USA; e-mail: Sridhar.Prasad@syrrx.com;
`fax: (858) 550-0526.
`1Present address: Celgene Corp., San Diego, CA 92121, USA.
`Abbreviations: DPPIV, dipeptidyl peptidase IV; NPY, Neuropeptide Y;
`tNPY, N-terminal decapeptide (residues 1–10) of Neuropeptide Y.
`Article published online ahead of print. Article and publication date are at
`http://www.proteinscience.org/cgi/doi/10.1110/ps.03460604.
`
`De Meester et al. 1999; Kahne et al. 1999). A soluble form
`of the enzyme was also found in plasma (Iwaki-Egawa et al.
`1998; Durinx et al. 2000). As a dipeptidyl peptidase, DPPIV
`plays a major role in the regulation of physiological pro-
`cesses including immune, inflammatory, CNS, and endo-
`crine functions. For example, DPPIV plays an important
`role in maintaining glucose homeostasis (Deacon et al.
`1998; Balkan et al. 1999; Pauly et al. 1999; Drucker 2003).
`These studies reveal that DPPIV helps regulate plasma glu-
`cose levels by controlling the activity of the incretins glu-
`cagon-like peptide 1 (GLP-1) and glucose-dependent insu-
`linotropic polypeptide (GIP). Inhibition of DPPIV in wild-
`
`412
`
`Protein Science (2004), 13:412–421. Published by Cold Spring Harbor Laboratory Press. Copyright © 2004 The Protein Society
`
`Page 1 of 10
`
`AstraZeneca Exhibit 2006
` Mylan v. AstraZeneca
` IPR2015-01340
`
`

`
`type and diabetic mice leads to increased levels of
`unprocessed GLP-1 and GIP in the circulation, enhanced
`insulin secretion, and improved glucose tolerance. Selective
`inhibitors of DPPIV improve plasma glucose levels in hu-
`man type II diabetics (Ahren et al. 2002). Independent of its
`dipeptidyl peptidase activity, DPPIV also binds adenosine
`deaminase (ADA; Morrison et al. 1993). This interaction
`has been shown to modulate immune function (Franco et al.
`1998; Morimoto and Schlossman 1998).
`The catalytic triad of DPPIV is composed of residues
`S630, D708, and H740, which are located within the last
`140 residues of the C-terminal region (Ogata et al. 1992).
`The enzyme specifically removes dipeptides from the N
`terminus of peptide substrates that contain on average 30
`residues and have a Pro or Ala in the penultimate position.
`In addition, a slow release has been observed for dipeptides
`composed of X-Ser or X-Gly (Bongers et al. 1992; De
`Meester et al. 1999; Hinke et al. 2000; Lambeir et al. 2002).
`Physiological peptides recognized by DPPIV that contain
`this specificity profile at their cleavage site include neuro-
`peptides like neuropeptide Y, circulating peptide hormones
`like peptide YY, glucagon-like peptides (GLP)-1 and -2,
`gastric inhibitory peptides, as well as paracrine chemokines
`like RANTES (De Meester et al. 1999; Mentlein 1999).
`Catalytic efficiencies for the cleavage by DPPIV of differ-
`ent physiological substrates were determined by mass spec-
`trometry-based protease assays (Lambeir et al. 2001a,b; Zhu
`et al. 2003). These studies demonstrated that residues sur-
`rounding the scissile bond mainly determine the substrate
`selectivity of DPPIV. However, there is supporting kinetic
`evidence that nonconserved residues along the entire length
`of the peptide are involved in long-range interactions that
`play a role in substrate binding and catalysis (Lambeir et al.
`2001a,b, 2002; Zhu et al. 2003).
`Crystal structures of DPPIV in complex with several
`small molecule inhibitors and substrates have been pub-
`lished (Engel et al. 2003; Hiramatsu et al. 2003; Oefner et
`al. 2003; Rasmussen et al. 2003; Thoma et al. 2003). How-
`ever, the exact molecular determinants that contribute to the
`substrate specificity of DPPIV and how substrate peptides
`access the active site remains unclear. To help understand
`the function of DPPIV, we crystallized and solved the X-ray
`crystal structure of the enzyme in both its free form and in
`the presence of the first 10 residues of Neuropeptide Y.
`Neuropeptide Y is a physiological substrate of DPPIV
`widely distributed in the nervous system (Mentlein 1999),
`and involved in cardiovascular homeostasis and the regula-
`tion of insulin release (Ahren 2000; Ghersi et al. 2001). The
`catalytic efficiciency for N-terminal dipeptide cleavage of
`Neuropeptide Y by DPPIV is 3.0 × 106 M−1sec−1 (Mentlein
`et al. 1993). The DPPIV/tNPY structure provides direct evi-
`dence that the decapeptide accesses the active site through a
`side opening, unique to DPPIV, and not through the ␤-pro-
`peller opening. The latter mechanism was suggested for the
`
`DPPIV/decapeptide crystal structure
`
`closely related enzyme prolyl oligopeptidase (POP; Fulop et
`al. 1998, 2000). Our work also provides a detailed under-
`standing of the molecular determinants that contribute to the
`substrate specificity of DPPIV. Moreover, in the DPPIV/
`tNPY crystal structure the peptide was trapped in a tetrahe-
`dral intermediate, and gives new insight into DPPIV en-
`zyme catalysis. Earlier studies provided evidence for the
`existence of a tetrahedral intermediate, which was based on
`structural studies on complexes with small molecule “tran-
`sition-state analog” inhibitors; ab initio quantum mechanics
`(QM), molecular mechanics (MM), and molecular dynam-
`ics (MD) simulations or combined time-resolved/pH jump
`crystallographic studies (Wilmouth et al. 2001; Topf et al.
`2002a,b). Until now, no direct structural evidence of a
`single discrete intermediate formed between a physiological
`substrate and a serine protease has been published.
`
`Results
`
`Structure and domain organization of DPPIV
`
`The crystal structure of the extracellular domain (residues
`39–766) of DPPIV was solved to a resolution of 2.1 Å. The
`structure consists of two domains: an N-terminal 8-bladed
`␤-propeller domain (residues 61–495) and a C-terminal ␣/␤
`hydrolase domain (Nardini and Dijkstra 1999; residues 39–
`55 and 497–766; Fig. 1). The propeller domain packs
`against the hydrolase domain, and the catalytic triad (S630,
`H740, and D708) is at the interface of the two domains. In
`vitro catalytic activity of recombinant DPPIV was mea-
`sured. The catalytic efficiency for the cleavage of the fluo-
`rogenic substrate H-Ala-Pro-7-amido-4-trifluromethylcou-
`marin (Ala-Pro-AFC) by DPPIV is 5.2 × 106 M−1sec−1.
`The asymmetric unit is composed of two homodimers,
`the monomers of which are related by a twofold dyad axis
`(Fig. 1). This dimeric structure correlates with the biologi-
`cally active form of DPPIV (Bednarczyk et al. 1991; De
`Meester et al. 1992; Gorrell et al. 2001; Ajami et al. 2003).
`The overall structures of the monomers are similar with
`root-mean-square deviations (RMSDs) from 0.64 Å to 0.98
`Å for all heavy atoms and from 0.28 Å to 0.56 Å for the C␣
`atoms. The dimer interface buries a total of 2188 Å2 acces-
`sible surface area per monomer and comprises: (1) the last
`␤-strand (␤8) of the peptidase central ␤-sheet, (2) the last
`two ␣-helices (␣G and ␣H), (3) the loop between ␤6 and
`␣E, and (4) the antiparallel ␤-strand subdomain (␤1* and
`␤2*; Fig. 1). ␤8 mainly contains hydrophobic residues
`forming hydrophobic interactions at the center of the dimer
`interface. ␣-Helix H forms hydrogen bonds with the loop
`between ␣G and ␤8 in the other monomer. The antiparallel
`␤-strand arm formed by ␤1* and ␤2* interacts with its
`related arm, ␣G, and the loop between ␤6 and ␣E in the
`other monomer.
`
`www.proteinscience.org
`
`413
`
`Page 2 of 10
`
`

`
`Aertgeerts et al.
`
`Figure 1. Ribbon diagram showing overall structure of the DPPIV homodimer, viewed perpendicular to the twofold dyad axis.
`Secondary structural elements that are involved in dimer formation are represented in red and in blue. The active site residues are shown
`as ball-and-stick representations. The ␣-helix comprising residues E205 and E206 is indicated in gold. The figure was made using the
`programs MOLSCRIPT (Kraulis 1991) and Raster3D (Merrit and Bacon 1997).
`
`The primary structure of DPPIV contains nine potential
`N-linked glycosylation sites: N85, N92, N150, N219,
`N229, N281, N321, N520, and N685. The first N-acetyl-
`glucosamine (NAG) sugar moiety is observed with clear
`electron density in all the nine predicted sites. Detailed
`structural and biochemical analysis revealed that the glyco-
`sylation of DPPIV is not important for catalytic activity,
`homodimer formation and ADA binding (Aertgeerts et al.
`2004).
`
`Substrate access to the active site
`
`Bioactive peptides recognized by DPPIV could theoretically
`access the active site in two possible ways: through an open-
`ing in the propeller domain or via a side opening formed at
`the interface of the ␤-propeller and hydrolase domains (Fig.
`2). The propeller opening is formed by the ␤-propeller do-
`main, which is composed of an unusual eightfold repeat of
`blades. Each blade is composed of a four-strand antiparallel
`␤-sheet. The ␤-propeller domain defines a funnel shaped,
`solvent-filled tunnel that extends from the ␤-propeller’s
`lower face to the active site. The lower face of the funnel,
`distal to the hydrolase domain, has a diameter of approxi-
`mately 15 Å. The closing of the circle between the first and
`the last blade of propeller proteins has been termed “Vel-
`cro” (Neer and Smith 1996), and unlike most of the other
`known propeller proteins, the “Velcro” is not closed be-
`tween the first and the last blades in the DPPIV structure.
`This is similar to the arrangement observed in POP (Fulop
`
`et al. 1998). The propeller opening connects to a larger side
`opening (∼21 Å) formed at the interface of the ␤-propeller
`domain and the hydrolase domain. This oval-shaped cavity
`creates a second entrance to the active site (Fig. 2). To
`understand which entrance/exit pathway substrate peptides
`use to access the active site of DPPIV, we cocrystallized the
`enzyme with YPSKPDNPGE (tNPY), corresponding to the
`first 10 residues of the physiological substrate, Neuropep-
`tide Y. (DPPIV used in the experiment contains a single
`mutation S716A. The catalytic efficiency of this mutant for
`cleavage of Ala-Pro-AFC is 41 × 106 M−1 sec−1, which is
`similar to the value measured for wild-type DPPIV. We also
`obtained crystals of wild-type DPPIV in complex with
`tNPY, but
`the crystals using DPPIV-S716A/tNPY dif-
`fracted to a higher resolution.) Clear continuous electron
`density was observed for the first six of the 10 residues of
`the peptide. Four of the six residues make molecular inter-
`actions (see below), with residues lining the side opening of
`DPPIV. No clear electron density was observed for the last
`four residues because they are solvent exposed and there-
`fore not ordered in the structure. In conclusion, the crystal
`structure of the DPPIV/tNPY complex suggests that physi-
`ological substrates may employ the side opening of DPPIV
`to access the active site.
`
`Substrate specificity
`
`DPPIV cleaves the amide bond after the penultimate N-
`terminal residue (P1, according to Berger and Schechter
`
`414
`
`Protein Science, vol. 13
`
`Page 3 of 10
`
`

`
`DPPIV/decapeptide crystal structure
`
`Figure 2. (A) Surface representation of the ␤-propeller domain only, showing the propeller opening to the active site. The view was
`taken from the interface with the ␣/␤-hydrolase domain and down the pseudo-eightfold axis. The four-strand antiparallel ␤-sheets of
`the eight blades are indicated (␤1–␤8). (B) Surface representation of whole DPPIV molecule, showing the side opening to the active
`site. Residues of DPPIV that make direct molecular interactions with tNPY are colored in both panels. Hydrophobic negatively charged
`and positively charged residues are shown in green, in red, and in blue, respectively. The figures were made with the program MOE
`(MOE, Chemical Computing Group).
`
`1970) of physiological peptides. Oligopeptide N termini are
`recognized by the negatively charged active site residues
`E205 and E206, and are anchored by hydrogen bond for-
`mation with the side chains of the two glutamates (Fig. 3).
`E205 and E206 reside on a short ␣-helix insertion (residues
`200–206) protruding from the ␤-propeller domain and
`pointing toward the active site. The two glutamic acid resi-
`dues are conformationally restrained by salt bridge forma-
`
`tion and hydrogen bond interactions with residues R125,
`Y662, D663, and N710.
`The best catalytic efficiencies for dipeptide cleavage by
`DPPIV was measured for peptides with a Pro or Ala at P1
`(Lambeir et al. 2003; Leiting 2003). The well-defined hy-
`drophobic S1 pocket lined by residues V656, Y631, Y662,
`W659, Y666, and V711 determines this specificity (Figs. 3,
`4). The S2 pocket is hydrophobic and determined by the
`
`Figure 3. Stereo drawing of first six residues of Neuropeptide Y (magenta) and the underlying active site residues of DPPIV (pink)
`that make direct molecular interactions with the peptide. The peptide and selected DPPIV residues are shown as ball-and-stick
`representations. The peptide is not cleaved and trapped in a tetrahedral intermediate by which the carbonyl carbon is covalently linked
`to the active site S630. Hydrogen bonds are indicated as green dashed lines. The figure is made using the programs MOLSCRIPT
`(Kraulis 1991) and Raster3D (Merritt and Bacon 1997).
`
`www.proteinscience.org
`
`415
`
`Page 4 of 10
`
`

`
`Aertgeerts et al.
`
`side chains of residues R125, F357, Y547, P550, Y631, and
`Y666. In this structure we observe two water molecules
`occupying this S2 pocket, and the P2 tyrosine is interacting
`with these waters and partially occupying the S2 site. The
`S1⬘ pocket is flat and not well defined, and the only inter-
`actions observed between the P1⬘ serine and the S1⬘ resi-
`dues are nonspecific van der Waals interactions. The car-
`bonyl oxygen of the P1⬘ serine makes a hydrogen bond with
`R125. The side chain of the P2⬘ lysine packs against the face
`of W629, completely occluding the tryptophane from sol-
`vent, and forms a hydrogen bond with the hydroxyl oxygen
`of Y752 (Fig. 3). Beyond P2⬘, no specific interactions are
`observed between the peptide and the underlying DPPIV
`residues.
`Ser 630 is located on “the nucleophilic elbow” formed
`by residues Gly-Trp-Ser630-Tyr-Gly. This sequence is
`essential for DPPIV activity (Ogata et al. 1992) and con-
`served in the ␣/␤ hydrolase family (Gly-X-Ser-X-Gly). The
`orientation of S630 is maintained by hydrogen bonds
`between the carbonyl oxygen of S630 and the amide of
`Y634, and the amide of S630 and the carbonyl oxygen of
`V653.
`
`Tetrahedral intermediate
`
`In the crystal structure of the DPPIV/tNPY complex, we
`observed that the peptide was not cleaved, but trapped in a
`tetrahedral intermediate (Fig. 5). As expected for tetrahedral
`intermediate formation, the O␥ atom of S630 was found in
`close contact (between 1.6–1.8 Å) with the carbonyl carbon
`
`of the scissile bond. The electron density map contoured at
`3␴ was continuous between the two atoms, and the electron
`density map contoured at 1␴ was discontinuous between the
`O␥ atom of S630 and N␦2 of H740. Comparison of this
`structure with a 2.1 Å structure of the free form of DPPIV
`shows that the hydroxyl group of the active site serine
`(S630) has moved significantly to optimally interact with
`the carbonyl carbon of the scissile bond (Fig. 5B). In addi-
`tion, the imidazole ring of H740 rotates by about 15° along
`the ␹2 torsion (Fig. 5B). The hydrogen bond distance be-
`tween S630 and H740 in the native enzyme is 2.8 Å,
`whereas this distance changes to 3.2 Å in the transition state
`structure. The oxyanion is stabilized by hydrogen bond for-
`mation with the main chain amide of Y631 (∼3.1 Å) and
`with the hydroxyl group of Y547 (∼2.2 Å; Fig. 5B). For-
`mation of such a short, very strong, low-barrier hydrogen
`bond is expected in transition states, which stabilizes inter-
`mediates in enzymatic reactions and lowers the energy of
`transition states.
`To verify our conclusion that the decapeptide was trapped
`in a tetrahedral intermediate, the peptide was omitted from
`the model, the active site S630 was changed to an alanine,
`and the structure was again refined using REFMAC (CCP
`1994). The resultant electron density maps showed unam-
`biguous density for the decapeptide, O␥ atom of S630 and
`the continuous electron density between the O␥ atom of
`S630 and the carbonyl carbon of the scissile bond. The
`asymmetric unit is composed of four independent DPPIV/
`tNPY complexes, and in all four structures, the peptide is
`trapped in the tetrahedral intermediate.
`
`Figure 4. Molecular surface representations showing the interaction of tNPY with DPPIV. Residues of the peptide are shown in ball-and-stick represen-
`tations and DPPIV is shown as a solid surface. (A) Colors represent positive and negative electrostatic potential from blue (electropositive; white, neutral)
`to red (electronegative). (B) Colors represent hydrophobicity (green, polar; yellow, hydrophobic; white, exposed). The figures were made with the program
`MOE (MOE, Chemical Computing Group).
`
`416
`
`Protein Science, vol. 13
`
`Page 5 of 10
`
`

`
`DPPIV/decapeptide crystal structure
`
`Figure 5. Schematic representations showing tetrahedral intermediate formation. (A) 2Fo−Fc electron density map contoured at 1␴ of
`the first six residues of tNPY and of the active site serine (S630). The peptide and the side chain of S630 are shown as ball-and-stick
`representations, and part of the DPPIV molecule is represented as a pink ribbon diagram. (B) Schematic representation showing the
`difference in conformation and hydrogen bond formation of active site residues S630 and H740 between the free form (green) of the
`enzyme and the tetrahedral intermediate (pink). The first three residues of the peptide are shown in gold as ball-and-stick represen-
`tations. Hydrogen bonds are represented as green dotted lines, and measured distances are indicated in angstroms. Part of the DPPIV
`molecule around the active site is represented in pink as a ribbon diagram. The figures were made using the programs MOLSCRIPT
`(Kraulis 1991) and Raster3D (Merrit and Bacon 1997) and XtalView (McRee 1999).
`
`Discussion
`
`DPPIV is an ectopeptidase implicated in the degradation of
`various peptides and hormones including glucagon family
`peptides, neuropeptides, and chemokines (Mentlein 1999).
`The enzyme selectively cleaves dipeptides at the N terminus
`when preferably Pro or Ala is present in the penultimate
`position. Several structural features that were observed from
`the crystal structure of the free form of the enzyme and the
`DPPIV/tNPY complex explain the substrate specificity of
`DPPIV and its mode of interaction with substrate peptides.
`Two channels give access to the active site: a propeller
`opening, and a side opening. Until now, no direct evidence
`has been described on which entrance/exit pathway is uti-
`lized by bioactive peptide substrates. No side opening was
`observed in the structure of POP, and a gating filter mecha-
`nism utilizing the propeller opening was proposed to ex-
`plain its narrow substrate specificity (Fulop et al. 1998).
`Because the first and the last blade of the ␤-propeller do-
`main are not closed, they might partially separate to facili-
`tate substrate access of peptides to the active site. This
`hypothesis was subsequently proven by mutagenesis and
`kinetic data on POP (Fulop et al. 2000). DPPIV and POP
`have the same structural organization of the ␤-propeller
`domain, and DPPIV substrates could therefore theoretically
`utilize the central tunnel formed by this domain to access
`the active site. However, in contrast to the structure of POP,
`a second and larger side opening that facilitates access to the
`active site was observed in the structure of DPPIV. This
`opening is characterized by an oval shaped groove and is
`sterically the most favorable way to enter to and exit the
`
`active site. The DPPIV/tNPY structure provides direct evi-
`dence for peptides to employ this opening to access the
`active site of the enzyme. Six of the 10 residues of the
`substrate were ordered in the crystal structure, and the first
`four residues interact with underlying amino acids present
`in the side opening of DPPIV. The last four residues of the
`decapeptide are solvent exposed, and this observation pro-
`vides further evidence that the propeller opening was not
`utilized, because based on the length and diameter of the
`tunnel, extensive interactions for all 10 residues would have
`been predicted.
`The DPPIV/tNPY structure gives a detailed understand-
`ing of the molecular mechanisms that determine the inter-
`action of the peptide with the residues present in the active
`site of DPPIV. The presence of two glutamates (E205 and
`E206) at the end of an ␣-helical segment that protrudes from
`the ␤-propeller domain into the active site of the enzyme
`determines the aminopeptidase function of DPPIV. Both
`residues are essential for enzyme activity (Abbott et al.
`1999). The Glu motif is conserved in the DPPIV-like gene
`family, and was not found in the structure of POP. The Glu
`motif functions as a recognition site for the N terminus of
`peptide substrates, and anchors the substrate so that only
`dipeptides can be cleaved off. The hydrophobic S1 groove
`is shaped to optimally accommodate and interact with a Pro
`or Ala residue, and explains the strong preference of DPPIV
`for peptides with these amino acids in the penultimate po-
`sition. The S2 subsite preferentially recognizes large hydro-
`phobic and aromatic side chains. With respect to the shape
`and chemical composition of the S1⬘ subsite, we observed
`that most side chains can be modeled into this pocket; how-
`
`www.proteinscience.org
`
`417
`
`Page 6 of 10
`
`

`
`Aertgeerts et al.
`
`ever, charged residues are not preferred due to possible
`unfavorable electrostatic interactions. Because of the ␲
`electron system of W629, the S2⬘ groove is well suited to
`accept large aliphatic side chains. We observed that residues
`P2–P2⬘ of the decapeptide are mainly recognized by DPPIV,
`and substrate recognition does not extend beyond P2⬘. This
`is in agreement with kinetic data measured in vitro on
`chemokines that are substrates for DPPIV (Lambeir et al.
`2001a). However, the adenylyl cyclase-activating peptides,
`pituitary adenylate cyclase-activating polypeptide (PACAP)-
`27, and PACAP38 contain the same 27 amino acids at the N
`terminus, but PACAP38 was processed 15-fold more effi-
`ciently than PACAP27 (Lambeir et al. 2001b; Zhu et al. 2003).
`The only difference between the two peptides is a basic C-
`terminal extension present in PACAP38. These data suggest
`that secondary sites on DPPIV remote from the active site are
`important for substrate binding and catalysis.
`In the crystal structure of the DPPIV/tNPY complex, the
`peptide is trapped in a tetrahedral intermediate that occurs
`during enzyme catalysis. From the electron density maps,
`we observed electron transfer between O␥ of S630 and the
`carbonyl carbon of the scissile bond; the measured distance
`was 1.75 ± 0.15 Å. The theoretical expected distance of this
`bond during the tetrahedral intermediate is ∼1.4 Å (Topf et
`al. 2002a). This suggests that the intermediate is forming,
`but has not yet proceeded to completion. The oxyanion is
`stabilized by hydrogen bond formation with the main chain
`amide group of Y631 and the hydroxyl group of Y547.
`While we were writing our manuscript, Thoma et al. pub-
`lished the crystal structure of DPPIV in complex with di-
`protin A (Ile-Pro-Ile). The tripeptide was also trapped in a
`tetrahedral intermediate and its conformation is comparable
`to the first three residues of tNPY in our structure (Thoma
`et al. 2003).
`The crystal structures of the free form of DPPIV and the
`DPPIV/tNPY complex suggest that physiological peptide
`substrates utilize the side opening, unique to DPPIV, to
`access the active site. The structures also provide a clear
`insight into the different molecular determinants that are
`responsible for the substrate specificity of DPPIV. Further-
`more, in the DPPIV/tNPY complex, the decapeptide was
`trapped in a tetrahedral intermediate and gives thereby di-
`rect structural evidence for its existence and provides a de-
`tailed understanding in the molecular mechanisms that are
`utilized to stabilize the intermediate. The availability of the
`DPPIV/tNPY structure will assist in the rational design of
`highly specific and potent inhibitors that can be used to
`better understand the role of DPPIV, and as potential treat-
`ments for diabetes and related disorders.
`
`Materials and methods
`Protein expression and purification
`The cDNA encoding human DPPIV was isolated by PCR from
`spleen cDNA (Clontech) and the extracellular domain (residues
`
`418
`
`Protein Science, vol. 13
`
`39–766) was cloned into the SmaI site of a modified pFastBacHTb
`vector (Invitrogen). The final construct contains a baculovirus
`gp67 signal peptide followed by a His6 tag fused to the coding
`sequence corresponding to residues 39–766 of DPPIV. Recombi-
`nant baculovirus was generated by transposition using the Bac-to-
`Bac system (Gibco-BRL). Large-scale production of recombinant
`protein was performed by infection of Trichoplusia ni (Hi5) insect
`cells (Gibco-BRL) for 48 h in 5-L Wave Bioreactors (Wave Bio-
`tech). The secreted glycosylated recombinant protein was isolated
`from the cell culture medium by diafiltration using crossflow ul-
`trafiltration followed by passage over a nickel chelate resin (Bind-
`ing buffer: 25 mM Tris [pH 7.9], 400 mM NaCl). The column was
`washed overnight (0.2 mL/min) with 50 mM K2HPO4 (pH 7.9);
`400 mM NaCl; 20 mM Imidazole-HCl, and 0.25 mM TCEP fol-
`lowed by five column volumes (1 mL/min) of 50 mM Tris HCl
`(pH 7.9), 400 mM NaCl and 0.25 mM TCEP. Protein bound was
`eluted with four column volumes of 50 mM Tris-HCl (pH 7.9),
`400 mM NaCl, 200 mM imidazole-HCl, and 0.25 mM TCEP. To
`remove oligomeric forms, the sample was further purified over a
`size-exclusion column (BioSep SEC S3000, 300 × 21.2 mm, Phe-
`nomenex) equilibrated with 25 mM Tris (pH 7.6); 150 mM NaCl;
`0.25 mM TCEP; and 1 mM EDTA. A yield of 16 mg/L culture
`(3 × 106 cells/mL) was obtained. DPPIV substituted with Se-Met
`was produced as described above with the following modifica-
`tions: 16 h following infection the Hi5-infected insect cells were
`centrifuged at 500 × g for 15 min and the pelleted cells resus-
`pended in an equal volume of protein-free methionine-free ESF-
`921 medium (Expression Systems). Following 4 h offurther
`growth Se-Met (Acros) was added to a final concentration of 50
`mg/L. The medium from the infected culture was harvested 64 h
`following viral infection. The Se-Met-substituted protein was pu-
`rified as described above. The incorporation of Se-Met was esti-
`mated to be in the region of 30%–40%. The complex of DPPIV
`with a decapeptide (YPSKPDNPGE; tNPY) (custom synthesized
`by Biopeptide Co., LLC) that corresponds to the first 10 amino
`acids of Neuropeptide Y was formed at pH 7.6 (25 mM Tris, 250
`mM NaCl, 0.25 mM TCEP, 1 mM EDTA) by incubation for 30
`min at room temperature using a 10-fold molar excess of tNPY
`(final concentration 1 mM) over DPPIV (final concentration
`0.1 mM).
`
`Determination of catalytic activity
`
`The determination of the catalytic constants of DPPIV and DPPIV-
`S716A for dipeptide cleavage was performed using a fluorescent
`assay. Enzyme (0.1 nM) was mixed with 0.4–400 ␮M of Ala-Pro-
`AFC (Bachem) in 20 mM Tris (pH 7.4) 20 mM KCl, 0.1 mg/mL
`BSA, and 1% DMSO in a 96-well half-area plate and monitored
`kinetically at Ex400 nm and Em505 nm using Molecular Devices
`SpectraMax Gemini. Assays were performed in duplicate for each
`sample. MDL data analysis toolbox was used for analysis of Mi-
`chaelis-Menten kinetics.
`
`Crystallization and data collection
`
`Wild-type DPPIV, Se-Met DPPIV, and the DPPIV/tNPY complex
`were crystallized at 4°C using Syrrx’s automated Nanovolume
`CrystallisationTM technology (Hosfield et al. 2003). In all cases,
`the reservoir solution was 20% PEG MME 2000, 100 mM Bicine
`(pH 8.0–8.5). Thick plate-shaped crystals appeared in about 5
`days, which grew to about 0.5 mm in longest dimension and vary-
`ing width and thickness. For X-ray data collection, crystals were
`flash-frozen at 100 K using 25% v/v ethylene glycol as a cryo-
`
`Page 7 of 10
`
`

`
`DPPIV/decapeptide crystal structure
`
`Table 1. Heavy atom and Se-Met data statistics for human DPPIV
`
`Unit cell parameters
`a (Å)
`b (Å)
`c (Å)
`␤ (°)
`Wavelength (Å)
`Resolution (Å)
`Total observations
`Unique reflections
`Completeness (%)
`Rsymm (I)
`l/␴ (I)
`Number of sites
`Phasing power (ano)
`Phasing power (iso)
`Figure of merit
`NCS correlation
`
`Native
`
`121.8
`124.1
`144.5
`114.7
`1.0
`2.1
`996,059
`218,087
`96.4 (95.0)
`0.062 (0.524)
`19.8 (2.3)
`
`Se-Met
`
`121.9
`123.0
`145.0
`114.9
`0.97913
`3.0
`656,502
`91,599
`95.8 (94.0)
`0.164 (697)
`10.4 (2.7)
`56
`0.46
`0.50
`
`0.91837
`3.0
`646,108
`91,218
`95.4 (93.0)
`0.170 (0.652)
`8.4 (2.2)
`
`0.61
`0.38
`
`EMTSa
`
`PIPb
`
`122.2
`123.1
`145.8
`114.9
`1.00720
`2.8
`914,648
`96,270
`99.8 (100)
`0.204 (0.716)
`12.4 (1.6)
`8
`1.01
`2.83
`
`121.4
`121.7
`144.2
`114.8
`1.0721
`3.0
`582,365
`76,748
`99.8 (98.3)
`0.119 (0.639)
`16.6 (2.5)
`16
`0.98
`1.35
`
`0.97893
`2.8
`1,323,896
`92,320
`96.9 (96.0)
`0.168 (0.701)
`15.0 (2.4)
`
`1.043
`
`0.385 (acentric), 0.345 (centric)
`0.44 (initial) 0.81 (final)
`
`a Ethylmercurithiosalicylate (EMTS).
`b Di-␮-iodobis(ethylenediamine)diplatinum (PIP).
`
`protectant. Data were collected at Advanced Light Source (ALS)
`and Stanford Synchrotron Laboratory (SSRL) beam lines and pro-
`cessed with both HKL2000 programs and MOSFLM (Otwinowski
`and Minor 1997; Leslie et al. 2002). For heavy atom derivatization,
`the native crystals were soaked in varying concentrations of heavy
`
`atom solutions made in synthetic mother liquor. Extensive screen-
`ing of a large number of heavy atom-soaked crystals resulted in
`isomorphous derivatives: di-␮-iodobis(ethylenedi-
`two useful
`amine)diplatinum (PIP), and ethylmercurithiosalicylate (EMTS).
`In addition, a three wavelength multiple wavelength anomalous
`
`Table 2. Data collection and refinement statistics for wild-type DPPIV and DPPIV/tNPY complex
`
`Crystals
`Space group
`Cell dimension
`
`Solvent content (%)
`Data processing statistics
`Wavelength (Å)
`Resolution range (Å)
`Total reflections
`Unique reflections
`I/␴(I)
`Completeness (%)
`Multiplicity
`Rmerge
`Refinement statistics
`Number of protein/sugar/water residues
`Reflection in working/free set
`Rfactor/Rfree
`Average B-factor (Å2)
`r.m.s. deviation of Bonds (Å)/angle (°) from ideality
`Ramachandran plot
`Residues in most favor region (%)
`Additionally allowed region (%)
`Generously allowed regio