FEBS 25913
`
`FEBS Letters 515 (2002) 165^170
`
`NMR studies of the aggregation of glucagon-like peptide-1: formation of
`a symmetric helical dimer
`Xiaoqing Changa;1, Danielle Kellera;2, Sea¤n I. O’Donoghueb;3, Jens J. Leda;(cid:3)
`
`aDepartment of Chemistry, University of Copenhagen, The H.C. (cid:210)rsted Institute, Universitetsparken 5, DK-2100 Copenhagen (cid:210), Denmark
`bEuropean Molecular Biology Laboratory, Meyerhofstrasse 1, D-69012 Heidelberg, Germany
`
`Received 12 February 2002; accepted 14 February 2002
`
`First published online 27 February 2002
`
`Edited by Thomas L. James
`
`Abstract Nuclear magnetic resonance (NMR) spectroscopy
`reveals that higher-order aggregates of glucagon-like peptide-1-
`(7^36)-amide (GLP-1) in pure water at pH 2.5 are disrupted by
`35% 2,2,2-trifluoroethanol (TFE), and form a stable and highly
`symmetric helical self-aggregate. NMR spectra show that the
`helical structure is identical to that formed by monomeric GLP-1
`under the same experimental conditions [Chang et al., Magn.
`Reson. Chem. 37 (2001) 477^483; Protein Data Bank at RCSB
`code: 1D0R], while amide proton exchange rates reveal a
`dramatic increase of the stability of the helices of the self-
`aggregate. Pulsed-field gradient NMR diffusion experiments
`show that the TFE-induced helical self-aggregate is a dimer. The
`experimental data and model calculations indicate that the dimer
`is a parallel coiled coil, with a few hydrophobic residues on the
`surface that may cause aggregation in pure water. The results
`suggest that the coiled coil dimer is an intermediate state towards
`the formation of higher aggregates, e.g. fibrils. (cid:223) 2002 Feder-
`ation of European Biochemical Societies. Published by Elsevier
`Science B.V. All rights reserved.
`
`Key words: Glucagon-like peptide-1-(7^36)-amide;
`Nuclear magnetic resonance; Aggregation; Intermediate;
`Modeling
`
`1. Introduction
`
`Glucagon-like peptide-1-(7^36)-amide (GLP-1) is a 30 ami-
`no acid peptide with the sequence: HAEGTFTSDVSSYLEG-
`QAAKEFIAWLVKGR-NH2. It is an important gluco-incre-
`tin hormone that can potentiate glucose-induced insulin
`secretion, stimulate insulin biosynthesis and inhibit glucagon
`secretion [1,2]. Consequently, it is a potential drug for the
`treatment
`of
`non-insulin-dependent
`diabetes mellitus
`(NIDDM or type 2 diabetes [3]). However, GLP-1 su¡ers
`
`*Corresponding author. Fax: (45)-3535 0609.
`E-mail address: led@kiku.dk (J.J. Led).
`
`1 Present address: Computational Chemistry, GLYCODesign Inc.,
`480 University Ave., Suite 400, Toronto, ON, Canada M5G 1V2.
`2 Present address: Department of Physics, MEMPHYS, University of
`Southern Denmark, Campusvej 55, 5230 Odense M, Denmark.
`3 Present address: LION Bioscience AG, Waldhoferstr. 98,
`Heidelberg 69123, Germany.
`
`Abbreviations: CSI, chemical shift index; GLP-1, glucagon-like pep-
`tide-1-(7^36)-amide; NOESY, nuclear Overhauser e¡ect spectrosco-
`py; PDB, Protein Data Bank at RCSB; PFG, pulsed-¢eld gradient;
`TFE, 2,2,2-tri£uoroethanol; water-sLED, water-suppressed longitudi-
`nal encoding decoding
`
`instability. Recently the
`from both metabolic and physical
`metabolic stability and the biological activity of a series of
`GLP-1 analogues were investigated [4^6]. As for the physical
`instability, it has been found that the conformation, aggrega-
`tion, and solubility of GLP-1 depend on the puri¢cation pro-
`cedure and the in-process storage and handling [7,8]. Still, the
`K-helix seems to be the predominant structural motif of GLP-
`1 in solution. It was found that GLP-1 can form oligomers
`with a high helical content [7^9], and that it is mainly helical
`in membrane-like environments (dodecylphosphocholine mi-
`celle) [10]. More recently it was found [11] that monomeric
`GLP-1 is a random coil in pure water but forms a helical struc-
`ture in aqueous 2,2,2-tri£uoroethanol (TFE) solution (Protein
`Data Bank at RCSB (PDB) code: 1D0R), similar to the struc-
`ture observed in the membrane-like environments. However,
`GLP-1 can also form less soluble L-sheet aggregates [7].
`To provide further insight into the self-association behavior
`of GLP-1 we have studied the conformation of aggregated
`GLP-1 in water/TFE mixtures using nuclear magnetic reso-
`nance (NMR) spectroscopy. TFE is known to stabilize helical
`conformations in proteins and peptides while disrupting the
`speci¢c tertiary interactions of native proteins. These e¡ects
`are ascribed to the ability of TFE to enhance internal hydro-
`gen bonding in polypeptides and to lessen hydrophobic inter-
`actions between residues distant in the amino acid sequence
`[12]. A study of the structure of GLP-1 in water/TFE mixtures
`seems, therefore, highly interesting in order to get more in-
`sight into the complex aggregation propensity of GLP-1 in
`solution.
`
`2. Materials and methods
`
`2.1. Sample preparation
`Samples of aggregated and monomeric recombinant GLP-1 were
`kindly provided by the pharmaceutical company Novo Nordisk
`A/S. The lyophilized peptide was dissolved in H2O (with 10% D2O)
`or in 99.96% D2O. In all samples the concentration of monomeric
`GLP-1 was approximately 1.4 mM and the pH was 2.5 (meter read-
`ing). The concentration of TFE in the water/TFE NMR samples was
`35%. Perdeuterated TFE (TFE-d3) was used in all experiments.
`
`2.2. NMR spectroscopy
`The NMR spectra were recorded at 300 K on Varian Inova Unity
`500, 750 and 800 spectrometers as described previously [11]. Qualita-
`tive discrimination between slowly and fast exchanging amide protons
`was achieved by recording a series of nuclear Overhauser e¡ect spec-
`troscopy (NOESY) spectra of GLP-1 in 99.96% D2O with 35% TFE
`at regular times after the dissolution, and de¢ning the amide protons
`still giving rise to cross peaks as slowly exchanging. A series of pulsed-
`¢eld gradient (PFG) experiments with water-suppressed longitudinal
`encoding decoding (water-sLED) pulse [13,14] was performed to de-
`
`0014-5793 / 02 / $22.00 (cid:223) 2002 Federation of European Biochemical Societies. Published by Elsevier Science B.V. All rights reserved.
`PII: S 0 0 1 4 - 5 7 9 3 ( 0 2 ) 0 2 4 6 6 - 3
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 1
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`166
`
`X. Chang et al./FEBS Letters 515 (2002) 165^170
`
`termine the self-di¡usion coe⁄cients and the relative molecular size of
`monomeric and aggregated GLP-1 in 35% TFE.
`
`2.3. Generation of the GLP-1 coiled coil dimer
`An initial model of the backbone coordinates of the GLP-1 dimer
`was obtained using the program ThreadCoil [15]. ThreadCoil takes as
`an input an amino acid sequence that is proposed to fold into a coiled
`coil, and returns a three-dimensional structure (K-carbons only). The
`most likely structure is found by a grid search over the three coiled
`coil parameters (radius, pitch, and phase). The likelihood of each
`structure is evaluated using a pairwise residue potential derived
`from the PDB database.
`The coordinates of the remaining side-chain heavy atoms were in-
`troduced by MODELLER4 [16], where the side-chains were energy
`minimized from initially random positions. The general form of the
`energy functions and the optimization in MODELLER4 are similar to
`CHARMM [17]. The hydrogen atoms were then included by the
`HBUILD feature in X-PLOR, and the full-atom structures were re-
`¢ned against the NMR-derived distance restraints (including the intra-
`monomeric NOEs and two distance restraints for each of the hydro-
`gen bonds observed in the aggregated GLP-1 spectrum, except the one
`involving the amide proton of residue A2, vide infra) using simulated
`annealing. The dimer symmetry was imposed using a non-crystallo-
`graphic symmetry term. The simulated annealing was based on the
`method described by Nilges and Bru«nger for automated modeling of
`coiled coils [18]. During the re¢nement, the positions of the K-carbons
`were restrained using a harmonic restraining potential. A calculation
`of 10 slightly di¡erent model structures was carried out starting with
`di¡erent side-chain coordinates, and the model with the lowest total
`energy was selected as the representative structure.
`
`Fig. 2. The amide and aromatic proton region of the NOESY spec-
`tra of (a) monomeric and (b) aggregated GLP-1, both in water with
`35% (v/v) TFE. The spectra were recorded at (a) 800 MHz and (b)
`500 MHz. The same correlations are present in the two spectra, and
`the chemical shift values are identical.
`
`3. Results and discussion
`
`3.1. Formation of a symmetric helical GLP-1 aggregate
`Fig. 1 shows the one-dimensional 1H NMR spectra of ag-
`gregated and monomeric GLP-1 dissolved in pure water and
`in water with 35% TFE, respectively. The excessive broaden-
`ing of the resonances of the aggregated form in pure water
`(Fig. 1a) clearly reveals an extensive aggregation.
`In sharp contrast to this, the one-dimensional 1H spectra of
`aggregated and monomeric GLP-1 in water with 35% TFE
`(Fig. 1b,c) are practically identical. This identity is con¢rmed
`by the two-dimensional NOESY spectra in Fig. 2, which show
`that the chemical shifts and the pattern of NOE signals of the
`two forms are almost identical. Only two extra dKL(i,i+3)
`NOEs in the N-terminal end of the peptide appear in the
`spectrum of the aggregated form (Fig. 3b). Moreover, all
`the observed signals can be assigned to the same monomeric
`
`Fig. 1. One-dimensional 1H NMR spectra of GLP-1. a: Aggregated
`GLP-1 in pure water. b: The same aggregated form of GLP-1 in
`water with 35% (v/v) TFE. c: Monomeric, non-aggregated GLP-1
`in water with 35% (v/v) TFE.
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 2
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`X. Chang et al./FEBS Letters 515 (2002) 165^170
`
`167
`
`Fig. 3. Summary of sequential and medium-range NOE connectivities, and schematic representation of the 1HK chemical shift index (CSI) of
`(a) monomeric GLP-1 and (b) aggregated GLP-1. In both cases the sample (1.4 mM, pH 2.5, 300 K) was dissolved in H2O with 35% TFE.
`The observed NOE connectivities are indicated by bars connecting the two involved residues, and the intensities of the NOEs are indicated by
`the thickness of the bars. The CSIs were calculated according to Wishart et al. [36]. Slowly exchanging backbone amide protons are indicated
`by ¢lled circles (b).
`
`structure, i.e. only one set of resonances is present and no
`long-range NOEs corresponding to a tertiary fold are ob-
`served. These results show, unambiguously, that the higher
`aggregates of GLP-1 have been disrupted by TFE. Further-
`more they show that the local geometry of the individual
`residues of the resulting structure is identical to that in mono-
`meric, helical GLP-1;
`i.e. the peptide has the same helical
`structure as the monomeric form under the same experimental
`conditions, except for a small extension of the K-helix towards
`the N-terminal end.
`However, the amide proton exchanges reveal that the helix
`is considerably more stable in the aggregated form than in the
`monomeric form. Thus, the number of slowly exchanging
`amide protons is increased dramatically in the aggregated
`form (27 versus 12 in the monomeric form) as shown in
`Fig. 3b. Furthermore, the intensities of the amide proton sig-
`nals were unchanged several days after dissolution in D2O. In
`fact, most of the signals still remained after months, i.e. the
`exchange rates of the amide protons are about two orders of
`magnitude slower in the helical structure obtained from the
`aggregated form than in the monomeric helix. Consequently,
`the helix has been stabilized dramatically through a strength-
`ening of its i,i+4 hydrogen bonds, while at the same time it
`has been extended all the way to the N-terminus of the pep-
`tide (Fig. 3b). Taken together, the spectra and the amide
`proton exchange rates of the structure derived from the ag-
`gregated form clearly indicate the formation of a highly sym-
`
`metric, highly stable and soluble self-associate consisting of
`GLP-1 monomers with basically the same extended helical
`structure as the monomeric GLP-1 described previously [11].
`Finally it was found that the helical self-associate can be
`formed also from monomeric GLP-1 by seeding it with the
`highly aggregated form. Thus addition of traces of the latter
`form to a solution of monomeric GLP-1 in water with 35%
`TFE resulted in the same dramatic decrease of the amide
`proton exchange rates and the extension of the helix towards
`the N-terminus of the peptide as described above. This result
`further emphasizes the stability of the helical self-associate
`and suggests a role as an intermediate toward the formation
`of higher aggregates of GLP-1.
`
`3.2. The size of the helical GLP-1 aggregate
`The size of the symmetric GLP-1 self-associate in TFE was
`evaluated by comparing its self-di¡usion coe⁄cient with that
`of monomeric GLP-1 using the PFG NMR self-di¡usion ex-
`periment, as described in Section 2. The identical experimental
`conditions that were applied in the two cases (1.4 mM GLP-1
`in D2O and 35% TFE, pH 2.5) and the identical nature of the
`two macromolecules to be compared ensure a reliable com-
`parison.
`The self-di¡usion coe⁄cient is sensitive both to the size and
`to the shape of the molecule, and the oligomerization of sev-
`eral well-characterized proteins has been studied using the
`approach [19,20] applied here. For the simplest dimerization
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 3
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`168
`
`X. Chang et al./FEBS Letters 515 (2002) 165^170
`
`Ds;dimer/Ds;monomer was estimated to be V0.85 using the sym-
`metric cylinder model where the self-di¡usion coe⁄cient is
`given by [21,22]
`Ds (cid:136) (cid:133)kBT=3ZR0L3(cid:134)(cid:133)lnp (cid:135) X(cid:134)
`Here kB, T, and R0 are the Boltzmann constant, the temper-
`ature, and the pure solvent viscosity, respectively, while L is
`the length and a the diameter of the cylinder. Further, Xis the
`end-e¡ect correction factor given by X= 0.312+0.565p313
`0.100p32, where p is the ratio between L and a. For the
`GLP-1 dimer the values L = 45 A(cid:238) and a = 7.4 A(cid:238) were applied.
`The latter value was obtained from the modeled GLP-1 coiled
`coil dimer (vide infra).
`
`(cid:133)1(cid:134)
`
`The results of the PFG self-di¡usion experiments performed
`on the monomeric and the aggregated GLP-1 are shown in
`Fig. 4. The obtained ratio of 0.85 (cid:254) 0.02 between the self-dif-
`fusion coe⁄cients, Ds, of the two forms is in good agreement
`with an extended dimer structure of the highly symmetric and
`stable GLP-1 aggregate indicated by the NMR spectra and
`the slowly exchanging amide protons. It is of interest here to
`note that also other amphipathic K-helical sequences have
`been reported to form highly stable, helical dimers [23,24].
`The results of the self-di¡usion measurements thus indicate
`that the highly stable and highly symmetric GLP-1 self-asso-
`ciate is a dimer consisting of two helical GLP-1 monomers
`with identical and completely overlapping signals. Moreover,
`lack of any intermonomeric NOEs, including NOEs between
`residues far from each other in the sequence, excludes an
`antiparallel arrangement of the two helical monomers of the
`dimer. Therefore, the only structure of the aggregated form
`that agrees with all of the experimental ¢ndings is a parallel
`
`Fig. 4. Measurement of self-di¡usion coe⁄cients. a: Stack plot of
`aggregated GLP-1 (1.4 mM, pH 2.5, 300 K) in D2O with 35% (v/v)
`TFE using water-sLED sequence. The signals used for the determi-
`nation of the self-di¡usion coe⁄cients are marked with an asterisk
`(*). A corresponding experiment was performed on monomeric
`GLP-1. b: Fit of the intensity variation in the water-sLED di¡usion
`experiments of monomeric (b) and dimeric coiled coil (a) GLP-1;
`the ¢ts correspond to the marked signal at the highest ¢eld. The
`data were analyzed using the equation [37] I = Ie+I0 exp(3Kx2),
`where x = G, G being the gradient strength; I0 and Ie are the nor-
`malized resonance intensities at zero and in¢nite time, respectively;
`K = (QN)2(v3N/3)Ds, where Q is the gyromagnetic ratio of the proton,
`N is the duration of the PFG, and v is the time between PFG
`pulses. The same values of v and N were used in both experiments.
`The K values obtained from the three signals were (in cm2 G32):
`0.0051 (cid:254) 0.0005, 0.0049 (cid:254) 0.0002, and 0.00465 (cid:254) 0.00008 for mono-
`meric GLP-1,
`and
`0.00398 (cid:254) 0.00015,
`0.00402 (cid:254) 0.00016,
`and
`0.0036 (cid:254) 0.0004 for dimeric coiled coil GLP-1. The corresponding
`weighted average of the K values from the three signals for each of
`the
`two
`forms
`of GLP-1 were
`0.004685 (cid:254) 0.000005
`and
`0.00397 (cid:254) 0.00008 cm2 G32 corresponding to the ratio Ds;b/Ds;a =
`0.85 (cid:254) 0.02, in close agreement with a dimeric coiled coil (see text).
`
`model, where the monomer^monomer interaction was ap-
`proximated by a hard-sphere molecular contact, the expected
`change in the self-di¡usion coe⁄cient, Ds, was estimated to be
`V25% (Ds;dimer/Ds;monomerW0.75). However, the shape of an
`aggregate consisting of extended single-strand GLP-1 helices
`[11] is more like a rod or a cylinder, and end-e¡ects that cause
`the molecule to di¡use at a slower rate must be taken into
`consideration. Thus, for a closely packed coiled coil, the ratio
`
`Fig. 5. Models for the coiled coil dimer of GLP-1. a: Helical wheel
`representation with the terminal residues H1 and R30 excluded.
`b: Side-chain hydrogen bond between the carboxylic acids of the
`two D9 that may impart the parallel orientation of the coiled coil.
`c: Interhelical packing model of the coiled coil showing the back
`bone and the CL carbons of a and d residues (prepared using MOL-
`MOL [38]).
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 4
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`X. Chang et al./FEBS Letters 515 (2002) 165^170
`
`169
`
`coiled coil dimer. As described below the parallel orientation
`is supported by a possible hydrogen bond between the carbox-
`ylic acid side-chains of D9 in the two monomers.
`
`3.3. Modeling the coiled coil dimer
`Coiled coil domains occur frequently in native protein (5^
`10% of all proteins). This supersecondary structure motif can
`consist of two to ¢ve K-helices wound around each other,
`allowing favorable side-chain meshing without much K-helix
`distortion. The coiled coil motifs found in oligomerized pro-
`teins share a characteristic heptad repeat (a^b^c^d^e^f^g) with
`hydrophobic interactions between a and d positions, and elec-
`trostatic interactions between e and g positions.
`The GLP-1 sequence ¢ts this pattern approximately (Fig.
`5a). Only the polar residues D9 and G16 are not usually
`found in position a in leucine zippers or ¢brous proteins
`and will, in general, destabilize the structure. On the other
`hand, buried polar residues appear to be important for the
`structure of coiled coils by favoring a speci¢c orientation of
`the strands through interstrand hydrogen bonding [25^27]. In
`the case of GLP-1 the D9 residue is uncharged at the exper-
`imental conditions (pH 2.5) since its side-chain carboxylic acid
`has a random coil pK value of 4.0 in the absence of any
`electrostatic interactions [28]. Therefore, the parallel packing
`of the two helices in the coiled coil GLP-1 dimer may be
`imparted by an interchain hydrogen bond between the side-
`chains of the two opposite D9 residues (Fig. 5b).
`The coiled coil fold has a relatively simple geometry, with
`only three unknown parameters ^ the pitch of the superhelix,
`the radial separation of the helices, and a phase angle, deter-
`mining which residues are on the interface. Due to this sim-
`plicity, various methods have been developed that can predict
`coiled coil backbone structures with an accuracy of about 1 A(cid:238) .
`Here, ThreadCoil [15] was used with the GLP-1 sequence to
`perform a wide grid search of the three coiled coil parameters,
`including the right-handed coiled coil possibility. The struc-
`ture with the lowest energy score was a left-handed coiled coil,
`with a pitch, radius, and phase of 136 A(cid:238) , 3.70 A(cid:238) , and 40‡,
`respectively. These values are quite similar to those for the
`leucine zippers. Thus, for example, the Jun homodimer has a
`pitch, radius, and phase of 147 A(cid:238) , 4.70 A(cid:238) , and 34‡, respec-
`tively [29]. The most signi¢cant di¡erence occurs in the super-
`coil radius, suggesting that the two helices are 2 A(cid:238) closer in
`GLP-1 than in the leucine zippers. This may be explained by
`the relatively small and £exible side-chains that characterize
`most of the a- and d-position residues, which allow the K-
`helices in coiled coils to pack more closely [30,31]. This sug-
`gestion is consistent with the very slowly exchanging amide
`protons observed for the GLP-1 dimer.
`The ThreadCoil energy score of the calculated coiled coil
`backbone structure of GLP-1 (30.3 kT) was signi¢cantly
`higher than that of leucine zippers (between 31 and 32.5
`kT) with the largest contribution stemming from two hydro-
`phobic residues (a leucine and a valine) that are solvent ex-
`posed. This indicates that not all of the hydrophobic residues
`of GLP-1 can be buried in a coiled coil structure, and hence
`that the structure is not likely to be stable in pure water but
`will lead to further aggregation. This is consistent with the
`experimental result obtained here that TFE is necessary for
`the stabilization of the helical structure.
`Starting from this initial backbone model, an all-atom mod-
`el was calculated as described in Section 2, In accordance with
`
`the experimental observation of the dimerized GLP-1 in solu-
`tion. The resulting dimer (Fig. 5c) consists of two parallel
`monomers wound around each other in a left-handed way.
`The symmetric, parallel arrangement of the two extended heli-
`ces is in agreement with the observation of only one set of
`resonances and the lack of long-range NOEs in the molecule.
`For the 15 amide protons, that are slowly exchanging only
`in the coiled coil dimer, the slow exchange can be attributed
`to their involvement in the helical hydrogen bonding, except
`for residue A2. The reason for the slow exchange of this res-
`idue is not immediately apparent, although side-chain hydro-
`gen bonding involving the amide proton of A2, and the loca-
`tion of the A2 residue on the dimer interface may limit the
`accessibility of solvent water to its amide proton.
`As mentioned above, most of the amide proton signals were
`still present after a month. However, the intensities of nine of
`these signals decreased signi¢cantly faster than those from the
`remaining amide protons. This clearly indicates a relatively
`higher solvent exposure of the nine residues. Seven of these
`residues (i.e. T7, S8, V10, S11, A18, E21, K28) are located at
`positions b, c, f and g in the helical wheel, i.e. outside the
`coiled coil dimer interface. The faster exchange rates of their
`amide protons thus support the conclusion that the GLP-1
`dimer is a coiled coil. Similar observations of a relatively
`faster amide proton exchanges rates for residues in positions
`b, c, and f have been observed for other coiled coil proteins
`[25,32,33]. The remaining two residues for which faster amide
`proton exchanges were observed, i.e. G29 and R30, are both
`at the C-terminal end and are located at positions g and a,
`respectively.
`Coiled coils may aggregate to higher-order structures [34].
`In the case of GLP-1 further aggregation could take place by
`intermolecular hydrophobic interactions between the hydro-
`phobic residues located at the b, c, and f positions. However,
`here the relatively large amount of TFE protects it against
`further aggregation. In accordance with this, Sykes and co-
`workers [35] reported that addition of 15% TFE decreased the
`dimerization and self-aggregation ability of muscle regulatory
`protein troponin C, without any signi¢cant e¡ect on the sec-
`ondary or tertiary structures. In a similar manner the high
`content of TFE does not destabilize the supersecondary coiled
`coil motifs, as observed here for GLP-1 in 35% TFE. How-
`ever, TFE prevents further aggregation of the peptide by less-
`ening the hydrophobic interaction of solvent-exposed non-po-
`lar side-chains [12].
`
`4. Conclusion
`
`The results here show that TFE stabilizes a highly symmet-
`ric, helical GLP-1 dimer, when aggregated GLP-1 is dissolved
`in TFE/water at pH 2.5, while disrupting the higher aggre-
`gates. The results indicate that the helical dimer is a parallel
`coiled coil. The formation of a helix and the absence of any
`tertiary structures are consistent with the general helix pro-
`moting e¡ect of TFE and its destabilizing e¡ect on the tertiary
`architecture of native proteins. The formation of a coiled coil
`dimer was indicated by (1) an increase of the number of
`slowly exchanging amide protons to more than twice the num-
`ber observed for the single-strand K-helix, and a decrease of
`the amide proton exchange rates by about two orders of mag-
`nitude, (2) insigni¢cant changes in chemical shifts and pat-
`terns of signals in the NMR spectra as compared with mono-
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 5
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`170
`
`X. Chang et al./FEBS Letters 515 (2002) 165^170
`
`meric helical GLP-1, (3) the size of the dimer as determined
`from its self-di¡usion coe⁄cient, and (4) molecular modeling.
`Although the coiled coil dimer has a propensity for further
`aggregation due to the hydrophobic side-chains that are lo-
`cated on its surface, the aggregation is prevented by the pres-
`ence of TFE molecules. Overall the results suggest that the
`dimeric coiled coil is an intermediate towards the formation of
`higher-order GLP-1 aggregates, e.g. ¢brils, in water.
`
`Acknowledgements: This work was ¢nancially supported by the Dan-
`ish Technical Research Council (J. Nos. 16-5028-1 and 9601137), the
`Danish Natural Science Research Council (J. 9502759, 9601648 and
`9801801), Direkt(cid:214)r Ib Henriksens Fond, Carlsbergfondet and Novo
`Nordisk Fonden. The 750 and 800 MHz NOESY spectra were ob-
`tained at The Danish Instrument Center for NMR Spectroscopy of
`Biological Macromolecules. We are grateful to Dr. S(cid:214)ren M. Kristen-
`sen for helpful discussions, and Dr. Jens Duus and Ms. Else Philipp
`for technical assistance.
`
`References
`
`[1] Drucker, D.J., Philippe, J., Mosjov, S., Chick, W.L. and Hab-
`ener, J.F. (1987) Proc. Natl. Acad. Sci. USA 84, 3434^3438.
`[2] Drucker, D.J. (1998) Diabetes 47, 159^169.
`[3] Adelhorst, K., Hedegaard, B.B., Knudsen, L.B. and Kirk, O.
`(1994) J. Biol. Chem. 269, 6275^6278.
`[4] Deacon, C.F., Knudsen, L.B., Madsen, K., Wiberg, F.C., Jacob-
`sen, O. and Holst, J.J. (1998) Diabetologia 41, 271^278.
`[5] Knudsen, L.B., Nielsen, P.F., Huusfeldt, P.O., Johansen, N.L.,
`Madsen, K., Pedersen, F.Z., Thogersen, H., Wilken, M. and
`Agerso, H. (2000) J. Med. Chem. 43, 1664^1669.
`[6] Xiao, Q., Giguere, J., Parisien, M., Jeng, W., St-Pierre, S.A.,
`Brubaker, P.L. and Wheeler, M.B.
`(2001) Biochemistry 40,
`2860^2869.
`[7] Kim, Y., Rose, C.A., Liu, Y., Ozaki, Y., Datta, G. and Tu, A.T.
`(1994) J. Pharmacol. Sci. 83, 1175^1180.
`[8] Sendero¡, R.I., Kontor, K.M., Kreilgaard, L., Chang, J.J., Patel,
`S., Krakover, J., He¡ernan, J.K., Snell, L.B. and Rosenberg,
`G.B. (1998) J. Pharmacol. Sci. 87, 183^189.
`[9] Lambert, W.J., Grucza, R.A., Stamper, G.F. and Chrunyk, B.
`(1994) Pharm. Res. 11, S-83.
`[10] Thornton, K. and Gorenstein, D.G. (1994) Biochemistry 33,
`3532^3539.
`[11] Chang, X., Keller, D., Bj(cid:214)rn, S. and Led, J.J. (2001) Magn.
`Reson. Chem. 39, 477^483.
`
`[12] Buck, M. (1998) Q. Rev. Biophys. 31, 297^355.
`[13] Altieri, A.S., Hinton, D.P. and Byrd, R.A. (1995) J. Am. Chem.
`Soc. 117, 7566^7567.
`[14] Gibbs, S.J. and Johnson Jr., C.S. (1991) J. Magn. Reson. 93,
`395^402.
`[15] O’Donoghue, S.I. and Nilges, M. (1997) Fold. Des. 2, S47^S52.
`[16] Sali, A. and Blundell, T.L. (1993) J. Mol. Biol. 234, 779^815.
`[17] Brooks, B., Bruccoleri, R., Olafson, B., States, D., Swaminathan,
`S. and Karplus, M. (1983) J. Comput. Chem. 4, 187^217.
`[18] Nilges, M. and Bru«nger, A.T. (1991) Protein Eng. 4, 649^659.
`[19] Dingley, A.J., Mackay, J.P., Chapman, B.E., Morris, M.B., Ku-
`chel, P.W., Hambly, B.D. and King, G.F. (1995) J. Biomol.
`NMR 6, 321^328.
`[20] Ilynia, E., Roongta, V., Pan, H., Woodward, C. and Mayo, K.
`(1997) Biochemistry 36, 3383^3388.
`[21] Eimer, W. and Pecora, R. (1991) J. Chem. Phys. 94, 2324^2329.
`[22] Tirado, M.M. and Garcia de la Torre, J. (1979) J. Chem. Phys.
`71, 2581^2587.
`[23] Brems, D.N., Plaisted, S.M., Kau¡man, E.W. and Havel, H.A.
`(1986) Biochemistry 25, 6539^6543.
`[24] Roongta, V., Powers, R., Jones, C., Beakage, M.J., Shields, J.M.
`and Gorenstein, D.G. (1989) Biochemistry 28, 1048^1054.
`[25] Junius, F.K., Mackay, J.P., Bubb, W.A., Jensen, S.A., Weiss,
`A.S. and King, G.K. (1995) Biochemistry 34, 6164^6174.
`[26] Lumb, K.J. and Kim, P.S. (1995) Biochemistry 34, 8642^8648.
`[27] Oakley, M.G. and Kim, P.S. (1998) Biochemistry 37, 12603^
`12610.
`[28] States, D.J. and Karplus, M. (1987) J. Mol. Biol. 197, 122^130.
`[29] O’Donoghue, S.I., King, G.F. and Nilges, M. (1996) J. Biomol.
`NMR 8, 193^206.
`[30] Gernert, K.M., Surles, M.C., Labean, T.H., Richardson, J.S. and
`Richardson, D.C. (1995) Protein Sci. 4, 2252^2260.
`[31] Nooren, I.M.A., Kaptein, R., Sauer, R.T. and Boelens, R. (1999)
`Nature Struct. Biol. 6, 755^759.
`[32] Goodman, E.M. and Kim, P.S. (1991) Biochemistry 30, 1615^
`1620.
`[33] Harbury, P.B., Plecs, J.J., Tidor, B., Alber, T. and Kim, P.S.
`(1998) Science 282, 1462^1467.
`[34] Creighton, T.E. (1993) Proteins: Structures and Molecular Prop-
`erties, Freeman, New York.
`[35] Slupsky, C.M., Kay, C.M., Reinach, F.C., Smillie, L.B. and
`Sykes, B.D. (1995) Biochemistry 34, 7365^7375.
`[36] Wishart, D.S., Sykes, B.D. and Richards, F.M. (1992) Biochem-
`istry 31, 1647^1651.
`[37] Stejskal, E.O. and Tanner, J.E. (1965) J. Chem. Phys. 42, 288^
`292.
`[38] Koradi, R., Billeter, M. and Wu«thrich, K. (1996) J. Mol. Graph.
`14, 51^55.
`
`FEBS 25913 28-3-02
`
`Novo Nordisk Ex. 2029, P. 6
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`