articles
`
`Crystal structure of human dipeptidyl
`peptidase IV/CD26 in complex with a substrate
`analog
`Hanne B. Rasmussen1, Sven Branner1, Finn C. Wiberg2,3 and Nicolai Wagtmann2
`
`Published online 16 December 2002; doi:10.1038/nsb882
`
`Dipeptidyl peptidase IV (DPP-IV/CD26) is a multifunctional type II transmembrane serine peptidase. This enzyme
`contributes to the regulation of various physiological processes, including blood sugar homeostasis, by cleaving
`peptide hormones, chemokines and neuropeptides. We have determined the 2.5 Å structure of the extracellular
`region of DPP-IV in complex with the inhibitor valine-pyrrolidide. The catalytic site is located in a large cavity
`formed between the ␣/␤-hydrolase domain and an eight-bladed ␤-propeller domain. Both domains participate in
`inhibitor binding. The structure indicates how substrate specificity is achieved and reveals a new and unexpected
`opening to the active site.
`
`The serine peptidase dipeptidyl peptidase IV (DPP-IV, CD26,
`EC 3.4.14.5) modulates the biological activity of several peptide
`hormones, chemokines and neuropeptides by specifically cleav-
`ing after a proline or alanine at amino acid position 2 from the
`N terminus1. Various genetic and pharmacological studies have
`revealed a prominent physiological role for this regulatory
`mechanism. Inhibitors of DPP-IV enzyme activity delay allo-
`graft rejection2 and lessen experimental arthritis3 and experi-
`mental autoimmune encephalomyelitis (EAE)4 in animals. The
`DPP-IV substrates involved in these models of chronic inflam-
`mation have not yet been defined but may include chemokines.
`For instance, DPP-IV–mediated cleavage abrogates the activity
`of the monocyte chemotactic peptides MCP-1, -2 and -3, and
`changes the receptor specificity of the chemokine RANTES5.
`DPP-IV is important in maintaining physiological glucose
`homeostasis. Mice lacking the gene for DPP-IV show enhanced
`insulin secretion and accelerated clearance of blood glucose,
`partly because of
`increased endogenous
`levels of active
`glucagon-like peptide-1 (GLP-1) and glucose-dependent
`insulinotropic polypeptide (GIP)6. GLP-1 and GIP are potent
`stimulators of insulin secretion, but their activity is rapidly abol-
`ished by DPP-IV–mediated truncation. Thus, DPP-IV con-
`tributes to the tight control of glucose levels by terminating
`GLP-1 and GIP signalling.
`A significant, rapidly growing fraction of the human popu-
`lation is affected by type 2 diabetes, a disease characterized by
`elevated blood glucose levels and relative insufficiency of
`insulin. Pharmacological
`inhibition of DPP-IV activity
`increases insulin secretion and improves glucose control in
`diabetic animals7–10 and in humans11. Thus, DPP-IV inhibition
`is a promising new strategy for treating type 2 diabetes, and
`DPP-IV inhibitors are now in clinical trials. In this context,
`knowledge of the three-dimensional structure of DPP-IV is
`important for designing new DPP-IV inhibitor drugs and
`understanding the structure–activity relationship of known
`inhibitors.
`
`DPP-IV is expressed as a 220 kDa homodimeric type II
`transmembrane glycoprotein on the surface of various cell types,
`including epithelial and endothelial cells, and lymphocytes5.
`DPP-IV functions as a binding partner for other proteins,
`including adenosine deaminase (ADA)12, the kidney Na+/H+
`exchanger13 and the T-cell antigen CD45 (ref. 14). Cell surface-
`bound DPP-IV is involved in T cell co-stimulation and tumor
`suppression15,16. Proteolytic cleavage of membrane bound
`DPP-IV results in a soluble form (amino acids 39–766) that
`circulates in the plasma. Both the membrane-bound and the
`soluble form show identical enzymatic activity17.
`Despite advances in understanding the biological functions of
`DPP-IV, the structural basis for the dipeptidyl peptidase activity
`of this enzyme and how it selectively chooses short peptides,
`such as GLP-1 and GIP, as substrates are still unknown. To
`address these questions, we determined the structure of a recom-
`binant, soluble form of human DPP-IV that begins at residue
`Ser39, corresponding to the predominant form found in human
`plasma but with a more homogenous glycosylation pattern (see
`Methods). The recombinant soluble DPP-IV retains enzyme
`activity (data not shown). A complex of DPP-IV and a substrate
`analog, the competitive inhibitor valine-pyrrolidide (Val-Pyr)
`(Ki = 2 µM), was crystallized and diffraction data collected at
`2.5 Å resolution. The structure was solved using the MAD
`method.
`
`Overall structure of DPP-IV
`DPP-IV is a dimer in the crystal (Fig.1), in agreement with previ-
`ous biochemical data reporting that the active enzyme is a
`dimer18. The N terminus of each subunit is located at the same
`site of the dimer (Fig. 1a),
`indicating that full-length,
`membrane-bound DPP-IV could exist as homodimers at the cell
`surface. The surface facing the membrane is positively charged
`and complements the negatively charged phospholipids at the
`cell-membrane surface. This indicates that DPP-IV, when bound
`to the membrane, is in close contact with the cell surface. Each
`
`1Protein Chemistry, Research and Development, Novo Nordisk A/S, Novo Allé, DK-2880 Bagsvaerd, Denmark. 2Biotechnology, Research and Development, Novo
`Nordisk A/S, Novo Allé, DK-2880 Bagsvaerd, Denmark. 3Present address: Symphogen A/S, Elektrovej building 375, 2800 Lyngby, Denmark.
`
`Correspondence should be addressed to H.B.R. e-mail: Hbrm@novonordisk.com
`
`nature structural biology • volume 10 number 1 • january 2003
`
`19
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`AstraZeneca Exhibit 2071
`Mylan v. AstraZeneca
`IPR2015-01340
`
`Page 1 of 7
`
`

`
`articles
`
`a
`
`c
`
`b
`
`␣/␤-hydrolase domain
`(Catalytic domain)
`
`
`Ser630
`
`Extended
`arm
`
`laded
`8-b
`␤-propeller domain
`
`1
`
`Bentblade
`
`Side
`opening
`
`Horizontalhelix
`(incl.Glu205andGlu206)
`
`Bottom
`opening
`
`90˚
`
`Fig. 1 Structure of DPP-IV. a, DPP-IV forms a homodimer (subunit A shown in green and subunit B in magenta). Each subunit consists of two
`domains: an α/β-hydrolase domain and a β-propeller domain. The full-length DPP-IV is a type II transmembrane protein in which amino acids 7–28
`constitute the membrane spanning region. The α/β-hydrolase domain, located closest to the membrane, consists of amino acids 39–51 and 506–766,
`and contains the active triad Ser630, Asp708 and His740. The eight-bladed β-propeller domain is formed by residues 55–497. The inhibitor Val-Pyr is
`shown in CPK and colored by element: carbon (gray), nitrogen (blue) and oxygen (red). Five S–S bridges (yellow) have been identified in each mole-
`cule: Cys328–Cys339 (blade 5), Cys385–Cys394 (blade 6), Cys444–Cys447 (blade 7), Cys454–Cys472 (blade 8) and Cys649–Cys762 (hydrolase).
`Carbohydrates (blue) have been located in the electron density map at seven of nine possible N-glycosylation positions in both subunits, holding up
`to three branched proximal glycoside units per position. The glycosylation sites are all situated in the β-propeller domain, except one, and clustered
`mainly to subdomain two (blade 2–5) of the β-propeller. b, Schematic illustration of DPP-IV showing the individual domains and important structural
`elements. c, The surface of molecule A of DPP-IV colored by electrostatic potential. The negatively charged surface is red and positively charged sur-
`face is blue, viewed from the side and the bottom of the propeller. The inhibitor, Val-Pyr is shown in CPK (yellow). The figure was generated by
`GRASP35.
`
`subunit consists of two domains, an α/β-hydrolase domain and
`an eight-bladed β-propeller domain (Fig. 1b). Between these two
`domains, a large cavity of ∼30–45 Å width is found. The Val-Pyr
`molecule is bound in a smaller pocket within the cavity, next to
`the serine-protease active triad, Ser630, Asp708 and His740. The
`large cavity is accessible via two openings. Substrates and prod-
`ucts may pass either through a funnel in the center of the pro-
`peller domain or through a much bigger opening in the side,
`between the hydrolase and propeller domains (discussed below).
`The α/β-hydrolase domain consists of a central β-sheet sand-
`wiched by α-helices, as in other members of the α/β-hydrolase
`family. In DPP-IV, the hydrolase domain is assembled by the
`C-terminal sequence (residues 506–766) and a short stretch of
`the N-terminal sequence (residues 39–51). The eight β-strands
`form a central β-sheet, with six arranged in a parallel and two in
`an antiparallel manner. The sheet is strongly twisted, with the
`last β-strand oriented at a 120° angle to the first strand. The six
`
`α-helices pack against the sheet, with two on one side and four
`on the other. Furthermore, two small α-helices extend the
`β-propeller domain and become part of the hydrolase domain.
`The closest structural homolog of the α/β-hydrolase domain of
`DPP-IV for which the three-dimensional structure is known is
`prolyl oligopeptidase (POP)19, an endopeptidase cleaving after
`proline (r.m.s. deviation is 1.84 Å for 224 Cα atoms in the
`α/β-hydrolase domain). DPP-IV and POP belong to distinct
`subfamilies of the S9 clan of serine peptidases20. The active
`Ser630 in DPP-IV is situated in a so-called ‘nucleophile elbow’ in
`the sequence Gly-Trp-Ser-Tyr-Gly, compared with Gly-Gly-Ser-
`Asn-Gly in POP, which are both in agreement with the con-
`served sequence Gly-X-Ser-X-Gly for the α/β hydrolase family.
`The N-terminal β-propeller domain of DPP-IV (amino acids
`55–497) consists of eight blades, each made up of four anti-
`parallel β-strands (Fig. 2a,b). Although the propeller is topolog-
`ically regular, it is structurally irregular. The propeller may be
`
`20
`
`nature structural biology • volume 10 number 1 • january 2003
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`Page 2 of 7
`
`

`
`articles
`
`a
`
`c
`
`b
`
`d
`
`Fig. 2 Structural comparison of DPP-IV and POP. The eight (DPP-IV) and seven (POP) bladed β-propeller domains are illustrated by solid ribbon and
`colored by blade number. The α/β-hydrolase domain is gray. a, DPP-IV viewed from the side. Glu205, Glu206 and Ser630 are illustrated by ball and
`stick (red). The extended arm is yellow, and the horizontal helix holding Glu205 and Glu206 is black. b, DPP-IV viewed from the bottom. c, POP
`viewed from the side. The N-terminal extension relative to DPP-IV is dark gray. d, POP viewed from the bottom, with Ser554 illustrated in ball and
`stick (red).
`
`divided into two subdomains (Fig. 2a), with blades 2–5 forming
`one subdomain; and blades 1 and 6–8, forming another.
`All the blades of subdomain one (blades 2–5) bend toward the
`cavity, whereas those of subdomain two (blades 1 and 6–8) bend
`away from the cavity, especially blade 1, which is strongly bent
`outward (Fig. 2a). This results in an ellipsoid shape of the inside
`of the propeller (the major axis going from blade 2 to blade 6)
`rather than the circular shape, which is more common for this
`fold. POP has a seven-bladed propeller19 that is much more reg-
`ular than the one observed in DPP-IV. It is not possible to make
`meaningful sequence or structure alignments, considering the
`full β-propeller domains of DPP-IV and POP.
`The inside of the propeller of DPP-IV forms a funnel-shaped
`tunnel between the active site and the bottom of the monomer.
`The bottom opening (defined according to the side-on view of
`the molecule, Fig. 2a), distal to the hydrolase domain, is
`∼7 Å × 14 Å; at the top, the opening widens to a diameter of
`∼24 Å. The length of the funnel is ∼28 Å; however, the funnel is
`not the only opening to the active site. A pronounced side open-
`ing is created by (i) the pronounced bending of blade 1 in the
`propeller (Fig. 2c), (ii) a cleft formed by the smaller bending of
`blades 2–4 and some shorter β-strands in these blades and the
`β-turn holding the active Asp708, and (iii) an extended arm
`from the β-propeller (Fig. 2a). A calculation of the surface elec-
`trostatic potential shows that the cleft, together with the binding
`site for the inhibitor, is negatively charged (Fig. 1c). The cavity
`itself and the bottom of the funnel are also negatively charged,
`
`although not as strongly as in the side-opening cleft. On the basis
`of size and electrostatic characteristics, both the funnel and the
`side opening may serve as an entrance/exit to the active site.
`The ADA-binding property of DPP-IV has been ablated by site
`directed mutagenesis of L294R or V341K21. L294 is situated in
`the small α-helix between blade 4 and 5, and V341 is situated in
`a long loop between β-strands 3 and 4 of blade 5. Both residues
`are exposed to solvent, located only 15 Å (Cα distance) apart
`from one another on the backside of the propeller looking at the
`molecule from the side (Fig. 2a).
`A sequence motif has been associated with the β-propeller
`proteins, the ‘WD motif ’22. The β-propeller of DPP-IV has some
`of the WD repeat characteristics in some of the eight blades;
`however, a consensus throughout the whole propeller is not
`found. The WD repeat is named from the motif found at the end
`of the third strand, namely amino acids Trp-Asp. This motif is
`not found in the propeller blades of DPP-IV, although homolo-
`gous residues are observed in several of the blades.
`
`The active dimer
`In POP the hydrolase domain has a 56-amino acid N-terminal
`extension relative to DPP-IV (Fig. 2c). Part of this extension
`(residues 1–10 and 39–54) forms a structural patch that, based
`on superimposition of the POP monomer on the DPP-IV dimer,
`aligns with a corresponding patch in DPP-IV that is formed by
`the C-terminal residues 745–761 (Fig. 3) and residues 729–737
`from the other subunit of the DPP-IV homodimer. Thus, the
`
`nature structural biology • volume 10 number 1 • january 2003
`
`21
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`Page 3 of 7
`
`

`
`articles
`
` 1/β1
`
`S R K
`40
`
`V T N
`280
`
` 1/β2
` 1/β4
` 1/β3
`
`
`
`
`
`
`W I S D H E Y L Y K
`T Y R L K L Y S L R
`T Y T L T D Y L K N
`N S T F D E F G H S
`E Y G N S S V F L E
`Q E N N I L V F N A
`70
`60
`50
`100
`90
`80
` 2/β3
` 2/β4
` 3/β3
` 3/β2
` 3/β1
`
`
`
`
`
`
`
`
`
`
`K Q W R H S Y T A S
`N Y V
`L I T E E R I P N N
`Y D I Y D L N K R Q
`I Y V K I E P N L P
`H K L A Y V W N N D
`T Q W V T W S P V G
`120
`130
`150
`140
`180
`170
`160
` 4/β1b
` 4/β2
`
`
`Extended/β1
`Extended/β2
`Horizontal helix
`
`
`
`
`T D W
`V Y E E E V F S A Y
`F L A Y A Q F N D T
`S A L W W S P N G T
`S D E S L Q Y P K T
`V R V P Y P K A G A
`E V P L I E Y S F Y
`200
`210
`230
`220
`250
`260
`240
` 5/β1
` 4/β4
` 5/β3
` 5/β2
`
`
`339
`328
`
`
`A T S I Q I T A P A
`S M L I G D H Y L C
`E S S G R W N C L V
`Y S V M D I C D Y D
`S L Q W L R R I Q N
`D V T W A T Q E R I
`290
`300
`340
`330
`320
`310
` 7/β1
` 6/β1
` 6/β2
` 7/β2
` 6/β4
` 6/β3
`
`
`
`
`
`
`
`
`
`
`385
`394
`
`
`
`
`
`
`P H F T L D G N S F
`P S E
`Y K I I S N E E G Y
`S D Y L Y Y I S N E
`W E V I G I E A L T
`K D C T F I T K G T
`R H I C Y F Q I D K
`370
`360
`380
`420
`410
`400
`390
` 7/β4
` 8/β1
` 8/β2
` 8/β3
` 8/β4
`
`
`
`
`
`
`472
`454
`447
`444
`
`
`
`
`
`
`Y T K
`V T C L S C E L N P
`E R C Q Y Y S V S F
`S K E A K Y Y Q L R
`L H S S V N D K G L
`C S G P G L P L Y T
`R V L E D N S A L D
`440
`450
`460
`470
`490
`480
`500
`
` 2/β2
`
`
`Q F I L L E Y
` 4/β1a
`
`
`D I I Y N G I
`
` 2/β1
`
`
`I N D Y S I S P D G
`110
` 3/β4
`
`
`S Y R I T W T G K E
`190
` 4/β3
`
`
`V N P T V K F F V V
`270
` 5/β4
`
`
`G W V G R F R
`A R Q H I E M S T T
`350
` 7/β3
`
`
`Y K G M P G G R N L
`Y K I Q L S D
`430
`
`N T D S L S S
`
`K M L Q N V Q M P S
`510
`
`K K L D F I I
`
`L N E
`520
`
`T K F W Y Q M I L P
`530
`
`P H F D K S K K Y P
`540
`
`G T F
`600
`
`E V E D Q I E A A R
`610
`
`Q F S K M G F V D N
`620
`
`N L D
`680
`
`H Y R N S T V M S R
`690
`
`649
`
`A E N F K Q V E Y L
`700
`
`L L L D V Y A G P C
`550
`*
`K R I A I W G W S Y
`630
`*
`L I H G T A D D N V
`710
`
`S Q K A D T V F R L
`560
`
`N W A T Y L A S T E
`570
`762
`
`G G Y V T S M V L G
`640
`
`S G A G V F K C G I
`650
`
`H F Q Q S A Q I S K
`720
`
`A L V D V G V D F Q
`730
`
`N I I V A S F D G R
`580
`
`G S G Y Q G D K I M
`590
`
`H A I N R R L
`
`A V A P V S R W E Y
`660
`*
`A M W Y T D E D H G
`740
`
`Y D S V Y T E R Y M
`670
`
`G L P T P E D
`
`I A S S T A H Q H I
`750
`
`Y T H M S H F
`
`I K Q
`C F S
`760
`Fig. 3 Secondary structure of DPP-IV shown over the amino acid sequence. Arrows indicate β-strands, and bars indicate helices. Color code is accord-
`ing to propeller blade number as in Fig. 2. The hydrolase domain is gray. Glycosylated residues located in the electron density map are marked with
`a cyan squared background. Cysteines involved in S–S bonds are marked with a yellow circular background and the corresponding residue number
`is listed above. The active triad residues are marked by a red asterisk, and other important residues for inhibitor binding are marked by an orange
`pentagonal background.
`
`monomeric POP in itself fills part of the dimeric structure
`observed in DPP-IV.
`The extended arm from the propeller participates strongly in
`the dimerization of DPP-IV (Fig. 1). β-strand 2 of blade 4 of the
`propeller extends into a small domain (amino acids 234–260)
`that includes an antiparallel two-stranded β-sheet. The function
`of this arm is clearly to stabilize the dimeric structure. POP also
`has an extended arm (residues 192–206) at the same spatial
`position relative to the hydrolase domain as in DPP-IV, but pro-
`truding from blade 3 rather than blade 4. Although the loop in
`POP is much shorter than that in DPP-IV, it is noteworthy that
`this loop is present in both structures despite their apparent
`monomeric versus dimeric structures and the low level of
`sequence and structural alignment between the propellers in the
`two peptidases.
`Besides participating in the dimerization of DPP-IV, the
`extended arm also gives the impression of a lid function, as it is
`situated close to the active site and has a size that matches the
`side opening. If the dimer dissociates, this arm could move
`towards the cavity, thereby closing the side opening of the active
`site. Thus, the extended arm may provide a structural explana-
`tion for the observations that DPP-IV is enzymatically active as a
`dimer18. In addition to the extended arm, residues 658–661,
`713–736, and 746–757 contribute to dimerization. The dimer
`interface is hydrophilic, indicating that DPP-IV might be able to
`exist in solution as a monomer. The dimer interface is 2,220 Å2,
`corresponding to only 7% of the total accessible surface area of
`one subunit. These structural observations, in combination with
`reported biochemical data, raise the possibility that DPP-IV may
`exist either as an open, enzymatically active dimer or as a closed,
`inactive monomer, and that dimerization-mediated loop move-
`ment may represent a new type of mechanism for regulating the
`enzyme activity.
`The topology of the propeller allows for some flexibility. All
`four S–S bridges present in the propeller are formed within the
`
`individual blades (Fig. 1a), rather than between the blades.
`Furthermore, the N and C termini of the propeller do not both
`participate in the same β-sheet — that is, there is no ‘Velcro’ clo-
`sure of the propeller23, a common stabilizing feature in many
`β-propellers. Finally, the crystallographic B-values of blade 1 (48
`and 64 Å2 for molecule A and B, respectively) are increased rela-
`tive to the B-values of the rest of the blades (30 and 32 Å2 for
`molecule A and B, respectively). This suggests flexibility for this
`bent blade, which may even straighten and thereby close the
`propeller. By itself or together with the extended arm, blade 1
`may regulate substrate entry or exit from the active site by an
`opening/closing mechanism.
`Several features of the DPP-IV structure suggest that the
`entrance to the active site is via the large side opening. This is the
`shortest and most easily accessible way to the active site. The
`negative potential in the cleft would attract the positively
`charged N terminus of the peptide substrate into the cavity. By
`entering this way rather than through the propeller, the peptide
`is orientated correctly for cleavage, as determined from analogy
`with the orientation of the pseudopeptide Val-Pyr. In contrast, a
`peptide entering through the propeller would have to make a
`turn inside the cavity. Once the dipeptide has been cleaved from
`the substrate, it may exit through the propeller funnel. Although
`we favor the hypothesis that substrate may access the active site
`via the side opening in DPP-IV, we can not rule out that entry
`may occur via the propeller funnel, as suggested for POP, whose
`structure revealed a single opening through the propeller funnel.
`As in POP, the bottom propeller opening of DPP-IV is negatively
`charged and, therefore, might attract the positively charged
`N terminus of peptides. Introduction of a S–S bridge by mutage-
`nesis between blade 7 and blade 1 in POP abolished enzyme
`activity24, supporting the theory of an entry through the pro-
`peller. POP was suggested to possibly be able to open the pro-
`peller even further than the 4 Å seen in the structure because of
`the lack of molecular ‘Velcro’. However, another plausible expla-
`
`22
`
`nature structural biology • volume 10 number 1 • january 2003
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`Page 4 of 7
`
`

`
`Tyr547
`
`Ser630
`
`His740
`
`Asp708
`
`Tyr666
`
`Tyr662
`Asn710
`
`Arg125
`
`Glu206
`
`Glu205
`
`a
`
`b
`
`articles
`
`c
`
`CH3
`
`CH3
`
`NH2
`
`N
`
`O
`
`Fig. 4 Active site of DPP-IV. a, Stereo draw-
`ing of the active site of DPP-IV. Val-Pyr is
`shown in CPK, with selected residues are
`shown in stick and colored by element type
`as in Fig. 1a. Hydrogen bonds/electrostatic
`interactions are shown with dotted lines.
`Glu205 and Glu206 coordinate the N termi-
`nus. Arg125 and Asn710 coordinate the car-
`bonyl group of the peptide bond before
`the cleavage point. The pyrollidide ring is
`packed between Tyr547, Tyr762 and Tyr666.
`The valine moiety of the inhibitor has no
`contacts but points towards the pocket.
`b, Val-Pyr is shown in thick sticks and sel-
`ected residues (as in a) in thin sticks. The ini-
`tial MAD electron density map after SHARP
`refinement at 2.5 Å resolution is overlaid
`contoured at 1 σ (blue) and 2 σ (red). c, The
`chemical structure of the inhibitor Val-Pyr.
`
`nation is that the flexible closure of the propeller may allow the
`enzyme to open its propeller on the side by bending blade 1, as
`seen in the structure of DPP-IV. Further functional studies are
`needed to address these possibilities.
`Although the side opening in DPP-IV is large, the active site is
`located in a small pocket within the large cavity. Thus, only elon-
`gated peptides, or unfolded or partly unfolded protein frag-
`ments, can reach the site. This explains why most natural
`DPP-IV substrates are peptides <80 amino acids. However,
`larger proteins might be substrates as long as they have an
`unfolded N-terminal region.
`
`The binding site of DPP-IV
`The structure of the active site reveals how substrate specificity is
`achieved in DPP-IV, which preferentially cleaves after Xaa-Pro or
`Xaa-Ala (Xaa being any amino acid). Residues from both the
`hydrolase domain and the propeller domain take part in the
`binding of the substrate analog inhibitor Val-Pyr (Fig. 4). Two
`glutamic acids, Glu205 and Glu206, form salt bridges to the free
`amino group of Val-Pyr, which corresponds to the N terminus of
`a peptide substrate. Glu205 and Glu206 are situated in a small
`horizontal helix (residues 201–207) in blade 4 of the propeller
`domain (Fig. 2a). This small horizontal helix narrows the active
`site itself, leaving room for only two amino acids before the pep-
`tide reaches the active-site serine, making DPP-IV a dipeptidyl
`peptidase. The spatial arrangement of the active-site Ser630 with
`respect to Glu205 and Glu206 in DPP-IV probably makes the
`two glutamic acids the most important feature for alignment of
`the peptide before cleavage. This notion is consistent with stud-
`ies showing that mutation of either one of these glutamic acids
`destroys the enzymatic activity of DPP-IV25. The pyrollidide
`moiety of Val-Pyr is buried in a hydrophobic pocket next to the
`
`active serine. Only amino acids with smaller side chains (proline,
`alanine or glycine1) will be able to fit into this narrow pocket,
`thereby restricting possible residues at the P1 position in sub-
`strates. Tyr662 and Tyr666 stack at each side of the pyrollidide
`ring of Val-Pyr — Tyr662 in a parallel fashion and Tyr666 in an
`orthogonal fashion — and are therefore part of the features
`determining the preference for a proline preceding the scissile
`bond. The valine side chain of Val-Pyr points into the large
`cavity and does not make specific contacts with DPP-IV, explain-
`ing why DPP-IV has no specific requirements for the N-terminal
`amino acid in the P2 position. The carbonyl oxygen of Val-Pyr
`forms hydrogen bonds to Asn710 and Arg125. The oxyanion
`hole is probably formed by Tyr547, as predicted21,26,27, together
`with the backbone NH of Tyr631.
`The residue of POP equivalent to Asn710 of DPP-IV is an argi-
`nine. The guanidine part of this arginine is situated close to the
`guanidine of Arg125 in DPP-IV, although it is distant in the Cα
`alignment. The side chain of Tyr662 in DPP-IV aligns struc-
`turally with the side chain of Trp595 in POP, whereas the back-
`bone of Trp595 of POP aligns structurally with Tyr666 of
`DPP-IV. The essential Cys255 in POP, sitting in the large P2
`pocket of DPP-IV in a superimposition based on the active triad
`of the two enzymes, does not have a homologous counter-
`residue in DPP-IV. Thus, DPP-IV and POP show some homolo-
`gy in the upper part of the catalytic cavity orienting the molecule
`(Fig. 4a). In contrast, significant differences are present in the
`bottom part of these molecules, explaining their different sub-
`strate specificities.
`
`Concluding remarks
`Several homology models of DPP-IV have been made based on
`the POP structure21,26,27. However, as described in this paper,
`
`nature structural biology • volume 10 number 1 • january 2003
`
`23
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`Page 5 of 7
`
`

`
`CD5/DPP-IV construct was inserted into the vec-
`tor pBlueBac4.5 (Invitrogen) and used to gener-
`ate a recombinant baculovirus. Hi5 insect cells
`were infected with this virus and grown in
`serum-free medium, resulting in secretion of
`soluble, active DPP-IV.
`
`Purification. The recombinant human DPP-IV
`(RhDPP-IV) (39–766) was purified using a vari-
`ant of the adenosine deaminase affinity chro-
`matography method28 in which the two initial
`purification
`steps are
`interchanged. The
`amount of adenosine deaminase was reduced
`to 25% of that reported28 when immobilizing
`ADA on the CNBr-activated sepharose. The cul-
`ture supernatant was adjusted to pH 8 and the
`ionic strength to ∼20 mS cm–1 conductance,
`after which it was filtered and applied directly
`to the adenosine deaminase affinity column.
`The buffer system was as described28. The elu-
`ate from the affinity column was applied
`directly to a pre-equilibrated Q Sepharose High
`Performance column (Amersham Biosciences)
`connected serially to the affinity column. The
`sepharose column was eluted with a 0–0.5 M
`NaCl gradient in 20 mM Tris buffer with 0.1%
`(w/v) Triton X-100, pH 8. DPP-IV eluted at
`∼0.1 M NaCl. Mono Q
`Ion Exchange
`Chromatography (Amersham Biosciences) was
`used as the final purification step with the
`same buffer as in the Q Sepharose step, except
`that the detergent was excluded. The enzyme
`eluate was concentrated on Centriprep YM-10
`and Centricon YM-10 (Millipore) to 40 mg ml–1.
`The DPP-IV was homogenous on SDS-PAGE
`with a molecular weight (Mw) of ∼90 kDa. N-ter-
`minal amino acid sequencing verified the
`sequence starting at position 39.
`The Mw of the purified recombinant DPP-IV
`(amino acids 39–766) was determined to be
`93,500 Da by MALDI-MS. The theoretical Mw for
`the non-glycosylated DPP-IV
`(amino acids
`39–766) is 84,400 Da. Therefore, glycosylation
`can be estimated to be 11%. For comparison, the
`Mw for purified human placental DPP-IV was
`104,400 Da, corresponding to 24% glycosylation.
`
`articles
`
`Space group
`Unit cell (Å)
`a
`b
`c
`Wavelength (Å)
`Resolution range (Å)2
`Measurements
`Unique reflections
`Completeness2
`I / σ (I)2
`Rmerge (%)2,4
`Phasing power
`Mean figure of merit5
`
`Table1 Data collection and refinement statistics1
`λ1
`λ2
`Native
`P212121
`P212121
`
`119.2
`119.0
`123.5
`123.2
`131.3
`130.7
`0.8265
`1.0085
`1.02660
`20–2.6 (2.64–2.60) 30–2.65 (2.71–2.65) 30.0–2.50 (2.56–2.50)
`231,482
`227,811
`223,026
`56,062
`104,278
`111,440
`94.0 (82.3)
`95.9 (83.0)3
`86.0 (64.8)3
`13.9 (3.0)
`11.5 (2.1)
`12.2 (2.3)
`6.5 (40.2)
`5.6 (32.8)
`1.3
`0.40 (SHARP)
`0.86 (SOLOMON)
`0.30 (SOLVE)
`
`1.6
`
`Rcryst / Rfree (%)6
`Number of atoms
`Non-hydrogen protein
`Ligand
`Carbohydrate atoms
`Hg ions
`Water molecules
`Average B-factor (all atoms, Å2)
`R.m.s. deviation
`Bond lengths (Å)
`Angles (°)
`
`21.0 / 26.1
`
`11,911
`24
`399
`4
`931
`33.4
`
`0.007
`1.6
`
`1Native data were collected at MaxLab, Lund, beamline 711, and the heavy atom derivative
`data were collected at ESRF, Grenoble, beamline BM14 at two wavelengths (peak (λ1) and
`remote (λ2)).
`2Values in parentheses are for the highest resolution shell.
`3Friedel pair was kept separate.
`4Rmerge (I) = Σhkl |Ihkl – <Ihkl>| / ΣhklIhkl, where Ihkl is the measured intensity of the reflections with
`indices hkl.
`5See refs. 32,33,31, respectively.
`6R-factors were calculated using data F > 0 σ. R-factor = Σhkl ||Fo| – |Fc| / Σ|Fo|, where |Fo| and |Fc|
`are the observed and calculated structure factor amplitudes for reflection hkl, applied to the
`work (Rcryst) and test (Rfree) (5 % omitted from refinement) sets, respectively.
`
`©2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology
`
`large structural variations are observed between these two
`enzymes. Two of the major differences are the larger access to
`the active side via the side opening and the localization of
`Glu205 and Glu206 as part of the binding site rather than as
`‘gatekeepers’ to the site. Furthermore, the full positioning of
`the individual amino acids in the β-propeller are different
`from predicted in the models. The crystal structure of DPP-IV
`will, therefore, allow new interpretations of earlier biochemi-
`cal data.
`The results reported in this paper explain the substrate speci-
`ficity of DPP-IV and its selectivity for elongated peptides, and
`provide a framework for further studies on how the enzyme
`activity of DPP-IV is regulated and its multiple biological func-
`tions. DPP-IV is a promising drug target for intervention in type
`2 diabetes, and the three-dimensional structure provides an
`excellent basis for structure-based rational drug design of phar-
`maceutically relevant inhibitors.
`
`Methods
`Expression of recombinant DPP-IV. A construct encoding the sol-
`uble, secreted form of human DPP-IV was made by fusing the cDNAs
`for the signal sequence of CD5 in-frame with a fragment coding for
`amino acids 39–766 of human DPP-IV (provided by B. Seed). This
`
`Activity assay. The activity of DPP-IV and inhibition by Val-Pyr was
`measured by cleavage of the substrate Gly-Pro-pNA in 50 mM Tris,
`0.15 M NaCl and 0.1 % (w/v) Triton X-100 pH 7.4 in a 96 well Elisa
`format assay.
`
`Crystallization and data collection. RhDPP-IV (39–766) was
`premixed with Val-Pyr (at 2× the molar concentration of DPP-IV)
`and crystallized using sitting drop vapor diffusion method. A
`0.8 µl sample of RhDPP-IV (40 mg ml–1 in 20 mM Tris, pH 8.0, and
`25 mM NaCl) was mixed with 0.8 µl reservoir (0.3 M sodium
`acetate, 17–18% (w/v) PEG 4000, 0.1 M Tris, pH 8.0). The crystals
`grew to a size up to 500 × 150 × 75 Å within 2–3 days using micro
`seeding. Before data collection, the crystal was cryo-cooled using
`a mixture of 0.4 M sodium acetate, 35% (w/v) PEG 4000 and 0.1 M
`Tris, pH 8.0. Native data were collected at MaxLab, Lund
`(Beamline 711). A MAD experiment was performed at ESRF,
`Grenoble (Beamline BM14) using mercury as an anomalous scat-
`terer. Diffraction data were collected at two wavelengths, the
`inflection wavelength and the remote wavelength, and processed
`using DENZO29 and SCALA (CCP4 program suite)30. Statistics are
`listed in Table 1.
`
`Structure determination and refinement. Using SOLVE31, four
`mercury sites were located. The positions, occupancies and B-values
`were further refined using SHARP32 including native as well as deriv-
`
`24
`
`nature structural biology • volume 10 number 1 • january 2003
`
`Page 6 of 7
`
`

`
`articles
`
`ative data. After applying solvent flattening using SOLOMON33 within
`SHARP, the resulting map was of high quality, showing continuous
`density through practically the whole backbone chain and clearly
`revealing the side chains, S–S bridges, glucoside residues, inhibitors
`and water molecules. All model building was performed using QUAN-
`TA (Accelrys). On the basis of the solvent-flattened electron density
`map an initial Cα trace was built using the X-POWERFIT within QUAN-
`TA. This trace was converted